Analytical Methods for Problems of Molecular Transport
FLUID MECHANICS AND ITS APPLICATIONS
Volume 83 Series Editor:...

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Analytical Methods for Problems of Molecular Transport

FLUID MECHANICS AND ITS APPLICATIONS

Volume 83 Series Editor: R. MOREAU

MADYLAM Ecole Nationale Supérieure d'Hydraulique de Grenoble Boîte Postale 95 38402 Saint Martin d'Hères Cedex, France

Aims and Scope of the Series The purpose of this series is to focus on subjects in which fluid mechanics plays a fundamental role. As well as the more traditional applications of aeronautics, hydraulics, heat and mass transfer etc., books will be published dealing with topics which are currently in a state of rapid development, such as turbulence, suspensions and multiphase fluids, super and hypersonic flows and numerical modeling techniques. It is a widely held view that it is the interdisciplinary subjects that will receive intense scientific attention, bringing them to the forefront of technological advancement. Fluids have the ability to transport matter and its properties as well as to transmit force, therefore fluid mechanics is a subject that is particularly open to cross fertilization with other sciences and disciplines of engineering. The subject of fluid mechanics will be highly relevant in domains such as chemical, metallurgical, biological and ecological engineering. This series is particularly open to such new multidisciplinary domains. The median level of presentation is the first year graduate student. Some texts are monographs defining the current state of a field; others are accessible to final year undergraduates; but essentially the emphasis is on readability and clarity.

For a list of related mechanics titles, see final pages.

Analytical Methods for Problems of Molecular Transport by

I.N. IVCHENKO University of Missouri-Columbia, Columbia, Missouri, U.S.A.

S.K. LOYALKA University of Missouri-Columbia, Columbia, Missouri, U.S.A.

and

R.V. TOMPSON, JR. University of Missouri-Columbia, Columbia, Missouri, U.S.A.

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN-10 ISBN-13 ISBN-10 ISBN-13

1-4020-5864-0 (HB) 978-1-4020-5864-6 (HB) 1-4020-5865-9 (e-book) 978-1-4020-5865-3 (e-book)

Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands. www.springer.com

Printed on acid-free paper

All Rights Reserved © 2007 Springer No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

Dedications

This book is dedicated to the memory of my son Yaroslav.

I. N. Ivchenko This book is dedicated to the memory of Joel H. Ferziger, Heinz Lang, and Lloyd B. Thomas who inspired me to work in this area.

S. K. Loyalka This book is dedicated to my family; my parents, my siblings, my nephews and nieces, and especially to my wife Elena and my son Arseniy.

R. V. Tompson, Jr.

Special Dedication

This book is especially dedicated to the memory of our very dear friend and colleague

Igor Nikolaevich Ivchenko without whom this book would not have been possible.

S. K. Loyalka R. V. Tompson, Jr.

Contents

Table of Tables

xiii

Table of Figures

xv

Preface

xvii

Acknowledgments

xxiii

Chapter 1. The General Description of a Rarefied Gas 1. Some Introductory Remarks. 2. Density and Mean Motion. 3. The Distribution Function of Molecular Velocities. 4. Mean Values of Functions of Molecular Velocities. 5. Transport of Molecular Properties. 6. The Pressure Tensor. 7. The Hydrostatic Pressure. 8. The Amount of Heat. 9. The Kinetic Temperature. 10. The Equation of State for a Perfect Gas. 11. The Thermal Flux Vector. 12. Summary. Problems References

Chapter 2. The Boltzmann Equation 1. 2. 3. 4. 5. 6.

Derivation of the Boltzmann Equation. The Moment Equations. Another Form of the Moment Equations. The Equations for a Continuum Medium. Molecular Encounters. The Relative Motion of Two Molecules.

1 1 2 3 4 4 6 8 9 9 10 10 11 12 14

15 15 16 18 18 19 25

x

Analytical Methods for Problems of Molecular Transport Problems References

Chapter 3. The Collision Operator

27 28

29

1. The Differential and Total Scattering Cross Sections. 2. The Statistics of Molecular Encounters. 3. The Transformation of Some Integrals. Problems References

29 31 35 35 37

Chapter 4. The Uniform Steady-State of a Gas

39

1. The Boltzmann H-Theorem. 2. The Maxwellian Velocity Distribution. 3. The Mean Free Path of a Molecule. Problems References

Chapter 5. The Non-Uniform State for a Simple Gas 1. Expansion in Powers of a Small Parameter. 2. The First Approximation. 3. A General Formal Solution for the Second Correction. 4. The Transformation of the Non-Homogeneous Term. 5. The Second Approximation. 6. The First-Order Chapman-Enskog Solution for Thermal Conduction. 7. The First-Order Chapman-Enskog Solution for Viscosity. 8. The Thermal Conductivity and Viscosity Coefficients. 9. The First-Order Approximation for Arbitrary Intermolecular Potential. 10. The Second-Order Approximation for Arbitrary Intermolecular Potential. Problems References

Chapter 6. Regimes of Rarefied Gas Flows 1. The Knudsen Number. 2. A General Analysis of the Different Gas Flow Regimes. 3. The Boundary Conditions. 4. The Boundary Dispersion Kernel. 5. Features of the Boundary Conditions for Small Knudsen number. Problems References

Chapter 7. The Free-Molecular Regime 1. The Free-Molecular Distribution Function. 2. The Force on a Particle in a Uniform Gas Flow. 3. Calculation of Macroscopic Values in the Free-Molecular Regime. 4. Thermophoresis of Particles in the Free-Molecular Regime. 5. Condensation on a Spherical Droplet. 6. Non-Stationary Gas Flows. Problems References

39 40 42 44 51

53 53 55 56 57 59 62 64 66 67 68 70 73

75 75 76 77 80 82 85 88

91 91 96 98 101 106 109 111 139

Contents

Chapter 8. Methods of Solution of Planar Problems 1. Maxwell’s Method. 2. Loyalka's Method. 3. The Half-Range Moment Method. 4. Features of the Boundary Conditions for the Moment Equations. 5. Solution of the Thermal-Creep Problem by the Half-Range Moment Method. 6. Influence of the Boundary Models on the Thermal-Creep Coefficient. Problems References

Chapter 9. The Variational Method for the Planar Geometry 1. 2. 3. 4. 5. 6.

Another Form of the Boltzmann Equation. The Variational Technique for the Slip-Flow Problem. Discussion of the Slip-Flow Results. The Variational Solution for the Thermal-Creep Problem. Discussion of the Thermal-Creep Results. Slip-Flow and Temperature-jump Coefficients for the Lennard-Jones (6-12) Potential Model. Problems References

Chapter 10. The Slip-Flow Regime 1. Basic Equations. 2. The Spherical Drag Problem. 3. The Thermal Force Problem. Problems References

Chapter 11. Boundary Value Problems for All Knudsen Numbers 1. The Moment Equations in Arbitrary Curvilinear Coordinates. 2. The Two-Sided Maxwellian Distribution Functions. 3. Moments of Discontinuous Distribution Functions. 4. Analytical Expressions for the Bracket Integrals. 5. Boundary Conditions for Moment Equations. 6. Thermal Conduction from a Heated Sphere. 7. Method of the ‘Smoothed’ Distribution Function. 8. The Polynomial Expansion Method. 9. Solution of One Classic Transport Problem. 10. A Simplification of Moment Systems for Curvilinear Problems. 11. The Torque Problem. Problems References

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture 1. 2. 3. 4. 5. 6.

The First-Order Chapman-Enskog Approximation for a Binary Gas Mixture. The Transport Coefficients for a Binary Gas Mixture. The Second-Order Chapman-Enskog Approximation for a Binary Gas Mixture. Analytical Methods of Solution for Planar Boundary Value Problems Involving Binary Gas Mixtures. The Slip Coefficients for a Binary Gas Mixture. Discussion of the Slip Coefficient Results.

xi 141 141 149 155 162 165 168 169 178

181 181 183 188 192 196 200 203 208

211 211 214 218 224 237

239 239 242 245 247 252 254 259 260 265 270 272 277 286

289 289 291 294 300 304 308

xii

Analytical Methods for Problems of Molecular Transport Problems References

Appendix 1. Bracket Integrals for the Planar Geometry 1. Bracket Integrals Involving Two Sonine Polynomials. 2. Bracket Integrals Containing Several Components of Molecular Velocity. 3. Bracket Integrals Containing Two Discontinuous Functions. 4. Bracket Integrals Containing One Discontinuous Function. References

Appendix 2. Bracket Integrals for Curvilinear Geometries 1. The Special Function of the First Kind for the Spherical Geometry. 2. The Special Function of the Second Kind for the Spherical Geometry. 3. The Special Function of the First Kind for the Cylindrical Geometry. 4. The Special Function of the Second Kind for the Cylindrical Geometry. 5. Approximate Expressions for the Special Functions. References

Appendix 3. Bracket Integrals for Polynomial Expansion Method 1. Calculation of the Bracket Integrals of the First Kind. 2. Analytical Expressions for the Bracket Integrals of the Second Kind. References

Appendix 4. The Variational Principle for Planar Problems 1. Some Definitions and Properties for Integral Operators. 2. The Variational Principle. References

Appendix 5. Some Definite Integrals 1. 2. 3. 4.

Some Frequently Encountered Integrals. Some Integrals Encountered in Boundary Problems. Some Integrals Connected with the Second-Order Chapman-Enskog Solution. Some Integrals Connected with Non-Linear Transport Problems.

Appendix 6. Omega-Integrals for Second-Order Approximation

312 331

335 335 338 339 348 350

351 351 353 355 358 360 361

363 363 367 368

369 369 370 372

373 373 374 376 378

379

References

393

Author Index

395

Subject Index

399

Table of Tables

Table 2-1. Transformation of the various derivatives to new variables. 5-1. Functions for calculating the second-order transport coefficients. 8-1. Values of the slip-flow coefficient. 9-1. The slip-flow coefficient. 9-2. The slip-flow coefficient. 9-3. The velocity defect at the wall. 9-4. The velocity defect at the wall. 9-5. The dimensionless velocity defect. 9-6. The mean velocity. 9-7. The thermal-creep coefficient. 9-8. The thermal-creep coefficient. 9-9. The velocity defect at the wall. 9-10. The velocity defect at the wall. 9-11. The dimensionless velocity defect. 9-12. Numerical values of the velocity defect. 9-13. Velocity defect data obtained during the solution of Problem 9.8. 10-1. The isothermal-slip coefficient for different values of the tangential momentum accommodation coefficient. 10-2. Comparison of experimental and theoretical values of drag on a sphere. 10-3. The reduced thermal force. 10-4. The reduced thermal force on NaCl aerosol particles. 11-1. Expressions for some bracket integrals. 11-2. Values of the reduced heat flux ratio. 11-3. Values of the reduced heat flux ratio. 11-4. Analytical and numerical values of reduced torque on a rotating sphere.

page 17 70 165 190 190 191 191 192 192 197 197 198 198 199 199 207 217 218 222 222 264 265 269 277

xiv

Analytical Methods for Problems of Molecular Transport

12-1. Values of the transport coefficients and of the slip coefficients for a selection of binary gas mixtures obtained using the rigid-sphere potential model and the firstand second-order Chapman-Enskog approximations. 12-2. Values of the transport coefficients and of the slip coefficients for a selection of binary gas mixtures obtained using the Lennard-Jones (6-12) potential model and the first- and second-order Chapman-Enskog approximations. 12-3. Relevant parameters for two of the most commonly used intermolecular potential models; the rigid-sphere model and the Lennard-Jones (6-12) potential model. 12-4. The tangential momentum accommodation coefficients in a binary gas mixture chosen for calculation of the slip coefficients and slip factors. 12-5. A comparison of diffusion-slip coefficient values. A-1. Values of the J-integrals for use in evaluating bracket integrals containing two discontinuous functions. A-2. Values of the I-integrals for use in evaluating bracket integrals for planar problems. B-1. Numerical values of the spherical special functions of the first- and second-kind. B-2. Numerical values of the cylindrical special functions of the first- and second-kind. B-3. Numerical values of the integrals of the cylindrical special functions of the firstand second-kind. C-1. Values of the bracket integrals of the first-kind for the polynomial expansion method. C-2. Analytical values of the bracket integrals of the second-kind for the polynomial expansion method. E-1. Values of the commonly encountered i-integrals. F-1. Values of some reduced omega-integrals. F-2. A Mathematica® program to compute the reduced omega-integrals for selected values of the reduced temperature.

309

310 311 312 322 342 348 353 357 359 366 368 374 382 386

Table of Figures

Figure 1-1. Passage of molecules across an arbitrary surface element. 1-2. Components of the pressure tensor. 2-1. Geometry of a generic encounter. 2-2. Parameters of a generic encounter. 2-3. The spherical coordinates of the relative velocity. 3-1. The scattering solid angle. 3-2. Encounters of rigid-sphere molecules. 4-1. The geometry to be used in Problem 4.8 in determining the rate at which molecules reflected from one differential surface element cross another. 6-1. The variation of the mean velocity of a gas near a moving wall. 6-2. The variation of the gas temperature near a wall. 6-3. The thermal-creep geometry. 7-1. The cone of influence for a sphere. 7-2. The geometry of the sphere drag problem. 7-3. The geometry of the thermal force problem. 7-4. The geometry to be used for Problem 7.7 in determining the force on a round disk perpendicular to a uniform gas flow. 7-5. The geometry to be used for Problem 7.8 in determining the drag on a plate parallel to a steady, free-molecular gas flow. 7-6. The geometry to be used for Problem 7.9 in determining the drag on a plate oriented at an angle to a steady, free-molecular gas flow. 7-7. The geometry to be used for Problem 7.10 in determining the force on a round disk in a rarefied gas when the two sides of the disk have different temperatures. 7-8. The geometry to be used for Problem 7.11 in determining the thermophoretic force on a thin disk located in a rarefied gas having a constant temperature gradient.

page 5 7 21 22 23 31 32 47 82 84 85 93 95 102 116 117 118 120 120

xvi

Analytical Methods for Problems of Molecular Transport

7-9. The geometry to be used for Problem 7.12 in determining the number density behind a round disk perpendicular to a free-molecular gas flow in which the mean gas velocity is much less than the thermal molecular velocity. 7-10. The geometry to be used for Problem 7.13 in determining the number density of a free-molecular flow exiting through a round hole. 7-11. The geometry to be used for Problem 7.14 in determining the number density of an effusing gas at large distances from a small slot. 7-12. The geometry to be used for Problem 7.16 in determining the frictional force on a plate due to the horizontal motion of a second, parallel plate. 7-13. The geometry to be used in Problem 7.22 in determining the gas temperature between two parallel plates in the free-molecular regime. 7-14. The geometry to be used for Problem 7.25 in determining the torque on a flat, rotating disk in the free-molecular regime. 8-1. The geometry of slip problems. 8-2. The geometry to be used in Problem 8.1 in determining the mean velocity profile, the pressure tensor component, and the apparent viscosity for Couette flow. 8-3. The geometry to be used in Problem 8.5 in determining the temperature distribution and apparent thermal conductivity coefficient between parallel plates having different temperatures. 9-1. The dimensionless, mean velocity profile. 9-2. The dependence of the slip coefficients on reduced temperature for the LennardJones (6-12) potential model. 9-3. The dependence of the temperature-jump coefficient on reduced temperature for the Lennard-Jones (6-12) potential model. 10-1. Comparison of the experimental and theoretical values of the drag on a sphere. 10-2. A comparison of experimental and theoretical reduced thermal forces for NaCl. 10-3. The geometry to be used in Problem 10.8 in determining the resistance force on a flat, round disk in the slip-flow regime due to its approach to an identical, parallel, stationary disk. 10-4. The geometry to be used for Problem 10.13 in determining the net mass transfer in a closed capillary loop with different radii and end temperatures. 10-5. The geometry to be used in Problem 10.14 in determining the torque on a rotating sphere in the slip-flow regime. 10-6. The geometry to be used for Problem 10.16 in determining the average speed of a steady gas flow between two, parallel, planar surfaces in the slip-flow regime. 11-1. The cone of influence in cylindrical coordinates. 11-2. A comparison of reduced heat flux values 11-3. A comparison between reduced heat flux data for Argon and analytical results. A-1. The relationship between various collision angles. B-1. Dependence of the spherical special functions of the first- and second-kind on the radial coordinate. B-2. Dependence of the cylindrical special functions of the first- and second-kind on the radial coordinate.

122 123 124 126 132 135 142 170

174 200 201 202 219 223

228 233 234 236 244 266 270 344 354 358

Preface

The transport of a given species (atoms, molecules, neutrons, photons, etc.), either through its own kind or through some other host medium, is a problem of considerable interest. Practical applications may be found in many technologically and environmentally relevant areas such as the transport of neutrons in a nuclear power reactor or in a nuclear weapon, the transport of ions and electrons in plasma, the transport of photons which constitutes radiative heat transfer in various industrial, environmental and space applications, the transport of atoms or molecules of one species either through itself or as one component of a multi-component gas mixture, and the interactions of such gas mixtures with various solid and liquid surfaces such as one might find associated with capillary tubes, aerosol particles, interstellar dust grains, etc.. These application areas are obviously quite broad and it is readily apparent that there are, indeed, few scientific activities that do not require some level of understanding of transport processes. One of the most important and influential texts in the area of transport theory has been The Mathematical Theory of Non-Uniform Gases by Sidney Chapman and T.G. Cowling that was first printed in 1939. This book, along with several other more recent texts (Hirschfelder, J.O., Curtiss, C.F. and Bird, R.B., Molecular Theory of Gases and Liquids, John Wiley and Sons, NY, 1954; Kogan, M.N., Rarefied Gas Dynamics, Plenum Press, NY, 1969; Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases, North-Holland, Amsterdam-London, 1972; Cercignani, C., Theory and Application of the Boltzmann Equation, Scottish Academic Press, Edinburg, UK, 1975.), have provided students and researchers in the area of transport theory with an excellent basis for pursuing in-depth research in the area and have made possible some limited, albeit very useful,

xviii

Analytical Methods for Problems of Molecular Transport

applications of the theory. These texts have focused primarily on the development and simplification of the Boltzmann equations and on the derivation of general expressions for the coefficients associated with the transport phenomena of viscosity, thermal conduction, and diffusion in essentially unbounded gases. In recent years, however, it has been recognized that not only do intermolecular collisions within the bulk of the gas play an important role in determining the various transport coefficients but that interactions at boundaries are also significant. Hence, a simultaneous understanding of both is required in order to obtain reasonably complete descriptions of, and ultimately solutions to, any given transport problem. The relative importance of a boundary in a given transport problem is generally specified by Knudsen number, Kn O d , where O represents the mean free path of the species of interest in the host medium while d is some characteristic dimension of the system. Thus, in transport problems in general, there exist several somewhat arbitrarily defined regimes of interest. These cover the range from Kn 1 where a given boundary exerts a strong influence on the transport processes to Kn 1 where a given boundary exerts an essentially negligible influence on the transport processes. Accounts of the effects of boundaries on transport problems have been discussed in a variety of texts on subjects such as Neutron Transport Theory, Radiative Heat Transfer, and the Kinetic Theory of Gases. These, and the basic transport theory books mentioned above, are generally written at a level suitable only for specialists in the field or very advanced students. Thus, even though many have tried to use them as teaching texts, their application in this capacity has remained somewhat constrained with a clear need continuing to exist for a more student-friendly text having a clearer emphasis on basic problem solving. At the University of MissouriColumbia, we became clearly aware of this need in the Winter of 1992 when one of us (INI), as a part of teaching obligations associated with his status as a visiting professor to Columbia from Russia under the Fulbright exchange program, taught a class on the subject of transport theory. Based on this experience, we felt that the time had arrived for a new text that would combine a concise, but basically complete, introduction to the field of transport theory with a fairly tight focus on a few recently successful analytical solution techniques. Further, we recognized that any successful text of this type would, of necessity, include a good selection of easily applicable, representative problems that were either fully solved or which included sufficient basic guidance to assist the problem solver in reaching the correct solution. In this context, we have been strongly influenced by the organization and style of the excellent series of texts by Landau and Lifshitz.

Preface

xix

The subject material in the book divides itself quite naturally into two parts: the non-equilibrium properties of an infinite expanse of gas and those of a bounded gas. The first part (Chapters 1-5) contains the basis of the Kinetic Theory of Gases developed along the traditional lines of Chapman and Cowling. Some simplifications have been made during this development but only to facilitate the reader’s understanding of the basic principles of the kinetic theory. For example, the derivation of the Boltzmann collision operator is substantially simpler than that described in the above mentioned texts. The analysis presented in these chapters is based on the framework of the scattering cross section being the basic quantity characteristic of molecular encounters. In description of the ChapmanEnskog method, we focus our attention on the first-order approximation for rigid-sphere molecules to facilitate one’s understanding of this method. Nevertheless, the second-order Chapman-Enskog solutions along with tabulated values of all quantities necessary for their complete evaluation, are also provided for the Lennard-Jones (6-12) intermolecular potential as this development is potentially very important for practical applications. The presentation of this material here is in the most complete form currently available. The second part of the book (Chapters 6-12) presents focused descriptions of a limited selection of analytical methods that may be used to solve boundary value problems of transport with reasonable accuracy. The typical features found in the statements of boundary value problems for various gas flow regimes are discussed in Chapter 6. Within the framework of the traditional Maxwellian boundary model, new expressions for the reflected distribution functions at surfaces are presented which are then adapted to numerous applied problems of a practical nature. The physical features associated with statements of the boundary conditions for condensable and non-condensable gases are discussed and appropriate definitions for the basic accommodation coefficients are described in detail. Chapter 7 contains a kinetic theory treatment of gas transport problems for extremely high Knudsen numbers. The theory presented here starts with a Boltzmann equation that takes its simplest form due to the absence of the collision operator. Under stationary conditions, the constancy of the distribution function along molecular trajectories results in its discontinuity in velocity space; an effect that may be described in terms of the classic ‘cone of influence.’ For the spherical geometry, a new analytical representation of the distribution function in the full velocity space is given. This representation is a generalization of that previously proposed for the planar geometry. A number of basic problems are considered which show, in detail, the features of a mathematical solution technique that is presented

xx

Analytical Methods for Problems of Molecular Transport

here, in monographic form, for the first time and which otherwise could only be found by perusing the original journal articles in which it was developed. The next two chapters, Chapters 8 and 9, present selected methods of solution of planar transport problems. Detailed analytical solutions of some applied problems are given for the special case when molecules can be approximated as rigid spheres. In our description of selected methods, we have concentrated on various moment approaches in which we have emphasized the importance of using conservation laws to construct exact moment solutions. In Chapter 8, our development of the various moment approaches starts with the classic Maxwellian method and proceeds on to a generalized form of the Maxwellian method, which we term the Loyalka method after the one of us (SKL) who developed it. A key feature of this latter method that we emphasize is the opportunity that it represents to construct an accurate theory of slip phenomena for arbitrary models of the intermolecular potential. We follow these with a description of the characteristic features of the half-range moment method which, in combination with the mathematical solution technique described in Appendix A, is presented here in monographic form for the first time. Also in Chapter 8 we discuss in detail various boundary models and introduce a new model that allows for the incorporation of two conservation laws. We continue our discussion of planar transport problems in Chapter 9 with a discussion of the variational method which is followed by a discussion of the accuracy of all the various solution methods described for planar transport problems. Features of the integral representation of the distribution function and of the construction of basic functionals for slip problems are discussed in detail. We touch briefly on a new approximate method (Prob. 9.8) that may be used to describe gas behavior in the Knudsen layer. This comparative analysis of various approximate methods and estimates of their accuracy are compiled here in monographic form for the first time allowing the reader to make easy comparisons between the various methods. Chapter 10 contains straightforward applications of the slip-flow theory to various classical problems of interest. Derivations of the basic equations and statements of the boundary conditions in the slip-flow regime are discussed as well. While the subject treated here may be found in various places in the literature, several gaps in the subject have been bridged in order to make possible a systematic description of the various linearized problems. Chapter 11 contains original material that has been worked out only recently and which, in some cases, is still under development (such as the sphere drag problem). In many cases, methods and results for arbitrary Knudsen number and different geometries are described for the first time in such a complete manner. For instance, analytical expressions for the bracket

Preface

xxi

integrals associated with each transport problem as well as numerical and various approximate methods of calculation of these bracket integrals for different curvilinear geometries are original and have not been reported previously. The authors were the first to suggest the use of the ChapmanEnskog solutions as the molecular properties used to construct moment systems. This approach yields a very simple procedure that may be used to calculate bracket integrals for arbitrary intermolecular potentials. The advantage of this method is shown in analyses of some classical spherical transport problems such as the torque and thermal conduction problems. In Chapter 12 the analytical methods presented earlier for the Maxwell and Loyalka methods are again used to solve planar boundary value problems involving the various boundary slip phenomena. Here, however, the emphasis is on binary gas mixtures where the preceding chapters have focused only on simple one-component gases. The relevant boundary slip phenomena are discussed in detail and the accuracy of the various methods is evaluated further in the context of gas mixtures. This chapter also contains all of the material necessary to completely specify the transport coefficients for a binary gas mixture for both the first- and second-order Chapman-Enskog approximations. In addition to the material in Chapters 1-12, a number of appendices have been included to facilitate the ease of the reader in finding and using such information as is necessary to effectively make use of the analytical methods described in this book. Appendices A-C contain descriptions of analytical methods needed to evaluate the necessary bracket integrals associated with the various transport problems in both the planar and curvilinear geometries. Of particular importance, the methods of calculation of bracket integrals containing the discontinuous distribution function have all been collected together in one monograph for the first time. Appendix D provides additional information about the variational principle for use in planar transport problems. Here, the reader can find greater detail regarding the general principles behind the construction of the necessary variational functionals needed for specific boundary value transport problems. Appendix E contains an extensive listing of definite integrals that are most frequently encountered in boundary value transport problems and Appendix F contains tabulated numerical values of the : -integrals that are encountered when using the Lennard-Jones (6-12) potential model. This set of tabulated : -integrals is the complete set that it is necessary to use when employing either the first-order or the second-order Chapman-Enskog approximations. The problems following each chapter may be useful to both students and instructors. While all of these problems have been solved, sometimes only an outline of the solution is given and the reader must supply the details.

xxii

Analytical Methods for Problems of Molecular Transport

The authors recommend that students of the subject seek to obtain their own solutions to these problems in order to gain familiarity with the techniques. In addition to basic experience, this will also help students of the subject to understand some of the subtleties of the analytical methods described. While some of the problems presented at the ends of the chapters only require the reader to utilize basic information that may be readily obtained in the necessary form directly from the book, other problems will require the reader to engage in some significant extension and generalization of the basic material presented in the book. This book is designed to serve a dual function. It is intended that it be capable of serving as a teaching instrument, either in a classroom environment or independently, for the study of basic analytical methods and mathematical techniques that may be used in the Kinetic Theory of Gases. It is primarily suitable for use in graduate level physics and engineering courses on the Kinetic Theory of Gases. This book should also prove to be useful as a reference for scientists and engineers working in the fields of Rarefied Gas Dynamics and Aerosol Mechanics. In addition, the material in this book may prove to be of interest to individuals working in the areas of Physical Chemistry, Chemical Engineering, or any other applied discipline in which gas-surface interactions can be expected to play a significant role. INI SKL RVT

Acknowledgments

The authors gratefully acknowledge the invaluable assistance that they have received from many sources including: Dr. Robert L. Buckley for extensive help in typing and proof reading. Dr. James L. Griffin for Mathematica® assistance in some calculations. Professor Mikhail N. Kogan for helpful discussions relating to arbitrary Knudsen number in Chapter 11. Professor Yuri I. Yalamov for helpful discussions relating to boundary conditions for the slip-flow regime in Chapters 8 and 10. Dr. David Gabis for discussions relating to the spinning sphere problem at the end of Chapter 10. Dr. Perapong Tekasakul for help with figures and equations. Mr. Earl L. Tipton for extensive help with Adobe Illustrator® and for proof reading the final version. Mr. Ryan Meyer, Mr. Zeb Smith, and Mr. Earl L. Tipton for contributions made in preparing the omega-integral program of Appendix F. The Council for the International Exchange of Scholars (CIES) for the Fulbright grant that supported INI during his first visit to Columbia Missouri, 1/15/92-7/4/92 and which sponsored two subsequent visits, 8/25/92-1/15/93 and 7/29/93-2/18/94. The U.S. National Aeronautics and Space Administration (NASA) for funding from grant NAG-3-1420 that was used to support INI during his second and third Fulbright visits to Columbia as well as during his fourth visit, 7/01/94-3/15/95. This manuscript was initially prepared on a Macintosh® IIci using System 6.0.5. The text was initially formulated using Microsoft Word® (ver.

xxiv

Analytical Methods for Problems of Molecular Transport

4.0). Equations were initially prepared using MathType® (ver. 2.03), plots using Cricket Graph® (ver. 1.3.2), and figures using MacDraft® (ver. 1.2b). The manuscript was subsequently converted to Microsoft Word® (ver. 2002) employing MathType® (ver. 5.2a) on a Dell Dimension® 2400 configured with Microsoft Windows® XP (ver. 2002 including Service Pack 2). Plots were converted to Microsoft Excel® (ver. 2002) and figures were converted to Adobe Illustrator® CS (ver. 11.0.0). Omega-integral calculations in Appendix F were performed using Compact Visual Fortran® 77 (ver. 6.4) and Mathematica® (ver. 5.2) on a number of different machines.

Chapter 1 THE GENERAL DESCRIPTION OF A RAREFIED GAS

1.

SOME INTRODUCTORY REMARKS.

Systems consisting of what are usually called rarefied gases will be considered in this book. Such systems contain large numbers of molecules; specifically, about 2.7 u 1019 cm–3 for a gas at standard temperature and pressure (STP) defined to be 0 °C (273.15 K) and 1.0 atm (760 torr). How may the behavior of such huge collections of particles be described? If one were to try to apply the methods of classical mechanics, one would have to construct and solve a system of equations of motion containing as many equations and initial conditions as there are numbers of interacting particles. Obviously, such a problem could not be solved today even with the help of the most advanced computers. On this basis, one might naively suspect that the greater the number of particles, the more difficult the problem. This is not strictly true, however, as it turns out that useful results can be obtained by applying statistical descriptions to such systems. The state of a gas can be analyzed by employing statistical laws which allow one to determine average values for the different macroscopic quantities that characterize the behavior of the gas. It has been proven in statistical mechanics [1-3] that the relative fluctuations of additive quantities (i.e. quantities whose values for the body as a whole are equal to the sum of the values for its separate parts) are proportional to N 1 2 , where N is the number of molecules of the gas. In accordance with this theorem, the additive quantities are really equal to their average values to an extremely high degree of accuracy and, therefore, deviations of the actual quantities

2

Analytical Methods for Problems of Molecular Transport

from the average values do not have a practical influence on the trustworthiness of the statistical description. The mean values mentioned above may be found by using a probability framework in which the main interest is in the statistical distribution function for the molecules. It is very important to note that the extreme accuracy of this type of probabilistic analysis, due to the relatively small deviations of the macroscopic quantities from their mean values, is much greater than the accuracy of actual experimental measurements and, hence, small deviations of the macroscopic quantities from their mean values are typically neglected.

2.

DENSITY AND MEAN MOTION.

First, some macroscopic quantities for a ‘simple’ gas composed entirely of identical molecules [4-6] will be determined. Let the mass of any molecule be m and let dr denote a small volume element surrounding the point, r . This volume element is assumed to be large enough to contain a great number of molecules while still possessing dimensions small compared to the scale of variation of the macroscopic quantities of interest. Let the mass contained in dr be averaged over a time interval, dt , which is long compared to the average time needed for a molecule to traverse dr yet short compared to the scale of the time variations in the macroscopic properties of the gas. Then, the average value of the mass contained by dr will be proportional only to its volume and will not depend upon its shape. This mass will be denoted by U dr , where U r ,t may be termed the mass density (or typically just ‘density’) of the gas at r ,t . The number density of the gas, n r , t , is identified by analogy. These densities are connected by the relationship:

U r , t mn r , t , where m is the mass of a single gas molecule. If the velocity of a molecule is denoted by v , then the mean velocity of a gas at r ,t may be denoted by u r ,t which is defined by the vector equation:

ndr u ¦ v

,

3

Chapter 1. The General Description of a Rarefied Gas

where the summation over v on the right-hand-side extends over all of the molecules in dr , and both ndr and ¦ v are averaged over the time interval, dt . The translational motion of an individual molecule in dr may be specified either by its ‘actual’ velocity, v , relative to some standard frame of reference, or by its velocity, V c , relative to axes moving with a velocity, uc , so that:

Vc

v uc .

If uc u , then the quantity, V molecule.

3.

v u , is called the peculiar velocity of the

THE DISTRIBUTION FUNCTION OF MOLECULAR VELOCITIES.

Within a statistical framework, the state of a rarefied gas is described by the distribution function, f v, r , t , so that the quantity, fdvdr , gives the probable number of molecules which, at time, t , are situated around r in the volume element, dr , and have velocities near v lying in the range, dv . It is also important to note that fdvdr is the number of molecules in dvdr averaged over the time interval, dt . The distribution function may be specified as the probability density of points in the phase space (i.e. the sixdimensional space of r and v ) of a molecule. The description of a gas by its distribution function is excessively detailed. For practical purposes this amount of detail is not always needed. For instance, the gas as a continuum may be characterized by various macroscopic quantities which may be experimentally measured. These quantities may be considered to be moments of the distribution function and these moment relationships will be discussed in detail later. It is clear that the number density may be obtained by integrating the distribution function throughout the entire velocity space. This integration may be performed over either the actual or peculiar velocities of the molecules since the distribution of velocity points is unaffected if the origin in the velocity space is shifted to the point, u . The same integration concept also applies with respect to all similar integrals over the distribution function (moments of the distribution function). Hence, the number density may be expressed in the form:

n r , t

³ f v, r , t dv ³ f V u, r , t dV

.

(1-1)

4

Analytical Methods for Problems of Molecular Transport

4.

MEAN VALUES OF FUNCTIONS OF MOLECULAR VELOCITIES.

Let I v, r ,t be any function of the molecular velocity, v , position, r , and time, t . This function will be called a molecular property. The mean value of I v, r ,t in the volume element, dr , is defined by the relation:

ndr 1 ¦ I ,

I

where ¦ I denotes the time average during dt of the sum of the values of I v, r , t for all the molecules in dr and can be expressed in terms of the distribution function, f v, r , t . For this purpose, consider the set of molecules in the volume element of the phase space, dvdr . Molecules of this set all have the same velocity and thus their contribution to the sum, ¦ I , is I fdvdr . The net contribution of all molecules in the volume element, dr , may then be expressed in the form:

¦I

dr ³ I fdv .

Hence I , the mean value of any molecular property, I v, r ,t , may be represented by a moment of the distribution function, i.e.:

I

n 1 ³ I fdv .

(1-2)

In particular, the mean velocity of a gas is given by:

u

n 1 ³ vfdv .

(1-3)

From Eq. (1-3), it then follows directly that the mean value of the peculiar molecular velocity is equal to zero, i.e.:

V

5.

v u

v u 0 .

TRANSPORT OF MOLECULAR PROPERTIES.

Consider the passage of molecules, as shown in Fig. 1-1, across a small element of surface, dS , moving in the gas with arbitrary velocity, uc . Let n be a unit vector positioned normal to the surface element and pointing in

5

Chapter 1. The General Description of a Rarefied Gas

Figure 1-1. Passage of molecules across an arbitrary surface element,

dS .

the direction from the negative side to the positive side. The velocity, V c , of a molecule relative to dS is equal to v uc , or V u uc . In the Kinetic Theory of Gases, a standard analysis is usually employed to calculate the fluxes of the various molecular properties. First, the contribution to the flux across the surface element, dS , is determined for each separate velocity group of molecules. Then, the net flux is calculated by integrating over all velocity groups. Consider the molecules of one velocity group in dr whose peculiar velocities lie in the range, dV . Let a molecule cross the element dS in a time dt which is so short that the possibility of molecular encounters may be ignored. In this case, the molecule must lie somewhere inside the region having dS as the base and having a length and direction determined by V c dt . Thus, if dr denotes the volume of this region, the number of molecules crossing dS during dt is fdVdr . From the above description, it follows that dr r V c cos T dSdt where T is the angle between V c and n and the sign, plus or minus ( r ), is chosen so as to make the expression for dr positive. Thus, the flux of I V for

6

Analytical Methods for Problems of Molecular Transport

this molecular group is expressed, both in magnitude and in sign, by the following relation:

I V Vnc f V dVdSdt .

(1-4a)

The net flux is then the integral over all of the velocity groups and is written as:

dSdt ³ I V Vnc f V dV .

(1-4b)

The rate of flow of property I per unit area is expressed in the form:

nVncI V .

(1-4c)

This represents the n -component of the corresponding vector quantity: n V cI V .

(1-5)

Substituting the definition of V c into Eq. (1-5), one can obtain: n V cI V n VI V n u uc I V .

(1-6)

The vector n VI V is typically termed the flux density vector for the molecular property, I V , while the second term on the right in Eq. (1-6) is connected with the number flow of molecules across dS ; each of which possesses the mean value of the property, I V .

6.

THE PRESSURE TENSOR.

Let I V be equal to some component of the molecular momentum, mv . At the boundary of some containment vessel, every molecule that rebounds from the surface imparts momentum to it. These impacts simulate a continuous force on the surface equal to the rate at which momentum is imparted to the surface. The force per unit area on the surface is called the pressure which is a vector that is not necessarily normal to the surface. Let an element of the surface, dS , move with a velocity, uc , and let the direction pointing into the gas from the surface be taken as negative. The pressure on this boundary surface element, Pn , is equal to the total rate of

7

Chapter 1. The General Description of a Rarefied Gas

Figure 1-2. Components of the pressure tensor. Here, the point, the center of the cube face nearest to the observer.

M , is physically located at

flow of the molecular momentum, mnv , per unit area. Using Eq. (1-4c), one obtains [4]:

Pn

(1-7)

UVncv .

If the surface is impenetrable to the gas molecules, then the mean velocity relative to the wall has no component normal to the wall. This may be expressed as: (1-8)

Vnc 0 . Using this condition, one can perform the following transformations:

Vncv Vnc V u

ª¬n V u uc º¼ V

n V V

,

after which the following expression for Pn may be obtained:

Pn

UVn V n U VV n P .

(1-9)

8

Analytical Methods for Problems of Molecular Transport

Here, P is a symmetrical tensor defined by:

P

U VV

V2 V V V V ½ x y x z ° x ° ° ° 2 U ®VyVx Vy VyVz ¾ . ° 2 ° °¯VzVx VzVy Vz °¿

(1-10)

It is important to note that this tensor depends only upon the distribution of the peculiar velocities. The same analysis may be used to define the pressure distribution at any point, M , within the gas. Let the surface element, dS , containing M share the mean motion of the gas so that uc u, V c V 0 and, therefore, Vnc 0 . Since the basic condition for deriving Eq. (1-9) is satisfied in this particular case, the pressure distribution across the surface element, dS , is defined by the same pressure tensor, P , given by Eq. (1-10). The physical sense of each pressure tensor component is illustrated in Fig. 1-2.

7.

THE HYDROSTATIC PRESSURE.

The sum of the normal pressures at the point, M , across the three planes parallel to the coordinate planes, is given by:

Pxx Pyy Pzz

UV 2 .

The mean of the normal pressures across these planes is called the hydrostatic pressure, p , or the pressure at M . Immediately from this definition, one can obtain:

p

1 3

Pik G ik ,

(1-11)

where G i k is the unit symmetrical tensor whose components are given by:

Gi k

1 ; i k , ® ¯0 ; i z k .

If Pik 0 when i z k and the diagonal elements are equal, the pressure, p , may be expressed in the form:

9

Chapter 1. The General Description of a Rarefied Gas

p

Pxx

Pyy

Pzz ,

(1-12a)

or:

Pik

pG ik .

(1-12b)

In this case, the pressure vector may be written as Pn i Pik nk pni and, therefore, the pressure on any surface element through the point, M , is normal to the surface and is independent of the orientation of the surface.

8.

THE AMOUNT OF HEAT.

The kinetic energy of translational motion of molecules in the volume element, dr , at time, t , is:

ndr 12 mv 2 , which may be expressed in the form:

dr 12 U u 2 V 2 .

(1-13)

Here, the first term is the kinetic energy associated with the motion of the volume element which has a net mass. The second term is the kinetic energy associated with the peculiar motions of the molecules in the volume element and which is one component of the thermal energy of the gas in the volume element (and is the only component of the thermal energy for monatomic gases). This energy is communicable between molecules at encounters. It should be noted that there are also other forms of communicable energy that occur in polyatomic gases. In this case, the total heat energy, E , of a molecule is the sum of its peculiar kinetic energy and any of these other forms of communicable energy that may be present.

9.

THE KINETIC TEMPERATURE.

In the Kinetic Theory of Gases, a temperature, T , is defined directly by the relationship:

10

Analytical Methods for Problems of Molecular Transport 1 2

mV 2

3 2

kT ,

(1-14)

where k 1.38066 u 1016 erg K–1 is Boltzmann’s constant. It is important to note that the temperature definition given by Eq. (1-14) is more general than that employed in Thermodynamics and Statistical Mechanics where only equilibrium states are considered. Eq. (1-14) can be used for non-steadystates of a gas as well.

10.

THE EQUATION OF STATE FOR A PERFECT GAS.

From the definitions of the hydrostatic pressure and the temperature given in Eqs. (1-11) and (1-14), one can obtain the following relationship between the pressure and the temperature:

p

nkT .

(1-15)

This equation is called the equation of state for a perfect gas.

11.

THE THERMAL FLUX VECTOR.

The thermal flux vector, Q , is an important quantity that expresses the rate of flow of heat energy across a unit surface element. This vector may be obtained from Eq. (1-6) by substituting the heat energy of a molecule (the energy which is communicable between molecules at encounters) in place of the generic molecular property, I V . After this substitution, one obtains:

Q

nEV .

(1-16)

If a molecule possesses only kinetic energy, then this heat flux vector is defined by:

Q

n 12 mV 2 V .

(1-17)

Chapter 1. The General Description of a Rarefied Gas

12.

11

SUMMARY.

Rarefied gases, since they are systems containing a huge number of molecules, can only be realistically described within a statistical framework. The probabilistic basis of this description is not readily apparent in many practical transport problems, however, because the accuracy of the results predicted by the Kinetic Theory of Gases is much greater than the realistically achievable accuracy in actual experimental measurements and because the relative deviations of the various macroscopic quantities from their mean values are extremely small. Within this framework, the state of a rarefied gas is described by the distribution function, f v, r , t , so that the quantity, fdvdr , gives the probable number of molecules at time, t , which are located in the spatial region at r in dr , and which have velocities of v in the velocity space interval, dv . The description of a gas by the distribution function is excessively detailed while the gas, as a continuum, may be characterized by certain macroscopic quantities which are relatively few in number and which can be experimentally measured. These macroscopic quantities, and any other molecular property fluxes of interest, may be calculated as moments of the distribution function. The most significant of these macroscopic values for a monatomic gas include the number density, the mean velocity, the temperature, the pressure, and the thermal flux, and are given by:

n

³ fdv

u

n 1 ³ vfdv ,

3 2

(1-19)

2

n 1 ³ 12 m v u fdv ,

(1-20)

m ³ vi ui vk uk fdv ,

(1-21)

kT

Pik

(1-18)

,

and:

Q

1 2

2

m ³ v u v u fdv .

(1-22)

12

Analytical Methods for Problems of Molecular Transport

Thus, a complete knowledge of the distribution function can be used to fully describe the behavior of a gas.

PROBLEMS 1.1. Prove that the relative fluctuation of an additive quantity is proportional to N 1 2 , where N is the number of molecules of a gas. Solution: Let L be an additive quantity for a gas that contains N molecules. The mean value of this quantity may be written as:

L

N

¦ Li

,

i 1

and, therefore, this value is proportional to N . The relative fluctuation of L is given by:

G

L1

¦ 'L i

i

2

.

The radicand may be expressed in the form:

§ · ¨ ¦ 'Li ¸ © i ¹

2 2

¦ 'Li ¦¦ 'Li 'Lk i

.

i k izk

The later term in this relation is equal to zero since separate molecules are statistically independent and, for one molecule, 'Li 0 . Hence, this sum containing N terms is proportional to N and one can obtain the following:

G ~ N 1 N

N 1 2 .

1.2. Derive the condition of applicability of the point-mass hypothesis for a gas molecule. Solution: A molecule in a gas may be considered as a point particle when the de Broglie wavelength of the molecule is much less than the mean distance between the gas molecules. This relationship may be expressed as:

13

Chapter 1. The General Description of a Rarefied Gas 13

2S ! §V · ¨ ¸ mv ©N¹

n 1 3 ,

where ! is Planck’s constant ( ! 1.054 u 1027 erg sec) and v is the mean heat velocity of a molecule. The following expression may be written:

2S ! 2mkT

n1 3 1 .

1.3. Determine the temperature for which the de Broglie wavelength of a molecule is the same order of magnitude as the mean distance between gas molecules. Consider the case of hydrogen having a number density of n 2.69 u 1019 cm–3. Solution:

2S 2 ! 2 n 2 3 ; T0 mk

T0

0.43 K.

1.4. Let the behavior of a mechanical system be described by exact equations of motion ( N is the number of molecules, V is the volume, and N !! 1 ). How might a statistical description be introduced for such a system? How might the mean values and the distribution function be specified? What accuracy does the statistical description have? Solution: Consider the state of a subsystem which is situated in a small volume element, dr , surrounding the point, r , during a time interval, dt . This interval is assumed to satisfy the following relations:

dr 1 3 v

dt

V1 3 , v

where v is the mean velocity of the particles. Let the time interval, dt , be divided into k smaller subintervals of width, dW , and the whole velocity space be divided into m smaller velocity ranges, 'v j 'v jx 'v jy 'v jz , so that ¦ mj 1 'v j is equal to the product of the maximum values of the velocity components. If the equations of motion are known, it is easy to calculate the number of molecules at time, t i dW , which have velocities lying in the range, 'v j . Let this quantity be denoted by nij n r , v j , t idW . The number density and the mean value of any molecular property may be defined, respectively, by:

14

Analytical Methods for Problems of Molecular Transport

n

1 dr

m

k

¦¦

nij

j 1i 1

k

, and nI

m

k

¦¦ j 1i 1

I v j nij k

.

The distribution function may be written as:

f

1 dr'v j

k

nij

¦k

,

i 1

The relative fluctuations of additive quantities for such a system are given by G ~ N 1 2 .

REFERENCES 1. Landau, L.D. and Lifshitz, E.M., Statistical Physics (Pergamon, London; Addison Wesley, Reading, Mass, 1958). 2. Levich, V.G., Theoretical Physics: An Advanced Text (North-Holland, Amsterdam, 1970). 3. Huang, K., Statistical Mechanics (John Wiley and Sons, New York, 1963). 4. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 5. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 6. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, New York, 1969).

Chapter 2 THE BOLTZMANN EQUATION

1.

DERIVATION OF THE BOLTZMANN EQUATION.

It has been established previously that knowledge of the distribution function gives all the necessary information for a gas. To obtain the basic equation for the distribution function, consider a balance of the number of molecules that are located in the element, dvdr , of the six-dimensional phase space for the time interval, dt . Consider a gas in which each molecule is subject to an external force, mF , that is a function of r and t , but not a function of v . For the time interval, dt , the position of this phase element will change from v, r to v Fdt , r vdt . It is important to note that the magnitude of this volume element is unaltered in accordance with the Liouville theorem of classical mechanics [1]. There are f v, r , t dvdr molecules which, at the time, t , are situated in the volume element, dr , and have velocities in the velocity range, dv . These molecules are assumed to belong to the first molecular set. After the time interval, dt , there is another set in which the number of molecules is given by:

f v Fdt , r vdt , t dt dvdr . The number of molecules in the second set will differ from that in the first set owing to molecular encounters. The net gain of molecules to the second set is proportional to dvdrdt and will be denoted by G f G t dvdrdt . Consequently, one can obtain:

16

Analytical Methods for Problems of Molecular Transport

ª¬ f v Fdt , r vdt , t dt f v, r , t º¼ dvdr

G f

G t dvdrdt .

(2-1)

Now, if dt tends to zero, it is easy to obtain the Boltzmann equation which may be written in the form [2]:

Df

Gf . Gt

(2-2a)

Here, the following notation is introduced:

Df

wf wf wf vi Fi , wt wxi wvi

(2-2b)

where xi are components of the position vector, r . The quantity G f G t dvdr is the rate of change of the number of molecules in the phase element, dvdr , owing to molecular encounters.

2.

THE MOMENT EQUATIONS.

Let I I v, r ,t be any molecular property. After multiplying the Boltzmann equation by I dv and integrating throughout the velocity space, one can obtain the following integral relation:

³ I Dfdv

n'I ,

(2-3)

where:

n'I

Gf

³I G t

dv .

In Eq. (2-3), all the integrals are assumed to be convergent and, moreover, products such as I f tend to zero if v o f . This equation is the general form of a moment equation in which n'I is a moment of the Boltzmann collision integral. The various terms on the left-hand-side of Eq. (2-3) may be transformed by means of the following relationships:

17

Chapter 2. The Boltzmann Equation

Table 2-1. Transformation of the various derivatives to new variables.

v, r , t

V, r , t

wI wv

wI wV

wI wt

wI wI wui wt wVi wt

vi

wf

wI wxi

vi

w wI I fdv ³ f dv ³ wt wt

³ I wt dv wf

³ I vi wxi dv

w wI nI n , wt wt

w wI I vi fdv ³ fvi dv ³ wxi wxi

wf

³ I wvi dv ³ ³ I f

vi f vi f

wI wI wuk vi wxi wVk wxi

dv j dvk ³ f

w wI , nI vi nvi wxi wxi

wI dv wvi

n

wI . wvi

In the last of the above relationships, the double integral vanishes because I f |vvii ff 0 in accordance with the notation made above. Substituting these expressions into Eq. (2-3), one can obtain [2]:

w nI wt

w wI wI ° wI °½ nI vi n ® vi Fi 'I ¾ . wxi wxi wvi °¯ wt °¿

(2-4)

This equation is usually called the moment equation or the transport equation of the molecular property, I .

18

3.

Analytical Methods for Problems of Molecular Transport

ANOTHER FORM OF THE MOMENT EQUATIONS.

The basic moment relationship given in Eq. (2-4) is suitable for the laboratory frame of reference relative to which the gas has a mean velocity, u r , t . To transform this equation into the variables, V , r , t , where V v u r ,t , one must consider that I V, r ,t depends upon r and t , both explicitly and implicitly through its dependence on V . Table 2-1 expresses the transformation of the various derivatives in Eq. (2-4). Substituting the derivatives from Table 2-1 into Eq. (2-4), one can obtain [2]:

wu D w nI nI i nI Vi Dt wxi wxi

° DI Dui · wI wI § wI wuk ½° n ® Vi ¨ Fi Vi ¾ n' I , ¸ wxi © Dt ¹ wVi wVk wxi °¿ ¯° Dt

(2-5)

where D Dt w wt ui w wxi is the ‘mobile operator’ or time derivative with respect to the frame of reference moving with velocity, u . By analogy, the Boltzmann equation for these variables may be expressed in the form:

Dui · wf wu wf § wf Df Vi k Vi Fi wxi ¨© Dt Dt ¸¹ wVi wVk wxi

4.

Gf . Gt

(2-6)

THE EQUATIONS FOR A CONTINUUM MEDIUM.

Some important moment equations can be obtained from Eq. (2-5) by using the molecular properties, I , which are conserved during encounters. It is very important to note that these equations may be derived without knowledge of the collision operator because 'I 0 for such molecular properties. The invariant quantities of molecular encounters in a monatomic gas are the number of molecules, momentum, and kinetic energy because the mutual potential energy of two molecules equals zero both before and after an encounter. Let the following notations be introduced with respect to these invariant quantities:

\ 1 1 , \ 2

3 mV , \

1 2

mV 2 .

Chapter 2. The Boltzmann Equation

19

The statement n'\ 0 (for i 1, 2,3 , respectively) expresses the principles of conservation of number (of molecules), momentum, and energy i during encounters. Substituting I \ into Eq. (2-5), one can obtain: i

wu Dn n i Dt wxi

0 ,

(2-7)

wPik Dui · § U ¨ Fi 0 , wxk Dt ¸¹ ©

DT Dt

2 3nk

§ wu wQ ¨ Pij i i ¨ wx j wxi ©

(2-8)

· ¸ . ¸ ¹

(2-9)

These equations are usually called the continuum equations. This system is not closed as it contains 13 unknown functions that cannot be specified from the five equations. This lack of closure is a general characteristic of all moment systems. Since the distribution function is unknown, the moment transport equations may be considered to be the most general relations between the time and space derivatives of the macroscopic quantities because this system cannot be solved without definite suppositions or hypotheses about the distribution function which forms the basis of any moment method. The distribution function being used to close a moment system must satisfy certain general conditions. It must describe the characteristic features of the gas flow and must contain as many unknown parameters as the number of moment equations being employed. Then, all moments will depend on these parameters and the moment system becomes closed.

5.

MOLECULAR ENCOUNTERS.

Knowledge of the nature of the molecular interaction during encounters is necessary to derive the expression for the Boltzmann collision integral. At present, there is not a theory that describes the details of molecular encounters exactly. Therefore, a definite model of the molecular interaction has to be introduced to complete this description. One of the simplest molecular models is that of a smooth, rigid, and perfectly elastic sphere; commonly referred to as the rigid-sphere model for

20

Analytical Methods for Problems of Molecular Transport

brevity. The force between two molecules at a collision appears only if these molecules are in close contact with each other. This model is approximate since the real force varies continuously with the distance between the molecules. This fact might be better interpreted if a molecule was considered as a point-center of force depending on the nature of the interacting molecules and on their separation distance. Molecules represented as either rigid spheres or as point-centers of force may be termed smooth molecules since their internal energy is not altered during encounters. In this book, only smooth molecules are considered and the analysis of encounters follows that described in [2]. Let the masses of the two molecules undergoing an encounter be m1 and m2 . The interaction force is directed along the line joining their centers and it depends only upon the distance between molecules. The velocities before and after the encounter are denoted by v1 , v 2 and v1c , vc2 , respectively. Some general relations between these velocities may be found without consideration of the details of the molecular encounter. To start, one introduces the following notations:

m0

m1 m2 ,

M1

m1 m0 ,

M2

m2 m0 .

Owing to the momentum conservation during encounters, the center of mass of the two molecules moves with a constant velocity, G , that is given by:

m0G

m1 v1 m2 v 2

m1 v1c m2 vc2 .

(2-10)

Let g 21 and gc21 denote the relative velocities of the two molecules before and after an encounter:

g 21

v 2 v1 ,

gc21

vc2 v1c .

(2-11)

which have the following magnitudes:

g 21

g ,

gc21

gc .

From Eqs. (2-10) and (2-11), one can obtain:

v1

G M 2 g 21 ,

v1c

G M 2 gc21 ,

(2-12a)

v2

G M 1g 21 ,

vc2

G M 1gc21 .

(2-12b)

21

Chapter 2. The Boltzmann Equation

Figure 2-1. Geometry of a generic encounter.

The mutual potential energy of the two molecules equals zero both before and after an encounter and, hence, the energy conservation equation may be written in the following very simple form: 1 2

m v

2 1 1

m2 v22

m vc 1 2

1 1

2

m2 v2c2 .

Using this relation and Eqs. (2-12a) and (2-12b), one is able to determine that the relative velocity during an encounter is changed only in direction and not in magnitude, i.e.:

g

gc .

Consideration of momentum and energy conservation laws alone does not suffice to determine the direction of gc21 which depends not only upon the initial velocities but also upon two geometric variables characterizing the spatial orientation of the two interacting molecules under consideration.

22

Analytical Methods for Problems of Molecular Transport

Figure 2-2. Parameters of a generic encounter.

Consider the motion of the center B of the second molecule relative to the center A of the first. This relative motion can formally be considered as the motion of an imaginary particle with a reduced mass, P m1m2 m0 , in a central force field and, therefore, it will be confined to the plane through A [1] (the curve LMN in Fig. 2-1 is a trajectory of the second molecule). The asymptotes of this trajectory, PO and OQ , are in the directions of g 21 and gc21 . Let PcA be a line parallel to PO so that the direction of PcA is fixed by g 21 . The orientation of the plane LMN is specified by the angle H between this plane and another plane containing both PcA and the arbitrary fixed axis, Ax , as shown in Fig. 2-2. Let the plane RAO1 ( O1 is the point of intersection of the line PO with the plane RAO1 ) be perpendicular to PcA and AR be the line of intersection of the plane RAO1 with the plane through A which contains both PcA and the axis, Ax . Let the polar coordinates of O1 in this plane be b and H . These geometric variables complete the specification of the encounter.

Chapter 2. The Boltzmann Equation

23

Figure 2-3. The spherical coordinates of the relative velocity.

The angle, T , through which g 21 is deflected, depends, in general, on the magnitude of g and on the parameter, b . To derive the geometrical connection between g 21 and gc21 , this angle is assumed to be known. Let the spherical coordinates of the vector g 21 g be g ,D , E . Then, the Cartesian and spherical coordinates shown in Fig. 2-3 are related by the following:

gx

g cos D ,

(2-13a)

gy

g sin D cos E ,

(2-13b)

gz

g sin D sin E .

(2-13c)

24

Analytical Methods for Problems of Molecular Transport

Let H be the angle between the two planes, POQ and POR , as shown in Fig. 2-3. The vectors, g and gc , and g and i , lie in the planes, POQ and POR , respectively, hence one can obtain [3]:

cos H

gc u g i u g

gc u g i u g

g cx g x cos T g sin D sin T

(2-14)

.

To simplify this vector expression, the following relation is employed:

a u b c u b a c b 2 a b c b

.

Let Z be the angle between the two planes, POR and POy , shown in Fig. 2-3. Then, one can obtain:

cos Z

i u g ju g i u g ju g

cos D cos E 1 sin 2 D cos 2 E

,

(2-15)

and:

cos Z H

j u g g u gc j u g g u gc

g c g sin D cos E cos T y

sin T 1 sin 2 D cos 2 E

(2-16)

.

Using Eqs. (2-14)-(2-16) and analogous relations for g cz , one can express the components of the relative velocity by: g cx

g ª¬ cos D cos T sin D sin T cos H º¼ ,

g cy

g ª¬sin D cos E cos T cos D cos E sin T cos H sin E sin T sin H º¼ ,

(2-17a)

(2-17b)

Chapter 2. The Boltzmann Equation

g cz

25

g ª¬sin D sin E cos T cos D sin E sin T cos H

(2-17c)

cos E sin T sin H º¼ . If the angle, T , is known, these relations together with Eqs. (2-12a) and (212b) permit one to determine the velocity components after an encounter. Some other relations for the molecular velocities before and after an encounter may be derived from Eqs. (12-12a) and (12-12b). The trajectory, LMN , of the second molecule relative to the first molecule is symmetrical about the line, OA (usually known as the apse-line), shown in Fig. 2-1. The vectors, g 21 and gc21 , differ by twice the component of g 21 in the direction of k . This difference is expressed as:

g 21 gc21

2 g 21 k k .

Using this relation and Eqs. (2-12a) and (2-12b), one can obtain:

v1c v1

2 M 2 g 21 k k

vc2 v 2

2 M1 g 21 k k

2 M 2 gc21 k k ,

(2-18a)

2 M 1 gc21 k k .

(2-18b)

These formulas form another set of relations between molecular velocities during an encounter.

6.

THE RELATIVE MOTION OF TWO MOLECULES.

The relative motion of two molecules is equivalent to the motion of one particle with the reduced mass, P m1m2 m0 , moving in a central force field, F r wU wr r r , where U r is the mutual potential energy of the two molecules [1]. This problem can be solved in a general form by employing the energy and angular momentum conservation laws which, in polar coordinates r ,\ with the origin at the point, A , as shown in Fig. 21, may be written as [1,4]:

E

1 2

P r 2 r 2\ 2 U r ,

(2-19)

26

Analytical Methods for Problems of Molecular Transport

M

P r 2\ ,

(2-20)

where E is the energy before an encounter and M is the constant angular momentum component which is normal to the motion plane. Using Eq. (219), one can obtain:

M2 ª¬ E U r º¼ 2 2 . P P r 2

r

(2-21)

The polar angle, \ , can be found from Eq. (2-20) which may be expressed in the form:

M dt . P r 2

d\

From this relation and Eq. (2-21), after integrating with respect to r , one can obtain the following expression for \ as a function of r :

\

M dr r 2

³

M2 2P ª¬ E U r º¼ 2 r

After substituting E

b³ r0

P g 2 and M

dr r 2

f

\0

1 2

const .

b 2 2U r 1 2 r Pg2

(2-22)

P bg , Eq. (2-22) becomes:

.

(2-23)

Here, the lower limit of integration, r0 , is defined as:

r02 b 2

2r02U r0

Pg2

.

and the scattering angle, T , can be found from:

T

S 2\ 0 .

(2-24)

27

Chapter 2. The Boltzmann Equation

PROBLEMS 2.1. Prove that, if molecules move in accordance with Newton’s equations of motion, the phase volume element is not altered for the time interval, dt . Solution: For the time interval, dt , the position of the phase element, d * dvdr , is changed from v, r to vc v Fdt , r c r vdt . Then:

d* c

w vc, r c d* w v, r

w vc, r c w v, r c d* w v, r c w v, r

w v , v , v

w x c, y c, zc w x , y , z

w vcx , vcy , vzc x

y

z

r c const

d*

d* .

v const

The external force, F , is assumed to be independent of v .

2.2. Prove that the relative motion of two interacting molecules with mutual potential energy depending only on the distance between the molecules may be considered as the motion of a single particle in a central force field. Solution: First, introduce the vector: r r2 r1 e r r . The equations of motion of these molecules may be written in terms of this vector as m1 r1 f r e r , and m2 r2 f r e r , where f r wU r wr and U r is the mutual potential energy. From these equations, it is easy to find the r f r e r . This equation is the equation of motion of a relationship, P single particle of mass, P m1m2 m0 , in a central force field. 2.3. Prove that the relative motion of two interacting molecules is planar in nature. Solution: The differential equation of the relative motion of two molecules is given by: P g f r e r , where P m1m2 m0 , e r r r , r r2 r1 , g v 2 v1 , f r wU r wr and U r is the mutual potential energy. The vector product of this equation and the position vector, r , may be expressed as P r u g 0 or dM dt 0 , where M r u P g . This shows that M const and, consequently, that the position vector lies in the plane which is perpendicular to the constant vector, M . 2.4. Determine the distribution function for a gas in which the magnitude of the velocity vector is the same for all of the molecules, in which all directions of velocity are equally probable, and where the number density of the gas is n . Solution: The number of molecules situated in a phase element, dvdr , may be expressed as: fdvdr n\ v G v v0 dvdr , where \ v is a

28

Analytical Methods for Problems of Molecular Transport

function that depends only on the magnitude of the molecular velocity and G v v0 is the Dirac delta function. The function, \ v , may be found from the condition:

n ³\ v G v v0 v 2 sin T dvdT dI

³ fdv

n .

Then, it is easy to obtain the following form for the distribution function:

f

n G v v0 . 4S v 2

2.5. Using the conditions given in Problem 2.4, determine the distribution function for a two-dimensional gas (in which all molecular velocity vectors are parallel to the same plane) with a number density, n . Answer:

f

n 2S v

G v v0 .

REFERENCES 1. Landau, L.D. and Lifshitz, E.M., Mechanics (Pergamon, Oxford, 1960). 2. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 3. Jeans, J., Dynamical Theory of Gases (Cambridge University Press, Cambridge, U.K., 1990). 4. Bird, G.A., Molecular Gas Dynamics (Clarendon Press, Oxford, 1976).

Chapter 3 THE COLLISION OPERATOR

1.

THE DIFFERENTIAL AND TOTAL SCATTERING CROSS SECTIONS.

One of the most important concepts relating to the statistics of molecular collisions is that of the scattering cross section which is the basic quantity characteristic of molecular encounters. Consider a uniform flux of molecules that have a constant velocity, g , towards a spherically symmetric force center, O . The parameters of importance to the scattering events are shown in Fig. 3-1. Let N g be the g -component of the uniform molecular number flux and d : sin T dT d H be the solid angle in which molecules are deflected after encounters. The differential scattering cross section, D T , g , is defined by [1]:

D T , g d :

dN , Ng

(3-1)

where dN is the number of molecules deflected per unit time by the central force field in d : . Since the incident molecules scattered into d : must primarily pass through that portion of the annular differential element represented by dS bdbd H , dN may be expressed in the form:

dN

N g bdbd H .

From Eqs. (3-1) and (3-2), one can obtain:

(3-2)

30

Analytical Methods for Problems of Molecular Transport

D T , g

b db . sin T dT

(3-3)

The absolute value of db dT must be included in this relation since the signs of db and dT are usually opposite. The integral or total scattering cross section is then defined as:

D tot g 2S ³ D T , g sin T dT .

(3-4)

The differential scattering cross section has a very simple physical sense. The quantity, D T , g d : bdbd H , is that part of the annular differential area through which the molecular trajectory at infinity must pass in order for a molecule to be scattered into the solid angle, d : , after an encounter. The total scattering cross section, D tot g , is the area of the circular region through which the incident molecular trajectory must pass for any scattering to occur. The range of scattering that would contribute to this total cross section would encompass all of the available scattering solid angle 4S and any incident trajectory that were to lie outside of this circular region would be deemed to have not been scattered at all. Now, for example, consider molecules which are rigid spheres with diameters, V 1 and V 2 . The relation between the collision parameters shown in Fig. 3-2 may be expressed in the form:

b V 12 sin \ ,

(3-5)

where V 12 12 V 1 V 2 and \ 12 S T . Since the relative motion of two molecules is equivalent to motion in a central force field, the differential scattering cross section specified by Eq. (3-3) is given by:

D T , g

1 4

V 122 .

(3-6)

The total scattering cross section may be written as:

D tot g SV 122 .

(3-7)

It is very important to note that this molecular interaction model is unique in that D T , g and D tot g are independent of the relative velocity of the two molecules, g . This significant property of this model allows one to facilitate many calculations. For other models of intermolecular

Chapter 3. The Collision Operator

31

Figure 3-1. The scattering solid angle.

interaction that involve forces acting across infinite distances, the integral given by Eq. (3-4) is divergent since molecules are scattered even if the impact parameter, b , tends to infinity. Consequently, in this case, definite assumptions must be made regarding a cutoff of this parameter for which, when b ! bmax , no encounter occurs.

2.

THE STATISTICS OF MOLECULAR ENCOUNTERS.

Consider encounters between two different molecules with masses, m1 and m2 , situated in the volume element, dr . To derive the collision operator, one should calculate the rate of change of the number of m1 molecules having velocities in the range, dv1 , due to collisions with all molecules of the second kind. First, encounters between the two molecular sets, dv1 and dv 2 , in the volume element, dr , must be analyzed. Then, this analysis must be generalized for all molecules of the second kind. Encounters in which more than two molecules take part are assumed to be negligible in number compared with binary encounters and, moreover, molecules of the two velocity sets are assumed to be distributed at random without any correlation between velocity and position in the neighborhood of the point, r . These two assumptions are fundamental to the classical

32

Analytical Methods for Problems of Molecular Transport

Figure 3-2. Encounters of rigid-sphere molecules.

kinetic theory. It is necessary to note that the physical sense of the second assumption cannot be understood within the framework of the one-particle distribution function and, therefore, it may be considered to be a hypothesis the applicability of which may only be confirmed by using a more general theory [2-4] or by comparison with experiment. Examine the scattering of molecules of the second set by a molecule of the first in dr . The molecular number flux in the direction of the relative velocity, may be expressed in the form:

gf 2 v 2 , r , t dv 2 .

(3-8)

Then, in accordance with Eq. (3-1), the number of molecules of the second set deflected into the solid angle, d : , during the time interval, dt , is given by:

gD12 T , g f 2 v 2 , r , t d : dv 2 dt ,

(3-9)

where the time interval, dt , is assumed to be short compared to the scale of the time variation of the macroscopic properties of the gas, but large compared to the duration of an encounter. The appropriate number of v 2 -

Chapter 3. The Collision Operator

33

molecules reflected by all v1 -molecules in dr and dv1 may be written as [5]:

gD12 T , g f1 f 2 d : dv1dv 2 drdt .

(3-10)

In Eq. (3-10) it is assumed that the number of v1 , v 2 -collisions is proportional to f1 f 2 such that any potential correlation between velocity and position of molecules in dr has been ignored. Moreover, this expression holds true only if binary encounters are exclusively considered. The total number of encounters of the first set molecules with all molecules of the second kind in dr during dt may be found by integration over all values of : and v 2 . Then, one can obtain:

N12 dv1drdt

ª ³ ³ f1 f 2 gD12 T , g d : dv 2 º dv1drdt . ¬ ¼

(3-11)

Such encounters change the velocity of the first set molecules and thus the molecules are lost to the first set. As a result, N12 may be interpreted as the rate of loss of molecules from the first set per unit volume and velocity interval caused by collisions with molecules of the second kind. In the same way, one can obtain the number of m1 -molecules that enter the velocity set, dv1 . For this purpose, the m1 -molecules whose velocities after encounters lie in the range, dv1 , must be considered. In such inverse collisions, the molecular velocities are changed from v1c , vc2 to v1 , v 2 . Using the above analysis, one can obtain the total gain of m1 -molecules to the first velocity set, dv1 , for the time interval, dt , in the volume element, dr , owing to inverse encounters. This is given by: N12 dv1drdt

ª ³ ³ f1cf 2cg cD12 T , g c d : dvc2 º dv1c drdt , ¬ ¼

(3-12)

where f1c and f 2c stand for f1 v1c , r , t and , f 2 vc2 , r , t respectively. Now, one transforms the integrand in Eq. (3-12). First, since g g c , one can obtain:

g cD12 T , g c

gD12 T , g .

(3-13)

Next, one considers two subsystems of molecules the velocities of which, before an encounter, lie in the ranges, dv1c and dvc2 . If the collision parameters of these molecules lie in the ranges, db and d H , then their velocities after encounters will lie in the ranges, dv1 and dv 2 . In

34

Analytical Methods for Problems of Molecular Transport

accordance with Liouville’s theorem [1], the phase elements of these subsystems are connected by the relationship:

dv1c dr1cdvc2 dr2c

(3-14)

dv1dr1dv 2 dr2 .

Since the collisions between the two molecular sets occur in the same volume element, dr , Eq. (3-14) may be written as:

dv1c dvc2

(3-15)

dv1dv 2 .

Taking into account Eqs. (3-13) and (3-15), one can express Eq. (3-12) in the form: N12 dv1drdt

ª ³ ³ f1cf 2cgD12 T , g d : dv 2 º dv1drdt , ¬ ¼

(3-16)

where N12 is equal to the rate of gain of the first set molecules, per unit volume and velocity interval, due to encounters. Now, it is straightforward to obtain the collision operator for encounters between molecules of the first set and of the second kind:

§ G f1 · ¨ Gt ¸ © ¹2

N12 N12

³ ³ f1cf 2c f1 f 2 gD12 T , g d : dv 2

.

(3-17)

This form of the collision operator is usually called the Boltzmann collision integral. For a simple gas m1 m2 m, v1 o v, v 2 o v1 , Eq. (3-17) may be written as:

Gf Gt

³ ³ f cf1c ff1 gD T , g d : dv1

,

(3-18)

where G f G t denotes the rate of variation of the distribution function due to encounters at a fixed point, r . The rate of variation of any molecular property, I v, r , t , per unit volume, due to encounters, is defined by:

n'I

³ ³ ³ I f cf1c ff1 gD T , g d : dv1dv

.

(3-19)

This very important quantity may be called the moment of the collision integral associated with the property, I .

35

Chapter 3. The Collision Operator

3.

THE TRANSFORMATION OF SOME INTEGRALS.

Let the variables of integration in Eq. (3-19) be changed from v, v1 to vc, v1c . Then, this integral may be written as:

n'I c

³ ³ ³ I c ff1 f cf1c g cD T , g c d : dvcdv1c

.

(3-20)

Since the integration over all values of vc, v1c is equivalent to integration over all encounters, one can conclude that n'I c n'I . Taking into account Eqs. (3-15), (3-19) and (3-20), one can obtain:

n'I

1 2

³ ³ ³ I I c f fc1c ff1 gD T , g d : dvdv1

.

(3-21)

Since the variables, v1 and v , both refer to the same molecules, an interchange of them does not alter the value of this integral. After making this interchange of variables, Eq. (3-21) may be expressed in the form:

n'I

1 2

³ ³ ³ I1 I1c f cf1c ff1 gD T , g d : dvdv1

.

(3-22)

Then, combining Eqs. (3-21) and (3-22), one can obtain the following integral relation:

n'I

1 4

³ ³ ³ I I1 I c I1c f cf1c ff1 gD T , g d : dvdv1

.

(3-23)

PROBLEMS 3.1. Evaluate the time duration of an encounter between two molecules. Solution: Let r0 be the radius of action of the molecular forces and let V be the mean velocity of a molecule. The time duration of an encounter, W coll , is then given by W coll ~ r0 V . For rigid-sphere molecules, this may be written as W coll ~ V V where V is a molecular diameter. 3.2. Derive the condition of applicability of the Boltzmann equation if n is the gas number density. Solution: The mean distance between molecules is proportional to n 1 3 . Let r0 be the radius of action of the molecular forces. The binary interaction

36

Analytical Methods for Problems of Molecular Transport

hypothesis holds true if the molecular force action sphere contains not more than one molecule. This condition may be expressed: r03 n | 1 . 3.3. Prove by direct calculation that dv1c dvc2 dv1dv 2 during molecular encounters. Solution: First, derive the expressions; dv1dv 2 dGdg 21 and dv1c dvc2 dGdgc21 in which dg 21 g 2 dgd : . The desired proof follows directly if one takes into account the relationship g g c . 3.4. For a rigid-sphere gas in which all of the molecules have velocities of the same magnitude, determine the molecular mean free path assuming that all velocity directions are equally probable (number density, n , molecular diameter, V ). Solution: The total number of collisions occurring between molecules per unit volume and time is given by:

³ ³ ³ ff1 gD T , g d : dv1dv

N11

,

(P-1)

12

where g v 2 v12 2 vv1 cos F and F is the angle between v and v1 . The distribution function (Problem 2.4) may be expressed in the form:

f

n G c 1 , 4S v03c 2

where c obtain:

N11

v v0 and c

S 21 2 V 2 n 2 v0

4S

2

(P-2)

c . Substituting Eq. (P-2) into Eq. (P-1), one can

12

³ ³ 1 cos F

d Z1d Z ,

(P-3)

where d Z sin F d F d E and d Z1 sin F1 d F1d E1 . After integration over all directions of c and c1 , Eq. (P-3) becomes N11 43 SV 2 n 2 v0 . The mean free path is then given by:

O v0

n N11

4 3

SV 2 n

1

.

37

Chapter 3. The Collision Operator

3.5. For a two-dimensional rigid-sphere gas (number density, n , molecular diameter, V ), determine the molecular mean free path using the conditions in Problem 3.4. Solution: All molecular velocity vectors for this gas lie in the same plane. Therefore, it is convenient to use polar coordinates in the velocity space. The distribution function, in dimensionless variables, is given by (Problem 2.5):

f

n G c 1 . 2S v02 c

The total number of molecular collisions per unit volume and time is then:

S 21 2 V 2 n 2 v0

N11

2S

2

2S

12

³ 1 cos I 0

2S

dI

³ dI1

4V 2 n 2 v0 ,

0

where I is the angle between v and v1 , and I1 is the polar coordinate of v1 . The mean free path may then be expressed as:

O v0

n N11

4V n 2

1

.

REFERENCES 1. Landau, L.D. and Lifshitz, E.M., Mechanics (Pergamon, Oxford, 1960). 2. Bogoliubov, N.N., Problems of a Dynamical Theory in Statistical Physics (State Technical Press, Moscow, 1946). English translation by Gora, E.K. in Studies in Statistical Mechanics, Vol. 1, Eds. de Boer, J. and Uhlenbeck, G.E. (Wiley, New York, 1964). 3. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 4. Lifshitz, E.M. and Pitaevskii, L.P., Physical Kinetics (Pergamon, Oxford, 1960). 5. Anselm, A.I., Principles of Statistical Physics and Thermodynamics (in Russian) (Nauka, Moscow, 1973).

Chapter 4 THE UNIFORM STEADY-STATE OF A GAS

1.

THE BOLTZMANN H-THEOREM.

Consider a simple gas whose molecules possess only energy of translation, and are subject to no external forces. Let the state of the gas be uniform so that the distribution function, f , is independent of r . For this case, the Boltzmann equation may be expressed in the following simple form: wf wt

³ ³ f cf1c ff1 gD T , g d : dv1

.

(4-1)

Let the following H-function be introduced:

H

³ f ln f dv

.

(4-2)

The time derivative of the H-function may be written as:

wH wt

wf

³ 1 ln f wt dv ³ ³ ³ 1 ln f f cf1c ff1 gD T , g d : dv1dv

Now, taking into account Eq. (3.23), one can obtain:

(4-3)

.

40

Analytical Methods for Problems of Molecular Transport

wH wt

1 4

³³³

ln ff1 f cf1c f cf1c ff1 gD T , g d : dv1dv .

(4-4)

In this relation, ln ff1 f cf1c is always opposite in sign to f cf1c ff1 and, therefore, the integral is either negative or zero. As a result of this: wH d0 . wt

(4-5)

The latter differential relation indicates that the H-function can never increase. This fundamental inference is known as Boltzmann’s H-theorem and gives the time direction of processes for the uniform state of a gas [1-7].

2.

THE MAXWELLIAN VELOCITY DISTRIBUTION.

A very important question arises regarding whether there exists a limiting state for a gas after it has been disturbed from an initial uniform state. To show that there is, in fact, a limiting state, it is necessary to prove that the Hfunction cannot decrease indefinitely (i.e., that the H-function is bounded below). It is clear that H o f only if the integral given by Eq. (4-2) is divergent. If v o f , then the distribution function tends to zero and ln f o f . Since the integral, ³ 12 mv 2 fdv , expressing the energy of translation of the molecules, must be convergent, the H-function would be divergent only if ln f tended to infinity more rapidly than c 2 , where 12 c m 2kT v is the dimensionless velocity. In this case, however, f must tend to zero more rapidly than exp c 2 and, if this occurs in the integral given by Eq. (4-2), then the integral is convergent and the Hfunction is bounded below. Since it is bounded below, the H-function cannot decrease indefinitely but must tend to some limit, corresponding to a state of the gas, in which wH wt 0 . From Eq. (4-4), one may then conclude that the distribution function for the gas in this state may be found from the relation, f cf1c ff1 . This equation is equivalent to:

ln f c ln f1c ln f ln f1 .

(4-6)

If the distribution function satisfies Eq. (4-6), then, from Eq.(4-1), one can obtain that wf wt 0 also, so that such a state of the gas is steady as well as uniform.

41

Chapter 4. The Uniform Steady-State of a Gas

Now, consider the form of this distribution function. Eq. (4-6) shows that ln f is a summational invariant of encounters and, therefore, must be a linear combination of the three summational invariants. It may be expressed in the form [1,2]: 1 2 3 ln f D a mv D 12 mv 2 ,

(4-7)

or:

f

D 0 exp D 3 12 m V c

2

a V c v 3 . D 2

;

(4-8)

The constants, D , a , and D , can be expressed in terms of the number density, n , the mean velocity, u , and the temperature, T [1,2]. Using definitions of these quantities given by Eqs. (1-18)-(1-20), one can obtain: 0

2

0 § 2S · n D ¨ 3 ¸ © mD ¹

3

32

2 a

, u

D

3

3 , D

1 . kT

Then, the uniform steady-state of a gas is described by:

f

§ m · n¨ ¸ © 2S kT ¹

32

§ m v u 2 exp ¨ ¨ 2kT ©

· ¸ . ¸ ¹

(4-9)

This function is usually called the Maxwellian distribution function. Some mean values for the Maxwellian state of a gas may be calculated quite easily as they involve only simple integrations. For example, the mean value of the magnitude of the peculiar molecular velocity is given by: 12f

V

4 § 2kT · ¨ ¸ S1 2 © m ¹

3 2 ³ C exp C dC 0

12

§ 8kT · ¨ Sm ¸ © ¹

.

(4-10)

The number of molecules per unit time crossing a unit area surface element that is normal to the axis Ox and which moves with the same mean velocity as the gas is given by:

Nx

³ Vx fdv

,

(4-11)

42

Analytical Methods for Problems of Molecular Transport

where the notation, , signifies that the range of integration is over all values of v for which Vx is positive. After integration, one can obtain:

Nx

1 4

(4-12)

nV .

The last expression is very frequently used in the elementary kinetic analysis of transport phenomena [8-10].

3.

THE MEAN FREE PATH OF A MOLECULE.

Consider collisions between pairs of molecules with masses, m1 and m2 , in a binary gas mixture having a uniform steady-state. The number of collisions per unit volume and time between the two molecular sets having collision parameters in the ranges, db and d H , is given by:

dN12

f1 f 2 gD12 T , g d : dv1dv 2 .

f1 f 2 gbdbd H dv1dv 2

(4-13)

The total number of collisions per unit volume and time, N12 , is obtained by integrating Eq. (4-13) over all values of the collision parameters and the molecular velocities. For a gas having a uniform steady-state, the Maxwellian distribution function is used in the integrand. Then, N12 for rigid-sphere molecules may be written as [1]: 32

N12

S n1n2 m1m2 V 122

2S kT 3

§ m1v12 m2 v22 · exp ³ ³ ¨ 2kT ¸ gdv1dv 2 . © ¹

(4-14)

Let the variables of integration be changed from v1 , v 2 to G , g g g 21 as introduced in Section 2-5. Due to the following relationships between Jacobians:

w G, g

w v1 M 2 g , g

w v1 , g

w v1 , v 2

w v1 , v 2

w v1 , v 2

w v1 , v 2 v1

w v1 , v 2

w v1 , v 2

w v1 , v 2

1,

the velocity space element, dv1dv 2 , may be replaced by dGdg . Then, Eq. (4-14) becomes:

43

Chapter 4. The Uniform Steady-State of a Gas 32

S n1n2 m1m2 V 122

N12

2S kT 3

§ m0 G 2 M 1M 2 g 2 u³ ³ exp ¨ ¨ 2kT ©

·¸ gdGdg .

(4-15)

¸ ¹

Having performed the integration in spherical coordinates for G and g , one obtains: 12

N12

§ 2S kTm0 · 2n1n2V 122 ¨ ¸ © m1m2 ¹

.

(4-16)

The number of encounters solely between molecules of the first-kind is given by: 12

N11

§ S kT · 4n12V 12 ¨ ¸ © m1 ¹

.

(4-17)

The average number of collisions of a molecule of the first-kind per unit volume and time is usually called the collision frequency and may be written as:

Q1

N11 N12 . n1

(4-18)

The mean time between collisions (the collision interval) is then defined by:

W1

1

Q1

n1 . N11 N12

(4-19)

The mean distance between collisions is called the mean free path. This quantity is:

O1 V1W 1

12 ª m1 · º 12 2 2 § « 2 S n1V 1 S n2V 12 ¨ 1 ¸ » « © m2 ¹ »¼ ¬

1

.

(4-20)

44

Analytical Methods for Problems of Molecular Transport

For a simple gas m1 m2 m, V 1 V 2 path can be expressed in the form:

ª 21 2 S nV 2 º ¬ ¼

O

1

V , n1 n, n2

.

0 , the mean free

(4-21)

Using Eq. (4-21) the mean free path in a simple gas can now be estimated for various conditions. Assuming that one has nitrogen under conditions of Standard Temperature and Pressure (STP; 0 °C and 1 atm), the appropriate density and molecular diameter are n ~ 2.7 u 1019 cmí3 and V ~ 3 u 108 cm, respectively. Using these values in Eq. (4-21), the calculated mean free path for nitrogen at STP is then approximately 105 cm.

PROBLEMS 4.1. For the Maxwellian gas state, determine the number of molecules per unit volume with velocities in the ranges, dvz and dvU , where vz is the molecular velocity component along the axis, OZ , and vU is that along any perpendicular axis. Solution: In cylindrical coordinates, the distribution function is given by:

§ m · n¨ ¸ © 2S kT ¹

f

32

§ mv 2 exp ¨ © 2kT

· ¸ vU . ¹

The number of molecules per unit volume in dvU and dvz may be written as:

dn vU , vz

ª 2S º « ³ fd E » dvU dvz «¬ 0 »¼

§ m · 2S n ¨ ¸ © 2S kT ¹

32

§ mv 2 · exp ¨ ¸ vU dvU dvz . © 2kT ¹

4.2. For the Maxwellian gas state, determine the number of molecules per unit volume, n vx t v0 x , having an x -velocity component in the interval, vx t v0 x . Solution: The number of molecules in the above range may be determined from:

45

Chapter 4. The Uniform Steady-State of a Gas

n vx t v0 x

§ m · n¨ ¸ © 2S kT ¹ f

n

S

32 f

³

f

dv y

f

³

f

³ exp cx dcx 2

§ mv 2 · exp ¨ ¸ dvx ³ © 2kT ¹ v0 x f

dvz 1 2

c0 x

n ª1 erf c0 x º ¬ ¼

1 2

n cerf c0 x .

4.3. For the Maxwellian gas state, determine the distribution function of the kinetic energy of translation. Solution: Starting with the Maxwellian distribution function for the magnitude of the molecular velocity use the relationship 12 mv 2 E to transform the differential elements of the relationship f v dv f E dE . Substitute this transformation and the original energy relationship into the molecular velocity distribution to obtain the following distribution for the kinetic energy of translation:

f E

2

S

n kT

3 2

§ E · exp ¨ ¸ E . © kT ¹

4.4. For the Maxwellian gas state, determine the number of molecules impinging upon a unit surface element per unit time and lying in the angular range, dT , where T is the angle with respect to the surface element normal. Solution: From the Maxwellian distribution function in spherical coordinates, after integration over all velocities and the azimuthal angle, one can obtain: 12

dNT

§ 2kT · n¨ ¸ cos T sin T dT . © Sm ¹

4.5. For the Maxwellian gas state, determine the number of molecules impinging on a unit surface element per unit time and having molecular velocity magnitudes in the range, dv . Solution: This flux of molecules is given by:

dN v

§ m · 2S n ¨ ¸ © 2S kT ¹ § m · ¸ © 2S kT ¹

S n¨

32

32

§ mv 2 exp ¨ © 2kT

§ mv 2 exp ¨ © 2kT

· 3 S2 ¸ v dv ³ cos T sin T dT ¹ 0

· 3 ¸ v dv . ¹

46

Analytical Methods for Problems of Molecular Transport

4.6. A narrow molecular beam exits a system of slots into a vacuum. Determine the mean velocity and the root mean square (rms) velocity of the beam molecules. Solution: The distribution function of the beam molecules may be expressed in the form:

f v dv

n

dN v , N

where dN v has been found in Problem 4.5 and N is defined by Eq. (4-12). Then, any mean value for the beam molecules may be determined by means of this distribution function with the following results: 12

V

§ 9S kT · ¨ 8m ¸ © ¹

12

and

V

§ 4kT · ¨ m ¸ © ¹

2

.

4.7. For an equilibrium gas, prove that the ratio of the mean velocity of molecules crossing a surface element to the mean velocity of other molecules is 83 S . Solution: Using Eq. (4-10) and the result from Problem 4.6 yields:

VS V

12

§ 9S kT · ¨ 8m ¸ © ¹

12

§ 8kT · ¨ Sm ¸ © ¹

3 8

S .

4.8. Consider molecules reflecting from a differential surface element, dS c , on one surface and subsequently passing through a differential surface element, dS , on another surface. Assume that the reflected molecules at the first surface are being diffusely reflected and that the gas has a number density of nc and a temperature of T c . Moreover, assume that no molecular collisions occur while a molecule is transiting the space between the two differential surface elements. Determine the rate at which molecules reflected from dS c cross dS . Use the geometry in Fig. 4-1. Solution: The number of molecules crossing dS per second in a given direction may be expressed in the form (see Problem 4.4): 12

dNT

§ 2kT · nc ¨ ¸ © Sm ¹

dS cos T sin T dT

ncV cdS cos T 4S

d:

where d : is the solid angle subtended by dS c at dS . Taking into account that d : dS c cos T c r 2 , one can obtain:

47

Chapter 4. The Uniform Steady-State of a Gas

Figure 4-1. The geometry to be used in Problem 4.8 in determining the rate at which molecules reflected from one differential surface element cross another.

dNT

ncV cdSdS c cos T cos T c 4S r 2

12

§ 8kT c ·

Vc ¨ ¸ © Sm ¹

.

4.9. Two vessels containing the same gas at the pressures and temperatures, p1 , T1 and p2 , T2 , respectively, are connected by a small slot possessing a typical dimension of, L , which is much less than the mean free path of the gas molecules in either chamber. Derive the condition of kinetic equilibrium. Solution: If O !! L , molecules are leaving the vessels independently of each other. The condition of kinetic equilibrium requires that the net number of molecules crossing the area of the slot per second be equal to zero. This condition may be expressed as 14 n1V1 14 n2 V2 0 . From this relationship, one can obtain:

p1

p2

T1

T2

.

4.10. A binary gas mixture exits a chamber through a small slot by means of effusion. The escaping gas is collected in another chamber without loss. Calculate the ratio of the gas densities in the collection chamber.

48

Analytical Methods for Problems of Molecular Transport

Solution: Let n1 , m1 and n2 , m2 be the number densities and molecular masses, respectively, of the gases in the first chamber. The total number of molecules of each constituent in the collection chamber may be expressed as:

N ic

1 4

ni Vi S 't ,

where Vi is the mean velocity of each constituent, S is the area of the slot, and 't is the collection time interval. From this, the ratio of the gas densities can be written as: n1c n2c

N1c N 2c

n1 V1 . n2 V2

Equation (4-10) then yields:

n1c n2c

12

n1 § m2 · ¨ ¸ n2 © m1 ¹

.

4.11. Electrons evaporating from a hot wire form an electron gas with a number density, n . These electrons exit the chamber in which they are generated by means of a system of slots which serve to direct the electrons in one specific direction forming a beam. While exiting the chamber, this electron beam must pass through a hindering potential, M , which stops some of the electrons. Determine the flux density of the electrons that are able to overcome this hindering potential. Solution: The flux density of electrons having velocities in the interval, dv , is given by (Problem 4.5):

§ m · ¸ © 2S kT ¹

dN v

S n¨

32

§ mv 2 · 3 exp ¨ ¸ v dv . © 2kT ¹

The flux density of electrons overcoming the hindering potential may be written as: f

N

³ dNv

v0

12

n § kT · § eM · § eM · 1¸ exp ¨ ¨ ¸ ¨ ¸ . m kT 2S © ¹ © ¹ © kT ¹

49

Chapter 4. The Uniform Steady-State of a Gas

where

1 2

mv02

eM .

4.12. The work function of a metal is W . Determine the current density of the emitted electrons if the temperature is T and the electron number density inside the metal is n . Solution: Let OX be the axis normal to the surface of the metal. The current density along this axis is given by: 12 f

jx

§ m · ne ¨ ¸ © 2S kT ¹

§ mvx2 v exp ¨ ³ x ¨ 2kT © v0 x

· ¸¸ dvx ¹

1 4

§ W · neV exp ¨ ¸ , © kT ¹

where 12 mv02x W . This is equivalent to the classical Richardson thermionic emission formula. 4.13. Determine the number density variation for the steady-state of a gas that is situated in a force field described by the potential, M r . Solution: Let the distribution function be represented in the form:

f

§ · m n r ¨ ¨ 2S kT r ¸¸ © ¹

32

§ mv 2 · exp ¨ . ¨ 2kT r ¸¸ © ¹

Substituting this function into the Boltzmann equation, one obtains:

w w · § ¨ v wr F wv ¸ ln f 0 . © ¹ This expression may be reduced to:

° w § n v ® ln ¨ 3 2 ¯° wr © T

2 · mv wT m °½ F¾ 0 . ¸ 2 ¹ 2kT wr kT °¿

which is an identity in v . As such, the following system of equations may be obtained: wT wr

0 and

w ª § n ln ¨ wr ¬« © T 3 2

º · m M r » ¸ ¹ kT ¼

0 .

From the above system of equations, it is easy to obtain:

50

Analytical Methods for Problems of Molecular Transport

§ mM r · n r n0 exp ¨¨ ¸ . kT ¸¹ © where n0 is the number density at the point for which M r 0 and T is the constant temperature. 4.14. A gas rotates around a fixed axis with an angular speed, Z . Determine the number density distribution in the gas. Solution: The rotation of a gas is equivalent to an external force field having the potential: M r 12 Z 2 r 2 , where r is the distance from the rotational axis. From Problem 4.13, the variation in the number density is known to be:

§ mM r · n r n0 exp ¨¨ ¸ . kT ¸¹ © Substituting the rotational potential into this expression yields:

§ mZ 2 r 2 · n r n0 exp ¨ ¸ . © 2kT ¹ 4.15. Determine the number density distribution, n r , of a gas contained in a cylindrical vessel of radius, R , and length, L , that is rotating about its axis with an angular velocity, Z . The total number of molecules in the vessel is N . Solution: Using the number distribution obtained in Problem 4.14, one can obtain:

n r

§ mZ 2 r 2 · ª § mZ 2 R 2 NmZ 2 exp ¨ exp ¸« ¨ 2S L © 2kT ¹ ¬« © 2kT

· º ¸ 1» ¹ ¼»

1

.

4.16. Evaluate the mass of the Earth’s atmosphere assuming that almost all of the mass is contained in the thin layer near the Earth’s surface. Solution: The potential energy of a molecule near the Earth’s surface is given by: U h mgh . The total mass of the atmosphere is then approximated by: f

§ mgh · M | 4S R02 mn ³ exp ¨ ¸ dh © kT ¹ 0

4S R02 nkTg 1 ,

Chapter 4. The Uniform Steady-State of a Gas

51

where R0 is the radius of the Earth. Using typical values for the quantities involved yields M 5.3 u 1018 kg.

REFERENCES 1. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 2. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 3. Huang, K., Statistical Mechanics (John Wiley and Sons, New York, 1963). 4. Cercignani, C., Theory and Application of the Boltzmann Equation (Scottish Academic Press, Edinburgh, U.K., 1975). 5. Brush, S.G., Selected Readings in Physics. Vol. 2. Irreversible Processes (Pergamon Press, New York, 1966). 6. Klein, M.J., Paul Ehrenfest (North-Holland, Amsterdam and Elsevier, New York, 1970). 7. Uhlenbeck, G.E. and Ford, G.W., Lectures in Statistical Mechanics (American Mathematical Society, Providence, RI, 1963). 8. Kennard, E.H. , Kinetic Theory of Gases (McGraw-Hill, New York and London, 1938). 9. Knudsen, M. , The Kinetic Theory of Gases, third edition (Methuen & Co., London, and Wiley, New York, reprinted 1952). 10. Loeb, L.B. , The Kinetic Theory of Gases (Dover, New York, 1961).

Chapter 5 THE NON-UNIFORM STATE FOR A SIMPLE GAS

1.

EXPANSION IN POWERS OF A SMALL PARAMETER.

Consider the unsteady, non-uniform state of an infinite gas. For this general case, the Boltzmann equation has the form [1]:

Df

J ff1 ,

(5-1)

where:

Df

wf wf wf v F , wt wr wv

and:

J ff1

³ ³ ³ ff1 f cf1c gbdbd H dv1

.

The Boltzmann equation may be readily converted into a dimensionless form by introducing the following quantities: 12

§ 2kT · § m v ¨ ¸ c , f n¨ © m ¹ © 2kT

t W t , r Lr , F wF ,

· ¸ ¹

32

f ,

b V 21 4 b , H

SH ,

54

Analytical Methods for Problems of Molecular Transport

where W and L are typical values for the time and length variations of the macroscopic quantities of a gas. In terms of the above dimensionless quantities, Eq. (5-1) becomes: 12

L § m · wf wf wmL wf c F ¨ ¸

W © 2kT ¹ wt 2kT wr wc 1 J f f1 ,

D f

(5-2)

where O L . The parameter, , is usually called the typical parameter of non-uniformity of a gas but, since L is not necessarily the typical dimension of a boundary problem, is not the Knudsen number [2]. Let the macroscopic values vary slightly in the mean free path length, so that the following relations are satisfied:

O ln n 1 , O ln T 1 , O ln u 1 . Moreover, the typical values W and w are assumed to be confined by the 12 and w d 2kT mL . Then, the basic conditions W t L m 2kT dimensionless parameter, 1 . This makes it quite natural to seek a solution to Eq. (5-2) in the form of a power series in . Taking into account only the linear terms of this expansion, the distribution function may be approximated by:

f

0

1 f f ,

(5-3a)

or:

f

0 1 f f ,

(5-3b)

in ordinary variables. Substituting Eq. (5-3a) into Eq. (5-2), and retaining only the main terms in , one obtains:

0 D f

1

0

0

0

1

1 0 J f f1 J f f1 J f f1

^

` .

(5-4)

After equating coefficients of like powers of , the following relationships may be written:

55

Chapter 5. The Non-Uniform State for a Simple Gas

0

0 J f f1

(5-5)

0 ,

^

0 D f

0

1

1 0 J f f1 J f f1

` .

(5-6)

These integral equations are the basic ones used to obtain analytical expressions for the first and second-order correction terms to the distribution function.

2.

THE FIRST APPROXIMATION. In the usual variables, Eq. (5-5) has the form:

³ ³ f

0

0 0 0 c f1 c f f1 gbdbd H dv1

0 .

The general solution of this integral equation is defined by [1]: 0 0 f c f1 c

0 0 f f1 ,

and therefore, ln f is a linear combination of the summational invariants of molecular encounters. By a simple transformation one can obtain: 0

f

0

§ m · n¨ ¸ © 2S kT ¹

32

§ m v u 2 exp ¨ ¨ 2kT ©

· ¸ , ¸ ¹

(5-7)

where n , u , and T denote arbitrary values related to each of the summational invariants. These values can be identified with number density, mean velocity, and temperature at a point, r ,t . This means that the Maxwellian distribution function in which n n r , t , T T r , t , and u u r ,t is chosen as the first approximation to f . This function is usually called the local Maxwellian distribution function, which, for the general, non-uniform, non-steady-state of a gas, may be written as: 0 f r , t

§ · m n r , t ¨ ¨ 2S kT r , t ¸¸ © ¹

32

§ m v u r , t 2 exp ¨ ¨ 2kT r , t ©

· ¸ . ¸ ¹

(5-8)

56

Analytical Methods for Problems of Molecular Transport

As a consequence of the identification of n , u , and T with the density, 1 mean velocity, and temperature at the point, r ,t , the function f must satisfy the following relations:

³f

1 i

\ dv 0 ,

(5-9)

where \ 1, mV , 12 mV 2 . These conditions must be satisfied in the formulation of any expressions for the second-order correction term to the distribution function. i

3.

A GENERAL FORMAL SOLUTION FOR THE SECOND CORRECTION. Eq. (5-6) in ordinary variables may be written in the form:

0 1 1 0 J f f1 J f f1

0 D f .

(5-10)

This is the linear non-homogeneous integral equation for the second-order 1 1 1 correction to the distribution function, f . Let F and F be the particular and general solutions of the non-homogeneous and homogeneous equations, respectively. Then, the general solution can be represented by:

f

F F .

1

1

1

(5-11)

Substituting this expression into Eq. (5-10), one can obtain the following form of the homogeneous integral equation:

I M 1

(5-12)

0 .

Here, the correction M is given by the relation, F f M and I M is the standard linearized collision operator which may be expressed as: 1

1

0 0 I F n 2 ³ ³ f f1 F F1 F c F1c gbdbd H dv1 .

0

1

1

(5-13)

Eq. (5-12) means that M must be the summational collision invariant which may be written as: 1

57

Chapter 5. The Non-Uniform State for a Simple Gas

M 1 D 1,1 D 2,1 mv D 3,1 12 mv 2 ,

(5-14)

where D are arbitrary functions of r and t . These functions may be found from the conditions given by Eq. (5-9). Now, to find the necessary conditions of solubility for Eq. (5-10), one uses the integral theorem [1] given by: i ,1

³\

i ª

0 1 1 0 J f f1 J f f1 º» dv 0 . ¬« ¼

Using this theorem in Eq. (5-10), one can obtain the following conditions:

³\

i

0 D f dv 0 .

(5-15)

These relations, when i 1, 2,3 , mean that the values n , u , and T 0 contained in f must satisfy the continuum equations obtained by means of the use of the first-order distribution function [1]. 1 0 1 1 Let F be represented in the form, F f ) . Then, Eq. (5-10) gives the following basic equation for this correction:

n2 I ) 1

0 D f .

(5-16)

From general conditions given by Eq. (5-9), one can conclude that the 1 function, ) , must satisfy the following additional conditions:

³\

i 0

4.

f

) M dv 1

1

0 .

(5-17)

THE TRANSFORMATION OF THE NONHOMOGENEOUS TERM.

If the variables V, r , t are used, the right-hand-side of Eq. (5-16) may be expressed in a form analogous to that given in Eq. (2.6): 0 Df

0 0 f ln f

D Du · w w § w wu ½ V: ¾ , u® V ¨ F ¸ Dt ¹ wV wV wr © wr ¿ ¯ Dt

(5-18)

58

Analytical Methods for Problems of Molecular Transport

where: D Dt

w w . u wt wr

The space and time derivatives of n , u , and T are connected by the equations of the continuous medium in which the pressure tensor and the heat flux must be calculated by use of the local Maxwellian distribution function given in Eq. (5-8). This results in the following expressions: 0 Pik

0 pG ik , q

nkT G ik

(5-19)

0 .

Substituting these quantities into Eqs. (2.7)-(2.9), one obtains: Dn Dt

n

Du Dt

F

DT Dt

23 T

w u , wr

(5-20)

1 wp , U wr

(5-21)

w u . wr

(5-22)

Eqs. (5-20) and (5-22) result in: D ln nT 3 2 Dt

0 , ln nT 3 2

const .

(5-23)

Using the second relationship in Eq. (5-23), one can write:

0 ln f

const

mV 2 . 2kT

Taking into account Eq. (5-22), the first term in Eq. (5-18) may be expressed in the form:

59

Chapter 5. The Non-Uniform State for a Simple Gas

D 0 ln f Dt

1 2

mV 2 DT kT 2 Dt

13

mV 2 w u kT wr

wu mV 2 G ik i . kT wxk

13

The last term in Eq. (5-18) may be written as:

V : wu

0 w ln f

wV

wu m ViVk i . kT wxk

wr

The sum of the first and last terms of Eq. (5-18) is then:

wui m ViVk 13 V 2G ik kT wxk

m $ wu VV : . kT wr

(5-24)

The sum of the two middle terms of Eq. (5-18) is given by: 0 w w § f § w · 0 V ¨ ln f ln nkT ¸ V ln ¨ wr wr ©¨ nkT © wr ¹

· ¸ ¸ ¹

§ w w ln T mV 2 wT · § 1 mV 2 5 · . V ¨ ln T 5 2 12 2¸V ¸ ¨2 2 wr kT wr ¹ © kT © wr ¹

(5-25)

Using Eqs. (5-24) and (5-25), Eq. (5-18) can be represented as: $ w ln T wu °½ 0 ° f ® C 2 52 V 2 CC : ¾ . wr wr °¿ °¯

0 Df

(5-26)

The basic relation expressed by Eq. (5-16) can now be written in the form:

n2 I ) 1

5.

$ w ln T wu °½ 0 ° f ® C 2 52 V 2 CC : ¾ . wr wr ¿° ¯°

(5-27)

THE SECOND APPROXIMATION.

Since Eq. (5-27) is a linear integral equation and the right-hand-side of this equation is linear in w ln T w r and wu w r , one may seek a solution of this equation in the form of a sum of the two terms each of which is

60

Analytical Methods for Problems of Molecular Transport

proportional to the appropriate space non-uniformity. Thus, one can write [1,3-5]:

) 1

A

w ln T wu , 2B : wr wr

(5-28)

where A is a vector and B is a tensor. The quantities, A and B , introduced here are identical to those obtained from the Chapman-Enskog theory for multiple gas mixtures in the limiting case of a simple gas [1]. Substituting Eq. (5-28) into Eq. (5-27), one can separate out equations for the unknown functions, A and B . These functions are then particular solutions for the following equations:

n2 I A

0 f C 2 52 V ,

(5-29)

n2 I B

$ 0 f CC ,

(5-30)

which, in the coordinate form, can be written as:

n 2 I Ai

n 2 I Bik

0 f C 2 52 Vi ,

(5-31a)

0 f Ci Ck 13 C 2G ik .

(5-31b)

One now assumes that the solutions of Eqs. (5-31a) and (5-31b) are of the following form: A

AC C ,

B

B C CC ,

$

(5-32)

(5-33)

where A C and B C are two new unknown functions of n , T , and C C . The general formal solution of Eq. (5-27) then has the form:

61

Chapter 5. The Non-Uniform State for a Simple Gas $ w ln T wu ° 0 2 B C CC : f ® A C C wr wr °¯

f 1

D

1,1

D

2,1

mV D

3,1

1 2

(5-34)

2½

mV ¾ dv . ¿

where the constants D in Eq. (5-34) are chosen such that the function 1 f satisfies the following equations: i ,1

³f

1 i

\ dv 0 ; i 1, 2,3 .

(5-35)

Neglecting vanishing integrals, these equations have the form:

³f ³f ³f

0

D

1,1

0 §

3,1 D 12 mV 2 dv 0 ,

¨¨ A C ©

0

D

1,1

w ln T 12 2,1 · 2mkT D ¸¸ C 2 dv 0 , wr ¹

3,1 D 12 mV 2

1 2

mV 2 dv 0 .

The first and third of the above relations show that D D 0 . The 2,1 second shows that D ~ w ln T wr and, hence, that this term may be 2,1 absorbed into the first term. In order that D 0 also, the function A C must satisfy the additional relation: 1,1

³ AC C

2

0 f dv 0 .

3,1

(5-36)

Now, the second approximation to the distribution function is given by:

f 1

0 1 1 f ) M

$ w ln T wu ½° ° 0 f ® A C C 2 B C CC : ¾ . wr wr ¿° ¯°

(5-37)

62

Analytical Methods for Problems of Molecular Transport

6.

THE FIRST-ORDER CHAPMAN-ENSKOG SOLUTION FOR THERMAL CONDUCTION.

Suppose that the function A can be expressed as a convergent series of the form: f

¦ ar a r

A

,

r a

r 0

where S3 2 C 2 defined by: r

S m x n

S3 2 C 2 C , r

(5-38)

are the Sonine polynomials. The polynomial S m x is n

n

¦ x p 0

p

m n n p p n p

,

where:

m n n p m n m n 1 ... ª¬ m n n p 1 º¼

,

and m n 0 1 . Two special cases exist where:

S m x 1 , 0

S m x m 1 x . 1

These polynomials satisfy the following integral relations:

f

p q m ³ exp x Sm x Sm x x dx 0

0 ; pzq , ° ° ® ° * m p 1 ; p q , °¯ p

where * k is the gamma function. The additional condition expressed by Eq. (5-36) gives:

63

Chapter 5. The Non-Uniform State for a Simple Gas 2 2 ³ exp C C A C dC f ° f ½° p 4S ³ C 4 exp C 2 ® ¦ a p S3 2 C 2 ¾ dC °¯ p 0 °¿ 0

f

f

p 0

0

2S ¦ a p ³ exp C 2 S3 2 C 2 S3 2 C 2 C 3 d C 2 2S a0 *

52

0

p

0 ,

and therefore a0 0 . To get acceptable accuracy [1], only one term of the series in Eq. (5-38) need to be taken into account, i.e.:

A

a1 CS3 2 C 2 . 1

1

Substituting this expression into Eq. (5-29), one obtains:

n 2 a1 I CS3 2 C 2 1

1

0 f C 2 52 V

1 0 f VS3 2 C 2 .

(5-39)

After multiplying this equation by CS3 2 C 2 and integrating over all values of v , Eq. (5-39) can be expressed in the form: 1

n 2 a1 ³ CS3 2 C 2 I CS3 2 C 2 dv D1 , 1

1

1

1 or a1 a11 D1 where:

a11

ªCS 1 C 2 , CS 1 C 2 º , 32 ¬ 32 ¼

12

§ 2kT · ¸ © m ¹

D1 n 2 ³ f 0 VS312 C 2 CS312 C 2 dv 154 n 1 ¨

.

The expression for the bracket integral a11 is calculated in Appendix A. Using Eq. (A-9), the following solution of Eq. (5-39) can be obtained for rigid-sphere molecules:

64

Analytical Methods for Problems of Molecular Transport 12

§ 2kT · 154 n 1 ¨ ¸ © m ¹

a1 1

1 a11

15 SO . 16

(5-40)

The vector, A , for the first-order Chapman-Enskog solution is given by:

A

15 S O CS3 2 C 2 . 16 1

7.

(5-41)

THE FIRST-ORDER CHAPMAN-ENSKOG SOLUTION FOR VISCOSITY.

To simplify the following calculations, consider the viscosity equation for only one component of the tensor, B . The equation for Bxy may be written as:

n2 I CxC y B C

0 f C x C y .

(5-42)

Since the tensor, B , is a non-divergent tensor of molecular velocity components, it can be expressed in the general form [1]: f

$ $ r 1 B CC B C CC ¦ br S5 2 C 2 , r 1

(5-43)

where the coefficients, br , are constants to be determined. When only one tensor component, Bxy , is being considered, it can be expressed as:

Bxy

f

r 1 C x C y ¦ br S5 2 C 2 . r 1

(5-44)

To obtain the first-order Chapman-Enskog solution (retaining only one term of the expansion), this tensor is given by:

Bxy

b1 C x C y . 1

Substituting Eq. (5-45) into Eq. (5-42) one can obtain:

(5-45)

65

Chapter 5. The Non-Uniform State for a Simple Gas

n 2b1 I C x C y

1

0 f Cx C y .

(5-46)

After multiplying both sides of Eq. (5-46) by C x C y dv and integrating over all velocities one obtains:

c b1 b11 1

(5-47)

E1c ,

where:

c b11

¬ªC x C y , C x C y ¼º ,

E1c n 2 ³ C x2C y2 f 0 dv

1 4

n 1 .

c and E1c introduced here differ It is important to note that the values of b11 from the corresponding values in [1] by a factor equal to 101 . Using Eq. (Ac , one obtains: 12) for b11 b1 1

12

1 cn 4b11

5 8

§ m · ¸ © 2kT ¹

(5-48)

.

SO¨

Both Eqs. (5-40) and (5-48) have been obtained for rigid-sphere molecules. From the preceding analysis, in general, the components of the tensor for the first-order Chapman-Enskog solution, Bik , have the form: 12

Bik

5 8

§ m · ¸ © 2kT ¹

SO¨

C C i

k

13 G ik C 2 .

(5-49)

Using Eqs. (5-41) and (5-49) and introducing the mean free path, the firstorder Chapman-Enskog distribution function is then given by:

f

w ln T 0 ° O S S312 C 2 Ci f ®1 15 16 wxi °¯

12

§ m · O S¨ ¸ © 2kT ¹ 5 4

C C i

k

1 2 C ik 3

G

wui ½° ¾ . wxi °¿

Here, molecules have been assumed to act as rigid spheres.

(5-50)

66

Analytical Methods for Problems of Molecular Transport

8.

THE THERMAL CONDUCTIVITY AND VISCOSITY COEFFICIENTS.

Since the distribution function is known, the coefficients of viscosity and thermal conductivity may be readily calculated. The thermal conductivity coefficient can be obtained from the relation:

qi

m ³ V 2Vi fdV

1 2

N

wT . wxi

(5-51)

Having performed the very simple integration in Eq. (5-51), one obtains the following expression for the thermal conductivity coefficient of a monatomic gas: 12

N

75 64

k § kT · ¨ ¸ V 2 ©Sm ¹

.

(5-52)

The molecules here have been assumed to behave as rigid spheres. To derive a similar expression for the viscosity coefficient, the simplest case of u ^0, u ,0` and u u x is considered. In this case:

Pxy

m ³ VxVy fdV

P

wu . wx

Using Eq. (5-50) in this expression and integrating yields: 12

P

5 16

V

2

§ mkT · ¨ S ¸ © ¹

.

(5-53)

The general expression for the pressure tensor [1] is given by:

Pik

ª§ wu wu · º pG ik P «¨ i k ¸ 23 G ik u » . «¬© wxk wxi ¹ »¼

(5-54)

Chapter 5. The Non-Uniform State for a Simple Gas

9.

67

THE FIRST-ORDER APPROXIMATION FOR ARBITRARY INTERMOLECULAR POTENTIAL.

In the first approximation, the thermal conductivity and viscosity coefficients depend only on the bracket integrals, a11 and b11 , respectively. These bracket integrals for arbitrary intermolecular potentials may be expressed in the form [1]:

a11

ªCS 1 C 2 , CS 1 C 2 º 32 ¬ 32 ¼

b11

$ ª $ º CC , CC « » ¬ ¼

2,2 4: ,

(5-55)

and: 2,2 4: .

(5-56)

where the b11 given here is in the same form as that used by Chapman and Cowling [1]. i, j The : -integrals can be evaluated only if the form of the intermolecular interaction is specified. All potential functions may be characterized by two parameters, H and V , and it is convenient to represent the mutual potential energy of two molecules in the form:

M r H f r V ,

(5-57)

where f r V is the same function for all gases described by the specific interaction model. The Lennard-Jones (6-12) potential model has found the widest use in applied transport problems. This model is described by:

M r

ª§ V ·12 § V ·6 º 4H «¨ ¸ ¨ ¸ » , © r ¹ »¼ «¬© r ¹

(5-58)

where H is the depth of the potential well and V r at the point where M r 0 . This parameter may be called an effective molecular diameter. The two parameters of this model, H k (K) and V , have been tabulated for many gases by a comparison of the theory with experimental data. All of the : -integrals are expressible in the form [3,5]:

68

Analytical Methods for Problems of Molecular Transport

: i, j

ª: i , j º : i , j å , ¬ ¼ r .s.

(5-59)

i, j where ª : º is the integral for rigid-sphere molecules that is given by: ¬ ¼ r .s.

ª: i , j º ¬ ¼ r .s.

12

§ kT · SV ¨ ¸ ©Sm ¹ 2

i 1 1 º » . 1 2 i 1 » «¬ ¼

j 1 «ª 2

(5-60)

i, j å The reduced integral, : , depends only on the reduced temperature which is defined as:

T

kT

H

.

(5-61)

The numerical values of the reduced : -integrals for various models of the intermolecular potential may be found in [3,5,6]. Taking into account Eq. (5-59), one can express the first approximation to the transport coefficients for a monatomic gas as follows:

> P @1

5 16

>N @1

25 32

S mkT 1 2

SV 2 : 2,2 å T

S mkT 1 2

SV 2 : 2,2 å T

,

cV ,

(5-62)

(5-63)

where cV is the specific heat per unit mass at constant volume which is (for a monatomic gas, for example) cV 3k 2m .

10.

THE SECOND-ORDER APPROXIMATION FOR ARBITRARY INTERMOLECULAR POTENTIAL.

The parameters of the second-order Chapman-Enskog solutions may be found from the following algebraic systems [1]:

Chapter 5. The Non-Uniform State for a Simple Gas

69

12

2 2 a1 a11 a2 a21

§ 2kT · 154 n 1 ¨ ¸ © m ¹

2 2 a1 a12 a2 a22

0 ,

,

and: 2 2 b1 b11 b2 b21

5 2

2 2 b1 b12 b2 b22

0 ,

n 1 ,

All of the parameters are expressible in terms of the bracket integrals defined in [1]. In the second approximation, the thermal conductivity and viscosity coefficients can be shown to be:

>N @2 >N @1 fN 2 T

,

(5-64)

> P @2 > P @1 f P 2 T

,

(5-65)

and:

2 2 where the functions, fN T and f P T , are given by:

77 112v 80t , 28 64v 2 80t

(5-66)

301 336v 240t . 154 48t

(5-67)

2 fN T

and: 2 f P T

Here, the following notations have been introduced:

2,3 å v : T

: 2,2 å T ,

(5-68)

70

Analytical Methods for Problems of Molecular Transport

Table 5-1. Functions for calculating the second-order transport coefficients. 2 2 2

fN

T

0.30 0.50 0.75 1.00 1.25 1.50 2.00 2.50 3.00 4.00 5.00 10.0 50.0 100.0 400.0

t

T

fP

1.00189 1.00009 1.00004 1.00000 1.00012 1.00055 1.00214 1.00380 1.00516 1.00737 1.00867 1.01096 1.01160 1.01167 1.01170

: 2,4 å T

T

1.01030 1.01020 1.02181 1.02775 1.03074 1.03149 1.02832 1.02580 1.02448 1.02105 1.01948 1.01582 1.01419 1.01410 1.01383

a2

T

0.02680 0.00588 –0.00369 –0.00030 0.00680 0.01419 0.02801 0.03719 0.04325 0.05158 0.05590 0.06279 0.06465 0.06483 0.06492

: 2,2 å T .

2 b2 T

0.01724 0.00376 –0.00237 –0.00019 0.00441 0.00922 0.01825 0.02426 0.02824 0.03370 0.03653 0.04103 0.04222 0.04233 0.04239

(5-69)

The second coefficients of the Chapman-Enskog solutions are found to be: 2 a2

32v 28 , 77 112v 80t

(5-70)

2 b2

96v 84 . 301 336v 240t

(5-71)

and:

2 2 2 2 2 2 where a2 a2 a1 and b2 b2 b1 . For the Lennard-Jones (6-12) 2 2 2 2 potential, the functions, fN T , f P T , a2 T , and b2 T , are given in Table 5.1.

PROBLEMS 5.1. Show that if:

f

0

§ · m n r, t ¨ ¨ 2S kT r, t ¸¸ © ¹

32

§ m v u r, t 2 exp ¨ ¨ 2kT r, t ©

· ¸ , ¸ ¹

71

Chapter 5. The Non-Uniform State for a Simple Gas

then the collision operator associated with this function is equal to zero. Solution: Very simple transformations result in the following expression which occurs in the collision operator: 0 0 0 0 f f1 f c f1 c

§ m v 2 v12 · m ° ¸ A r, t ®exp ¨ v v u 1 ¨ ¸ kT 2kT °¯ © ¹

§ m vc2 v1c2 ·½ m ° vc v1c u ¸ ¾ 0 . exp ¨ ¨ ¸° 2kT kT © ¹¿ That this is equal to zero (and hence the collision operator is equal to zero) follows from the conservation of energy and momentum during molecular encounters.

5.2. Obtain the second-order Chapman-Enskog solution for the thermal conduction problem by using the following polynomial expansion: A

^

` .

1 2 2 2 a1 C S3 2 C 2 a2 S3 2 C 2

Assume that the molecules in this problem are rigid spheres. 1 Solution: Multiplying Eq. (5-29) alternately by the terms, CS3 2 C 2 2 2 and CS3 2 C , one can obtain the following algebraic system:

2 2 a1 a11 a2 a21

12

§ 2kT · 154 n 1 ¨ ¸ © m ¹

where a12 a21 14 a11 and a22 solution of this system yields: 2 a1

45 1 a 44 1

2 , a2

1 1 a 11 1

2 2 , a1 a12 a2 a22

45 a 16 11

2 , a2

0 ,

for rigid-sphere molecules [1]. The

4 45

.

5.3. Derive the expression for the thermal conductivity coefficient for the second-order Chapman-Enskog solution for rigid-sphere molecules. Solution: From Eqs. (5-37) and (5-51), one can obtain:

72

Analytical Methods for Problems of Molecular Transport

qi

§ 2kT · 12 mnS 3 2 ¨ ¸ © m ¹ 3 2

mnS

1 3

§ 2kT · mnS 1 2 ¨ ¸ © m ¹ 12

2kT · ¸ © m ¹

5§ 4¨

kna1

1 wT C 2Ci2 A C exp C 2 dC T wxi ³

32

1 wT T wxi

32

1 wT T wxi

§ 2kT · ¨ m ¸ © ¹

1 6

32

wT wxi

N

2 2 2 ³ 52 C C A C exp C dC f

³x

32

0

§ f · p 1 S3 2 x ¨ ¦ a p S3 2 x ¸ exp x dx ¨ ¸ ©p 1 ¹

wT . wxi

This implies that the thermal conductivity coefficient is proportional to a1 only and may be written as: 12

§ 2kT · 54 ¨ ¸ © m ¹

N

kna1 .

For the second-order Chapman-Enskog solution (Problem 5.2), one can obtain:

N

2

12

2kT · ¸ © m ¹ 5§ 4¨

2 kna1

45 44

N 1 ,

where the molecules have been assumed to be rigid spheres.

5.4. Obtain the second-order Chapman-Enskog solution for the viscosity problem by using the following polynomial expansion:

Bxy

^

` .

2 2 1 b1 C x C y 1 b2 S5 2 C 2

Assume that the molecules in this problem are rigid spheres. Solution: Multiplying Eq. (5-30) alternately by the terms, C x C y and 1 Cx C y S5 2 C 2 , one can obtain the following algebraic system:

2 c b2 2 b21 c b1 b11

4n 1

2 c b2 2 b22 c b1 b12

0 ,

,

Chapter 5. The Non-Uniform State for a Simple Gas

c b21 c 14 b11 c and b22 c where b12 solution of this system yields: 2 b1

205 1 b 202 1

2 , b2

1 6 b 101 1

205 c b 48 11

73

for rigid-sphere molecules [1]. The

2 , b2

12 205

.

5.5. Derive the expression for the viscosity coefficient for the second-order Chapman-Enskog solution for rigid-sphere molecules. Solution: Taking into account the definition of the pressure tensor and Eq. (5-37), one can obtain:

Pxy

4S 3 2 nT

wu C x2C y2 B C exp C 2 dC wx ³

S 1 2 nkT 16 15

wu f 6 C B C exp C 2 dC wx ³0

158 S 1 2 nkT

§ f · wu f 5 2 0 p 1 x S x ¨¨ ¦ bp S5 2 x ¸¸ exp x dx 52 ³ wx 0 ©p 1 ¹

nkTb1

wu . wx

This implies that the viscosity coefficient is proportional to b1 only and does not contain any other coefficients from the polynomial expansion. Then the viscosity coefficient can be written as P nkTb1 . For the second-order Chapman-Enskog solution (Problem 5.4), one can obtain:

P 2

2 kTb1

205 202

P 1 .

REFERENCES 1. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 2. Uhlenbeck, G.E. and Ford, G.W., Lectures in Statistical Mechanics (American Mathematical Society, Providence, RI, 1963). 3. Hirschfelder, J.O., Curtiss, C.F. and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954). 4. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, New York, 1969). 5. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 6. Maitland, G.C., Rigby, M.A., Smith, E.B., and Wakeham, W.A., Intermolecular Forces (Oxford University Press, Oxford, 1981).

Chapter 6 REGIMES OF RAREFIED GAS FLOWS

1.

THE KNUDSEN NUMBER.

Consider the steady-state flow of an infinite stream of a rarefied gas over a body having a characteristic dimension, R , in the absence of external forces. In this case, the Boltzmann equation may be expressed in the form:

v

wf wr

³ ³ ³ f cf1c ff1 gbdbd H dv1

.

(6-1)

Let this equation be transformed into dimensionless variables by means of the following: 12

§ 2kT · v ¨ ¸ c, f © m ¹

§ m · n¨ ¸ © 2kT ¹

32

f , r

21 4 V b , H

Rr , b

SH .

In these variables, Eq. (6-1) becomes:

c

wf wr

R

O

J f f1

Kn 1 J f f1 ,

(6-2)

where Kn O R is the Knudsen number and J f f1 is the standard dimensionless collision operator. The Knudsen number is the natural parameter characterizing the degree of rarefaction of a gas for a given boundary problem. The Knudsen number is defined as the ratio of the mean free path of a molecule to a characteristic external geometrical dimension of

76

Analytical Methods for Problems of Molecular Transport

the boundary problem. The Knudsen number is not a property of the gas alone as it explicitly involves the system dimension, R . It is, however, the basic parameter for any given boundary value transport problem.

2.

A GENERAL ANALYSIS OF THE DIFFERENT GAS FLOW REGIMES.

The problem of the steady-state flow of a rarefied gas past a body has a characteristic dimensionless parameter. This parameter is the Knudsen number described in Section 6.1. Large or small values of the Knudsen number permit simplification or modification of the Boltzmann equation for each particular case. Corresponding to certain ranges of the Knudsen number, there are four regimes in the Kinetic Theory of Gases. This classification has a schematic character. For conditions when Kn o 0 ( Kn 0.01 , the continuum regime), one may use the Navier-Stokes equations with the usual hydrodynamic boundary conditions (Eqs. (2-7)-(2-9) together with Eqs. (5-51) and (5-54)). If Kn 1 ( 0.01 Kn 0.1 , the slip-flow regime), the gas may be described by the Navier-Stokes equations but the boundary conditions should be modified by the introduction of slip conditions on the surface of the body. The slipflow regime will be described in detail in Chapter 10. The other regions are those in which the Knudsen number is greater than unity. Two sub-regions can be distinguished here. The limiting regime, where Kn o f , is usually called the free-molecular regime. If the dimensions of the body are insignificant compared with the mean free path, one says that the flow past the body is of the free-molecular type. For such flows the presence of the body does not disturb the distribution function of the incident molecules of the gas in the neighborhood of the body. In this case, the molecules that rebound from the surface of the body collide with other molecules at distances on the order of the mean free path from the body. Therefore, in a free-molecular flow, the collisions between the molecules that rebound and those coming from the gas occur at a great enough distance from the body that the influence of these collisions can be neglected. The free-molecular distribution function is constant along the molecular trajectories. On the surface of the body the distribution function of incident molecules is the same as that in the surrounding volume of gas far from the body. The distribution function of molecules rebounded from the body can be found from the boundary conditions which will be considered in Section 6.3. The region where Kn 1 is usually called the near free-molecular regime. For flow in this regime, Knudsen iteration [1] is a well known

Chapter 6. Regimes of Rarefied Gas Flows

77

approximate solution of the Boltzmann equation. The distribution function is assumed to have the form:

f

0 1 f Kn 1 f .

(6-3)

The function, f , may be found from the non-homogeneous linear partial differential equation that contains the free-molecular distribution function inside the collision integral. The solutions of several specific problems of interest in this regime may be found in [2-9]. The region where Kn ~ 1 is the most difficult to describe theoretically. In this region the flow has features typical of both the free-molecular and continuum regimes. In this transition region, it is often convenient to use the various moment methods that will be discussed in Chapter 11 in order to find the solutions to boundary problems of interest. 1

3.

THE BOUNDARY CONDITIONS.

Both the mathematical formulation and the solution of all boundary value transport problems, by both the direct and the moment methods, requires knowledge of the applicable microscopic boundary conditions. These conditions on the surface of a body can be formulated for the distribution function of the reflected molecules by means of a dispersion kernel [10-15]. Within this framework only a general description may be provided at present, which cannot be used to solve practical boundary problems because no theory currently exists from which the necessary kernels can be derived. The absence of a strict theory for the molecular interaction of a gas with a solid surface results in the need to use definite models to postulate the distribution function for the reflected molecules. This function contains a certain number of free parameters which are usually called accommodation coefficients. These coefficients may be determined by a comparison of the theory constructed from a given model with an appropriate experiment. Maxwell was the first to propose the boundary model that has been widely used in various modified forms. At present, this model still appears to be the most convenient and correct formulation for the various boundary value transport problems; particularly those in which one assumes a noncondensable gas for which the boundary surface is impenetrable. The Maxwellian boundary model is constructed on the assumption that some fraction 1 DW of the incident gas molecules are reflected from the surface specularly, while the remaining fraction, DW , are reflected diffusely with a Maxwellian distribution. This supposition can be expressed as:

78

Analytical Methods for Problems of Molecular Transport

f v, rS , t DW f r 1 DW f vc, rS , t ,

(6-4)

where f r are the distribution functions of the molecules reflected and incident on the surface at a point, rS , vc v 2 v n n , n is the unit normal vector to the surface at the given point, and f r is the Maxwellian distribution function given by:

fr

§ · m nr rS , t ¨ ¨ 2S kT r , t ¸¸ r S © ¹

32

§ · mv 2 exp ¨ . ¨ 2kT r , t ¸¸ r S © ¹

(6-5)

There are two parameters in this distribution function, nr and Tr , which correspond to the number density and temperature of some fictitious gas emitted by the surface. Also, in this form of the reflected distribution function, the surface element is assumed to be stationary. Now, consider the physical meaning of the coefficient, DW . The tangential momentum transferred to a unit surface element per second by incident molecules is given by:

PW

m

f vW v n dv ,

³

(6-6)

vn 0

and that carried away by the reflected molecules may be expressed as:

PW

m

³

f vW v n dv

1 DW PW

.

(6-7)

vn ! 0

From Eqs. (6-6) and (6-7), the following relation can be written:

DW PW

PW PW ;

0 d DW d 1 .

Therefore, DW gives the fraction of the tangential momentum of incident molecules transmitted to the surface by all molecules. This parameter is usually called the tangential momentum accommodation coefficient. The unknown parameters, nr and Tr , in Eq. (6-5) may be found if the impenetrability condition for the surface and the energy accommodation coefficient, DT , are introduced by means of the relations:

N N

0,

(6-8)

Chapter 6. Regimes of Rarefied Gas Flows

79

and:

DT

Q Q Q Qeq

;

0 d DT d 1 ,

(6-9)

where N # and Q # are the fluxes of the number of molecules and energy for the incident and reflected molecules across a unit element of the surface at the point rS per unit time. The value Qeq is the energy flux which would be carried away by the reflected molecules if the gas were in equilibrium with the surface, i.e., when Tr Tw , where Tw is the temperature of the surface at the point rS . The energy accommodation coefficient can be interpreted as the fraction of the equilibrium heat flux transmitted to the surface by the gas molecules. The details of the calculational technique associated with this formulation of the boundary problem are given in detail in Problem 6.3. The accommodation coefficients, DW and DT , may be found via experiment. The classical Millikan experiments [16], for instance, have shown that for most cases DW # 1 . This would seem to indicate that the model of diffuse reflection is generally the most appropriate model to employ in practice. The boundary conditions needed for a condensable gas are significantly different from those considered above. On a liquid surface, for instance, the evaporation (condensation) coefficient, D m , should be introduced for the gas (the vapor phase of the liquid comprising the surface) via the relationship:

Dm

N N N NS

,

(6-10)

where N S is the reflected flux of molecules which have the parameters of a saturated vapor at the applicable surface temperature. This coefficient should be interpreted to mean that a fraction of the incident flux of molecules at the surface, D m , is reflected from the surface having the parameters of a saturated vapor and that the remaining fraction of the incident molecules, 1 D m , leaves the surface without having undergone a phase transition. This latter fraction forms what is, in essence, a non-condensable gas for which the parameters may be specified by an impenetrability condition for the surface and an energy accommodation coefficient, DT , as was done above for the case of a purely non-condensable gas. One must keep in mind, however, that the use of an energy accommodation coefficient in such a scenario makes sense only for the non-

80

Analytical Methods for Problems of Molecular Transport

condensable part of the gas. The evaporating molecules must have the equilibrium parameters of a saturated vapor. To be consistent with the definition of the evaporation coefficient, the reflected distribution function can be expressed in the form:

f v, rS , t § · m D m nS Tw rS , t ¨¨ ¸¸ © 2S kTw rS , t ¹

32

§ · m 1 D m nr rS , t ¨ ¨ 2S kT r , t ¸¸ r S © ¹

§ · mv 2 exp ¨ ¨ 2kT r , t ¸¸ w S © ¹

32

(6-11)

§ · mv 2 exp ¨ , ¨ 2kT r , t ¸¸ r S © ¹

where nS Tw rS , t is the saturated vapor density at the given surface temperature Tw rS , t which, in the general case, is a function of the specific location on the surface, rS , and the time, t . The unknown parameters associated with the non-condensable fraction of the reflected molecules may be found using the standard impenetrability and energy accommodation conditions. A detailed description of this technique is given in Problem 6.4.

4.

THE BOUNDARY DISPERSION KERNEL.

The Maxwellian model of the boundary conditions is frequently used to analyze boundary value transport problems. This model gives a simple expression for the distribution function of molecules reflected from the surface of the wall. For a non-condensable gas, the diffusely reflected part of this distribution function contains two unknown parameters which may be defined by introducing the energy accommodation coefficient and the impenetrability condition for the surface. Note that the expression for the energy accommodation coefficient contains an additional unknown parameter connected with the number density of the molecules that are reflected and which are in thermal equilibrium with the wall which is assumed to be located at x 0 and to have a constant temperature, T0 . Hence, three additional integral conditions must be used to obtain the unknown parameters. For the linearized case, if evaporation (condensation) is absent at the wall, a general form of the boundary condition may be derived in which these additional relations are taken into account by the dispersion kernel, W c, c1 :

81

Chapter 6. Regimes of Rarefied Gas Flows

) f c,0

³ W c, c1 ) f c1 ,0 dc1

,

(6-12)

where f r c, x f 1 ) rf c, x and ) rf c, x are the full corrections to the distribution function which, in general, contain the Chapman-Enskog distributions. For the correction to the reflected distribution function, the Maxwellian boundary condition may be written as: 0

1 DW ) f cc,0 DW ª¬Q r c 2 32 W r º¼

) f c,0

,

(6-13)

where:

cc c 2 c n n Using Eq. (6-13) and Eqs. (6-8) and (6-9), one can obtain:

Qr

Wr

^ `^2

S 1 ³ cx exp c 2

1 2

1 DT 2 c 2 `) f dc

1 DT S 1 ³ cx 2 c 2 exp ^c 2 `) f dc

,

.

Then, Eq. (6-13) may be written as:

1 DW ) f cc,0 DW S 1 ³ c1 x exp ^c12 `

) f c,0

(6-14)

^

u 2 1 DT 2 c

2

2 c12

`

) f c1 ,0 dc1 ,

Allowing for Eq. (6-12), one can express the dispersion kernel in the form:

W c, c1

1 DW G ª¬c1 c 2 c n n º¼

^ `^2 1 D 2 c 2 c ` ,

S 1DW c1 x exp c12

2

T

where G x is the Dirac delta function [17].

2 1

(6-15)

82

Analytical Methods for Problems of Molecular Transport

Figure 6-1. The variation of the mean velocity of a gas near a moving wall. Here, uw is the velocity of the wall, 'u is the slip velocity, and u 0 is the real mean velocity at the wall.

5.

FEATURES OF THE BOUNDARY CONDITIONS FOR SMALL KNUDSEN NUMBER.

Consider the gas flow along a plane wall under conditions in which the mean velocity has only a tangential component that is a linear function of the normal coordinate beyond the wall. This linear dependence may be obtained from the solution of both the Boltzmann equation and Navier-Stokes equations. Since the Navier-Stokes distribution function is a solution of the Boltzmann equation only at large distances from the wall, the gas state in the Knudsen layer must be investigated in order to establish the boundary conditions for Navier-Stokes equations [18-26]. Let the solid curve (Fig. 6-1) be the variation of the mean velocity of a gas that has been obtained by the solution of the Boltzmann equation. Then, the linear extrapolation of the asymptotic solution (outside of the Knudsen layer) of this equation gives a fictitious value of mean velocity at the wall, uw 'u . If one uses this fictitious velocity, uw 'u , on the wall as a boundary condition for the Navier-Stokes equations, then the same velocity profile may be obtained as from the solution of the Boltzmann equation outside of the Knudsen layer. The difference between the fictitious velocity and the actual wall velocity is called the slip velocity. The most reliable expression

Chapter 6. Regimes of Rarefied Gas Flows

83

for the slip velocity has been obtained by Ivchenko, Loyalka and Tompson [27]:

'u

§ wu · cm OP ¨ ¸ , © wx ¹f

(6-16a)

where:

cm

2 DW

DW

5 8

ª 0.6690 0.1775 DW º », «¬ 0.7549 0.09714 DW »¼

S«

(6-16b)

is the isothermal-creep coefficient, the molecules have been assumed to be 205 rigid spheres, and OP 202 O for the second-order Chapman-Enskog solution. In the very important case when DW 1 , one obtains:

'u

wu · ¸ . © wx ¹f

1.1006 OP §¨

(6-17)

The temperature-jump may be introduced in the same way (Fig. 6-2). The most reliable analytical expression for the temperature-jump has been obtained by Loyalka [18]. Loyalka’s formula, if the tangential momentum accommodation coefficient is taken into account, can be expressed as [28]:

'T

§ wT · cT ON ¨ ¸ , © wx ¹f

(6-18a)

where:

cT

75 128

S

2 DW DT ª DW DT 2 DW § 52 H T 1 · º », «1 DW DT ¬« 2 DW DT ¨© 25 S 2 ¸¹ »¼

(6-18b)

45 is the temperature-jump coefficient, ON 44 O , and H T 0.9378 for rigidsphere molecules if the second-order Chapman-Enskog solution is employed. Consider a gas having a tangential temperature gradient located over a non-uniform heated plane surface (Fig. 6-3). Under these conditions, the gas has a mean velocity in the direction of the temperature gradient, i.e. the gas

84

Analytical Methods for Problems of Molecular Transport

Figure 6-2. The variation of the gas temperature near a wall. Here, Tw is the wall temperature, 'T is the temperature-jump, and T 0 is the real temperature of the gas at the wall

slips along the wall. The gas velocity beyond the wall is the thermal-creep velocity or the thermal-creep velocity. The most accurate analytical expression for the thermal-creep velocity has been obtained by Ivchenko, Loyalka and Tompson [28,29]:

usl

cTslQ

w ln T , wy

(6-19)

where:

cTsl

3 2

0.4354 0.2179 DW 0.8518 0.1096 DW

,

is the thermal-creep coefficient, Q ON OP Q , Q P U is the kinematic viscosity, and molecules have been assumed to be rigid spheres. For two specific cases of interest, this coefficient may be written as:

cTsl

1.0193 ; DW ® ¯0.7667 ; DW

1, 0 .

(6-20)

85

Chapter 6. Regimes of Rarefied Gas Flows

Figure 6-3. The thermal-creep geometry.

The slip boundary conditions may be obtained only from the solution of the Boltzmann equation with microscopic boundary conditions for the distribution function. These problems will be considered in Chapter 8.

PROBLEMS 6.1. Using the Maxwellian boundary model, determine the form of the reflected distribution function if the boundary surface is moving and uc is the wall velocity. Solution: The mean velocity of the diffusely reflected molecules is equal to uc . Therefore, the reflected distribution function may be written as:

f

v, rS

§ · m DW nr rS ¨¨ ¸¸ © 2S kTr rS ¹

32

§ m v uc 2 exp ¨ ¨ 2kTr rS ©

1 DW f vx u cx , v y , vz , rS

· ¸ ¸ ¹

,

where x is a normal coordinate.

6.2. Determine the form of the dispersion kernel which is specified by:

f v, rS

³

W v1 , v f v1 , rS dv1 ,

v1 n 0

for the Maxwellian boundary model. Solution: The distribution function of reflected molecules is given by Eqs. (6-4) and (6-5). The condition of wall impenetrability gives:

86

Analytical Methods for Problems of Molecular Transport 12

§ · m nr rS 2 S ¨ v1 n f v1 , rS dv1 . ³ ¨ 2kT r ¸¸ r S ¹ © v1 n 0 Using this expression, one can obtain:

W v1 , v

D W 2S

2

§ m · § mv 2 · ¨¨ ¸¸ exp ¨¨ ¸¸ v1 n © kTr rS ¹ © 2kTr rS ¹ 1 DW G ª¬ v1 v 2 v n n º¼ .

6.3. Write down the full system of boundary conditions at the solid impenetrable wall if the wall temperature is T0 . Solution: The number density, nr , and the temperature, Tr , in Eqs. (6-4) and (6-5) are defined by N N 0 and:

DT

Q Q Q Qeq

,

where N r and Q r are normal components of the relative fluxes. The flux, , is determined by means of the reflected distribution function in which Qeq nr neq and Tr T0 . The unknown equilibrium number density, neq , may be found from N eq N 0 .

6.4. At a liquid surface, write down the boundary conditions for reflected molecules of the same liquid vapor if the temperature of the surface is slightly different from the equilibrium temperature T0 , Tw T0 1 W w . Solution: The reflected distribution function may be presented in the form:

f

v,0

§ m · D m nS Tw ¨ ¸ © 2S kTw ¹

32

§ mv 2 · exp ¨ ¸ © 2kTw ¹

§ m · 1 D m nr ¨ ¸ © 2S kTr ¹

32

§ mv 2 · exp ¨ ¸ , © 2kTr ¹

where the first term describes molecules evaporated from the liquid surface while the second term is related to molecules reflected from the surface without a phase transition. It should be noted that, for this latter group of

87

Chapter 6. Regimes of Rarefied Gas Flows

molecules, it is appropriate to treat the liquid surface as if it were impenetrable and to employ an appropriate surface condition. The unknown parameters nr and Tr , may be determined from the relations 1 D m N N r 0 and:

DT

1 D m Q Qr 1 D m Q Qreq

,

where Qreq contains an additional unknown parameter related to the density of a fictitious gas reaching equilibrium with the surface if Tr Tw . This parameter may be found from 1 D m N N req 0 . The fluxes, N r , N req , Qr , and Qreq , are calculated by using the second term of the reflected distribution function. The number density of the saturated vapor at the temperature, Tw , may be found by using the Clausius-Clapeyron equation [30] which yields nS Tw nS T0 ª¬1 ] 1 W w º¼ where ] h fg kT0 and h fg is the latent heat of condensation for a vapor molecule.

6.5. Generalize Eq. (6-14) to obtain an expression for the distribution function of the rebounding molecules for a case when the equilibrium temperature of the wall is given by Tw T0 'T where 'T T0 . Solution: The boundary condition for a correction to the distribution function may be written as:

) f c,0

1 DW ) f cx , c y , cz ,0 DW ª¬Q r c 2 32 W r º¼

,

where the two boundary quantities, Q r and W r , may be found by introducing the energy accommodation coefficient, DT , together with the condition of impenetrability of the wall for molecules of the gas. This yields:

) f c,0

1 DW ) f cx , c y , cz ,0 ª 2G · º 2G ½° §G ° DW ® 2 c 2 « DT W w 1 DT ¨ 2 1 ¸ » 1 ¾ , S ¹ ¼ S ¿° ©S ¬ ¯°

where:

Ww G1

'T , T0 2 ³) f c,0 cx exp c dc

,

G2

³) f c,0 cx c

2

exp c 2 dc .

88

Analytical Methods for Problems of Molecular Transport

6.6. Generalize Eq. (6-14) to obtain an expression for the distribution function of the rebounding molecules for a case when the evaporation/condensation processes at the wall are taken into account and the equilibrium temperature of the surface is given by Tw T0 'T where 'T T0 and T0 is the equilibrium vapor temperature. Solution:

) f c,0 D m ªQ w c 2 32 W w º 1 D m ªQ r c 2 32 W r º , ¬

¼

¬

¼

where Q w is the correction to the number density of the saturated vapor at the temperature, Tw T0 1 W w , and Q r and W r , may be found by introducing the energy accommodation coefficient, DT , and using the impenetrability condition for molecules reflected without a phase transition. This yields:

) f c,0 D m ªQ w c 2 32 W w º ¬

¼

ª 2G · º 2G °½ §G ° 1 D m ® 2 c 2 « DT W w 1 DT ¨ 2 1 ¸ » 1 ¾ , S ¹ ¼ S °¿ ©S °¯ ¬

where:

Ww G1

'T , T0 2 ³) f c,0 cx exp c dc

,

G2

³) f c,0 cx c

2

exp c 2 dc .

The correction, Q w , may be found by using the Clausius-Clapeyron equation [30] which yields nS Tw nS T0 ª¬1 ] 1 W w º¼ where ] h fg kT0 and h fg is the latent heat of condensation (evaporation) for a vapor molecule.

REFERENCES 1. Keller, J.B., “On the Solution of the Boltzmann Equation for Rarefied Gases,” Comm. Pure App. Math. 1(3), 275-284 (1948). 2. Rose, M.H., “Drag on an Object in Nearly-Free-molecular Flow,” Phys. Fluids 7(8), 12621269 (1964).

Chapter 6. Regimes of Rarefied Gas Flows

89

3. Liu, V.C., Pang, S.C., and Jew, H., “Sphere Drag in Flows of Almost Free Molecules,” Phys. Fluids 8(5), 788-796 (1964). 4. Brock, J.R., “The Thermal Force in the Transition Region,” J. Colloid Interface Sci. 23(3), 448-452 (1967). 5. Brock, J.R., “Highly Non-equilibrium Evaporation of Moving Particles in the Transition Region of Knudsen Number,” J. Colloid Interface Sci. 24, 344-351 (1967). 6. Brock, J.R., “The Diffusion Force in the Transition Region of Knudsen Number,” J. Colloid Interface Sci. 27(1), 95-100 (1968). 7. Ivchenko, I.N. and Yalamov, Yu.I., “On the Thermophoresis of Aerosol Particles in the Almost-Free-molecular Regime,” Izv. AN SSSR, M. Zh. G. (3), 3 (1970). 8. Kelly, G.E. and Sengers, J.V., “Droplet Growth in a Dilute Vapor,” J. Chem. Phys. 61(7), 2800-2807 (1974). 9. Barrett, J.C. and Shizgal, B., “Condensation and Evaporation of a Spherical Droplet in the Near Free Molecule Regime,” in Rarefied Gas Dynamics: Physical Phenomena, Progress in Astronautics and Aeronautics Series, V. 117, edited by Muntz, E.P., Weaver, D.P., and Campbell, P.H. (AIAA, Washington, D.C., 1989) pp. 447-459. 10. Gross, E.P., Jackson, E.A., and Ziering, S., “Boundary Value Problems in Kinetic Theory of Gases,” Annals of Physics 1(2), 141-167 (1957). 11. Cercignani, C., Theory and Application of the Boltzmann Equation (Scottish Academic Press, Edinburgh, U.K., 1975). 12. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, NY, 1969). 13. Barantsev, R.G., Rarefied Gas Interactions with Streamlined Surfaces (Nauka, Moscow, 1975). 14. Goodman, F.O. and Wachman, H.Y., Dynamics of Gas-Surface Scattering (Academic Press, New York, 1976). 15. Kušþer, I., “Phenomenology of Gas-Surface Accommodation,” in Proceedings of the Ninth International Symposium on Rarefied Gas Dynamics, edited by Becker, M. and Fiebig, M. (DFVLR Press, Pozz-Wahn, 1974). Vol. 2, p. E.1-1. 16. Millikan, R.A., “Coefficients of Slip in Gases and the Law of Reflection from the Surfaces of Solids and Liquids,” Phys. Rev. 21(1), 217-238 (1923). 17. Dirac, P.A.M., The Principles of Quantum Mechanics, 4-th edition revised (Clarendon Press, Oxford, reprinted 1993). 18. Ivchenko, I.N. and Yalamov, Yu.I., “The Hydrodynamic Method of Calculation of the Thermophoresis of Aerosol Particles,” J. Phys. Chem. (Russia) 45(3), 577-582 (1971). 19. Loyalka, S.K., “Momentum and Temperature-Slip Coefficients with Arbitrary Accommodation at the Surface,” J. Chem. Phys. 48(12), 5432-5436 (1968). 20. Hidy, G.M. and Brock, J.R., Topics in Current Aerosol Research, part 2 (Pergamon Press, 1972). 21. Derjaguin, B.V., Ivchenko, I.N., and Yalamov, Yu.I., “About Construction of Solutions of the Boltzmann Equation in the Knudsen Layer,” Izv. AN SSSR, M. Zh. G. (4), 167-172 (1968). 22. Derjaguin, B.V. and Yalamov, Yu.I., “The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and Their Experimental Testing,” in Topics in Current Aerosol Research, part 2, edited by Hidy, G.M. and Brock, J.R. (Pergamon Press, 1972). 23. Loyalka, S.K. and Lang, H., “On Variational Principles in the Kinetic Theory,” in Proceedings of the Seventh International Symposium on Rarefied Gas Dynamics (Editrice Tecnico Scientifica, Pisa, Italy, 1970).

90

Analytical Methods for Problems of Molecular Transport

24. Loyalka, S.K., “Slip and Jump Coefficients for Rarefied Gas Flows: Variational Results for Lennard-Jones and n(r)-6 Potentials,” Physica A163, 813-821 (1990). 25. Rolduguin, V.I., Application of the Non-Equilibrium Thermodynamics Method in Boundary Problems of the Kinetic Theory of Gases (M.S. Thesis, Moscow, 1979). 26. Savkov, S.V., The Slip-Flow Boundary Conditions for the Non-Uniform Binary Gas Mixture and an Application of Them in Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 27. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “The Precision of Boundary Models in the Gas Slip Problem,” High Temp. 31(1), 127-129 (1993). 28. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “On the Use of Conservation Laws in Plane Slip Problems,” Teplofizika Vysokikh Temperatur (in Russian), 33(1), 66-72 (1995). 29. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “On One Boundary Model for the Thermal Creep Problem,” Fluid Dynamics 28(6), 876-888 (1993). 30. Landau, L.D. and Lifshitz, E.M., Statistical Physics (Pergamon, London, Addison Wesley, Reading, Mass., 1958).

Chapter 7 THE FREE-MOLECULAR REGIME

1.

THE FREE-MOLECULAR DISTRIBUTION FUNCTION.

The theoretical treatment of gas flow problems for extremely high Knudsen numbers (the free-molecular regime) is very different from that typically encountered in conjunction with the hydrodynamics of media having the higher number densities most people are used to working with. The theoretical frameworks encountered at these extremes are connected via various fundamental relationships between the Kinetic Theory of Gases and the mechanics of media that exhibit continuum type behaviors (the continuum regime). The continuum equations may be obtained from the Boltzmann equation if one uses the summational invariants of encounters as molecular properties with which to construct a system of moment equations. This moment system, however, is indeterminate unless one knows the relationships between the basic hydrodynamic values and the stress tensor or the thermal flux vector. These relationships may be derived by using the Chapman-Enskog method of solution of the Boltzmann equation. It is very important in the continuum regime that such gas characteristics as the viscosity and thermal conductivity coefficients be determined independently of the boundary conditions. The boundary problems for high Knudsen numbers are quite different. The kinetic analysis may only be used for the description of gas flows. The general solution of the Boltzmann equation requires usage of appropriate boundary conditions for the distribution function itself which ultimately results in a dependence of the hydrodynamic functions on the boundary parameters of the specific problem under consideration. Hence, in this

92

Analytical Methods for Problems of Molecular Transport

regime there are no general relationships between the various macroscopic quantities. The free-molecular regime will be examined in this chapter. A general characteristic of gas flows in this regime is that the presence of a body does not effectively disturb the distribution function of the impinging molecules since the collisions between molecules being reflected and those coming from infinity occur at distances on the order of O from the surface of the body. Since O R , where R is the typical dimension of the body, the influence of these collisions can be neglected. Owing to this simplification, the theory describing free-molecular flows has been sufficiently developed that several useful solutions to applied transport problems have been obtained [1-6]. A rigorous theory of transport problems must start with the Boltzmann equation which, for stationary conditions, when Kn o f , may be written as:

v

wf wr

vx

wf wf wf vy vz wx wy wz

0 ,

or:

0 .

(7-1)

Eq. (7-1) is a linear partial differential equation of the first-order. The auxiliary system of ordinary differential equations, usually called the characteristic or Lagrange system, is:

dx vx

dy vy

dz vz

df 0

dt ,

(7-2)

where t is an auxiliary parameter used to ease the solution of this system of equations. The characteristics of these equations are the straight lines specified by:

r r0 v t t0 ,

(7-3)

where t denotes the value of the parameter on the characteristic (i.e. t is not time). These characteristics may be interpreted to be the molecular trajectories associated with this problem [7]. Eq. (7-2) means that:

Chapter 7. The Free-Molecular Regime

93

Figure 7-1. The cone of influence for a sphere.

df dt

0 .

(7-4)

which implies that f is constant along the molecular trajectories. For the molecular trajectories that start at infinity and at the surface of the body the distribution functions are, therefore, equal to the values of the distribution

94

Analytical Methods for Problems of Molecular Transport

functions at infinity in the undisturbed flow and at the point, P rS , on the surface. These functions are not the same and, hence, the free-molecular distribution function is discontinuous in velocity space. Consider the form of the distribution function at the point, M r , in various velocity domains when a spherical model of the particle is assumed. At each spatial point, M r , there exists a cone of influence, the surface of which is formed by the extensions of the tangents to the surface of the particle passing through the point M r (see Fig. 7-1). This conical surface in the velocity space is described by F F 0 , F 0 arcsin R r arcsin 1 r where R is the radius of the particle and F is the angle between vectors c and r . In the velocity space, the surface of the cone of influence is the surface of discontinuity for the distribution function. This surface divides the molecules into two groups with distribution functions f1 and f 2 ( f 2 describes the molecules for which vr ! vr where vr v cos F 0 , and f1 describes the remaining molecules). The free-molecular distribution function in the full velocity space may be expressed as [8]:

f

1 2

f1 f 2 12 f 2 f1 sign vr vr

.

(7-5)

This relationship is a generalization of expression previously proposed for the planar geometry [9,10]. On the particle surface, vr 0 , and the distribution functions of the reflected and incident molecules may be written as:

° f v, Rn f v, Rn ® °¯ f v, Rn

f2

;

vr ! 0 ,

f1

;

vr 0 ,

where n is a unit vector normal to the surface element and pointing into the gas. These distribution functions are connected by the boundary condition given in Eq. (6-4). Having derived the distribution function one may calculate the flux of any molecular property across the particle surface. For example, the force, F , on the particle is given by integrating the net momentum transferred to the particle by the gas molecules over the particle surface, S , per unit time:

F

³ dS ¦ ³ mv v n f r v, rS dv , r r

(7-6)

95

Chapter 7. The Free-Molecular Regime

Figure 7-2. The geometry of the sphere drag problem.

where dS is the area of the surface element, f and f indicate the distribution functions at the particle surface for both incident v n 0 and reflected v n ! 0 molecules, respectively, and rS is the position vector of an arbitrary point on the surface. For evaporation (condensation) problems one should know the flux of the number of molecules across the particle surface, S , per unit time. This flux may be given by:

N

r ³ dS ¦r ³ v n f v, rS dv

,

(7-7)

r

where N is positive or negative depending on whether the process is connected with evaporating or condensing particles.

96

Analytical Methods for Problems of Molecular Transport

2.

THE FORCE ON A PARTICLE IN A UNIFORM GAS FLOW.

Consider a small spherical particle R O in a uniform gas flow for which the velocity at infinity is a constant value, u (Fig. 7-2). A coordinate system is introduced in which the polar axis has the same direction as the vector, u . Then, u ur e r uT eT where ur u cos T and uT u sin T . From the symmetry of this problem, it is clear that the force has only one component in the direction of the vector u . This component may be given by:

Fu

S ° ½° 2S R 2 m ³ sin T ®¦ ³ vr vr cos T vT sin T f r dv ¾ dT , (7-8) 0 ¯° r r ¿°

where the integration over all values of the azimuthal angle has been performed. 12 Suppose that u 2kT m . In this case of low velocity of the undisturbed flow, the following approximation for the distribution function of the incident molecules may be used:

f

0 f 1 2u c

^

`

0 f 1 2u ª¬cr cos T cT sin T º¼ ,

(7-9)

where: 12

u

12

§ m · § m · ¨ 2kT ¸ u , c ¨ 2kT ¸ © ¹ © ¹

0 v , f

§ m · n0 ¨ ¸ © 2S kT ¹

32

exp c 2 .

For the distribution function of the reflected molecules, one can use the Maxwellian model of the boundary conditions given by Eq. (6-4) in which the density of molecules reflected diffusely may be written as: nr rS n0 1 Q rS , where Q rS is proportional to the small parameter u which is a linearization parameter. The linearized boundary condition may be expressed in the form:

97

Chapter 7. The Free-Molecular Regime

f v, rS

1 DW f cr , cT , cI , rS § m · DW n0 ¨ ¸ © 2S kT ¹

32

exp c 2 1 Q rS .

(7-10)

The unknown function Q rS may be found from the condition that there is no evaporation or condensation of molecules at the surface element surrounding the point, rS (i.e. the particle surface is impenetrable for the gas molecules). This condition may be written as:

n ³ cf dv n ³ cf dv 0 .

(7-11)

Having performed the integration in Eq. (7-11), one obtains:

Q rS S u n .

(7-12)

Substituting Eqs. (7-9), (7-10), and (7-12) into Eq. (7-8), the following expression may be written:

Fu

2S R

2

U § 2kT ·

S

¨ ¸ u sin T S 3 2 © m ¹ ³0

° u ®2 1 DW ³ cr ª¬cr2 cos 2 T cT2 sin 2 T º¼ exp c 2 dc °¯

DW S ³ cr ª¬ cr cos 2 T cT sin T cos T º¼ exp c 2 dc

½° 2 2 ³ cr ª¬cr cos T cT sin T º¼ exp c 2 dc ¾ dT . °¿

After some straightforward integrations, one obtains:

F

8 3

12

R 2 n0 2S mkT

1 18 DW S u .

(7-13)

The same expression for the free-molecular force has been obtained by Waldmann [11].

98

Analytical Methods for Problems of Molecular Transport

3.

CALCULATION OF MACROSCOPIC VALUES IN THE FREE-MOLECULAR REGIME.

Each macroscopic quantity may be expressed as the mean value of a molecular property, I v, r , by the relation:

³ I fdv

nI

.

(7-14)

For the free-molecular regime, this integral contains the discontinuous distribution function given by Eq. (7-5). Substituting Eq. (7-5) into Eq. (714), one obtains:

nI

³

I f1dv

1 2

³ I f 2 f1 dv

.

(7-15)

2

where, in the first integral, the integration extends over all values of the molecular velocity while, in the second integral, the integration extends only over that part of the velocity range for which vr ! vr (region 2 in the velocity space). For spherical geometries it is convenient to use spherical coordinates in the velocity space to perform the integration over region 2. These coordinates are given by:

cr

c cos F ,

cT

c sin F sin E ,

cI

c sin F cos E .

The integration over region 2 may be represented as:

³

2

dc

f

2S

w

0

0

0

2 ³ c dc ³ d E ³ sin F d F ,

(7-16)

where w arcsin 1 r and r r R . The main subtleties here are connected with the integration of the diffusely and specularly reflected parts of the distribution function. These parts always contain the scalar product of n S n rS and the typical flow vector, u . To complete the integration over region 2 one must find the connection between the unit vector n S and the molecular velocity vector represented by the characteristic PM (see Fig. 7-1). The vector n S can be expressed as:

99

Chapter 7. The Free-Molecular Regime

nS

r c PM . R c R

The value PM may be found from the triangle OPM (see Fig. 7-1) by solution of the quadratic equation:

PM 2 2r cos F PM r 2 R 2

0 ,

that may be written as: PM

r cos F r R 1

r 2 sin 2 F . R2

In this formula, only the minus sign may be used in order that, as r o f and F o 0 , then PM r R . Then, the expression for n S n S r , c may be written as:

n S r , c

º r c ª r r 2 « cos F 1 2 sin 2 F » . R c «R R »¼ ¬

(7-17)

Now, as an example, the integration required for the problem of freemolecular flow past a sphere that was discussed in Section 7.2 will be considered in order to calculate the number density of the flow. The distribution function may be written as:

f

f1 ° ° ® f2 ° ° ¯

1 D f 1 2cc u D f ª1 S n u º ; ¬ ¼ 0 f 1 2cr ur 2cT uT ; 0

cr cr ,

(7-18)

W

0

W

S

cr !

cr

,

where cc c 2n S n S c and for integration of the reflected distribution function one should use Eq. (7-17) for n S . To simplify what would otherwise be a very complicated set of integrations the conclusion of this calculation will be shown for the particular case when molecules are assumed to undergo only diffuse reflection at the surface ( DW 1 ). Then, the number density is expressed by:

100

Analytical Methods for Problems of Molecular Transport

n

f1dv

³

1 2

³ f 2 f1 dv

,

(7-19)

2

where the integrand of the second term has the form:

0 f ª S n S r, c u 2cr ur 2cT uT º . ¬ ¼

f 2 f1

After integration of the first term in Eq. (7-19) one obtains:

i1

f1dv

³

n0 .

1 2

The second integral may be written as:

i2

f

n0

2S

w

0

0

2 2 ³ c exp c dc ³ d E ³ sin F

S3 2

0

^ S

ur

ª r sin ¬«

2

F cos F

1 r 2 sin 2 F º» ¼

(7-20)

`

2ur c cos F d F . Basic integration gives:

i2

n0ur F r ,

(7-21)

where:

F r

1 2

^

`

S r 1 a 13 r 1 a 3 r 2 ª¬ 13 2S 1 º¼ ,

and a 1 r 2 . The vanishing integrals have been eliminated from Eq. (720). Now, the number density may be expressed in the form:

n r n0 ª¬1 u F r cos T º¼ .

(7-22)

For the limiting case when r o f the number density n r o n0 which is equivalent to an undisturbed flow.

101

Chapter 7. The Free-Molecular Regime

4.

THERMOPHORESIS OF PARTICLES IN THE FREE-MOLECULAR REGIME.

It has been observed that a temperature gradient in a gas causes small particles suspended in the gas to migrate in direction of decreasing temperature. The force arising from the temperature gradient causing the migration of the particles is commonly termed the thermal force and the phenomenon is usually known as thermophoresis. Previously, several theoretical descriptions of the thermophoresis of particles in the free-molecular regime, Kn o f , have been given [11-16]. In these papers, calculations are made of the thermal force from the net momentum transferred by gas molecules striking and reflecting from the particle surface. The physical model that is used is a single spherical particle of radius, R , in an infinite gas with a temperature gradient, T , which is constant at large distances from the particle. It is assumed that the gas as a whole is in mechanical equilibrium so that at large distances from the particle there is no pressure gradient. Introduce a spherical coordinate system r ,T , I with its origin at the center of the stationary spherical particle and a polar axis, x , in the direction of the vector, T f (see Fig. 7-3). Then, the temperature distribution at large distances from the particle has the form:

T

T r T0 T x r cos T T0 1 q r ,

(7-23)

where q (T )f T0 . The condition, O ln T 1 , is assumed to be satisfied at large distances from the particle. Then, the distribution function is the ChapmanEnskog distribution, that is given by:

f

^

` ,

f eq 1 a1 q c S3 2 c 2 1

1

(7-24)

where a1 15 S O for rigid-sphere molecules and f eq is the local 16 Maxwellian distribution that may be written as: 1

f eq

§ · m n r ¨ ¨ 2S kT r ¸¸ © ¹

32

§ mv 2 · exp ¨ . ¨ 2kT r ¸¸ © ¹

(7-25)

102

Analytical Methods for Problems of Molecular Transport

Figure 7-3. The geometry of the thermal force problem.

From the condition, n0 kT0 represented as:

n r

n0T0 T r

n r kT r , n r in Eq. (7-25) may be

n0 . 1 q r

For r ~ O the term q r 1 and Eq. (7-25) may be represented as a power series in q r . Considering only the linear term in this expansion, Eq. (725) may be written as:

f eq where:

0 f ª1 c 2 52 q r º , ¬ ¼

(7-26)

103

Chapter 7. The Free-Molecular Regime 0 f

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

§ mv 2 · exp ¨ ¸ . © 2kT0 ¹

From the symmetry of the problem under consideration, it is clear that the thermal force has only one component in the direction of the vector, q . This component may be given by: S

Fq

2S R 2 m ³ sin T dT 0

(7-27)

° °½ u ®¦ ³ vr vr cos T vT sin T f r dv ¾ , °¯ r r °¿

where f r are the distribution functions of the reflected and incident molecules, at the surface of the particle, respectively. Using Eqs. (7-24) and (7-26), and the boundary condition given by Eq. (6-4), one can obtain the following expression for the distribution function at the surface of the particle:

f

f ° ° °° f ® ° ° ° °¯

1 0 1 f ª1 c 2 52 q n R a1 q c S3 2 c 2 º ¬ ¼ 0 1 DW f ª¬1 c 2 52 q n R

1 1 a1 g cr cos T cT sin T S3 2 c 2 º ¼

(7-28)

0 DW f ª1 Q r rS c 2 32 W r rS º . ¬ ¼

The unknown functions, Q r rS and W r rS , are determined from the two boundary conditions given by:

° ½° n ® ³ vf dv ³ vf dv ¾ 0 , ¯° ¿°

(7-29)

and:

§ wTp · ° °½ n ® ³ 12 mv 2 vf dv ³ 12 mv 2 vf dv ¾ N p ¨ ¸ °¯ °¿ © wr ¹ r

, R

(7-30)

104

Analytical Methods for Problems of Molecular Transport

where Tp is the temperature of the particle and N p is the thermal conductivity coefficient of the particle. The distribution of temperature inside the particle, in spherical coordinates, may be expressed in the form:

Tp

f

T0 ¦ Cn r n Pn cos T .

(7-31)

n 1

Here, Cn are the coefficients of the Legendre polynomials, Pn cos T . The left-hand-side of Eq. (7-30) contains the terms that are proportional only to cos T , and therefore there is only one term in this sum and the temperature distribution inside the particle may be written as:

Tp

T0 1 W p r

§ C r cos T · T0 ¨¨1 1 ¸¸ , T0 © ¹

(7-32)

where C1 is an unknown constant. Before using the boundary conditions, one can transform the distribution function given by Eq. (7-28). Small terms of O > R O @ may be neglected in this regime. Eliminating these small terms, the following expression for the distribution function at the surface of the particle is obtained:

f

f ° ° °f ® ° ° ° ¯

1 0 1 f ª1 a1 q c S3 2 c 2 º , ¬ ¼

1 DW f 0

(7-33)

1 1 u ª1 a1 g cr cos T cT sin T S3 2 c 2 º ¬ ¼ 0 ª 2 3 1 Q r rS c 2 W r rS º . DW f ¬ ¼

The parameters of diffusely reflected molecules are assumed to deviate only slightly from the equilibrium state. The boundary condition, Eq. (7-29), gives:

Q r 12 W r

(7-34)

0 .

Eq. (7-29) is not altered in the case when the gas molecules are reflected from the surface of the particle under equilibrium conditions W r W p . In this case one has:

105

Chapter 7. The Free-Molecular Regime

Q p 12 W p

(7-35)

0 .

After performing the integrations in Eq. (7-30), one obtains:

Q r 32 W r 85 S a11 q cos T

N p C1 cos T , DW SD

(7-36)

where: 1 2

D

§ 2kT0 · ¸ © m ¹

S 3 2 n0 m ¨

32

.

Using the definition of DT in Eq. (7-30) yields:

DT Q p 32 W p 85 S a11 q cos T

N p C1 cos T , DW SD

(7-37)

where:

Wp

C1 R cos T . T0

From Eqs. (7-34)-(7-37) it follows that:

ª ¬

W r rS 85 S a11 q n «1 DT

1 º , 1 H »¼

(7-38)

where:

H

DW DT N g R , 75SN p O

and N g is the thermal conductivity coefficient for the gas. Molecules are assumed to be rigid spheres. In the free-molecular regime the quantity H may be neglected. Then the following expressions specify the boundary parameters, Q r rS and W r rS :

106

Analytical Methods for Problems of Molecular Transport

Q r rS

5 16

S 1 DT a11 q n ,

(7-39)

and:

W r rS 85 S 1 DT a11 q n .

(7-40)

Using these relations, one may transform Eq. (7-27) into the form:

4n0 kT0 R 2 a1 q S 1

Fq

S f

³ sin T dT 0

° 1 u ®1 DW ³ cr cr2 cos 2 T cT2 sin 2 T S3 2 c 2 exp c 2 dc °¯ 0

165 S DW 1 DT cos 2 T

(7-41)

f ªf º u « ³ cr2 exp c 2 dc 2 ³ cr2 c 2 32 exp c 2 dc » «¬ 0 »¼ 0 f °½ 1 ³ cr cr2 cos 2 T cT2 sin 2 T S3 2 c 2 exp c 2 dc ¾ . °¿ 0

Having performed the necessary integrations, the expression for the thermal force may be written as:

F

54 S R 2 n0 k O T f ª¬1 DW 1 D T 325 S º¼ .

(7-42)

This expression contains a new term, namely DW 1 DT 325 S , as compared with the well known results [11-13]. This term for the particular case when DW 1 has been obtained by Ivchenko and Yalamov [13] and later by Talbot et al. [14].

5.

CONDENSATION ON A SPHERICAL DROPLET.

Consider the growth of a single droplet in an infinite expanse of gas that consists only of the gas phase of the liquid which forms the droplet. At large distances from the droplet the gas is assumed to have a constant temperature and to be supersaturated. The distribution function for these conditions has the form:

107

Chapter 7. The Free-Molecular Regime 0 f 1 Q 0 ,

f

(7-43)

where:

f

0

§ m · n0 T0 ¨ ¸ © 2S kT0 ¹

32

§ mv 2 · exp ¨ ¸ , © 2kT0 ¹

n0 T0 is the number density of the saturated vapor at temperature, T0 , and

Q 0 is the supersaturation. The distribution function on the surface of the

droplet is assumed to be the Maxwellian function for the reflected molecules with the unknown number density and temperature. Let these parameters be deviated only slightly from the equilibrium state, and therefore one may use the linearized form of the distribution function. The distribution functions of incident and reflected molecules at the droplet may be expressed in the form:

f

f ° °f ° ® ° ° ° ¯

0 f 1 Q 0 , 0 f ª¬1 ) c, rS º¼ D m f 0 ª1 Q S c 2 32 W S º ¬ ¼

(7-44)

0 1 D m f ª1 Q r c 2 32 W r º . ¬ ¼

where Q S is the correction to the saturated vapor density at the temperature, TS T0 1 W S . The unknown parameters, Q r and W r , in the reflected distribution function may be found from the energy accommodation and impenetrability conditions which have sense only for those molecules reflected from the particle surface without a phase transition. The perfect representation of the reflected distribution function was obtained previously from Problem 6.6 and is expressible in the form:

) c, rS D m ªQ S c 2 32 W S º ¬

¼

ª 2G · º 2G ½° ° §G 1 D m ® 2 c 2 « DT W S 1 DT ¨ 2 1 ¸ » 1 ¾ , S ¹ ¼ S ¿° ©S ¬ ¯°

(7-45)

where:

G1

12 SQ 0 and G 2

SQ 0

(7-46)

108

Analytical Methods for Problems of Molecular Transport

The correction to the equilibrium number density may be represented by:

QS

2J m aW S , U1kT0 R

(7-47)

where J is the surface tension coefficient, U1 is the density of the drop, R is the drop radius, a 9 1 , 9 h fg kT0 , and h fg is the latent heat of condensation per molecule. This equation is the sum of the linearized Kelvin and Clausius-Clapeyron terms [17,18]. The temperature, W S , may be found from the equation of energy conservation, which may be written as: Qr

(7-48)

h fg N r ,

or:

DT Qr , S

(7-49)

h fg D m N r , S .

Eq. (7-48) expresses the fact that, for stationary conditions, the heat of condensation must be transferred to the ambient gas by thermal conductivity. It also shows that the growth rate of the particle is controlled by the rate at which the liberated heat of condensation can be dissipated into the environment. Eq. (7-49) gives:

WS

D mQ 0c

9 2 , 2D m a D m9 a 12 2DT 1 D m 3 2

(7-50)

where:

Q 0c

2J m Q . U1kT0 R 0

The total flux of the number of molecules across the droplet surface may be written as:

N

4S R 2 N r

4S R 2D m N r , S 12

4S R 2D m

n0 § 2kT0 · ¨ ¸ 2 S © m ¹

ªQ 0c a 12 W S º . ¬ ¼

(7-51)

109

Chapter 7. The Free-Molecular Regime

Substituting W S in Eq. (7-51), one obtains: 12

N

§ 8S kT0 · ¨ m ¸ © ¹

R 2 n0Q 0c[ D m ,DT , 9 ,

(7-52)

where:

[ D m ,DT , 9 4D m

D m DT 1 D m

2

D m 29 39 2 4D T 1 D m

.

The growth rate of the drop is determined by the equation:

dR dt

mN . 4S R 2 U1

(7-53)

Integration of this equation gives the dependence, R form:

R t , that has the

12

t

U1 § 2S · ° b § Q 0 R b · °½ ¨ ¸ ® R R0 ln ¨ ¸¾ , Q 0 © Q 0 R0 b ¹ ¿° n0Q 0[ © mkT0 ¹ ¯°

(7-54)

where:

b

6.

2J m . U1kT0

NON-STATIONARY GAS FLOWS.

The properties of non-stationary flows of a rarefied gas may be described by solving the Cauchy problem for Boltzmann’s equation that may be written in the form: wf wf wf wf vx vy vz wt wx wy wz

0 .

The initial condition for this problem is given by:

(7-55)

110

Analytical Methods for Problems of Molecular Transport

f v, r ,0

fi v, r .

(7-56)

The solution of this mathematical formulation begins with the construction of a general solution of Lagrange’s auxiliary system of ordinary differential equations which may be expressed in the form:

dt 1

dx vx

dy vy

dz vz

df . 0

(7-57)

The general solution of Eq. (7-57) is:

f

D

r vt

const ,

r0

(7-58)

const .

(7-59)

By virtue of Eqs. (7-58) and (7-59), the distribution function may be shown to be expressible in the following general form:

f v, r , t < r0 , where < is an arbitrary function. The initial condition then yields:

f v, r ,0 < r0

fi v, r0 .

(7-60)

Allowing for this relation, one can obtain the following solution of the Cauchy problem for the distribution function:

f v, r , t

fi v, r vt .

(7-61)

The knowledge of the distribution function allows one to calculate any macroscopic quantity characteristic of the gas flow. It is, however, very important to note that the calculation technique is very different from that described in Section 7.3. For a non-stationary flow, there is no cone of influence in the velocity space since the molecular trajectories passing through a fixed point, M r , at various moments of time, t , must start from different initial points, r0 . The moments of the distribution function may be found by means of direct integration of Eq. (7-61) over all velocity space. The mean value of any molecular property, I v , is given by:

111

Chapter 7. The Free-Molecular Regime

I

n 1 ³ I v f i v, r vt dv .

(7-62)

To calculate this integral, it is convenient to change the variable of integration from v to a new variable, r c , which is related to v by:

r c r vt .

(7-63)

The volume elements of this transformation are connected through the Jacobian in the following manner:

dv

w vx , v y , vz

dr c

w xc, y c, z c

t 3 dr c .

(7-64)

Finally, the mean value of I v may be expressed by [6]:

nI

§ r r c · § r r c · t 3 ³ I ¨ , r c ¸ dr c , ¸ fi ¨ © t ¹ © t ¹

(7-65)

where integration is extended over the domain occupied by the gas at t

0.

PROBLEMS 7.1. Determine the temperature of a sphere located in an infinite stream of a monatomic gas the velocity of which, at large distances from the body, is given by a vector, u . Consider the case of arbitrary values of the ratio, 12 u 2kT m , if the particle is assumed to be a perfect heat conductor and molecules of a gas are considered to be reflected by the sphere surface with equilibrium conditions. Solution: The distribution function at the surface of the particle can be expressed in the form: f

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

§ m v u 2 exp ¨ ¨ 2kT0 ©

· ¸ , ¸ ¹

112

Analytical Methods for Problems of Molecular Transport

f

m · ¸ © 2S kTS ¹

§

32

§ mv 2 exp ¨ © 2kTS

DW n rS ¨

· ¸ 1 DW f cr , cT , cI , rS ¹

,

where n rS denotes the density of a fictitious gas at the point, rS , and TS is the constant temperature of the particle. The unknown function, n rS , and the quantity, TS , can be specified from two conditions expressing the impenetrability of the surface for gas molecules and the heat balance:

n S ³ vfdv

0 ,

³ dS ³ 12 mv

2

vfdv

0 .

S

From the first condition, one can obtain: 12

n rS

§T · n0 ¨ 0 ¸ © TS ¹

^exp u

2 r

` ,

S ur 1 erf ur

where ur u n S u cos T . The direction of the vector, u , is assumed to be along the polar axis. Taking into account both conditions, one can obtain:

TS

T0

1

1 2 u 2

1 2

S

5 u 2

u 3 Q u

1 S u Q u

,

where: 1

Q u

³ t erf u t dt 0

1

³ exp u 0

2 2

t

dt

, and u

§ 2kT0 · u¨ ¸ © m ¹

1 2

.

A more detailed analysis is contained in [3,19].

7.2. Using the same conditions as in Problem 7.1, determine the drag force on the sphere.

113

Chapter 7. The Free-Molecular Regime

Solution: Substituting the distribution function described in Problem 7.1 in Eq. (7-8), one obtains:

FD

8 3

12

2

R n0 2S mkT0

° u ®Q 1 u DW °¯

ª § T ·1 2 º½ « 81 S ¨ S ¸ Q 2 u » °¾ , « © T0 ¹ »° ¬ ¼¿

where u is the dimensionless velocity of the gas, DW is the tangential momentum accommodation coefficient and the following notations have been introduced: 1

³ ª«¬t

Q1 u

3 2

2

0

Q 2 u

exp u 2t 2

1 2

º § t · 2u t 3 ¸ erf u t » dt ,

©u ¹ ¼

1 ª 32 ³ « t 2 12 exp u 2 t 2 0¬

S¨

º § t · 12 S ¨ 12 u t 2u t 3 ¸ erf u t » dt . © u ¹ ¼

The temperature of the sphere, TS , was determined in Problem 7.1.

7.3. Determine the mean velocity distribution (both components) for a linearized uniform flow of a gas past a sphere if the mean gas velocity far from the sphere is u . Consider the specific case when DW 1 and 12 u 2kT m . Solution: For this problem the distribution function given in Eq. (7-18) is employed. The component of the mean velocity vector may be calculated from Eq. (7-15). For example, the radial component may be determined in the following manner:

ur r ,T f

1 cr f1dv ur S 3 2 ³ n0 1 2

^

³ dE 0

w

³ cos F sin F

u³ c3 exp c 2 0

2S

0

12

u S r sin 2 F cos F ª¬1 r 2 sin 2 F º¼

2c cos F `dcd F ,

114

Analytical Methods for Problems of Molecular Transport

where w arcsin 1 r and ur u cos T . An analogous relationship exists for the other component, uT r ,T . Having performed the indicated integrations, one obtains:

ur r ,T u 1 g r cos T , and uT r ,T u 1 h r sin T . Here, the functions, g r and h r , may be written as:

g r

1 4

ª § 1 r 1 · º 1 3 r 3 r « 14 r 3 18 r 1a 2 18 a 4 ln ¨ ¸ » 2 1 a , and: a © ¹ ¼» ¬«

hr

1 2

° ½° ª 1 1 1 2 § 1 r 1 · º 3 ¸ » 2 1 a ¾ , ® g r r « 2 r 2 a ln ¨ «¬ © a ¹ »¼ ¯° ¿°

where a

12

1 r 2

. More detailed analyses may be found in [7,20].

7.4. Prove that the velocity field, as defined in Problem 7.3, is not the potential field. Solution: For the potential flow field, the following equation must be satisfied: u u r 0 . This quantity, for the free-molecular velocity field described in Problem 7.3, may be expressed in the form:

er r sin T

eT r sin T

eI

w wr ur r ,T

w wT ruT r ,T

w wI 0

2

u u r ,T

r

ª h r g r dh º eI u sin T « »z0 r dr ¼ ¬ From this one can see that the velocity field has no potential.

7.5. In Problem 7.3, determine the pressure and temperature distribution. Check the validity of the relationship: p nkT .

115

Chapter 7. The Free-Molecular Regime

Solution: The same distribution function and calculational techniques as in Problem 7.3 yield:

^

`

p0 1 12 S ur \ 1 r \ 2 r ,

p

¼

\ 1 r r ª1 a 13 1 a 3 º , ¬

\ 2 r

1 3

r 2 1 8S 1 , and:

T r ,T T0 1 13 S 1 2ur r 2 ,

12

where a 1 r 2 and ur u cos T . It is easy to see that the equation of state is not altered for this free-molecular flow.

7.6. A sphere suspended in a gas has radius, R , and temperature, Determine the heat flux and the temperature T0 'T 'T T0 . distribution of a gas if its number density and temperature far from the sphere are n0 and T0 , respectively. Solve this problem for the freemolecular regime. Solution: A distribution function should be sought in the form:

f1

0 f ,

f2

1 DW f 0 DW f 0 ^1 Q r c 2 32 W r `

,

where: f

0

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

§ mv 2 · exp ¨ ¸ , © 2kT0 ¹

is the distribution function at large distances from the heated sphere. The quantities, Q r and W r , may be determined from the boundary relationships

116

Analytical Methods for Problems of Molecular Transport

Figure 7-4. The geometry to be used for Problem 7.7 in determining the force on a round disk perpendicular to a uniform gas flow.in determining the force on a round disk perpendicular to a uniform gas flow

N r 0 , N r eq 0 , and Qr DT Qr eq . Performing some simple integrations and algebraic transformations, one can obtain: 12

Q

'T § 2kT0 · DW DT 4S R 2 n0 kT0S 1 2 ¨ , ¸ T0 © m ¹

where Q is the heat flux through the sphere surface. The temperature distribution can be expressed in the form:

T r T0 ®1 34 DW DT ª1 1 r 2 ¬« ¯

12

º 'T ½ . ¾ ¼» T0 ¿

7.7. Calculate the force on a round disk the surface of which is perpendicular to the velocity vector of a uniform gas flow at infinity. The radius of the disk, R , is assumed to be much less than the mean free path 12 where u and T are the mean gas velocity and and u 2kT m temperature, respectively. Use the geometry in Fig. 7-4. Solution: The force on the left and right sides of the disk may be expressed as:

F1,2

# S R 2 ¦ ³ mv vn1,2 f1,2 v dv ,

(P-1)

# #

where n i is the external (pointing into the gas) normal vector to the disk surface. The reflected distribution functions are given by:

§ m · f1 v DW n1 ¨ ¸ © 2S kT ¹

32

exp c 2 1 DW f1 cx , c y , cz ,

117

Chapter 7. The Free-Molecular Regime

Figure 7-5. The geometry to be used for Problem 7.8 in determining the drag on a plate parallel to a steady, free-molecular gas flow.

§ m · f 2 v DW n2 ¨ ¸ © 2S kT ¹

32

exp c 2 1 DW f 2 cx , c y , cz ,

where the notation (±) refers to those regions in the velocity space in which cx ! 0 and cx 0 , respectively. The incident distribution function may be written as:

f1 v

§ m · f 2 v n ¨ ¸ © 2S kT ¹

32

exp c 2 1 2cx u .

The number density, ni n 1 Q i , may be found from the impenetrability condition of the disk surface which yields: Q 1,2 r S u . Then, after some simple integration, the expression for the force becomes:

F

12

4 R 2 n 2S kTm

ª1 DW 18 S 12 º u . ¬ ¼

7.8. A plate of area, S , is situated in a steady gas flow and is oriented parallel to the velocity vector, u , as shown in Fig. 7-5. In the freemolecular regime, determine the drag on the plate for arbitrary values of the 12 speed ratio, u 2kT m , and of the tangential momentum accommodation coefficient. Solution: In this problem, the drag is given by:

F 2S ¦ ³ mvvx f r v dv . r r

The incident and reflected distribution functions associated with the upper side of the plate are specified by:

118

Analytical Methods for Problems of Molecular Transport

Figure 7-6. The geometry to be used for Problem 7.9 in determining the drag on a plate oriented at an angle to a steady, free-molecular gas flow.

f

f

32

v

§ m · n¨ ¸ © 2S kT ¹

v

§ m · DW n ¨ ¸ © 2S kT ¹

§ m v u 2 exp ¨ ¨ 2kT ©

32

m · 1DW n ¨§ ¸ © 2S kT ¹

· ¸ , and: ¸ ¹

§ mv 2 · exp ¨ ¸ © 2kT ¹ 32

ª m vx 2 v y u exp « ¬ 2kT

2

º vz2 » . ¼

Using these, the drag is found to be: 12

§ 2m · F DW pS ¨ ¸ u. © S kT ¹ 7.9. A plate of area, S , is situated in a steady gas flow and is oriented such that the gas stream is incident on the plate at an angle T as shown in Fig. 76. In the free-molecular regime, determine the drag on the plate for arbitrary 12 values of the speed ratio, u 2kT m , and of the tangential momentum accommodation coefficient. Solution: The net drag component in direction of the flow vector, u , is given by: Fu

S ¦ ³ m vx cos T v y sin T vx f1r f 2r d v , r r

119

Chapter 7. The Free-Molecular Regime

where the indices 1 and 2 refer to the upstream and downstream sides of the plate, respectively. The distribution functions of incident and reflected molecules may be expressed in the form:

§ m · f 2 v n ¨ ¸ © 2S kT ¹

f1 v

32

§ m exp ¨ v u 2 ·¸ , kT 2 © ¹

32

f1

v

32

§ m · § m 2· § m · v ¸ 1 DW n ¨ DW n1 ¨ ¸ exp ¨ ¸ © 2S kT ¹ © 2kT ¹ © 2S kT ¹ 2 2 m ª ½ u exp ® vx u cos T v y u sin T vz2 º ¾ , « » ¼¿ ¯ 2kT ¬

and:

§ m f 2 v DW n2 ¨ © 2S kT m u exp ® ¯ 2kT

· ¸ ¹

32

§ m 2· § m · exp ¨ v ¸ 1 DW n ¨ ¸ 2 kT © ¹ © 2S kT ¹

ª v u cos T 2 v u sin T y ¬« x

2

32

½ vz2 º ¾ . ¼» ¿

The values, n1 and n2 , are specified by the impenetrability condition. The drag is found to be: 12

F

§ 2m · pS ¨ u 2 DW cos T ª cos T exp u 2 cos 2 T ¸ ¬ © S kT ¹ º S 1 2 2 u cos T erf u cos T » DW ª¬ 12 S cos 2 T 2

u »¼

^

`

sin 2 T exp u 2 cos 2 T S u cos T erf u cos T º» , ¼ where u

m

12

2kT

u.

7.10. Determine the force on a round disk located in a rarefied gas if the surfaces of the disk have different temperatures, T1 and T2 . The disk has a radius of R which is assumed to be much less than the mean free path and T1 T2 T0 where T0 is the gas temperature. Use the geometry in Fig. 7-7.

120

Analytical Methods for Problems of Molecular Transport

Figure 7-7. The geometry to be used for Problem 7.10 in determining the force on a round disk in a rarefied gas when the two sides of the disk have different temperatures.

Figure 7-8. The geometry to be used for Problem 7.11 in determining the thermophoretic force on a thin disk located in a rarefied gas having a constant temperature gradient.

Solution: Using the same technique as in Problem 6.5, one can obtain:

; W 1w

T1 T0 , T0

; W 2w

T2 T0 , T0

)1

DW DT 2 c 2 W 1w

) 2

DW DT 2 c 2 W 2 w

where )1 and ) 2 are the corrections to the reflected distribution functions associated with the left- and right-hand surfaces of the disk, respectively, and )1 ) 2 0 . The force on the disk may be calculated from the equation given in the solution to Problem 7.7 [Eq. (P-1)]. This gives:

Fx

1 4

S R 2 p0DW DT

T1 T2 . T0

7.11. Determine the thermophoretic force on a thin absolutely thermal conductive disk located in a rarefied gas with a constant temperature gradient, T . The radius of the disk is R and is assumed to be much less than the mean free path. Use the geometry in Fig. 7-8. Solution: The corrections to the reflected distribution functions may be written as (Problem 6.5):

121

Chapter 7. The Free-Molecular Regime

)1 c,0

1 DW )1 cx , c y , cz ,0 ° DW ® 2 c 2 ¯°

ª

««1 D ¬

§ G 21

T ¨ ¨

© S

1 1 2G1 · º 2G1 ½° ¸» ¾ , S ¹¸ »¼ S ° ¿

and:

) 2 c,0

1 DW ) 2 cx , c y , cz ,0 ° DW ® 2 c 2 ¯°

ª § G 2 2G 2 «1 DT ¨ 2 1 ¨ S S «¬ ©

· º 2G1 2 ½° ¸» ¾ . ¸» S °¿ ¹¼

Here, )1 c,0 , ) 2 c,0 , G1 , and G 2 , are defined by: i

i

0 , G 2

2 0 , G 2

)1 c,0 a11 qcx S312 c 2 , G11

) 2 c,0 a11 qcx S312 c 2 , G1 2

1

5 8

S 3 2 a11 q , and:

5 8

S 3 2 a11 q ,

with:

q T01

wT , wx

where the indices 1 and 2 are associated with the left- and right-hand surfaces of the disk, respectively. Substituting these into the force equation from Problem 7.7 [Eq. (P-1)], one can obtain:

F

158 S R 2 n0 k O T f ª¬1 12 DW DW 1 DT 325 S º¼ .

7.12. Determine the number density of a gas at a point, A , located behind a round disk of radius, R , which is much less than the mean free path, i.e. R O . The mean gas velocity, u , is assumed to be much less than the thermal molecular velocity. Use the geometry in Fig. 7-9. Solution: Since n S const for all molecular trajectories beginning at the disk surface, vc vx , v y , vz . Therefore, the distribution function has a very simple form for arbitrary values of the tangential accommodation

122

Analytical Methods for Problems of Molecular Transport

Figure 7-9. The geometry to be used for Problem 7.12 in determining the number density behind a round disk perpendicular to a free-molecular gas flow in which the mean gas velocity is much less than the thermal molecular velocity.

coefficient. The distribution function outside and inside of the cone of influence is expressed as:

f1

0 f 1 2cx u ,

f2

DW f 0 1 S u 1 DW f 0 1 2cx u .

The number density is then given by:

° 2 n x n0 ®1 S °¯

f

F0

2 2 ³ c exp c dc ³ sin F d F 0

0

½ ª DW S u 2 2 DW u c cos F º °¾ ¬ ¼° ¿ 12 ° § m · ª S§ x n0 ®1 u ¨ ¨¨1 ¸ «DW 2 © R2 x2 © 2kT0 ¹ ¬« ¯°

2 DW

· ¸¸ ¹

R 2 º °½ »¾ . S R 2 x 2 ¼ ¿°

1

7.13. A gas with a number density, n0 , is flowing into vacuum across a round hole of radius, R O . Determine the number density at a point, A . Use the geometry in Fig. 7-10.

123

Chapter 7. The Free-Molecular Regime

Figure 7-10. The geometry to be used for Problem 7.13 in determining the number density of a free-molecular flow exiting through a round hole.

Solution: The distribution function outside and inside of the cone of influence may be expressed in the form:

f1

0 ,

f2

§ m · n0 ¨ ¸ © 2S kT ¹

32

exp c 2 .

From this, the number density at A is determined to be:

n x

1 2

ª x n0 «1 2 «¬ R x2

º » . »¼

7.14. A gas with number density, n0 , passes outward through a small slot by means of effusion (pressure driven motion of a gas from a vessel to a vacuum through a small hole). Determine the gas number density at a large distance, r , from the slot ( r S0 , where S0 is the slot area ). Use the geometry in Fig. 7-11. Solution: The distribution functions outside (1) and inside (2) of the cone of influence are given by:

§ m · f1 0 and f 2 n0 ¨ ¸ © 2S kT ¹

32

exp c 2 , respectively.

124

Analytical Methods for Problems of Molecular Transport

Figure 7-11. The geometry to be used for Problem 7.14 in determining the number density of an effusing gas at large distances from a small slot.

Taking into account the condition, r R , after integration one can obtain:

n r

n0 R 2 4 r 2

n0 S 4S r 2

n0 S0 cos T 4S r 2

.

7.15. Determine the number density of a gas at a point, A , behind a round disk of radius, R O . The mean gas velocity, u , is assumed to be arbitrary. The flow geometry is given in Problem 7.12 (see Fig. 7-9). Consider the particular case for which: 12

u 2kT0 m

12

and cos F 0 2kT0 m u 1 .

Solution: The distribution functions outside and inside of the cone of influence are given by:

f1

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

exp §¨ c u ©

2

· , ¸ ¹

125

Chapter 7. The Free-Molecular Regime

f2

§ m · ¨ ¸ © 2S kT0 ¹

^

32

u DW nr exp c 2 1 DW n0 exp §¨ cx u ©

2

`

c 2y cz2 ·¸ . ¹

The number density, nr , of reflected molecules may be specified from the impenetrability condition which yields:

nr

n0 ª« exp u 2 S u 1 erf u º» . ¬ ¼

Having employed standard integration techniques (see Section 7.3), one can obtain:

n x n0

^

1 2

2 DW 12 DW 1 cos F 0

u ª« exp u 2 S u 1 erf u º» ¬ ¼

12 DW cos F 0 exp u 2 sin 2 F 0 1 2

2 DW ª¬cos F 0 erf u cos F 0

`

u exp u 2 sin 2 F 0 erf u º , ¼ where:

cos F 0

x 2

R x2

.

In the particular case when u cos F 0 1 , this expression reduces to:

n x n0 cos F 0 exp u 2 sin 2 F 0 . 7.16. Determine the frictional force on the lower plate in a system where the upper of two parallel plates is moving horizontally. Use the geometry in 12 Fig. 7-12 and assume that u 2kT m .

126

Analytical Methods for Problems of Molecular Transport

Figure 7-12. The geometry to be used for Problem 7.16 in determining the frictional force on a plate due to the horizontal motion of a second, parallel plate.

Solution: The distribution function for vx ! 0 and vx 0 may be 0 expressed as f r f 1 a r c y where a r const . Using the Maxwellian boundary condition for each plate, one can obtain:

a

2 1 DW 2 DW

u , a

12

§ m · ¨ 2kT ¸ u . © ¹

2 u , u 2 DW

The force of friction per unit surface area is given by: 12

Fy

¦ ³ mvx v y f r dv r r

DW § m · 2 DW ¨© 2S kT ¸¹

pu ,

where p is the hydrostatic pressure.

7.17. Assume that a gas is flowing in the y -direction between two infinite planes separated in the x -direction by a distance, d , where d O and O is the mean free path of the gas molecules. Assume that the number density of the gas is maintained at different steady values at two different locations along the y -axis and that the temperature of this system is uniform. Determine the number of molecules per unit area per second flowing in y direction. Solution: The distribution function may be chosen in the form:

f

0 n ) c, x , where g n f 1 yg

1 n0 dn

dy

The correction to the distribution function, ) c, x , which describes the gassurface interactions may be specified from the Boltzmann equation such that:

127

Chapter 7. The Free-Molecular Regime

cx

w) cy gn 0 . wx

The correction, ) c, x , may be expressed in the form:

) c, x

1 2

a

x a x c y 12 a x a x c y sign cx

The half-range moment equations are obtained by multiplying the Boltzmann equation by c y 1 r sign cx exp c 2 dc and integrating over all velocities. These equations are found to be:

da r x r S gn . dx The boundary conditions are given by a r # d 2 simple calculations one finds: 12

Jy

n0u y

1 2

§ 2kT · n0 ¨ ¸ © m ¹

a

a

2 DW

DW

1 DW a # # d 2 . 12

1 8

§ 8kT · ¸ © Sm ¹

Sd¨

After

wn . wy

7.18. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at different temperatures, T c and T cc with T c ! T cc . At the beginning t 0 the vessels are filled with different gases to the same pressure. When the capillary is opened a pressure difference, G p pcc pc , will arise. Determine the time dependence of this pressure difference. Solution: To describe this transport problem one should start with the basic equations of balance for the number of molecules of each of the two gases which are given by: V c

dnic dt

J zi ,

(P-2)

V cc

dnicc dt

J zi ,

(P-3)

128

Analytical Methods for Problems of Molecular Transport

where J zi is the i -th constituent molecular number flux across the tube cross section per second. For the free-molecular regime, J zi , may be expressed in the form [21-23]:

J zi

2 D iW

D iW

2 3

S R3

8k d n T , S mi dz i

where D iW is the tangential momentum accommodation coefficient of the i th constituent. This flux may be approximately represented by:

J zi

ai

ai nicc T cc nic T c , where:

2 D iW 2S R 3 D iW 3L

8k

S mi

.

Then, the basic equations describing kinetics of this system take the form:

dnic a Tc a T cc nic i nicc , i dt Vc Vc

(P-4)

dnicc ai T c a T cc nic i nicc . dt V cc V cc

(P-5)

The necessary initial conditions are given by:

n1c 0 0 , n1cc

p 0 p 0 , n2c 0 , and n2cc 0 . kT cc kT c

This formulation then yields:

Chapter 7. The Free-Molecular Regime

129

G p t k ª¬T cc n1cc n2cc T c n1c n2c º¼ p 0 °§ V c T cc · § T c T c · ¸ ¨ ® 1 G 0 °¯©¨ V cc T c ¹¸ ©¨ T cc T cc ¹¸ · °½ § V c T c · § T cc exp Z2t exp Z1t ¸¸¾ , ¨ ¸ ¨¨ © V cc T cc ¹ © T c ¹ ¿° where:

G0

ª Tc Tc Vc T cc º and Zi ai « ». V cc »¼ T cc V cc ¬« V c

7.19. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at different temperatures, T c and T cc with T c ! T cc . At the beginning t 0 both vessels are filled with a simple gas to the same pressure. When the capillary is opened a pressure difference, G p t pcc pc , will arise. Determine the time dependence of this pressure difference and the steady-state pressure difference that develops. Solution: The pressure difference may be obtained from solution of Problem 7.18. in which one should substitute Z1 Z2 Z . This yields:

G p t

p 0 § T cc · § V c T c · ª1 exp Zt ¼º , where: ¨¨ 1 ¸ G0 © T c ¹¸ ¨© V cc T cc ¸¹ ¬ 1/ 2

2 DW 2S R 3 § 8kT cc · Z 3L ¨© S m ¸¹ DW

§ 1 1 Tc · ¨¨ ¸¸ . © V cc V c T cc ¹

The steady-state (or stationary) pressure difference is then given by:

'pstat

G0

G p f

Tc Vc . T cc V cc

p 0 § T cc · § V c T c · , where: ¨¨ 1 ¸ G0 © T c ¸¹ ¨© V cc T cc ¸¹

130

Analytical Methods for Problems of Molecular Transport

7.20. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at the same constant temperature, T c T cc T and are filled at the beginning t 0 with a simple gas to different initial pressures, pc 0 p and pcc 0 p 'p . When the capillary is opened the pressure difference, G p t pcc pc , will change as a function of time. Determine the time dependence of this pressure difference. Derive a formula for the tangential momentum accommodation coefficient if the pressure difference, G p t , is known at a particular time, t . Solution: Using the following initial conditions:

nc 0

p p 'p and ncc 0 kT kT

in Eqs. (P-4) and (P-5) (Problem 7.18), one can obtain:

G p t 'p exp Zt where:

Z

2 DW

DW

12

Z d , Zd

2S R 3 § 8kT · 3LV ¨© S m ¸¹

and V

V cV cc . V c V cc

The expression for 'p yields:

DW

§ 'p 1 ° ¨ 2 ®1 ln

¨ G p t t Z d °¯ ©

1

·½ ¸ °¾ . ¸° ¹¿

7.21. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at the same constant temperature, T c T cc T and are filled at the beginning t 0 with different gases to the same initial pressure. When the capillary is opened a pressure difference, G p pcc pc , will develop as a function of time. Determine the time dependence of this pressure difference and the maximum value of the pressure difference that develops.

131

Chapter 7. The Free-Molecular Regime

Solution: The time dependence of the number densities in the two vessels is described by Eqs. (P-4) and (P-5) (Problem 7.18). The initial conditions are given by:

p 0

n1c 0 0 , n1cc 0

kT

, n2c 0

p 0 kT

, and n2cc 0 0

This formulation yields a pressure difference of:

G p t p 0 ª¬exp Z1t exp Z2t º¼ where: 12

Zi

2 D iW 2S R 3 § 8kT · ¨ ¸ D iW 3LV © S mi ¹

V cV cc . V c V cc

and V

The subtraction of the two decreasing exponential terms results in the existence of a pressure difference maximum, 'p | G p |max . This phenomenon is usually called the diffusion-pressure effect. The value, 'p , is found to be:

'p p 0

f Z

1 · § ¨1 ¸ exp u . © Z ¹

The time at which the maximum pressure difference, 'p , occurs is given by t u Z2 . Here, the following notations have been introduced:

Z

D 2W 2 D1W m2 and u D1W 2 D 2W m1

.

ln Z

Z 1

7.22. For the free-molecular regime, determine the gas temperature between two parallel plates having temperatures T1 and T2 . Molecules are assumed to be reflected diffusely with arbitrary accommodation of energy at each surface. Use the geometry in Fig. 7-13: Solution: The distribution functions for the sets of molecules having positive ( vx ! 0 ) and negative ( vx 0 ) motions are expressible as:

f

§ m · n1 ¨ ¸ © 2S kT1r ¹

32

§ mv 2 · exp ¨ ¸ and: © 2kT1r ¹

132

Analytical Methods for Problems of Molecular Transport

Figure 7-13. The geometry to be used in Problem 7.22 in determining the gas temperature between two parallel plates in the free-molecular regime.

f

§ m · n2 ¨ ¸ © 2S kT2 r ¹

32

§ mv 2 · exp ¨ ¸. © 2kT2 r ¹

The number densities, n1 , and n2 , may be found from the relations:

Nx

³ vx f

dv ³ vx f dv 0 and n n1 n2 ,

where n is the gas number density. These yield: n1

a n and n2 1 a

12

§T · 1 n where a ¨ 2 r ¸ . 1 a © T1r ¹

The quantities, T1r , and T2r , are specified by the boundary conditions:

DT

Qx >Q1x @ eq

Qx where: >Q2 x @ eq

>Q1x @eq

1 2

ª º m « ³ v 2 vx f eq dv ³ v 2 vx f dv » and: «¬ »¼

>Q2 x @eq

1 2

ª º m « ³ v 2 vx f dv ³ v 2 vx f eq dv » . ¬« ¼»

133

Chapter 7. The Free-Molecular Regime

The additional unknown parameters, [n1 ]eq and [n2 ]eq , which are included in f eq and f eq can be expressed in terms of T1r , and T2r using the impenetrability conditions [ N1x ]eq 0 and [ N 2 x ]eq 0 . The gas temperature is found to be:

T

m v 2 fd v 3nk ³

T rT r 1 2 1

2

12

1

2 DT ª¬T1 1 DT T2 T2 1 DT T1 º¼

.

7.23. Determine the free-molecular heat flux in a monatomic gas between two coaxial cylinders the radii and temperatures of which are R1 , T0 'T , and R2 , T0 respectively, and where R1 R2 . Assume that 'T T0 and that the number density of the gas is n0 at r R1 . The tangential momentum and energy accommodation coefficients are assumed to be different for the inner and outer cylinders. Solution: The zone of influence around a cylindrical body is analogous to the cone of influence found around a spherical body or associated with a circular orifice. It is formed by the intersection of two planes that lie tangent to the cylinder in question. The distribution functions outside and inside the zone of influence of the inner cylinder may be chosen in the form:

f1

0 f 1 Q 1 c 2 32 W 1 and f 2

0 f 1 Q 2 c 2 32 W 2 .

The boundary conditions are given by:

D1T

Qr 1 Qr 1 ª¬Qr 1 º¼

N r 1 0 , n 1

N r 1 ª¬ N r 1 º¼ eq

, D 2T eq

Qr z Qr z ª¬Qr z º¼

eq

n0 , N r z 0 ,

0 , and N r z ª¬ N r z º¼

eq

0,

where z R2 R1 . In calculating the fluxes using this formulation, one has to allow for the discontinuity of the distribution functions. For example, the heat flux, if the reflected molecules have the same temperature as the surface of the outer cylinder, may be calculated in the following way:

134

Analytical Methods for Problems of Molecular Transport 12

Qr z ª¬Qr z º¼

p0 § 2kT0 · D 2W S 3 2 ¨© m ¸¹

eq

f

f

° u ®SQ eq ³ cU2 exp cU2 dcU ³ cU2 cz2 exp cz2 dcz °¯ 0 f

S D ªD u « ³ )1 c, z sin T dT ³ ) 2 c, z sin T dT «¬ 0 D S º ½° ³ )1 c, z sin T dT »¾ , S D ¼» °¿

where ) i c, z Q i ª cU2 cz2 32 º W i and D arccos z 1 . The heat flux ¬ ¼ per unit length of the cylinders is found to be: 1

QFM

12

ª 1 · º § 2k · 1 1 § 1 1 ¸ » ¨ « ¨ ¸ «¬ 2S R1 D1W D1T 2S R2 © D 2W D 2T ¹ »¼ © S mT0 ¹

p0 'T .

7.24. Determine the free-molecular heat flux in a monatomic gas between two concentric spheres the radii and temperatures of which are R1 , T0 'T , and R2 , T0 , respectively, and where R1 R2 . Assume 'T T0 and that the number density of the gas is n0 at r R1 . The tangential momentum and energy accommodation coefficients are assumed to be different for the inner and outer spheres. Solution: The distribution functions outside and inside the cone of influence of the inner sphere may be chosen in the form:

f1

0 f 1 Q 1 c 2 32 W 1 and f 2

0 f 1 Q 2 c 2 32 W 2 .

The same boundary conditions are used as were used in Problem 7.23, but the integration technique is necessarily different from that used with the cylindrical geometry. An example of this integration is given by:

135

Chapter 7. The Free-Molecular Regime

Figure 7-14. The geometry to be used for Problem 7.25 in determining the torque on a flat, rotating disk in the free-molecular regime.

12

Qr

p 2kT z ª¬Qr z º¼ eq 302 §¨ 0 ·¸ D 2W S © m ¹ w f ° ª 5 2 u ® « SQ eq 2S ³ c exp c dc ³ ) 2 c, z sin F cos F d F °¯ «¬ 0 0

S 2

º ½°

³ )1 c, z sin F cos F d F »» ¾ ,

¼ °¿

w

where ) i c, z Q i c 2 32 W i and w arcsin z 2 . The heat flux is then found to be: 1

QFM

12

ª 1 · º § 2k · 1 1 § 1 1 ¸ » ¨ « ¸ 2 2 ¨ ¬« 4S R1 D1W D1T 4S R2 © D 2W D 2T ¹ »¼ © S mT0 ¹

p0 'T .

7.25. A flat disk of radius, R O , revolves in a rarefied gas with a constant 12 angular speed, Z , where Z R 2kT m . Determine the torque on the disk, K . Use the geometry in Fig. 7-14. Solution: The torque, K , in this geometry has only one component, K x , which may be expressed as:

136

Analytical Methods for Problems of Molecular Transport R

Kx

R

4S ³ r 2 dr ¦ ³ mvI vx f r v dv ,

2³ rdFI 0

r r

0

where the (±) notation refers to vx ! 0 and vx 0 , respectively, and dFI is the force component acting on one side of the disk only. The distribution function of reflected and incident molecules may be written as:

12 0 0 f c DW f 1 2m kT cI Z r 1 DW f cr , cI , cx ,

f c

0 f .

Substituting these functions into the above expression, one can obtain:

Kx

12

12 DW Z R 4 n 2S mkT

.

7.26. An infinite cylinder of radius, R O , revolves in a rarefied gas with a 12 constant angular speed, Z , where Z R 2kT m . Determine the torque per unit length of the cylinder, K . Solution: The x -component of the torque is specified by:

Kx

2S R 2 ¦ ³ mvr vI f r v dv , r r

where:

f

f

12

DW f 0 1 2m kT cI Z R 1 DW f 0 cr , cI , cx ,

0 f .

and the direction of the x -axis is the same as the direction of Ȧ . Simple integrations then yield: Kx

12

DW Z R 3 n 2S mkT

.

137

Chapter 7. The Free-Molecular Regime

7.27. A sphere of radius, R O , revolves in a rarefied gas with a constant 12 angular speed, Z , where Z R 2kT m . Determine the torque, K , on the sphere if the polar axis, x , is in the same direction as Ȧ . Solution: The vector, K , has one component, K x , which may be found from: S

Kx

2S R 3 ³ sin 2 T dT 0

f

f

¦ ³ mvr vI f r v dv

, where:

r r

12 DW f 0 1 2m kT cI Z R sin T 1 DW f 0 cr , cT , cI ,

0 f .

After integration, one obtains:

Kx

12

43 DW Z R 4 n 2S mkT

.

7.28. An equilibrium gas occupies a semi-infinite expanse, x 0 and is contained by an infinite planar surface at x 0 . Determine the number density of the gas as a function of time in the neighborhood of the surface ( x O ) if the gas is expanding into a vacuum (the region of x ! 0 ) after the sudden removal of the surface (which disappears at t 0 ). Solution: If x O , molecular encounters may be neglected. Therefore, the Boltzmann equation is given by: wf wf vx wt wx

0 .

The Lagrange auxiliary system is then:

dt 1

dx vx

df . 0

The two independent solutions of this system may be written as x vx t x0 , and f D , where x0 and D are arbitrary constants. The general solution of the Boltzmann equation is then expressed as ) x0 ,D 0 where ) x0 ,D is an arbitrary function. This yields:

138

Analytical Methods for Problems of Molecular Transport

< x0 < x vx t .

f

Taking into account the initial condition, one can represent the distribution function in the form:

° f 0 ® °¯ 0

f

x0 0 x0 ! 0

; vx ! x t ; vx x t

, ,

where:

f

0

v

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

§ mv 2 · exp ¨ ¸ . © 2kT0 ¹

Now, the number density may be calculated according to the following:

n x, t

§ m · n0 ¨ ¸ © 2S kT0 ¹ n0

S

f

³

32 f f f

2 ³ ³ ³ exp c dv y dvz dvx

f f x t

exp cx2 dcx

x m t 2 kT0

§x n0 § m ¨1 erf ¨ ¨ t 2kT 2 ¨© 0 ©

·· ¸¸ ¸ . ¸ ¹¹

7.29. A sphere of radius, R O contains a gas with a constant density, n0 , and constant temperature, T0 . At some initial time, t 0 , the sphere vanishes releasing the gas into an infinite vacuum. Determine the number density, n r , t . Solution: Given Eq. (7-65) in the text, the number density may be expressed as: § r r c · n r , t t 3 ³ f i ¨ , r c ¸ dr c , © t ¹ where:

fi

§ m · n0 r c ¨ ¸ © 2S kT0 ¹

32

§ m § r r c · 2 · exp ¨ ¸ , ¨ 2kT0 ¨© t ¸¹ ¸ © ¹

139

Chapter 7. The Free-Molecular Regime

n n0 r c ® 0 ¯0

; rc d R , ; rc ! R .

This integration may be performed easily in spherical coordinates when the polar angle is taken as the angle between r and r c . Having integrated over all of the angular variables, one obtains:

n r, t

n0 § m ¨ r ¨© 2S kT0t 2

12R

· ¸¸ ¹

³ rc 0

ª § § m m 2· 2 ·º u «exp ¨¨ r r c ¸¸ exp ¨¨ r r c ¸¸ » dr c . 2 2 © 2kT0t ¹ © 2kT0t ¹ ¼» ¬« Further integration with respect to r c yields:

n r, t

1 2

where D

n0 ®erf D r R erf D r R ¯ 1

ª exp D 2 r R 2 exp D 2 r R 2 º ½ , ¾ ¼» ¿ D S r ¬«

12

m 2kT t 2

0

.

REFERENCES 1. Kennard, E.H., Kinetic Theory of Gases (McGraw-Hill, New York, 1938). 2. Heineman, M., “Theory of Drag in Highly Rarefied Gases,” Comm. Pure Appl. Math. 1(3), 259-273 (1948). 3. Patterson, G.N., Molecular Flow of Gases (Wiley and Sons, New York, 1956). 4. Schaaf, S.A. and Chambré, P.L., Flow of Rarefied Gases (Princeton University Press, Princeton, New Jersey, 1961). 5. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, NY, 1969). 6. Bird, G.A., Molecular Gas Dynamics (Clarendon Press, Oxford, 1976). 7. Szymanski, Zd., “Some Flow Problems of Rarefied Gases,” Arch. Mech. Stos. (Warsaw) 8, 449-470 (1956). 8. Ivchenko, I.N., “Generalization of the Lees Method in Boundary Problems of Transfer,” J. Coll. Interface Sci. 135(1),16-19 (1990). 9. Wang-Chang, C.S. and Uhlenbeck, G.E., “Transport Phenomena in Very Dilute Gases,” Report CM-579, UMH-3-F, University of Michigan, 1949. 10. Ziering, S., On Transport Theory of Rarefied Gases (Ph.D. Dissertation, Syracuse University, 1958).

140

Analytical Methods for Problems of Molecular Transport

11. Waldmann, L., “Uber die Kraft Eines Inhomogenen Gases auf Kleine Suspendierte Kugeln,” Z. fur Naturforsch. 14a, 589-599 (1959). 12. Brock, J.R., “The Thermal Force in the Transition Region,” J. Coll. Interface Sci. 23(3), 448-452 (1967). 13. Ivchenko, I.N. and Yalamov, Yu.I., “On the Thermophoresis of Aerosol Particles in the Almost-Free-molecular Regime,” Izv. AN SSSR, M. Zh. G. (3), 3-7 (1970). 14. Talbot, L., Cheng, R.K., Schefer, R.W., and Willis, D.R., “Thermophoresis of Particles in a Heated Boundary Layer,” J. Fluid Mech. 101(4), 737-758 (1980). 15. Williams, M.M.R., “On the Motion of Small Spheres in Gases, II: Thermo-phoresis, Diffusio-phoresis and Related Phenomena,” Z. fur Naturforsch. 27a(12), 1804-1811 (1972). 16. Mason, E.A. and Malinauskas, A.P., Gas Transport in Porous Media: The Dusty-Gas Model (Elsevier, Amsterdam, Oxford, N.Y., 1983). 17. Fuchs, N.A., Evaporation and Droplet Growth in Gaseous Media (Pergamon Press, Oxford, 1959). 18. Mason, B.J., Clouds, Rain and Rainmaking (Cambridge University Press, 1962). 19. Williams, M.M.R., “On the Motion of Small Spheres in Gases III: Drag and Heat Transfer,” J. Phys. D: Appl. Phys. 6, 744-758 (1973). 20. Wang-Chang, C.S., “Transport Phenomena in Very Dilute Gases, II,” Report CM-654, NORD 7924, UMH-3-F, University of Michigan, 1950. 21. Smoluchowski, M., “Zur Kinetischen Theorie der Transpiration und Diffusion Verdünnter Gase,” Annalen der Physik 33(16), 1559-1570 (1910). 22. Waldmann, L. and Schmitt, K.H., “Über das bei der Gasdiffusion auftretende Druckgefälle,” Z. fur Naturforsch. 16a, 1343-1354 (1961).

Chapter 8 METHODS OF SOLUTION OF PLANAR PROBLEMS

1.

MAXWELL’S METHOD.

For the last thirty years, the most progress in investigating transport problems has been achieved for one-dimensional, stationary transport problems. In this context, several analytical methods have been developed which have been applied to the classical slip problems. This has resulted in several analytical expressions for the slip coefficients. The recent development of direct numerical solutions for these same problems allows one to estimate the accuracies of the various analytical methods. For simplicity, only planar transport problems are described in this chapter and all molecules are assumed to act as rigid spheres. Although the methods that are discussed in this chapter may be generalized for arbitrary intermolecular potential models, this chapter does not include the specific details necessary for such generalizations. However, some generalization details for certain selected methods are included in Chapter 9. The reason that the details of the generalization process have been limited in this book to selected methods is that the details necessary for the generalization of each method will typically be specific to that method and some of these generalizations can be very complex and difficult to achieve. The half-range moment method is a case in point and no attempt has been made in this book to generalize it although the basic method is described in Section 8.3. In the case of planar transport problems there exist some simple approximate methods based on the use of conservation laws as exact moment solutions of the Boltzmann equation. First, consider the approximate method proposed by Maxwell to solve some boundary transport

142

Analytical Methods for Problems of Molecular Transport

Figure 8-1. The geometry of slip problems.

problems [1]. The main features of this method will be analyzed by the solving of the slip boundary problems. First of all, the velocity-slip and thermal-creep problems will be examined. One considers a semi-infinite expanse of a gas bounded by a flat plate located at x 0 , and lying in the y z plane. Far from the plate the gas is maintained at a constant velocity gradient wu wx f , normal to the plate and at a constant temperature gradient wT wy f , tangential to the wall. The geometry of these slip-flow problems is illustrated in Fig. 8-1. These gradients are assumed to be small such that the relations Ou u 1 and O ln T 1 are satisfied. Then, at large distances from the wall, the gas is described by the Chapman-Enskog distribution function. The distribution function near the wall is a discontinuous function in the velocity space. Maxwell was the first to notice this feature of the distribution function and to show that by a simple consideration of the constancy of the stress and heat flux, some meaningful results for the slip terms could be obtained. The influence of the wall near its surface may be taken into account by the introduction of the discontinuous term, ) r v, r , in the distribution function. This distribution function, for all planar and linearized transport problems, may be written as:

fr

^

`

f eq 1 < v, r ) r v, r ,

(8-1)

where f eq is the local Maxwellian distribution function given by Eq. (5-8) and < v, r is the Chapman-Enskog correction to the distribution function. For the slip-flow and thermal-creep problems the distribution functions f eq and < v, r may be written as:

f eq where:

0 y 2c y u x º , f ª1 c 2 52 yq ¬ ¼

(8-2)

Chapter 8. Methods of Solution of Planar Problems 0 f

§ m · n0 ¨ ¸ © 2S kT0 ¹

143

32

exp c 2

and:

< c, r < 1 c, r < 2 c, r c yIT c q y cx c yIu c h , where q y T01 wT wy f , T0 is the temperature in the plane y h wu wx . The remaining quantities are given by:

(8-3)

0 , and

f

12

§ m · u x ¨ ¸ u x , © 2kT0 ¹

IT c A c , and: 12

§ 2kT ·

Iu c 2 ¨ ¸ B c . © m ¹ The first approximations to the functions, A c and B c , have been found in Chapter 5 to be:

15 A c 16 S O S3 2 c 2 , 1

and: 12

B c

5 8

§ m · ¸ © 2kT ¹

SO¨

.

where molecules are assumed to be rigid spheres. The Navier-Stokes equations are used in the Maxwellian analysis to obtain the dependence of the gas velocity on the normal coordinate. These equations are the moment equations which for the current problem may be written as:

144

Analytical Methods for Problems of Molecular Transport

w vx fdv wx ³

0 ,

(8-4a)

w vx2 fdv 0 , ³ wx

(8-4b)

w vx v y fdv 0 . wx ³

(8-4c)

If the influence of the wall could be neglected ( ) r 0 in Eq. (8-1)), then from Eq. (8-4c) one would have a linear dependence of the mean velocity of the gas on the x -coordinate. Then, allowing for the velocity of the wall equal to zero, one would obtain: Pxy

const ,

wu

wx

u x

const ,

f

x wu wx

f

.

(8-5)

Now, the equation for the correction to the distribution function, Substituting Eq. (8-1) into the Boltzmann equation, one obtains:

) r c, x , may be derived.

c y c 2 52 q y 2cx c y h cx

w) r wx

J < 1 < 2 ) r .

(8-6)

Moreover, one has the following equations for the functions, IT c and Iu c :

c y c 2 52

2cx c y

J c I

c

y T

,

J cx c yIu c ,

(8-8)

where: 12

J M

n2 § m · 00 ¨ ¸ f © 2kT ¹

(8-7)

I M ,

Chapter 8. Methods of Solution of Planar Problems

145

and I M is the standard collision operator given by Eq. (5-13). Allowing for these equations, one may derive the basic equation for the correction, ) c, x :

cx

w) wx

J ) .

(8-9)

Now, consider the boundary condition for this equation. From the general form of the boundary conditions given by Eq. (6-4) one obtains for these slip problems:

) c,0 DW< 1 2 DW < 2 1 DW ) cc,0 ,

^

(8-10)

`

where cc cx , c y , cz . It is useful here to introduce a Hilbert space defined by the scalar product:

U1 x, c , U2 x, c ³ U1 x, c exp^c2 ` U2 x, c dc .

(8-11)

The Maxwellian analysis takes into account the influence of the wall by the choice of a definite form for the distribution function. Maxwell used the following approach for the distribution function of incident molecules:

) c,0 a0 c y ,

) c, f ) c, f a0 c y , where a0 const . This says that the incident distribution function at the wall is identical to the distribution function far from the wall which implies that, in the Knudsen layer, collisions between the incident and reflected molecules have been neglected. Now, to find an approximate solution for a0 , one proceeds in the following way. Taking the scalar product of Eq. (8-9) with c y (in the manner of Eq. (8-11)) gives:

w cx c y , ) c, x wx

i.e.:

0 ,

146

Analytical Methods for Problems of Molecular Transport

c c

, ) c, x

x y

const .

(8-12)

This constant can be evaluated by the use of the asymptotic values, a0 c y , for ) r c, f and thus, the relationship expressing conservation of tangential momentum may be written as:

c c

, ) r c,0

x y

c c

x y

, ) r c, f

0 .

(8-13)

The constant a0 may be found by the use of the boundary condition, Eq. (8-10), in the scalar product, Eq. (8-13). It is the main supposition of the Maxwellian analysis that the discontinuity of the distribution function is allowed for only by using the boundary distribution function for reflected molecules. The Maxwellian distribution function does not, itself, satisfy the boundary condition exactly, but rather, the boundary condition is satisfied only in an integral sense. Having performed the integration in Eq. (8-13), one obtains:

a0

2 DW

DW

5 8

§ wu · 1 § wT · 15 ¸ 32 S OT0 ¨ ¸ . © wy ¹f © wx ¹f

SO ¨

(8-14)

This expression results in relations for the slip-flow and thermal-creep coefficients. In accordance with the definitions of these coefficients, the mean velocity of the gas at large distances from the wall may be presented in the form:

c

u f

§ wT · § wu · cTslQ T01 ¨ ¸ , ¸ © wx ¹f © wy ¹f

O x ¨

m P

(8-15)

where Q QON OP and Q is the kinematic viscosity. Here, two other representations of the mean free path have been introduced; specifically OP and ON . Later, it will be useful to have the following expressions for these: 12

OP and:

8 5

P § 2kT · ¨ ¸ S p© m ¹

OP P 1 ,

(8-16a)

Chapter 8. Methods of Solution of Planar Problems 12

ON

N § mT · ¨ ¸ S p © 2k ¹

64 75

ON N 1 ,

147 (8-16b)

where P and N are the viscosity and thermal conductivity coefficients obtained from the first-order approximation in the Chapman-Enskog theory. This choice for the Maxwellian analysis provides independence of the slipflow coefficients on the order of the Chapman-Enskog solution. Since only the first-order Chapman-Enskog solution is employed in this section ( OP ON O ), the slip factors in Eq. (8-15) are given by: 1

1

cm

2 DW

DW

5 16

S ,

(8-17a)

and: 3 4

cTsl

,

(8-17b)

Next, consider the temperature-jump problem. One may use the Maxwell method to solve this problem. Consider a semi-infinite expanse of gas bounded by a plate oriented in the y z plane and located at x 0 . Assume that, far from the plate, the temperature of the gas depends only on the x -coordinate and that the temperature gradient is small such that the relationship O ln T 1 is satisfied. If this condition holds one may use the Chapman-Enskog distribution function at large distances from the wall and w ln T wx may be expressed as qx T01 wT wx , where T0 is the equilibrium temperature. The temperature is assumed to deviate only slightly from the equilibrium value. To write the correct form of the distribution function far from the wall one should take into account the conservation law obtained from the Boltzmann equation which, for this problem, is given by:

w vx v 2 fdv 0 , wx ³

(8-18)

This implies that:

³ vx v

2

fdv const .

(8-19)

148

Analytical Methods for Problems of Molecular Transport

From this equation, if x o f such that f o f eq 1 cxIT c qx , then (wT / wx )f const . The distribution function for this problem may be written as:

fr

^

`

0 x cxIT c qx c 2 52 a1r x , f 1 c 2 52 xq

(8-20)

where a1r x is a correction to the distribution function which takes into account the influence of the wall. The following approach has been used in the Maxwellian analysis:

a1r f a1 , a1 0 a1 .

(8-21)

The distribution function of the molecules incident on the wall is assumed to be the same as that far from the wall. The distribution function of the reflected molecules on the wall may be found by the use of a boundary condition, which may be written as:

f v,0

1 DW f vx , v y , vz ,0

(8-22)

0 DW f 1 Q r c 2 32 W r ,

where Q r and W r are unknown functions which may be determined by using the impenetrability condition and introducing the thermal accommodation coefficient. For the equilibrium conditions at the wall, W w 0 , but Q w z 0 . This correction to the equilibrium distribution function may be found from the condition:

³ vx f v,0 dv

°

°½

DW ® ³ vx f v,0 dv ³ vx f 0 1 Q w dv ¾ 0 . ¯°

(8-23)

¿°

From Eq. (8-23) one may obtain:

Qw

12 a1 .

(8-24)

The conservation law given by Eq. (8-19) may be written as: 2 2 ³ vx v f v,0 dv ³ vx v f v, f dv

,

149

Chapter 8. Methods of Solution of Planar Problems

and, using the boundary conditions, one can transform this relation to the form:

°

½°

DW D T ® ³ vx v 2 f v,0 dv ³ vx v 2 f 0 1 Q w dv ¾ ¯°

¿°

2

³ vx v f v, f dv

(8-25)

.

Performing the integrations in Eq. (8-25), one can obtain:

'T

a1T0

2 DW DT

DW DT

75 128

§ wT · ¸ . © wx ¹f

(8-26)

SO ¨

Thus, the Maxwellian method may be used successfully to obtain expressions for the slip coefficients. Within the framework of the moment theory, this method may be classified as a one-moment approach because only one moment has been used. The distribution function for the incident molecules at the wall is the same as that in the ambient gas at large distances from the wall and, therefore, the distribution function is independent of the normal coordinate. This does not permit one to obtain spatial distributions of macroscopic values for the gas in the Knudsen layer near the wall; a serious situation which comprises the main deficiency of the Maxwellian method.

2.

LOYALKA'S METHOD.

The Maxwellian method for determining the distribution function of incident molecules at the wall may be generalized by a simple modification suggested by Loyalka [2]. Consider the details of this generalization in the solutions of the slip-flow and temperature-jump problems. To obtain simple solutions for the slip coefficients, one considers a gas flow bounded by a flat wall for the conditions described in Section 8.1. The distribution function for these slip problems has the form:

f

^

0 y 2c y xh c yIT c q y f 1 c 2 52 yq

`

cx c yIu c h ) r c, x ,

(8-27)

150

Analytical Methods for Problems of Molecular Transport

where q y T01 wT wy f , h wu wx , and c yIT c q y and cx c yIu c h f are the Chapman-Enskog corrections to the distribution function for thermal conductivity and viscosity, respectively. The correction that allows for the influence of the wall may be found from the equation:

cx

w) r wx

J )r .

(8-28)

For x o f , the functions ) r c, x have the form: lim ) r c, x a0 c y ,

x of

where a0 is a constant. If the two-moment approach is used, then the correction, ) r c, x , may be expressed as:

) r c, x a0r x c y .

(8-29)

The Maxwellian analysis is constructed on the assumption that a0r f a0 0 a0 . Loyalka has proposed a simple generalization of Maxwell’s method which assumes that the distribution function of incident molecules has the following form at the wall:

) c,0 a0 0 c y

(8-30)

aw c y ,

where aw is an unknown constant. From a physical viewpoint, the relation aw z a0 implies that the incident stream has a mass velocity different from that of the hydrodynamic solution at the wall. The unknown constants, a0 and aw , may be found by the consideration of some scalar products of Eq. (8-28). First, the scalar product given by Eq. (8-13) is used. Next, the scalar product of Eq. (8-28) with cx c yIu c gives:

w 2 cx c yIu c , ) r c, x wx

c c I

x y u

J ª¬c c I c º¼ , ) x y u

r

c, x

c , J ª¬) r c, x º¼

2 cx c y , ) r c, x

0 ,

and, therefore:

c c I

2 x y u

c , ) r c, x

const .

(8-31)

151

Chapter 8. Methods of Solution of Planar Problems

To obtain this expression, Eq. (8-8) and the commutative property of the standard bracket integrals (which is due to the self-adjointness of the collision operator) have been used. Now there are two equations that may be represented as:

c c

x y

, ) r c,0

0 ,

(8-32)

and:

c c I

2 x y u

c , ) r c,0

c c I

2 x y u

c , ) r c, f

.

(8-33)

Using the boundary condition given by Eq. (8-10) with Eqs. (8-32) and (8-33), one can obtain:

aw

15 32

S Oqy

2 DW

DW

5 8

(8-34)

SO h ,

and:

a0

1 12 DW 1532

S Oqy

2 DW § 4 S · 5 1 DW ¸ 8 SO h . DW ¨© 2S ¹

(8-35)

Taking into account Eq. (8-15), the following expressions for the slip coefficients may be written:

cm

cTsl

2 DW § 4 S · 5 1 DW ¸ 16 S , DW ¨© 2S ¹

3 4

1 12 DW .

In particular, for the very important case when DW given by:

'u

wu · ¸ , © wx ¹f

1.116 O §¨

(8-36)

(8-37)

1 , the slip velocities are

(8-38)

152

Analytical Methods for Problems of Molecular Transport

usl

9 8

§ wT · ¸ . © wy ¹f

Q T01 ¨

(8-39)

The analytical expressions for the slip coefficients are not altered for arbitrary models of the intermolecular potential. This assumption is correct within the framework of this method if one uses the first-order ChapmanEnskog solutions for thermal conductivity and viscosity to derive these formulas. Next, consider the solution of the temperature-jump problem by Loyalka’s method. The same conditions are used for the gas at large distances from the wall as were used in the Maxwellian analysis. Therefore, the distribution function may be written as:

fr

^

`

0 x cxIT c qx ) r c, x , f 1 c 2 52 xq

(8-40)

where qx T01 wT wx f and ) r c, x is the correction that allows for the influence of the wall. This correction is assumed to have the form:

) r c, x a1r x c 2 52 ,

(8-41)

where:

lim ) r c, x constu c 2 52 .

x of

The following approach has been used in the Loyalka analysis:

a1r f a1 ,

a1 0 a1w ,

a1 z a1w ,

such that there are now two independent constants, a1 and a1w , instead of the single constant, a1 , in the Maxwellian analysis. This generalization gives a more correct form for the incident molecular distribution function at the wall than that used by Maxwell. The unknown constants, a1 and a1w , may be obtained by the use of two scalar products of the Boltzmann equation, Eq. (8-28), and the boundary conditions for the correction to the distribution function. First, the scalar product given by Eq. (8-18) results in one equation for the two unknown constants:

153

Chapter 8. Methods of Solution of Planar Problems

c c

2

x

, ) r c,0

c c x

2

, ) r c, f

0 .

(8-42)

Next, the scalar product of Eq. (8-28) with cxIT c yields:

w 2 cx IT c , ) r c, x wx

c I c , J ª¬) c, x º¼ ) c, x , J ª¬c I c º¼ ) c, x , c c c c , ) c, x 0 . r

x T r

x

T

2

r

5 2

x

2

r

x

This implies that:

c I

2 x T

c , ) r c, x

const .

(8-43)

Thus, the second equation for the unknown functions, a1 and a1w , may be written as:

c I

2 x T

c , ) r c,0

c I

2 x T

c , ) r c, f

.

(8-44)

The corrections to the distribution function of the reflected molecules at the wall may be found by using the Maxwellian model of the boundary conditions which, for this problem, may be expressed in the form:

) c,0 2 DW cxIT c qx 1 DW ) cc,0

DW ªQ r c 2 32 W r º , ¬ ¼

^

(8-45)

`

where cc cx , c y , cz , and Q r and W r are corrections to the number density and temperature for the diffusely reflected molecules. The corrections, Q r and W r , may be obtained from the three additional conditions: ³ vx f v,0 dv ³ vx f v,0 dv

0 ,

(8-46)

154

Analytical Methods for Problems of Molecular Transport ³ vx f v,0 dv ³ vx f w v,0 dv

DT

0 ,

(8-47)

Qx . Qxw

(8-48)

In these expressions, one must use the general form of the boundary condition (Eq. (8-22)). Then, using Eq. (8-41), after integration, one obtains:

Qr

75 256 S 1 DT O qx 12 2 DT a1w ,

Wr

75 128

(8-49)

S 1 DT O qx 1 DT a1w .

(8-50)

Now, Eqs. (8-42) and (8-44) may be employed to derive the relations for a1 and a1w . Substituting Eqs. (8-49) and (8-50) into Eq. (8-45), and performing the integrations indicated by Eqs. (8-42) and (8-44), one can obtain:

a1w

a1

75 128

S

2 DW DT

DW DT

O qx ,

ª D D 2 DW a1w «1 W T 2 DW DT «¬

(8-51)

52 25

º

»

S 1 12 » .

(8-52)

¼

Taking into account these expressions, one can obtain the following formula for the temperature-jump:

'T

§ wT · cT O ¨ ¸ . © wx ¹f

(8-53)

Here, cT is the temperature-jump coefficient given by:

cT

75 128

S

2 DW DT ª DW DT 2 DW «1 DW DT ¬« 2 DW DT

52 25

º

»

S 1 12 » . ¼

(8-54)

Chapter 8. Methods of Solution of Planar Problems

155

This expression for the temperature-jump coefficient given in Eq. (8-54) is more general than that reported in [2]. For the specific case when DT 1 , however, both expressions are identical. Eq. (8-54) is applicable for arbitrary models of the intermolecular potential if one uses the first-order Chapman-Enskog solution. A more complete description of the temperature-jump phenomenon may be obtained by solving Problem 8.6. The preceding analysis of the slip-flow transport problems shows that Loyalka’s method may be classified as a two-moment approximation. Although this method eliminates the main deficiency of the Maxwellian analysis, it does not permit one to obtain the dependence of macroscopic values on the normal coordinate in the Knudsen layer. Nevertheless, it is very important to note that the conservation laws used in Loyalka’s analysis are the exact moment solutions of the Boltzmann equation. Each exact solution for the correction to the distribution function must satisfy these moment relationships which are the most general properties of the Boltzmann equation. The use of exact moment solutions allows one to anticipate highly accurate results based on what is a reasonably simple analytical analysis. Moreover, for this method there would appear to be no difficulties in generalizing the analysis for arbitrary models of the intermolecular potential. To this end, it is necessary only that ChapmanEnskog solutions of sufficiently high-order be employed.

3.

THE HALF-RANGE MOMENT METHOD.

Another approximate method was first proposed by Wang-Chang and Uhlenbeck [3] and was then further developed by Gross and Ziering, among others, to solve planar boundary transport problems [4-13]. Known initially as the half-range moment method it is also now commonly referred to as the Gross-Ziering method after those principally responsible for its further development. To demonstrate the various features of this method as described in [13], the velocity-slip problem is examined. The conditions for the gas are assumed to be the same as described in Section 8.1. For the linearized problem the distribution function may be given by:

fr

½° § wu · § wu · 0 ° r f ®1 2c y x ¨ c c c I ¸ ¨ ¸ ) c, x ¾ , x y u © wx ¹f © wx ¹f ¯° ¿°

(8-55)

where ) r c, x is the correction to the distribution function which accounts for the influence of the wall and Iu c is specified by Eq. (8-3). The correction, ) r c, x , may be found from the linearized Boltzmann equation:

156

Analytical Methods for Problems of Molecular Transport

cx

w) wx

J ) ,

(8-56)

where:

J )

³ bdbd H ³ c1 c

0 f1 )1c ) c )1 ) dv1 .

In the half-range moment method, the distribution function is approximated by half-range polynomials in velocity space and the spacedependent coefficients are determined by solving half-range moment equations. It might appear that the most natural systematic scheme for construction of half-range moment equations is to introduce a complete orthonormal set of polynomials, defined over the half-range in velocity space, such that the distribution function may then be expanded in terms of these polynomials and characterized by the space-time dependent expansion coefficients. Unfortunately, this scheme requires the use of a large number of expansion terms which results in great mathematical difficulty. The complexity of the calculations increases rapidly with the number of terms and, therefore, this calculation scheme is not always practical for use in transport problems. However, the distribution function can sometimes satisfy the boundary conditions exactly such that very accurate results can be obtained by low-order approximations employing only a few terms in the expansion. Such approximations can give the exponential variation of the macroscopic values in the Knudsen layer near the wall. When the boundary conditions are not satisfied exactly, it is necessary to use the integral forms of the boundary conditions. For the current slip-flow problem, assume that the correction, ) r c, x , may be represented in the half-range velocity space as:

) r c, x a0r x c y a1r x cx c y .

(8-57)

If the following function is introduced:

1 ; cx ! 0 , sign cx ® ¯1 ; cx 0 , the correction to the distribution function, ) c, x , is conveniently expressed for the full velocity space by:

157

Chapter 8. Methods of Solution of Planar Problems

) r c, x ) a0

1 sign cx 2 2

a0

cy

a0

) 2

a0

1 sign cx 2

c y sign cx

a1 a1 cx c y 2

(8-58)

a1 a1 cx c y sign cx . 2

Substituting this correction to the distribution function into Eq. (8-56), multiplying both sides alternately by:

c y 1 r sign cx exp c 2 dc , and:

cx c y 1 r sign cx exp c 2 dc , and, finally, integrating over all velocities, one obtains the following system of half-range moment equations:

da0 dx

a0 A1 a0 A1 a1 A2 a1 A3 ,

(8-59a)

da0 dx

a0 A1 a0 A1 a1 A3 a1 A2 ,

(8-59b)

da1 dx

a0 B1 a0 B1 a1 B2 a1 B3 ,

(8-59c)

da1 dx

a0 B1 a0 B1 a1 B3 a1 B2 ,

(8-59d)

where the coefficients, Ai and Bi , may be expressed as linear combinations of some of the moments of the collision integral:

158

Analytical Methods for Problems of Molecular Transport

A1

b ª¬ I1 12 S I 2 º¼ ,

A2

b ª¬ I 2 12 S I 3 I 4 º¼ ,

A3

b ª¬ I 2 12 S I 3 I 4 º¼ ,

B1

b ª¬ I 2 12 S I1 º¼ ,

B2

b ª¬ I 3 I 4 12 S I 2 º¼ ,

B3

b ª¬ I 3 I 4 12 S I 2 º¼ ,

where:

b

4 . S 4 S

Introducing the following notation: ª¬) v ,< v º¼

2 ³) v J < v exp c dc .

the bracket integrals, I j , may be written as:

I1

ª¬ c y sign cx , c y sign cx º¼ ,

I2

ª¬c y sign cx , cx c y º¼ ,

159

Chapter 8. Methods of Solution of Planar Problems

I3

ª¬cx c y , cx c y º¼ ,

I4

ª¬cx c y sign cx , cx c y sign cx º¼ .

The following additional bracket integrals that one might expect to encounter are identically zero and do not appear in this moment system:

ª¬ cx c y , cx c y sign cx º¼

ª¬c y sign cx , cx c y sign cx º¼

0 .

The bracket integrals, I j , correspond to the original notation used by Gross and Ziering [5] and are different from the Chapman-Enskog bracket integrals usually encountered in kinetic theory. The relationship between these two kinds of bracket integrals is: 12

>) ,< @

§ m · nS 3 2 ¨ ¸ © 2kT ¹

>) ,< @

,

(8-60)

where >) ,< @ is the standard bracket integral. The values of the integrals, I j , are dependent on the intermolecular interaction potentials. A method of calculation of these integrals, for rigidsphere molecules, is given in Appendix A and is substantially simpler than previous methods [14, 15]. For rigid-sphere molecules, this method results in the following bracket integrals:

I1

1 8 2 S , I2 12 O

I3

52

S , I4 O

3S 8 2 S , S O 32

46 17 2 S . O 120

The value for the integral, I 4 , originally reported by Gross and Ziering [5] is in error. The absolute value of the Gross-Ziering integral is about 9.3 times greater than that reported here. The same analytical expressions being used in this analysis were previously reported by Porodnov and Suetin [9]. Now, the solution of the moment system, Eqs. (8-59a)-(8-59d), takes the form:

160

Analytical Methods for Problems of Molecular Transport

a0 x

A D 0 B exp D x ,

(8-61a)

a0 x

A B exp D x ,

(8-61b)

a1 x D1 B exp D x ,

(8-61c)

a1 x D1 B exp D x ,

(8-61d)

where:

4.19294 , 0.52413 ,

D 0 1

D

D1 D

4.12697 , 2.20153 O 1 .

The constants of integration, A and B , may be found from the boundary conditions (Eq. (8-10)) which may be written as:

a0 0

1 DW a0 0

,

§ wu · a1 0 2 DW Iu c ¨ ¸ 1 DW a1 0 . w x © ¹f

(8-62)

From these boundary conditions one then obtains:

§ wu · 2 DW 1 DW D 0 , ¸ © wx ¹f DW D1 1 DW D1

A

5 4

SO¨

B

5 4

SO¨

§ wu · 2 DW , ¸ x w © ¹f D1 1 DW D1

The solution of the moment system, Eqs. (8-61a)-(8-61d), can now be employed in the following expression for the mean velocity of the gas:

161

Chapter 8. Methods of Solution of Planar Problems

u x n 1 ³ v y f v, x dv 12

§ 2kT · ¨ ¸ © m ¹

º ½° 1 ° § wu · 1 ª a1 a1 » ¾ . ® x ¨ ¸ 4 « a0 a0 S ¬ ¼ °¿ °¯ © wx ¹f

(8-63)

which, at large distances from the wall, has the linear profile: 12

u x

x of

§ wu · § 2kT · x ¨ ¸ 14 ¨ ¸ w x © ¹f © m ¹

a

0

f a0 f

.

(8-64)

The solution of the Navier-Stokes equations, if one uses the slip boundary conditions, has the following form:

u x

wu · ¸ . © wx ¹f

cm O x §¨

(8-65)

A comparison of Eqs. (8-64) and (8-65) indicates that: cm

2 DW

DW

5 8

S

3.193 DW 3.603 0.524 DW

.

(8-66)

For the particular case when DW 1 , one has cm 1.125 . Additionally, for this case, the mean velocity of the gas may be expressed as:

u x

ª § 2.202 x · º § wu · « x 1.125 O 0.345 O exp ¨¨ ¸¸ » ¨ ¸ . O © ¹ ¼» © wx ¹f ¬«

(8-67)

In summary, the four-moment approximation to the half-range moment method described in this section satisfies the boundary conditions exactly and results in an exponential profile for the mean velocity of the gas in the Knudsen layer (near the wall). This equates to substantially more information about the state of the gas near the wall than is obtained from the methods of Maxwell and Loyalka described in Sections 8.1 and 8.2.

162

Analytical Methods for Problems of Molecular Transport

4.

FEATURES OF THE BOUNDARY CONDITIONS FOR THE MOMENT EQUATIONS.

In Sec. 8.3, the boundary slip-flow problem was solved by using the Maxwellian boundary conditions for the distribution function. Unfortunately, these boundary conditions have an essential deficiency connected with the execution of the conservation laws for the Boltzmann equation. It was stated in Section 8.2 that a correction to the distribution function is needed to satisfy the two conservation laws expressed by Eqs. (832) and (8-33). The satisfaction of these conditions is a general requirement of the Boltzmann equation in planar boundary transport problems. If the exact form of the boundary conditions is used, the conservation law expressed in Eq. (8-33) is not satisfied. Therefore, one must use another form of the boundary conditions. To explore this problem in more detail, consider a microscopic boundary condition with the following form: f

Af .

(8-68)

The Boltzmann equation near the flat wall may be represented by [11]:

vx

wf wx

Gf G x ª¬K vx vx f K vx vx Af º¼ , Gt

(8-69)

where G x is the Dirac delta function and: 1 ; v x ! 0 , ¯0 ; v x 0 .

K vx ®

Introducing the boundary condition into the right-hand-side of Eq. (8-69) and performing the integration of this equation over x from 0 to H ! 0 gives: H

lim ³ vx

H o0

0

wf dx vx f wx

° vx Af ; vx ! 0 , ® ; vx 0 . °¯vx f

(8-70)

From this relation one may conclude that to obtain the boundary conditions to the moment equations one should use the following form:

163

Chapter 8. Methods of Solution of Planar Problems

³ vxIi v f

³ vxIi v Af

dv

dv ,

(8-71)

where Ii are molecular properties which are used to construct the moment system. It would appear that there is now a complete set of boundary conditions for the moment equations. Unfortunately, these boundary conditions do not simultaneously satisfy both conservation laws which are expressed as Eqs. (8-32), (8-33) and (8-42), (8-44), for the slip-flow and thermal transport problems, respectively. These conservation equations are general properties of the Boltzmann equation and, therefore, to obtain the correct results, the two conservation laws must be taken into account. One possible solution to this problem is to use these conservation laws as integral expressions of the boundary conditions. To examine this idea, consider the slip-flow problem for which there exists a solution by the half-range moment method. Instead of the boundary conditions expressed by Eq. (8-62) there are two integral relations that may be written as:

c c

x y

, ) r c,0

c c I

2 x y u

0 ,

c , ) r c,0

(8-72)

c c I

2 x y u

c , ) r c, f

.

(8-73)

The correction to the reflected distribution function at the wall surface has the following form:

§ wu · ¸ © wx ¹f

) c,0 2 DW cx c yIu c ¨ 1 DW

ª a0 ¬

0 c y 0 cx c y º¼ a1

(8-74)

,

12

where Iu c 2 2kT m B c and the function, B c , corresponds to the second-order Chapman-Enskog solution. The second-order terms in this quantity improve the accuracy of the analysis. This function, for rigidsphere molecules, may be written as [17]:

¼

Iu c 54 S OP ª1 b2 2 S512 c 2 º , ¬

(8-75)

164

Analytical Methods for Problems of Molecular Transport

205 where b2 0.05854 , and OP 202 O. The correction to the distribution function has been obtained in Section 8.4 and is given by Eqs. (8-61a)-(8-61d). The use of the new boundary conditions, however, results in new values for the constants, A and B . By using the new boundary model of Eqs. (8-72)-(8-73), one obtains: 2

A

2 DW '1 DW '

5 4

S OP h ,

'2 '

B

5 4

S OP h ,

(8-76)

where '1 , ' 2 , and ' may be represented for rigid-sphere molecules by the following expressions: '1

ª¬ 0.6690 DW 0.1775 º¼ ,

(8-77a)

'2

0.3413 DW 0.1706

(8-77b)

'

,

ª¬ 0.7549 DW 0.09714 º¼ .

(8-77c)

The mean velocity of the gas is expressed by Eq. (8-63). Using Eqs. (877), the relation for the mean velocity may be given by:

§ wu · u x ¬ª cm OP x OP ud x ¼º ¨ ¸ © wx ¹

,

(8-78)

f

where cm is the isothermal-creep coefficient and ud x is the dimensionless velocity defect that describes the deflection of the mean velocity from the linear profile in the Knudsen layer. The expressions for these quantities, with an arbitrary tangential momentum accommodation coefficient, may be written as: cm

2 DW

DW

5 8

S

'1 , '

ud x 1.4229

§ 2.2015 x · '2 exp ¨¨ ¸¸ . ' O © ¹

(8-79)

(8-80)

165

Chapter 8. Methods of Solution of Planar Problems

Table 8-1. Values of the slip-flow coefficient,

cm

DW 0.2 0.4 0.6 0.8 1.0 †

G%

aa

Eq. (8-79) 9.0710 4.1310 2.4650 1.6185 1.1006

cm

Eq. (8-81) 9.0277 4.0976 2.4400 1.6006 1.0884

a u 100 ; a

cm .

G

†

0.48 0.81 1.01 1.11 1.11

cm

G

Eq. (8-66) 9.1235 4.1759 2.5027 1.6496 1.1254

†

0.80 1.51 2.14 2.71 3.22

cm (from Eq. (8-79).

Table 8-1 contains the comparison of results which are obtained by the use of the various methods for several values of the coefficient, DW . Results obtained by Loyalka’s analysis take into account the correction of the second-order to the Chapman-Enskog distribution function, and therefore the expression for the slip coefficient differs from that represented by Eq. (835). For this case, Loyalka’s formula has the form:

cm

2 DW § 4Z1 S · 5 1 DW ¸ 16 S ; Z1 1 b2 2 174 b2 2 ¨ DW © 2S ¹

2

.

(8-81)

The Gross-Ziering analysis does not permit the use of the second-order Chapman-Enskog solution owing to the form of the boundary conditions. The comparison of these results shows a good agreement between the values of cm calculated by the use of Eq. (8-79) and by Loyalka’s formula. Equation (8-66), which was obtained using the common forms of the boundary conditions, yields results that do not agree so well with Loyalka’s formula. However, this comparison does not allow one to draw any rigorous conclusions regarding the accuracy of any these models because the exact analytical solutions are not known. It should be possible, however, to draw conclusions regarding the accuracy of the different methods by comparing their results to reasonably exact direct numerical solutions of this slip-flow problem. This will be considered in Section 9.3, where a complete analysis of the accuracy of the various methods and boundary models will be made.

5.

SOLUTION OF THE THERMAL-CREEP PROBLEM BY THE HALF-RANGE MOMENT METHOD.

The flow of a gas in the y -direction, over a plane surface lying at x 0 , is considered for the conditions described in Section 8.1. Far from the plate

166

Analytical Methods for Problems of Molecular Transport

the gas is maintained at a constant temperature gradient, wT wy f . For these conditions, the thermal-creep problem is described by the following distribution function:

f

0 y c y A c q y ) r c, x º , f ª1 c 2 52 yq ¬ ¼

(8-82)

where A c may be written as [17]:

1 2 2 2 A c a1 ª S3 2 c 2 a2 S3 2 c 2 º , ¬ ¼ 45 where a2 0.08889 , a1 15 S ON , and ON 44 O . In this formula 16 molecules are assumed to be rigid spheres. The integral form of the boundary conditions allows one to employ the same correction to the distribution function for the thermal-creep problem as in the slip-flow problem. This form of the distribution function takes into account the viscosity effects in a sufficiently correct form. Let the correction to the distribution function be assigned by the relation: 2

2

) r c, x a0r x c y a1r x cx c y ,

(8-83)

where the functions, air may be written as:

a0 x C1 D 0 C2 exp D x ,

(8-84a)

a0 x C1 C2 exp D x ,

(8-84b)

a1 x D1 C2 exp D x ,

(8-84c)

a1 x D1 C2 exp D x .

(8-84d)

The values, D ir , D , for rigid spheres are given by:

D 0 D1

4.19294 , 0.52413 ,

D1 D

4.12697 ,

2.20153 O 1

.

167

Chapter 8. Methods of Solution of Planar Problems

The constants C1 and C2 may be obtained from boundary conditions which have the following form:

c c

x y

, ) r c,0

c c I

2 x y u

0 ,

(8-85a)

c , ) r c,0

c c I

2 x y u

c , ) r c, f

.

(8-85b)

where ) c,0 is represented by:

¼

) c,0 DW a1 2 q y c y ª S312 c 2 a2 2 S3 22 c 2 º ¬

(8-86)

1 DW ª¬ a0 0 c y a1 0 cx c y º¼ , and the standard notation for the scalar product from Eq. (8-11) has been used. In the case of Iu c , Eq. (8-75) is used. This integral form of the boundary conditions makes allowance for the features of the Boltzmann equation near a flat wall. Having performed integration in Eqs. (8-85) one obtains: C1

'1 , '

C2

'2 , '

where '1 , ' 2 , and ' may be written as:

'1

2 a1 q y ª¬ 0.4354 0.2179 DW º¼ ,

'2

0.3022 DW a1 2 q y

'

,

0.8518 0.1096 DW

.

For this problem it is convenient to represent the mean velocity of the gas in the form:

168

Analytical Methods for Problems of Molecular Transport

u x cTslQ q y ½ ª º' 1 D1 D1 » 2 exp D x ¾ , u ®1 12 « D 0 1 S ¬ ¼ '1 ¯ ¿

(8-87)

where cTsl is given by: 3 2

cTsl

0.4354 0.2179 DW 0.8518 0.1096 DW

(8-88)

.

The dimensionless velocity defect may be defined as:

ud x

u f u x

Q q y

(8-89)

.

This quantity may then be expressed by:

ud x DW

§ 2.2015 x · 0.5822 exp ¨¨ ¸¸ O 0.8518 0.1096 DW © ¹

.

(8-90)

For the particular case when DW 1 , one obtains cTsl 1.0193 . The full analysis of results obtained here will be described later. It is of considerable interest to investigate the influence of the various forms of the boundary conditions on the thermal-creep coefficient and the velocity defect. This analysis will be described in the next section.

6.

INFLUENCE OF THE BOUNDARY MODELS ON THE THERMAL-CREEP COEFFICIENT.

Here, expressions are derived for the thermal-creep coefficient and the velocity defect for the case when one uses the ordinary moment form of the boundary conditions that may be represented by: 2 2 ³ cx c y) c,0 exp c dc ³ cx c y A) c,0 exp c dc

,

(8-91a)

169

Chapter 8. Methods of Solution of Planar Problems 2 2 2 2 ³ cx c y) c,0 exp c dc ³ cx c y A) c,0 exp c dc

,

(8-91b)

where:

1 2 2 2 A) c,0 DW a1 q y ª S3 2 c 2 a2 S3 2 c 2 º ¬ ¼

(8-92)

1 DW ª¬ a0 0 c y a1 0 cx c y º¼ . Substituting the functions given by Eqs. (8-84a)-(8-84d) into Eqs. (8-91a) and (8-91b), one obtains the constants C1 and C2 which determine the correction to the distribution function. This analysis yields: cTsl

3 2

1.5464 1.1741 DW 3.0927 0.4499 DW

ud x DW

,

§ 2.2015 x · 3.4149 exp ¨¨ ¸¸ O 3.0927 0.4499 DW © ¹

(8-93)

.

(8-94)

For example, if DW 1 , then cTsl 1.1519 . This value is about 15% greater than that given by Eq. (8-96). But a reliable analysis of the accuracy of the model considered here, as was indicated previously, may only be performed by making a comparison with the numerical solution for this boundary transport problem. The accuracy of the various methods and boundary models for the thermal-creep problem will be investigated in Section 9.5.

PROBLEMS 8.1. Determine the mean velocity profile, u x , and the pressure tensor component, Pxy , for Couette flow by means of the Maxwell method. Derive an expression for the apparent viscosity coefficient, P c , that is determined by:

Pc

uw d

P

wu . wx

170

Analytical Methods for Problems of Molecular Transport

Figure 8-2. The geometry to be used in Problem 8.1 in determining the mean velocity profile, the pressure tensor component, and the apparent viscosity for Couette flow.

Use the flow geometry in Fig. 8-2. Solution: For this analysis the distribution function should be chosen in the form: 0 y , f 1 cx c yIu c A 2 Axc

^

f

`

which provides a linear profile for the mean velocity of the gas. Conservation of tangential momentum at the wall may be written as:

c c

x y

, ) r c, 12 d

0 .

Using the Maxwellian boundary condition for the rebounded distribution function, one can obtain: uw x] Kn , d

u x

] Kn

§ 2 DW 5 O · S ¸ ¨1 DW 8 d ¹ ©

Pxy

uw , d

P c

and P c P] Kn .

1

,

171

Chapter 8. Methods of Solution of Planar Problems

8.2. Determine the mean velocity profile, u x , and the pressure tensor component, Pxy , for Couette flow by means of Loyalka’s method. Derive an expression for the apparent viscosity coefficient, P c , that is determined by:

Pc

uw d

P

wu . wx

Use the same flow geometry as in Problem 8.1 (see Fig. 8-2). Solution: One should use a distribution function of the form:

f

^

`

0 f 1 ) r c, x ,

where ) r c, x 12 a0 x a0 x c y 12 a0 x a0 x c y sign cx . As a result of this choice, the boundary conditions are satisfied exactly. The conservation laws may then be expressed in the form cx c y , ) r c, x A1 and cx2 c yIu c , ) r c, x 2 A1 x A2 where A1 and A2 are constants which may be found from the two boundary conditions. From the conservation laws, one obtains:

a0 x uw

5 8

1

S

O

2

d 2 DW

5 8

DW

x d

S

,

O d

and:

a0 x uw

5 8

1

S

O

2

x d

5 8

S

d 2 DW

DW

O

.

d

Using these expressions, one can obtain the same relations for u x and Pxy as in Problem 8.1. However, the distribution function obtained here describes the Knudsen limit regime better than that obtained in Problem 8.1. 8.3. Determine the mean velocity profile, u x , and the pressure tensor component, Pxy , for Couette flow by means of the Gross-Ziering method for the two-moment approach. Take the correction to the distribution function in the form ) r c, x a0r x c y . Derive an expression for the apparent viscosity coefficient, P c , that is determined by:

172

Analytical Methods for Problems of Molecular Transport

Pc

uw d

Pxy .

Use the same flow geometry as in Problem 8.1 (see Fig. 8-2). Solution: For this two-moment approach, the half-range moment system is obtained by multiplying the Boltzmann equation, Eq. (8-56), by c y ª¬1 r sign cx º¼ exp c 2 dc and integrating over all velocities. The solution of this moment system yields:

uw x] Kn , d

u x

] Kn

§ 2 DW S · ¨1 ¸ DW I1d ¹ ©

12

p S § m · I1 ¨© 2kT ¸¹

Pxy

I1

1

,

uw ] Kn , d

1 8 2 S , 12 O

and P c P] Kn where P

0.2909 U v O .

8.4. Using the full-range moment equations given by: d vx Q v fdv dx ³

n' Q v ,

where Q v vx and vx v y , determine the mean velocity, u x , and the pressure tensor component, Pxy , for the Couette flow. Solve this problem for the two-moment approach using the following form of the correction to the distribution function ) r c, x a0r x c y . Derive an expression for the apparent viscosity coefficient, P c , that is determined by:

Pc

uw d

Pxy .

Chapter 8. Methods of Solution of Planar Problems

173

Use the same flow geometry as in Problem 8.1 (see Fig. 8-2). Solution: The moment equations are obtained by multiplying the Boltzmann equation, Eq. (8-56), by the terms, c y and cx c y , and integrating over all velocities. The solution of the moment system then yields: uw x] Kn , d

u x

§

] Kn ¨1 ©

2 DW S 3 2 · ¸ DW 2 I 2 d ¹ 12

Pxy

I2

1 2

pS § m · I 2 ¨© 2kT ¸¹

1

,

uw ] Kn , d

3S 8 2 S , S O 32

and P c P] Kn where P 0.5100 U v O . In this solution, molecules have been assumed to be rigid spheres. 8.5. Determine the temperature distribution between parallel plates by the Maxwell method. Derive an expression for the apparent thermal conductivity coefficient N c , that is given by:

Nc

2'T d

N

wT . wx

Use the flow geometry in Fig. 8-3. The correction, 'T , to the temperature, T0 , is assumed to be small such that the relationship 'T T0 is satisfied. Solution: A distribution function should be sought in the form:

f

^

1 0 f 1 cx S3 2 c 2

15 16

`

S O A c 2 52 Ax .

Using the boundary conditions obtained in Problem 6.5 and conservation of the heat flux as given by:

174

Analytical Methods for Problems of Molecular Transport

Figure 8-3. The geometry to be used in Problem 8.5 in determining the temperature distribution and apparent thermal conductivity coefficient between parallel plates having different temperatures.

c c x

2

, ) c, 12 d

0 ,

one can obtain W x qx] T Kn ,

q

]T

2'T , T0 d

§ 75 2 DW DT O · ¨1 64 S ¸ DW DT d ¹ ©

1

,

and N c N] T Kn where W x is a correction to the temperature. 8.6. Derive analytical formulas for the boundary temperature and pressure jumps for a planar surface in a monatomic non-condensable gas using Loyalka’s method. Use the second-order Chapman-Enskog solution for rigid-sphere molecules. Assume that molecules reflect diffusely at the surface. Solution: The following distribution function should be used to describe this boundary problem:

f

0 x cx A c qx ) r c, x º . f ª1 c 2 52 xq ¬ ¼

The correction, ) r c, x , should be taken in the form:

175

Chapter 8. Methods of Solution of Planar Problems

) r c, x a r x c 2 52 b r x . The limiting values of the functions introduced here are given by: lim a r x a f , lim b r x b f ,

x of

x of

a 0 aw , and b 0 bw . These limiting values are specified by the boundary-jump effects. The unknown constants, a f , b f , aw , and bw , may be determined from the conservation laws of Eq. (8-9) which are given by:

c

x

, ) r c,0

c c x

2

0 , cx2 , ª¬) r c,0 ) r c, f º¼

, ) r c,0

0 ,

0 , and cx2 A c , ª¬) r c,0 ) r c, f º¼

0 .

The distribution function of reflected molecules (see Problem 6.5) is given by:

ª 2G · º 2G §G ) c,0 cx A c qx 1 DT « 2 c 2 ¨ 2 1 ¸ » 1 , S ¹¼ S ©S ¬

where:

G1 12 S aw 12 bw , and G 2 S

5 8

S a1 2 qx aw 12 bw .

This formulation yields:

Tasy 0 T0 T0

cT ON qx , and

pasy 0 p0 p0

The jump coefficients are then given by:

c p ON qx .

176

Analytical Methods for Problems of Molecular Transport

cT

75 128

S

2 DT ª DT § 52 H T 1 · º » , «1 DT ¬ 2 DT ¨© 25 S 2 ¸¹ ¼

cp

75 256

S

2 DT ª D T § 8 H1 1 · º » , «1 DT ¬ 2 DT ¨© 5 S 2 ¸¹ ¼

and:

2

,

2 23 2 433 where H T 1 26 a2 208 a2 for rigid-sphere molecules.

H1 1 14 a2 2 , and a2 2

0.08889

8.7. Derive analytical expressions for the boundary temperature and pressure jumps in a monatomic vapor over a planar, liquid surface using Loyalka’s method. Use the second-order Chapman-Enskog solution for rigid-sphere molecules. Consider the case of an arbitrary evaporation coefficient. Solution: The distribution function used should be:

0 x 2cx u cx A c qx ) r c, x º , f ª1 c 2 52 xq ¬ ¼

f

12

where u m 2kT0 u x . The correction, ) r c, x , should be taken in the The limiting values of the form ) r c, x a r x c 2 52 b r x . functions introduced here are given by:

lim a r x a f , lim b r x b f ,

x of

x of

a 0 aw , and b 0 bw . These limiting values are specified by the boundary-jump effects. The unknown constants, a f , b f , aw , and bw , may be determined from the conservation laws of Eq. (8-9) which are given by:

c

x

, ) r c,0

c c x

2

, ) r c,0

0 , cx2 , ª¬) r c,0 ) r c, f º¼

0 ,

0 , and cx2 A c , ª¬) r c,0 ) r c, f º¼

0 .

Chapter 8. Methods of Solution of Planar Problems

177

The boundary conditions (see Problem 6.6) for the reflected distribution function may be expressed as:

) c,0 cx A c qx 1 D m ª 2G · º 2G °½ §G ° u ® 2 c 2 «1 DT ¨ 2 1 ¸ » 1 ¾ , S ¹ ¼ S ¿° ©S ¬ ¯°

where:

G1

12 S aw 12 bw S u

and:

G2 S

5 8

S a1 2 qx aw 12 bw 54 S .

This formulation yields:

Tasy 0 T0

T u cT ^D m ` ON qx cT ^D m ` u ,

T0 and:

pasy 0 p0 p0

T u cp ^D m ` ON qx cp ^D m ` u ,

where the jump coefficients are given by: T cT ^D m `

75 128

S

D § 52 H T 1 · º 2 D ª , «1 D ¬ 2 D ¨© 25 S 2 ¸¹ »¼

ª 4 3D 16 H1 º u cT ^D m ` 18 S « 5 » , S¼ ¬ D

178

Analytical Methods for Problems of Molecular Transport T cp ^D m `

75 256

S

D § 8 H1 1 · º 2 D ª , «1 D ¬ 2 D ¨© 5 S 2 ¸¹ »¼

and:

ª 4 3D m 1 § 4 3D u cp ^D m ` 12 S « 8¨ © D ¬ Dm Here, D 2 and a2

·º ¸» . ¹¼ 2

,

2 23 2 433 1 26 a2 208 a2 0.08889 for rigid-sphere molecules.

D m DT 1 D m , H T

H1 1 14 a2 2 ,

REFERENCES 1. Maxwell, J.C., “On Stresses in Rarefied Gases Arising from Inequalities of Temperature,” Phil. Trans. Roy. Soc. (London) 170, 231 (1879). 2. Loyalka, S.K., “Approximate Method in the Kinetic Theory,” Phys. Fluids 14(11), 22912294 (1971). 3. Wang-Chang, C.S. and Uhlenbeck, G.E., Transport Phenomena in Very Dilute Gases (VMH-3-F, University of Michigan, 1949). 4. Gross, E.P., Jackson, E.A., and Ziering, S., “Boundary Value Problems in Kinetic Theory of Gases,” Ann. Phys. 1(2), 141-167 (1957). 5. Gross, E.P. and Ziering, S., “Kinetic Theory of Linear Shear Flow,” Phys. Fluids 1(3), 215-224 (1958). 6. Gross, E.P. and Jackson, E.A., “Kinetic Theory of the Impulsive Motion of an Infinite Plane,” Phys. Fluids 1(4), 318-328 (1958). 7. Gross, E.P. and Ziering, S., “Heat Flow Between Parallel Plates,” Phys. Fluids 2(6), 701712 (1959). 8. Bakanov, S.P. and Derjaguin, B.V., “On the State of a Gas Moving Near a Solid Surface,” Dokl. AN SSSR 139(1), 71-74 (1961). 9. Porodnov, B.T. and Suetin, P.E., “Rarefied Gas Flow Between Two Parallel Plates,” Izv. AN SSSR, M. Zh. G. (6), 93-98 (1967). 10. Ivchenko, I.N. and Yalamov, Yu.I., “The Kinetic Theory of a Gas Flow Over a Solid Wall in a Velocity Gradient Field,” Izv. AN SSSR, M. Zh. G. (6), 139-143 (1968). 11. Rolduguin, V.I., Application of the Non-Equilibrium Thermodynamics Method in Boundary Problems of the Kinetic Theory of Gases (M.S. Thesis, Moscow, 1979). 12. Savkov, S.A., Slip Boundary Conditions for Non-Uniform Binary Gas Mixtures and their Application to Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 13. Yalamov, Yu.I. and Ivchenko, I.N., “Isothermal Heat Flux in the Knudsen Layer,” J. Phys. Chem. (Russian) 43(10), 2622-2624 (1969). 14. Derjaguin, B.V., Ivchenko, I.N., and Yalamov, Yu.I., “About Construction of Solutions of the Boltzmann Equation in the Knudsen Layer,” Izv. AN SSSR, M. Zh. G. (4), 167-171 (1968). 15. Derjaguin, B.V. and Yalamov, Yu.I., “The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and Their Experimental Testing,” In Topics in

Chapter 8. Methods of Solution of Planar Problems

179

Current Aerosol Research, vol. 3, part 2, edited by Hidy, G.M. (Pergamon Press, Oxford, 1972). 16. Loyalka, S.K. and Lang, H., “On Variational Principles in the Kinetic Theory,” in Proceedings of the Seventh International Symposium on Rarefied Gas Dynamics (Editrice Tecnico Scientifica, Pisa, Italy, 1970). 17. Loyalka, S.K., “Slip and Jump Coefficients for Rarefied Gas Flows: Variational Results for Lennard-Jones and n(r)-6 Potentials,” Physica A 163, 813-821 (1990).

Chapter 9 THE VARIATIONAL METHOD FOR THE PLANAR GEOMETRY

1.

ANOTHER FORM OF THE BOLTZMANN EQUATION.

In the previous chapter the basic analytical methods for the planar boundary transport problems were considered. All of these schemes are considered to be moment methods. It is very important to note that the simple analysis of some general properties of the Boltzmann equation related to the conservation of moments results in sufficiently accurate expressions for the velocity-slip and temperature-jump coefficients. The Maxwell method and its generalization proposed by Loyalka allow one to analyze some integral properties of a gas, for which one can expect reliable results. The physical basis of these methods is not questionable. But within the framework of these methods, as has already been noted, it is impossible to describe the gas behavior in the Knudsen layer. Nevertheless, Loyalka’s method of taking into account two conserving moments may be used to calculate the gas parameters both at large distances from the wall and directly on its surface. In this chapter a simple and accurate method for dealing with planar boundary transport problems will be described by applying a variational technique to the linearized Boltzmann equation. The analysis will be confined to a consideration of the slip and thermal-creep problems which are the classical problems used to analyze the main features and accuracy of this approach. Since the main interest is in linearized transport problems, the distribution function can be written as:

182

Analytical Methods for Problems of Molecular Transport 0 f ª¬1 < c, r ) r c, x º¼ ,

f

(9-1)

where < c, r is the Chapman-Enskog correction for the appropriate problems and x is the normal coordinate. The correction to the distribution function, ) r c, x , taking into account the influence of the wall, may be specified from the linearized Boltzmann equation that is given by Eq. (8-9). The specific features of the variational method are related to the integral form of Boltzmann’s equation. To derive the integral equation for the distribution function it is convenient to introduce another form of Boltzmann’s equation that, for stationary problems, is defined by:

wf wr

v

1 2 J J ,

(9-2)

where the collision operator is presented as a sum of two separate terms which are given by:

J 1

2 J

³ bdbd H ³ gf cf1cdv1

,

(9-3)

f ³ bdbd H ³ gf1dv1 .

(9-4)

For molecules of infinite range of interaction these integrals diverge, since they include among the colliding molecules some which interact at arbitrarily large distance with infinitely small changes of state [1]. These integrals converge for the rigid-sphere molecules, and, therefore, can be calculated separately. If the interaction potential falls off rapidly enough, distant collisions may be neglected by means of the use of the definite model for a ‘truncated’ potential having a finite interaction range. For planar, stationary, linearized transport problems, Eq. (9-2) may be expressed in the form:

cx

w ) c, x V c ) c, x wx

H) c, x .

(9-5)

Here, V c is a function depending only on the magnitude of a molecular velocity vector and H H c, cc is an operator defined by the relation:

H) c, x H ³ dcc exp cc2 K c, cc ) cc, x ,

(9-6)

183

Chapter 9. The Variational Method for the Planar Geometry

in which K c, cc is a symmetric kernel. The parameter H is defined as:

H

1

1

2S O

(9-7)

.

It is very important to emphasize that both the function, V c , and the kernel, K c, cc , can be defined by analytical expressions if molecules are assumed to be rigid spheres. A very simple integration in Eq. (9-4) gives:

V c HQ c

1

ª º 1·1 § 2 «exp c ¨ 2c c ¸ 2 S erf c » . O 2S ¬ © ¹ ¼

(9-8)

One can find an analytical expression for K c, cc in [2, 3]. This expression for the kernel, K c, cc , is not utilized in the variational analysis and, thus, it is sufficient to know only the general definition for the operator, H .

2.

THE VARIATIONAL TECHNIQUE FOR THE SLIP-FLOW PROBLEM.

In this section the classical slip-flow problem will be analyzed by means of a variational technique that was described by a consideration of the planar boundary transport problems in Loyalka’s works [4-7]. The statement of this problem is given in Sec. 8.1 and hence, here, the main focus will be on investigating the features and the accuracy of the variational technique. Let the distribution function be described by:

fr

^

`

0 cx c yIu c h ) r c, x . f 1 2c y xh

(9-9)

12

where Iu c 2 2kT m B c . The correction, ) r c, x , to the distribution function, which takes into account the perturbation of a gas flow by the wall, has the following asymptotic behavior: lim ) r c, x 2c y u0 ,

x of

(9-10)

where u0 is a dimensionless constant. Then, one uses the Maxwellian boundary model for a description of the gas-surface interaction. This boundary model is given by:

184

Analytical Methods for Problems of Molecular Transport

) c,0 2 DW cx c yIu c h 1 DW ) cx , c y , cz ,0 .

(9-11)

Now, one can use the form of the Boltzmann equation given in Eq. (9-5) to derive the integral equation for ) r c, x . Taking into account Eqs. (910) and (9-11) and performing a formal integration of this linear nonhomogeneous differential equation of the first-order, one obtains the following integral equation:

) c, x L) c, x pu c, x ,

(9-12a)

where L is an integral operator and:

§ V c · pu c, x K cx 2 DW cx c yIu c h exp ¨¨ x ¸¸ , © cx ¹

(9-12b)

1 ; x ! 0 , ¯0 ; x 0 .

K x ®

The operator, L , is defined by the relation:

° 1 x § V c · L ) c, x K cx ® ³ H) c, xc exp ¨¨ xc x ¸¸ dxc © cx ¹ ¯° cx 0 1 DW

f § V c · ½ 1 H ) c, xc exp ¨¨ xc x ¸¸ dxc¾° ³ cx 0 © cx ¹ °¿

(9-13)

f § V c · ½ ° 1 K cx ® ³ H) c, xc exp ¨¨ xc x ¸¸ dxc°¾ , © cx ¹ ¿° ¯° cx x

where H c, cc RH H cx , c y , cz , cc and R is the reflection operator. If the operator, H , is applied to the function, ) c, x , one obtains another function, H ) c, x , that is given by:

H ) c, x H ³ dcc exp cc2 K c, cc ) cc, x , where the following notation is introduced in this integrand:

(9-14)

185

Chapter 9. The Variational Method for the Planar Geometry

K c, cc

RK c, cc .

Since K c, cc is a symmetric function of c, cc one can conclude that:

K c, cc K cc, c .

(9-15)

From both the symmetry of this kernel and Eq. (9-14), it is shown [8] that the operator, H , is self-adjoint, i.e. H H , in the spaces defined by the scalar products:

] 1 c, x , ] 2 c, x ³ dc] 1 exp c2 ] 2

,

(9-16)

f

ª] 1 c, x , ] 2 c, x º ¬ ¼

³ dx ] 1 , ] 2 .

(9-17)

0

Now, in order to construct a variational principle, one must first express u0 as an integral of ) c, x . For this purpose one employs the conservation of moments of the Boltzmann equation as was done in the previous analysis. These properties of the Boltzmann equation can be expressed as:

c c

x y

, ) c, x

c c I

2 x y u

(9-18)

0 ,

c h , ) c,0

c c I

2 x y u

c h , ) c, f

.

(9-19)

Using the asymptotic solution and Eq. (9-19) one can obtain:

u0

^

1 2

cx2 c y u

I c h , cy

(9-20)

`

u 2 DW cx3c 2yIu2 c h 2 , K cx ª¬ H ) c, x , pu c, x º¼ . To obtain an integral relation for the dimensionless velocity, the first-order Chapman-Enskog solution for Iu c is used. This simplifies the algebraic calculations in Eq. (9-20) and results in the following integral relation:

186

Analytical Methods for Problems of Molecular Transport

u0

2 DW 85 O h

8

5S 2 O h ³

exp c 2 dc (9-21)

f

u ³)

c, xc pu c, xc dxc

,

0

where ) c, x H ) c, x . Now, one may construct the functional:

ª¬)

I )

, ) L) 2 pu º ¼

F ) ,) .

As is shown in Appendix D, this functional is stationary for ) ) , where ) is a solution of Eq. (9-12a). If the stationary value of I ) is introduced in Eq. (9-21) one can obtain:

u0

2 DW 85 O h

8 I st ) 5S 2 O h

(9-22)

cm O h .

Consider this functional in detail. It does not involve the real function,

) c, x , and, therefore, the variational principle can be immediately

employed to determine the best value of a given trial function. Since the main purpose is to obtain a solution for the mean velocity, u0 , which is directly related to the stationary value taken by I ) , one may expect that a sufficiently accurate approximation for u0 can be determined by considering even some simple trial functions. One simple trial function that takes into account the asymptotic behavior of ) c, x is:

) c, x D c y ,

(9-23)

) c, x DV c c y ,

(9-24)

where D is a ‘so-called’ variational parameter. Eq. (9-24) can be obtained directly when the trial function, D c y , is substituted into Eq. (9-5). Having inserted this trial function into the functional and performed a very simple integration, one can obtain the following expression:

I D

1 4

S ª¬D 2DW D 2 DW 54 SO h º¼ .

(9-25)

187

Chapter 9. The Variational Method for the Planar Geometry

The stationary value for I D occurs if dI dD

D

2 DW

DW

5 8

0 , so that:

SO h .

(9-26)

From Eq. (9-22) one can obtain the expression for the slip coefficient that may be written as:

cm

2 DW § 4 S · 5 1 DW ¸ 16 S . DW ¨© 2S ¹

(9-27)

It is very important to note that this formula is exactly the same as the expression derived by Loyalka’s method. Now, consider how one might describe the gas flow in the Knudsen layer. It should be emphasized that the variational method, even employed in its simplest form, permits one to perform this analysis. The main interest here is to investigate the velocity defect expression that is derived by taking a moment of the distribution function. The correction to the distribution function, ) c, x , obtained from Eq. (9-12) has the following form for a given trial function:

§ V c · x ¸¸ © cx ¹

°

) c, x K cx ® 2 DW cx c yIu c h exp ¨¨ °¯

ª § V c · º °½ x ¸¸ » ¾ K cx D c y . D c y «1 DW exp ¨¨ © cx ¹ ¼» °¿ ¬«

^

(9-28)

`

To obtain the right asymptotic form for this correction, insert D 2u0 into Eq. (9-28) instead of the value of D given by Eq. (9-26). This gives:

§ V c · x ¸¸ © cx ¹

) c, x 2u0 c y K cx exp ¨¨ u ª¬ 2 DW cx c y

5 4

S Oh

(9-29)

2u0 c y W

D º¼ .

Allowing for Eq. (9-29), one can obtain the following expressions for both the mean velocity and the velocity defect of a gas:

188

Analytical Methods for Problems of Molecular Transport

2c x 5 º § wu · ª u x ¨ ¸ O «cm DW 3m2 J1 x 2 DW J 2 x » , O 4S S © wx ¹f ¬ ¼

(9-30)

º § wu · ª 2c ud x ¨ ¸ O «DW 3m2 J1 x 2 DW 54 S 1J 2 x » . © wx ¹f ¬ S ¼

(9-31)

Here, the functions, J1 x and J 2 x , are defined by: f ª § V c · § V c ·º J1 x S ³ c 4 exp c 2 « E2 ¨¨ x ¸¸ E4 ¨¨ x ¸¸ » dc , ¹ © c ¹ ¼» 0 ¬« © c

(9-32)

f ª § V c · § V c ·º J 2 x S ³ c5 exp c 2 « E3 ¨¨ x ¸¸ E5 ¨¨ x ¸¸ » dc , c c « ¹ © ¹ »¼ 0 ¬ ©

(9-33)

where the functions, En z , are the exponential integrals that can be expressed in the form: 1

En z

³t 0

n2

§ z· exp ¨ ¸ dt . © t¹

(9-34)

As can be easily seen, the variational method allows one to investigate the Knudsen layer. The accuracies of this method and the other methods described previously will be considered in the next section.

3.

DISCUSSION OF THE SLIP-FLOW RESULTS.

There are several expressions for the mean velocity and slip coefficient that have been obtained here by employing various methods and using various boundary models. To establish the accuracy of these methods it is necessary to compare the approximate analytical results with direct numerical results. For this purpose, we have found it convenient to use the numerical results of Loyalka and Hickey [9]. The analytical results that we are comparing are those that take into account the second-order corrections to the Chapman-Enskog distribution functions and, moreover, the molecules are assumed in the analysis to be rigid spheres.

Chapter 9. The Variational Method for the Planar Geometry

189

It is easy to see that both Loyalka’s method and the variational method give the same results for the slip coefficient if the trial function given by Eq. (9-23) is employed in the variational analysis. These methods result in the following relation:

cm

2 DW

DW

5 16

S ª¬1 0.1086 DW º¼ .

(9-35)

The velocity defect cannot be obtained by Loyalka’s method with the exception of one point for which x 0 . Since the expression for aw given by Eq. (8-34) is not altered, if the second-order Chapman-Enskog solution is used, one can obtain the velocity defect at the wall given by:

ud 0

5 32

ª 4Z1

S 2 DW «

¬ S

2

1 S

1 2 b 2 2

º»¼

0.2759 2 DW

. (9-36)

2

where Z1 1 b2 174 b2 . The half-range moment method (four-moment approach) gives different results for various boundary models. The boundary model proposed by Ivchenko, Loyalka, and Tompson [10] (the ILT boundary model) results in the following expressions for cm and ud x : 2

cm

2 DW

DW

ud x

5 8

2

ª 0.6690 0.1775 DW º » , ¬« 0.7549 0.09714 DW ¼»

(9-37)

0.3413 0.1706 DW 0.7549 0.09714 DW

(9-38)

S«

1.4229

x · § exp ¨ 2.2015 ¸ . O¹ ©

The same method for the Maxwellian boundary model gives:

cm

2 DW

ud x

DW

5 8

º 3.1929 DW » «¬ 3.6028 0.5241 DW »¼

ª

S«

1.4229

,

2 DW x · § exp ¨ 2.2015 ¸ . O¹ 3.6028 0.5241 DW ©

(9-39)

(9-40)

190

Analytical Methods for Problems of Molecular Transport

Table 9-1. The slip-flow coefficient,

Four-Moment Approach ILT Maxwellian Boundary Model [10] Boundary Model [10] cm G cm G †

Numerical Values [9]

†

DW

cm

0.2 0.4 0.6 0.8 1.0

9.0458 4.1147 2.4531 1.6096 1.0940

G%

a anum

cm .

9.0710 4.1310 2.4650 1.6185 1.1006

DW

cm

0.2 0.4 0.6 0.8 1.0

9.0458 4.1147 2.4531 1.6096 1.0940

a anum

0.86 1.49 2.02 2.48 2.87

cm .

Numerical Values [9]

G%

9.1235 4.1759 2.5027 1.6496 1.1254

anum u 100 ; a cm

Table 9-2. The slip-flow coefficient,

†

0.28 0.40 0.49 0.55 0.60

Loyalka Method & Variational Results, Eq. (9-35) G † cm 9.0277 4.0976 2.4400 1.6006 1.0884

0.20 0.42 0.53 0.56 0.51

anum u 100 ; a cm

To compare the accuracy of the various expressions for the slip coefficient the analytical and numerical results are presented together in Tables 9-1 and 9-2. It can be seen that a very good agreement exists between analytical and numerical values of cm for the ILT boundary model. The discrepancy of these results is less than 0.6% for all DW . The maximum deviation of the other analytical data amounts to 0.56% and 2.87% for the variational method and the Maxwellian boundary model, respectively, and, therefore, Eqs. (9-35) and (9-37) are the most accurate analytical formulas for the slip coefficient. Another macroscopic parameter which has a significant influence on the description of the gas flow in the Knudsen layer is the velocity defect at the wall. The values of this parameter are presented in Tables 9-3 and 9.4. For this parameter the best agreement with numerical values in the full region of DW is obtained from the ILT boundary model. The maximum difference for this model, for all DW , is about 5.5%, while the discrepancies associated with the Loyalka and variational methods are 8.4% and 18.5% respectively for DW 1.0 , and range up to 16.9% and 42.2% respectively for DW 0.2 . The numerical and analytical data for the dimensionless velocity defect are presented in Table 9-5.

191

Chapter 9. The Variational Method for the Planar Geometry

Table 9-3. The velocity defect at the wall, Numerical Values [9]

†

Four-Moment Approach ILT Maxwellian Boundary Model [10] Boundary Model [10]

DW

ud 0

G †

ud 0

G

ud 0

0.2 0.4 0.6 0.8 1.0

0.5976 0.5180 0.4422 0.3701 0.3013

0.5645 0.4895 0.4181 0.3500 0.2851

5.5 5.5 5.5 5.4 5.4

0.6908 0.5972 0.5085 0.4245 0.3448

15.6 15.3 15.0 14.7 14.5

G%

a anum

anum u 100 ; a ud 0

Table 9-4. The velocity defect at the wall, Numerical Values [9]

†

ud 0 .

ud 0 .

Loyalka Method Eq. (9-36)

Variational Method Eq. (9-31)

DW

ud 0

G †

ud 0

G

ud 0

0.2 0.4 0.6 0.8 1.0

0.5976 0.5180 0.4422 0.3701 0.3013

0.4967 0.4414 0.3863 0.3311 0.2759

16.9 14.8 12.6 10.5 8.4

0.3452 0.3283 0.3061 0.2784 0.2455

42.2 36.6 30.8 24.8 18.5

G%

a anum

anum u 100 ; a ud 0

An analysis of Table 9-5 shows that for the ILT boundary model a satisfactory agreement with the numerical results occurs if 0 xc 0.5 . The maximum deviation from the numerical values for this domain of xc is about 11.5%. When xc ! 0.5 the velocity defect decreases more rapidly than the corresponding numerical value. In this region the discrepancy of the results is very large. This occurs because the asymptotic velocity defect behavior for this approach is described by one exponential term which does not characterize a real dependence on the normal coordinate [11]. This dependence is more complicated than a simple exponential drop, as one can see from Eq. (9-12). Moreover, when xc ! 0.5 , the velocity defect values become extremely small so that the moment approach employed does not have a sufficient accuracy to describe this very subtle quantity. The variational method does not accurately describe the velocity defect because a very simple trial function has been used. The results for the mean velocity profile are shown in Table 9-6. For the ILT model the maximum deviation is 2.92%. For the Maxwellian boundary model this difference is 3.77%.

192

Analytical Methods for Problems of Molecular Transport

Table 9-5. The dimensionless velocity defect,

ud xc , ( DW

Dimensionless Normal Coordinate

Numerical Values [9]

Four-Moment Approach ILT Maxwellian Boundary Model [10] Boundary Model [10]

xc

ud xc

ud xc

ud xc

0.3013 0.1997 0.1560 0.0888 0.0430 0.0233 0.0133

0.2851 0.2226 0.1738 0.0827 0.0240 0.0070 0.0020

0.3448 0.2692 0.2102 0.1000 0.0290 0.0084 0.0024

†

0.0 0.1 0.2 0.5 1.0 1.5 2.0 †

5 8

x

1.12423 O xc

S OP xc

Table 9-6. The mean velocity,

†

0.0 0.1 0.2 0.5 1.0 1.5 2.0 †

x

#

G%

5 8

4.

S OP xc

u xc , ( DW

1 ).

Four-Moment Approach ILT Maxwellian Boundary Model [10] Boundary Model [10]

Numerical Values [9]

xc

1 ).

u xc

u xc

G #

u xc

G

0.7924 1.0049 1.1594 1.5589 2.1586 2.7322 3.2961

0.8155 0.9888 1.1484 1.5718 2.1844 2.7553 3.3142

2.92 1.60 0.95 0.83 2.58 2.31 0.55

0.7806 0.9670 1.1368 1.5793 2.2042 2.7787 3.3386

1.49 3.77 1.95 1.31 2.11 1.70 1.29

a anum

1.12423 O xc

anum u 100 ; a u xc

THE VARIATIONAL SOLUTION FOR THE THERMAL-CREEP PROBLEM.

The thermal-creep problem has been stated in detail in the previous chapter, and therefore, only a brief outline will be given here. Let the distribution function be described by:

f

^

`

0 y c yIT c q y ) r c, x . f 1 c 2 52 yq

Here, IT c A c and may be written as:

(9-41a)

Chapter 9. The Variational Method for the Planar Geometry

IT c A c

15 16

¼

S ON ª S312 c 2 a2 2 S3 22 c 2 º , ¬

193 (9-41b)

45 where ON 44 O and a2 0.08889 . The correction, ) r c,0 , to the distribution function can be specified from Eq. (9-5) and its boundary conditions which can be expressed as: 2

) c,0 DW c yIT c q y 1 DW ) cx , c y , cz ,0 ,

(9-42)

lim ) r c, x 2c y u0 .

(9-43)

x of

As it is shown in Section 9.2, one can obtain the following integral equation for ) r c,0 :

) c, x L) c, x pT c, x .

(9-44)

Here the integral operator, L) c, x , is given by Eq. (9-13) and pT c, x is defined by the relation:

ª § V c ·º pT c, x K cx «DW c yIT c q y exp ¨¨ x ¸¸ » . «¬ © cx ¹ »¼

(9-45)

This function represents the external non-uniformity of a gas in Eq. (9-44). Now, taking into account the conservation of moments of the Boltzmann equation given by Eqs. (9-18) and (9-19) one can derive the following integral equation for the mean velocity:

u0

DW q y h cx2 c 2yIu c IT c , K cx I

2h c y , cx2 c yIu c

,

(9-46)

where I ª¬ H ) c, x , pu c, x º¼ and pu c, x is the known function given by Eq. (9-12a) that is proportional to the non-uniformity of a gas for the slip-flow problem. The standard notations given by Eqs. (9-16), (9-17), and (9-14) are used in this expression. The functional contained in Eq. (9-46) is a linear form with respect to the trial function ) c, x and, therefore, it cannot be immediately used for solution of a variational problem. A new functional must be constructed on

194

Analytical Methods for Problems of Molecular Transport

the basis of the square form of the trial functions and which has a stationary value equal to I . Let this new functional be the sum of two functionals, the first of which is equal to [() , pu (c, x ))] . Since pu is proportional to O h and ) ~ V (c)O q y ~ q y , the integrand of the second functional must be proportional to O q y h , and moreover this term is equal to zero if the trial function coincides with the real correction, ) c, x . This indicates that the integrand of the second functional has to contain the factor ) L) pT , which is proportional to O q y . Only the unique square form for this integrand, which is the product of ) u , where ) u is a trial function for the slip-flow problem, and ) L) pT , can be constructed so that it should be proportional to V c O 2 hq y . This very simple dimensional analysis shows that the following form can be used for this variational problem:

F ) u ,)

ª ) , pu º ª ) u , ) L) pT º . ¬ ¼ ¬ ¼

(9-47)

The stationary value of this functional is equal to I . Now, the asymptotic solutions of Eqs. (9-12) and (9-44) can be chosen as the trial functions for F ) u ,) which can be expressed in the form:

) D1c y

, )

V c D1c y ,

(9-48a)

) u D 2 c y , ) u V c D 2 c y .

(9-48b)

It is very important to note that different constants, D1 and D 2 , are used for these trial functions as they are asymptotic solutions of different equations. The first-order Chapman-Enskog solutions for Iu c and IT c are assumed to be employed in a later analysis and one can use unity instead of Iu c as it is easy to see from Eq. (9-46). In accordance with this suggestion one can obtain the following relationship:

F D1 , D 2

(9-49)

A1D1 A2D1D 2 A3D 2 .

Here, the following notations are introduced: A1

2 DW 18 S 3 2

, A2

DW 14 S , A3

DW

15 128

S 3 2O q y .

(9-50)

The desired values of D1 and D 2 for which this functional is stationary can be defined from the following equations:

Chapter 9. The Variational Method for the Planar Geometry

wF wD1

wF wD 2

0 ,

195

0 .

The stationary value of the functional can be expressed in the form:

Fst ) u ,)

15 S 2O q y 2 DW 254

.

(9-51)

Using Eq. (9-51) in Eq. (9-46), one can obtain:

u0

(9-52)

cTslQ q y .

The thermal-creep coefficient in this formula is given by:

cTsl

3 4

1 12 DW .

(9-53)

This expression coincides exactly with the one obtained by Loyalka’s method. Substituting the asymptotic value of the mean velocity 2u0 c y instead of D1c y into an expression of the distribution function, one can obtain the relation for the velocity defect defined by:

ud x

ud x

2cTsl

DW ®

Q qy

¯S

32

J1 x

3

S

32

½ J 3 x ¾ . ¿

(9-54)

Here, J1 x is given by Eq. (9-32) and J 3 x is defined by: f

J 3 x S ³ c 4 0

5 2

c 2 exp c 2

ª § V c · § V c ·º x ¸¸ E4 ¨¨ x ¸¸ » dc , u « E2 ¨¨ ¹ © c ¹ ¼» ¬« © c

(9-55)

where the notations in Eq. (9-55) are the same as in Eqs. (9-32) and (9-33).

196

Analytical Methods for Problems of Molecular Transport

5.

DISCUSSION OF THE THERMAL-CREEP RESULTS.

The accuracy of the thermal-creep problem results obtained by various approximate methods will be discussed here. This analysis can be performed by means of a comparison of analytical values with the numerical results obtained by Loyalka [12]. First, the accuracy of the various expressions of the thermal-creep coefficient will be examined. All of the analytical expressions for this coefficient and the velocity defect will be given here in the most convenient form for this analysis. Loyalka’s method and the variational solution give the same results for the thermal-creep coefficient. If the second-order Chapman-Enskog solution is used, these methods both yield:

cTsl

3 4

ªH1 H 2 12 H1 DW º , ¬ ¼

(9-56)

The variational where H1 1 14 a2 and H 2 1 72 b2 72 a2 b2 . velocity defect obtained by means of the simplest trial functions is given by Eq. (9-54). The velocity defect at the wall obtained by Loyalka’s method can be written in the form: 2

ud 0

3 4

2

2

2

(9-57)

DW H 2 .

In the previous chapter the various solutions of the thermal-creep problem were obtained within the framework of the half-range moment method for the different boundary models. The boundary model proposed by Ivchenko, Loyalka, and Tompson [13, 14] results in the following expressions for cTsl and ud x : cTsl

ud x

3 2

0.4354 0.2179 DW 0.8518 0.1096 DW

(9-58)

,

0.5822 DW x · § exp ¨ 2.2015 ¸ O¹ 0.8518 0.1096 DW ©

.

(9-59)

For the integral boundary model that is usually employed in transport problems, these relations can be written as:

197

Chapter 9. The Variational Method for the Planar Geometry

Table 9-7. The thermal-creep coefficient,

Numerical Values [12] #

† #

DW

cTsl

0.2 0.4 0.6 0.8 1.0

0.8234 0.8733 0.9207 0.9657 1.0089

a anum

G%

cTsl .

Four-Moment Approach ILT Integral Boundary Model [13] Boundary Model [13] Eq. (9-58) Eq. (9-60) G %† G% cTsl cTsl 0.8223 0.8752 0.9255 0.9735 1.0193

0.1 0.2 0.5 0.8 1.0

anum u 100 ; a cTsl

cTsl . Loyalka Method & Variational Results, Eq. (9-56) G %† cTsl

Numerical Values [12] #

#

1.7 5.8 9.1 11.8 14.2

These additional numerical values have been obtained by Loyalka.

Table 9-8. The thermal-creep coefficient,

†

0.8395 0.9240 1.0041 1.0799 1.1519

DW

cTsl

0.2 0.4 0.6 0.8 1.0

0.8234 0.8733 0.9207 0.9657 1.0089

G%

a anum

0.8120 0.8573 0.9027 0.9480 0.9933

1.4 1.8 2.0 1.8 1.5

anum u 100 ; a cTsl

These additional numerical values have been obtained by Loyalka.

cTsl

ud x

3 2

1.5464 1.1741 DW 3.0927 0.4499 DW

,

3.4149 DW x · § exp ¨ 2.2015 ¸ O¹ 3.0927 0.4499 DW ©

(9-60)

.

(9-61)

Now, consider the accuracy of the various analytical results for this problem. The numerical and analytical data for the thermal-creep coefficient versus the accommodation coefficient, DW , are presented in Tables 9-7 and 9-8. It is easy to see that the most accurate values of cTsl are predicted by the moment method for the ILT boundary model. In this case, the maximum discrepancy between the numerical and analytical results is about 1.0%, while Loyalka’s method and the variational method give results that differ from the numerical results by 1.4% for DW 0.2 up to 2.0% when DW 0.6 .

198

Analytical Methods for Problems of Molecular Transport

Table 9-9. The velocity defect at the wall,

Numerical Values [12] #

† #

ud 0 .

Four-Moment Approach ILT Integral Boundary Model [13] Boundary Model [13] Eq. (9-59) Eq. (9-61)

DW

ud 0

ud 0

G %†

ud 0

G%

0.2 0.4 0.6 0.8 1.0

0.1596 0.3078 0.4458 0.5748 0.6960

0.1333 0.2600 0.3807 0.4958 0.6056

16.5 15.5 14.6 13.7 13.0

0.2146 0.4174 0.6093 0.7913 0.9640

34.5 35.6 36.7 37.7 38.5

G%

a anum

anum u 100 ; a ud 0

These additional numerical values have been obtained by Loyalka.

Table 9-10. The velocity defect at the wall, Numerical Values [12] #

† #

ud 0 .

Loyalka Method Eq. (9-57)

Variational Method Eq. (9-54)

DW

ud 0

ud 0

G %†

ud 0

G%

0.2 0.4 0.6 0.8 1.0

0.1596 0.3078 0.4458 0.5748 0.6960

0.1220 0.2440 0.3660 0.4880 0.6100

23.6 20.7 17.9 15.1 12.8

0.0825 0.1800 0.2925 0.4200 0.5625

48.3 41.5 34.4 26.9 19.2

G%

a anum

anum u 100 ; a ud 0

These additional numerical values have been obtained by Loyalka.

The integral boundary model values deviate greatly from the numerical results (from 1.7% when DW 0.2 up to 14.2% when DW 1.0 ). In Tables 9-9 and 9-10 values of the velocity defect at the wall are presented. A comparison of these velocity defect values shows that the discrepancy between the analytical and numerical results is noteworthy for all of the analytical methods discussed. Loyalka’s method and the ILT boundary model have the best agreement with the numerical values. The maximum deviations of these values are 23.6% and 16.5%, respectively, when DW 0.2 and are seen to decrease to only 12.8% and 13.0%, respectively, when DW 1.0 . Values of the velocity defect versus the normal coordinate for DW 1.0 are presented in Table 9-11. It can be seen that the relative deviations of all of the analytical values from the corresponding numerical values are very large for xc t 0.5 . The velocity defect behavior inside the Knudsen layer seems to be more complicated [15] than the behavior predicted by these analytical formulae and, moreover, the accuracy of these methods appears insufficient overall to describe this extremely subtle quantity.

199

Chapter 9. The Variational Method for the Planar Geometry

Table 9-11. The dimensionless velocity defect, Numerical Values [12]

xc

ud xc

ud xc

ud xc

0.6960 0.3840 0.2539 0.1747 0.1229 0.0632 0.0349

0.6056 0.3052 0.0933 0.0366 0.0143 0.0022 0.0003

0.9640 0.3783 0.1485 0.0583 0.0229 0.0035 0.0005

†

x

15 16

S ON xc

1.69944 O xc

Table 9-12. Numerical values of the velocity defect,

ud xc .

Velocity Defect,

xc

†

1 ).

Four-Moment Approach ILT Boundary Integral Boundary. Model [13] Model [13] Eq. (9-59) Eq. (9-61)

Dimensionless Normal Coordinate 0.00 0.25 0.50 0.75 1.00 1.50 2.00 †

ud xc , ( DW

x

†

DW =0.2

DW =0.4

DW =0.6

0.00 0.25 0.50 0.75 1.00 1.50 2.00

0.1596 0.0840 0.05499 0.03768 0.02645 0.01355 0.007017

0.3078 0.1639 0.1076 0.0738 0.05184 0.02659 0.01404

0.4458 0.2400 0.1579 0.1085 0.07629 0.03915 0.02069

15 16

S ON xc

ud xc

DW =0.8 0.5748 0.3126 0.2062 0.1419 0.09984 0.05127 0.02712

DW =1.0 0.6960 0.3840 0.2539 0.1747 0.1229 0.0632 0.0349

1.69944 O xc

The results for the mean velocity profile are shown in Fig. 9-1. For the ILT model the maximum deviation of the mean velocity is approximately 13% which occurs when xc 0 , while for the integral model this discrepancy is 40%. It is clearly necessary to use even more exact approaches if one is to succeed in obtaining a good overall agreement between the analytical and numerical results for the mean velocity profile. In Table 9-12 the numerical results for the velocity defect are presented. These data have been obtained by Loyalka by the use of the numerical technique reported in [12]. The accuracy of any approximate method can be examined by using these data.

200

Analytical Methods for Problems of Molecular Transport

1.2

Dimensionless Mean Velocity

1.0

0.8

0.6

0.4

0.2

0.0 0.0

0.5

1.0

1.5

2.0

Dimensionless Normal Coordinate

Figure 9-1. The dimensionless, mean velocity profile, u ( xc) u ( xc)/Q q as a function of the dimensionless normal coordinate, xc , where x (1.69944)O xc . Numerical values [12] are shown as black circles, ILT Boundary Model results [13] are shown as gray diamonds, and Integral Boundary Model results [13] are shown as open squares. Curves are shown for continuity only in order to help the reader visualize the shape of the profile.

6.

SLIP-FLOW AND TEMPERATURE-JUMP COEFFICIENTS FOR THE LENNARD-JONES (612) POTENTIAL MODEL.

In the framework of the variational and generalized Maxwellian methods all the slip and temperature-jump coefficients depend upon a potential model if one uses the second- and higher-order Chapman-Enskog approximations to describe the gas state beyond the wall. For the second-order ChapmanEnskog approximation both these methods yield:

201

Chapter 9. The Variational Method for the Planar Geometry

Slip Coefficient Values (dimensionless)

1.14

1.12

1.1

cm

1.08

1.06

1.04

c Tsl 1.02 0.1

1

10

100

1000

Reduced Temperature (dimensionless)

Figure 9-2. The dependence of the slip coefficients, cm and cTsl , on the reduced

temperature, T , for the Lennard-Jones (6-12) potential model )(assuming that DW 1 ).

cm

cTsl

cT

2 DW

DW

3 4

5 16

§

S ¨1 ©

4Z1 S DW 2S

· ¸, ¹

ªH1 H 2 12 H1 DW º , ¬ ¼

2 DW DT

DW DT

75 128

ª

§

¬

©

S «1 ¨ 52 25

H T 1 · DW DT 2 DW º » S 2 ¸¹ 2 DW DT ¼

(9-62)

(9-63)

(9-64)

where the following notations have been introduced:

Z1 1 b2 2 174 b2 2

2

,

(9-65)

202

Analytical Methods for Problems of Molecular Transport

Temperature Jump Coefficient (dimensionless)

2.15

2.10

cT

2.05 0.1

1

10

100

1000

Reduced Temperature (dimensionless)

Figure 9-3. The dependence of the temperature-jump coefficient, cT , on the reduced

temperature, T , for the Lennard-Jones (6-12) potential model (assuming that DW DT 1 ).

H1 1 14 a2 2 ,

(9-66)

H 2 1 72 b2 2 72 a2 2 b2 2 ,

(9-67)

2 23 2 433 H T 1 26 a2 208 a2

2

.

(9-68)

The slip and jump coefficients depend upon the reduced temperature (see 2 2 Section 5.9), T * , through their dependence on a2 and b2 . Using the data given in Tab. 5-1, one can calculate the appropriate values of cm , cTsl , and cT for the Lennard-Jones (6-12) potential model.

203

Chapter 9. The Variational Method for the Planar Geometry

The variation of cm , cTsl , and cT with the reduced temperature, T * , is illustrated in Figs. 9-2 and 9-3. All the curves have a maximum when T * is about unity. The relative deviations of the maximum values from those, when T * is large, are about 10%, 3% and 2% for cTsl , cT , and cm , respectively. This analysis shows that the thermal-creep coefficient which is dependent on the cross effects (viscosity and thermal conductivity) is quite sensitive to the parameters of the Lennard-Jones (6-12) potential. Now, examine dependence of the slip and jump coefficients on the intermolecular potential model. The analysis is restricted to consideration of the rigid-sphere and Lennard-Jones (6-12) models. For rigid-sphere molecules the potential well depth, H kT T , is equal to zero and, therefore, all results depending on intermolecular interactions are very close to those for the Lennard-Jones (6-12) model in which T o f . This shows that only the thermal-creep coefficient is fairly sensitive to the potential model. The maximum relative discrepancy between cTsl for these two models is about 12% if DW 1 and T is about unity. It is very important to note that cTsl is independent of the temperature, T , for the rigid-sphere model.

PROBLEMS 9.1. Determine the slip and thermal-creep coefficients using the Maxwellian method for the second-order Chapman-Enskog solution. Solution: The distribution function may be written as:

f

^

`

f eq 1 cx c yIu c h c yIT c q y ) r c, x ,

where ) c,0 ) r c, f a0 c y and the functions Iu c and IT c are given by Eqs. (8-75) and (9-41b), respectively. The constant, a0 , may be found from the relation cx c y , K cx A) c,0 K cx ) c,0 0 . Having performed some simple integrations, one can obtain:

cm

2 DW

DW

5 16

S ,

and cTsl 34 H1 , where H1 1 14 a2 1.02222 . Thus, for a Maxwellian analysis, the usage of the second-order Chapman-Enskog solution results in an increase of the thermal-creep coefficient by 2.2% while the slip-flow coefficient is not altered. 2

204

Analytical Methods for Problems of Molecular Transport

9.2. Determine the temperature-jump at the wall using the Maxwellian method for the second-order Chapman-Enskog solution. Consider the case when DW 1 . Solution: The distribution function is given by:

f

^

`

0 x c 2 52 cxIT c qx ) r c, x , f 1 xq

where IT c

and:

S ON S312 c 2 a2 2 S3 22 c 2

15 16

) r c, f ) c,0 a1 c 2 52 . The temperature-jump may be calculated by employing the general moment solution of the Boltzmann equation given by cx c 2 , ) r c,0 0 . Since:

2 2 k 2 2 ³ cx c S3 2 c exp c dc

0 ;

kt2 ,

the temperature-jump is not altered for arbitrary orders of the ChapmanEnskog solution and may be expressed in the form:

'T

2 DT

DT

75 128

§ wT · ¸ © wx ¹f

SON ¨

;

ON

45 44

O .

9.3. Determine the temperature-jump coefficient using the Maxwellian method and the general form of the boundary conditions given by Eqs. (612) and (6-15). Use the second-order Chapman-Enskog solution. Solution: Following the solution of Problem 9.2, one obtains:

cT

2 DW DT

DW DT

75 128

S .

9.4. Derive expressions for the velocity defect at the wall for both the slipflow and thermal-creep problems by using the Maxwellian method. Solution: The velocity defect at the wall for these problems is given by:

ud 0 OP wu wx f

cm

u 0

OP wu wx f

,

205

Chapter 9. The Variational Method for the Planar Geometry

and:

ud 0

Q qy

cTsl

u 0

Q qy

; Q Q

ON . OP

Using Maxwellian boundary conditions to calculate the mean velocity at the wall, u 0 , one can obtain for the slip and thermal-creep problems:

ud 0 OP wu wx f

5 32

ª ¬

S 2 DW «1

2

S

1

1 2 b 2 2

º»¼ ,

and:

ud 0

Q qy

3 8

DW H1 ,

respectively, where:

H1

1

1 4

2 a2

9.5. Determine an expression for the temperature defect at the wall by using the Maxwellian method for the first- and second-order Chapman-Enskog solutions. Obtain the result for DW 1 and compare with the values from [12]. Solution: The temperature defect at the wall is given by:

§ wT · Td 0 Tasy 0 T 0 cT ON ¨ ¸ T0W 0 , © wx ¹f

where W 0 23 S 3 2 ³ c 2 32 ) r c,0 exp c 2 dc . Having performed the necessary integrations and algebraic transformations, one obtains:

Td 0

Td 0 ON wT wx f

75 256

¼

2 28 1 a º» . S 2 DW ª«1 15 S 1 13 28 2

¬

If DW 1 in this expression, then Td 0 0.3509 , while the value reported in [12] is Td [12] 0 0.6374 . The temperature defect for the first-order

206

Analytical Methods for Problems of Molecular Transport

Chapman-Enskog approximation may be formally obtained from the general 2 expression if a2 0 . This then results in Td 0 0.3735 which is an improvement in agreement with the value reported in [12]. 9.6. Derive a formula for the temperature-jump coefficient by Loyalka’s method for the second-order Chapman-Enskog solution and using the general form of the boundary conditions given by Eqs. (6-12) and (6-15). Solution: The correction to the distribution function is given by:

) r c, f a1 c 2 52 , and ) c,0 aw c 2 52 , where a1 z aw . The two constants, a1 and aw , may be found from the two conservation laws given by Eqs. (8-42) and (8-43). Boundary conditions are described by Eqs. (6-12) and (6-15). Integration yields the following expression:

cT

75 128

where H T

S

2 DW DT ª DW DT 2 DW «1 DW DT ¬ 2 DW DT

23 2 1 26 a2

433 208

a 2 2

2

52 25

º

S 1H T 12 » ,

0.9378

¼

for rigid-sphere molecules.

9.7. Generalize Loyalka’s expression for the temperature defect at the wall using two accommodation coefficients and the second-order ChapmanEnskog solution. Compare your analytical results with the numerical results [12] for the first- and second-order Chapman-Enskog approximations when DW 1 . Solution: Employing the same techniques as in Problem 9.5, one can obtain:

Td 0

75 256

¼

2 28 1 S 2 DW ª« 104 S 1H T 15 S 1 13 a º» , 25 28 2

¬

where H T 0.9378 . If DW 1 , then the values obtained for the first- and second-order Chapman-Enskog approximations are Td 0 0.6719 and Td 0 0.5734 , respectively. These values represent deviations from the numerical value reported in [12] of 5.4% and 10%, respectively.

207

Chapter 9. The Variational Method for the Planar Geometry Table 9-13. Velocity defect data obtained during the solution of Problem 9.8.

DW

1

0.8

0.6

0.4

0.2

ud >[email protected] 0

0.3013

0.3701

0.4422

0.5180

0.5976

0.2850

0.3420

0.3990

0.4560

0.5130

5.4

7.6

9.8

12.0

14.2

ud

0

deviation, %

9.8. Using the integral representation of the distribution function given by Eq. (9-12a), derive an expression for the velocity defect at the wall for the slip-flow problem if a trial function containing two independent constants is chosen in the form:

) K cx ª¬D c y º¼ K cx ª¬D c y E c y exp V x cx º¼ . For different values of DW , compare these results with those obtained in [9]. Use the second-order Chapman-Enskog solution. Solution: To obtain an approximate solution for the distribution function one can employ the approximate method described in [16,17]. Having substituted the trial function into Eq. (9-12a), one can obtain:

) c, x K cx ª¬ 2 DW cx c yIu c h exp V x cx D c y 1 DW exp V x cx 1 DW E c y exp V x cx º¼ K cx ª¬D c y E c y exp V x cx º¼ . The velocity defect at the wall may then be written as:

ud 0 OP wu wx f

5 8

2 DW Z1 12

,

where:

Z1 1 b2 2 174 b2 2

2

0.95602

.

For various values of DW , the velocity defect data are shown in Table 9-13.

208

Analytical Methods for Problems of Molecular Transport

9.9. For Problem 9.8, derive an expression for the velocity defect. Solution: The mean velocity is given by: 12

§ 2kT · 3 2 u x ¨ c y) c, x exp c 2 dc . ¸ S ³ m © ¹

After integration of the distribution function presented in Problem 9.8, one can obtain:

ud x OP wu wx f

5 4

ª J1 x

2 DW «

¬S

32

2Z1

J 2 x º » , S ¼

where Z1 0.95602 and the functions, J1 x and J 2 x , are given by Eqs. (9-32) and (9-33), respectively. 9.10. Using the same trial function as in Problem 9.8, derive an expression for the velocity defect for the thermal-creep problem. In particular, determine the velocity defect at the wall. Solution: Integration of the distribution function given in Problem 9.8 results in:

ud x

ud x

Q q

ª J x J x º 3DW « 3 3 2 H 2 1 3 2 » , S ¼ ¬S

where the functions, J1 x and J 3 x , are given by Eqs. (9-32) and (9-55), respectively. The velocity defect at the wall may then be written as 2 2 2 This ud 0 34 DW H 2 where H 2 1 72 b2 72 a2 b2 0.81332 . expression is exactly the same as that obtained via Loyalka’s method.

REFERENCES 1. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, NY, 1969). 2. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “On the Collision Kernels for Gas Mixtures,” Ann. Nucl. Energy 23(18), 1489-1495 (1996). 3. Loyalka, S.K. and Hickey, K.A., “Plane Poiseuille Flow: Near Continuum Results for a Rigid Sphere Gas,” Physica A160, 395-408 (1989). 4. Loyalka, S.K., “Momentum and Temperature-Slip Coefficients with Arbitrary Accommodation at the Surface,” J. Chem. Phys. 48(12), 5432-5436 (1968). 5. Loyalka, S.K. and Ferziger, J.H., “Model Dependence of the Temperature Slip Coefficient,” Phys. Fluids 11(8), 1668-1671 (1968).

Chapter 9. The Variational Method for the Planar Geometry

209

6. Loyalka, S.K., “Slip in the Thermal Creep Flow,” Phys. Fluids 14(1), 21-24 (1971). 7. Loyalka, S.K., “The Slip Problems for a Simple Gas,” Z. Naturforsch. 26a, 964-972 (1971). 8. Morse, P.M. and Feshbach, H., Methods of Theoretical Physics (McGraw-Hill, NY, 1953). 9. Loyalka, S.K. and Hickey, K.A., “The Kramers Problem: Velocity Slip and Defect for a Hard Sphere Gas with Arbitrary Accommodation,” ZAMP 41, 246-253 (1990). 10. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “The Precision of Boundary Models in the Gas Slip Problem,” High Temperature 31(1), 127-129 (1993). 11. Loyalka, S.K., “Velocity Profile in the Knudsen Layer for the Kramer’s Problem,” Phys. Fluids 18(12), 1666-1669 (1975). 12. Loyalka, S.K., “Temperature Jump and Thermal Creep Slip: Rigid Sphere Gas,” Phys. Fluids A 1(2), 403-408 (1989). 13. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “A Boundary Model for the Thermal Creep Problem,” Fluid Dynamics 28(6), 876-878 (1993). 14. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “On the Use of Conservation Laws in Plane Slip Problems,” Teplofizika Vysokikh Temperatur (Russian) 33(1), 66-72 (1995). 15. Loyalka, S.K., “Velocity Profile in the Thermal Creep Slip Problem,” Phys. Fluids 19(10), 1641-1642 (1976). 16. Loyalka, S.K., “An Approximate Method in Transport Theory,” Phys. Fluids 24(10), 1912-1914 (1981). 17. Loyalka, S.K. and Cipolla, J.W.Jr., “On Choice of Trial Functions in Integro-Differential Variational Principles of Transport Theory,” Nucl. Sci. Engineering 99, 118-122 (1988).

Chapter 10 THE SLIP-FLOW REGIME

1.

BASIC EQUATIONS.

In this chapter, the description of a particle in a rarefied gas is considered under conditions where the Knudsen number is small. The analysis is restricted to the usual conditions assumed for aerosol particle motion in nonuniform gases. These conditions will be discussed later in detail. The classical sphere drag and thermal force problems are solved as important practical applications of the theory and techniques described here. In this regime, an aerosol particle disturbs the surrounding gas out to a distance on the order of the particle radius. The behavior of this particle can be described using the Navier-Stokes continuum equations except for the thin Knudsen layer in the neighborhood of the particle where the disturbance is greatest. Since, in the slip-flow regime, the typical dimension of the system is much greater than the mean free path, the continuum equations are assumed to be suitable for all distances. The local features of the gas in the Knudsen layer are then taken into account in the boundary conditions which have a special form discussed in Section 6.4. The closed system of continuum equations derived in Chapters 2 and 5 may be written as:

wu Dn n i Dt wxi

0 ,

(10-1)

212

Analytical Methods for Problems of Molecular Transport

wPij

Dui · § 0 , U ¨ Fi Dt ¸¹ wx j ©

DT Dt

2 § wui wqi · ¨ Pij ¸ , Nk n ¨© wx j wxi ¸¹

Pij

§ wu wu j 2 wuk · pG ij P ¨ i G ¸ , ¨ wx j wxi 3 ij wxk ¸ © ¹

Qi

N

wT , wxi

(10-2)

(10-3)

(10-4)

(10-5)

where P and N are the viscosity and thermal conductivity coefficients that are given by Eqs. (5-53) and (5-52), respectively, D Dt is a time-derivative following the motion, as in hydrodynamics, and N is the number of degrees of freedom of a molecule. For monatomic gases, N 3 , while for other gases, N ! 3 . Each of these equations is a non-linear partial differential equation, and therefore, there are many difficulties in solving this system in the most general form. Thus, one considers possible methods that might be used to linearize the equations. In this analysis, the non-linear and linear terms are estimated using the technique described in [1]. The force on a stationary particle is proportional to the external parameters of the surrounding nonuniform gas flow. Such parameters are the mean gas velocity for the sphere drag problem and the temperature gradient for the thermal force problem The following conditions are assumed to be satisfied in the statement of the above mentioned problems: 1. the gas mean velocity is much less than the heat velocity of a molecule; 2. the relative alteration of the temperature is much less than unity for a distance of the same order as the mean free path; 3. the relative temperature variation is small for a typical dimension distance, R . These conditions governing the external parameters can be expressed mathematically as:

213

Chapter 10. The Slip-Flow Regime

U 1, V

O T f T

1 ,

R T f T

1 .

(10-6)

As mentioned above, to evaluate the order of the partial derivatives contained in Eqs. (10-1)-(10-3) the technique described in [1] is used. It is necessary to note that the typical distance of a variation of all the disturbed parameters is the characteristic dimension of the suspended particle, R . The disturbance of the pressure in the sphere drag problem, and that of the local thermal flow velocity in the thermal force problem, can be evaluated by dimensional analysis which gives:

'p ~ UU 2 ,

U th ~

P T f . UT

(10-7)

The ratio of the terms in the continuity equation for each problem can be expressed as:

ui wn wxi U 'p R U 2 1 , ~ ~ n wui wxi kTR nU V 2 and:

R T f ui wn wxi U th p0 T f kTR 1 , ~ ~ n wui wxi p0U th T kT 2 where the pressure in the thermal force problem is assumed to be constant. By analogy, one obtains the following for the appropriate ratio in Eq. (10-2):

ui wui wxi

Q 2ui

~

U 2 R 2 UR ~ Q R QU

Re 1 ,

where Q is the kinematic viscosity and Re is the Reynolds number. For the thermal force problem, one needs to use the third moment equation which contains two non-linear terms. The first ratio of the nonlinear to linear terms can be evaluated by the relation:

n kui wT wxi

N 2T

~

nkTU th R 2 U th R R T f 1 , ~ ~ R NT VO T

214

Analytical Methods for Problems of Molecular Transport

where N ~ UV O cv ~ nkV O and U th ~ V O T f T . For the other non-linear term the appropriate evaluation is given by:

Pij wui wx j

N 2T

~

R T f nkTU th R 2 1 . ~ R nkV OT T

As one can see, if the basic assumptions defined by Eq. (10-6) are satisfied, the ratio of the non-linear to the linear terms is much less than unity for all the moment equations and, therefore, this system can be linearized by the simple expedient of neglecting the non-linear terms. The stationary linearized form of this system can be expressed as:

u 0 ,

(10-8)

P 2u p ,

(10-9)

2T

0 ,

(10-10)

where the external force, Fi , is assumed to be equal to zero. Having obtained the basic equations describing the gas flow over the particle, one is now ready to consider, in detail, the various transfer problems of the slipflow regime.

2.

THE SPHERICAL DRAG PROBLEM.

An understanding of translational motion of small single particles in a gas is required in disciplines as diverse as nano-phase materials synthesis, environmental physics, clean-room technology, cloud physics, and nuclear reactor safety [2,3]. While these particles, excepting liquid drops, are generally non-spherical, studies of spherical particles are necessary for the intrinsic fundamentals involved as well as for applied reasons. The drag problem is of great significance and, since the work of Millikan [4], there has been a substantial body of related experimental [5] and theoretical work [6-9] generated. Consider the short statement of the spherical drag problem. The origin of a spherical coordinate system is assumed to be at the center of a stationary

Chapter 10. The Slip-Flow Regime

215

spherical particle and the direction of the polar axis is the same as that of the mean velocity of the uniform flow at large distances from the particle, U (Fig. 7-2). The gas state can be described by Eqs. (10-8) and (10-9) which, for the spherical geometry and azimuthally symmetric problems, are given by [1]: 2 1 w rvr 1 w 2 vr cot T wvr 2vr 2 2 r wr 2 r wT 2 r 2 wT r 2 § wv · R wp 2 ¨ T vT cot T ¸ , T w r © ¹ P wr

(10-11)

2 1 w r vT 1 w 2 vT cot T wvT 2 r wr 2 r wT 2 r 2 wT v 2 wv R 1 wp 2 r 2 T2 , r wT r sin T P r wT

(10-12)

wvr 2vr 1 wvT vT cot T 0 . r r wT r wr

(10-13)

The special features of the problem associated with the slip-flow regime are accounted for in the boundary conditions specified at the particle surface. For large distances from the particle, the radial and tangential components of the mean velocity of the gas are given by:

ur f U cos T and uT f U sin T .

(10-14)

At the sphere surface, however, the tangential component of the mean velocity is not equal to zero as it is for the continuum regime. This implies that the gas is slipping along the surface. Since this slipping is the consequence of momentum conservation at the surface, the slip boundary conditions take the form:

vT

cm

vr

0 ,

O P , P rT

(10-15)

(10-16)

216

Analytical Methods for Problems of Molecular Transport

where cm is the isothermal-slip coefficient. This coefficient, from a physical point of view, is the relative rate of communication of the tangential momentum, per unit area of the particle surface, to the surrounding gas. For this regime the force on the particle is given by:

Fu

³ Prr cos T PrT sin T nr dS ;

r

R ,

(10-17)

S

where nr is a unit normal vector at the surface of the sphere which points into the gas. The pressure tensor components, Prr and PrT , are defined by [1]:

wvr , wr

Prr

p0 2P

PrT

§ 1 wvr wvT vT · P ¨ ¸ . © r wT w r r ¹

(10-18a)

(10-18b)

Again, the standard method involving separation of variables [1] is used to obtain the following solution for the system of Eqs. (10-11)-(10-13):

vr

§ A B · ¨ 3 r U ¸ cos T , ©r ¹

(10-19)

vT

B § A · ¨ 3 2r U ¸ sin T , 2 r © ¹

(10-20)

p

p0 P

B cos T , r 2

(10-21)

where the two constants, A and B , may be found from boundary conditions given by Eqs. (10-15) and (10-16). Simple calculations result in the following expression for the force on the particle:

Fu

6SP R U

1 2cm Kn , 1 3cm Kn

where Kn OP R and OP is given by Eq. (8-16a).

(10-22)

217

Chapter 10. The Slip-Flow Regime

Table 10-1. The isothermal-slip coefficient,

DW

1.00 1.1006

cm

0.95 1.2106

cm , for different values of DW . 0.90 1.3321

0.85 1.4673

0.80 1.6185

The most reliable analytical expression for the slip coefficient is given by Eq. (8-79) that can be presented in the form: cm

2 DW

DW

5 8

S

0.6690 0.1775 DW 0.7549 0.09714 DW

.

(10-23)

Table 10.1 gives the slip coefficient, cm , for a number of values of the tangential momentum accommodation coefficient, DW . Compare the analytical expression given by Eq. (10-22) with the Millikan experimental data for this regime. It is convenient to use the ‘socalled’ drag ratio value, D , which is given by: D{

Fu Kn F fm

,

where F fm is the drag force in the free-molecular regime that is defined by Eq. (7-13). From Eq. (10-22) one can obtain:

Dth

45 8

S Kn 1 2cm Kn . 8 DW S 1 3cm Kn

(10-24)

The Millikan data for the full range of Kn can be described by means of the use of the Cunningham slip correction factor the most exact form of which has been proposed by Buckley and Loyalka [8]. The use of this factor gives: Dexp

1.617 Kn CC Kn

,

(10-25)

where CC Kn is defined by [8]:

ª § 0.425 · º CC Kn 1 Kn «1.099 0.518 exp ¨¨ ¸» . Kn ¸¹ »¼ «¬ ©

(10-26)

218

Analytical Methods for Problems of Molecular Transport

Table 10-2. Comparison of experimental and theoretical values of drag on a sphere.

Kn

1

1 2 4 6 8 10 12 14 16 18 20

Dth

Dexp

DW =1.00

DW =0.95

DW =0.90

DW =0.85

DW =0.80

0.6633 0.4870 0.3113 0.2265 0.1774 0.1456 0.1234 0.1071 0.0946 0.0847 0.0766

1.1803 0.6284 0.3368 0.2331 0.1790 0.1455 0.1227 0.1061 0.0935 0.0836 0.0756

1.1883 0.6315 0.3384 0.2344 0.1802 0.1466 0.1237 0.1070 0.0943 0.0844 0.0763

1.1970 0.6348 0.3401 0.2358 0.1814 0.1477 0.1247 0.1079 0.0952 0.0852 0.0771

1.2063 0.6383 0.3417 0.2371 0.1825 0.1487 0.1257 0.1089 0.0961 0.0860 0.0778

1.2163 0.6420 0.3434 0.2384 0.1837 0.1498 0.1266 0.1098 0.0969 0.0868 0.0786

The Millikan experimental data and analytical results are presented in Table 10-2. A comparison of experimental results with the theory for the slip-flow regime is illustrated by Fig. 10-1. It can be easily seen that the best agreement between the theoretical and experimental data for this regime is reached when DW 0.95 . For this case the discrepancy of results is less than 0.5% for Kn 0.1 . This excellent agreement of data is an additional confirmation of the sufficiently high accuracy of Eq. (10-23).

3.

THE THERMAL FORCE PROBLEM.

In Section 7.4 the problem of thermal forces (thermophoresis) on aerosol particles in the free-molecular regime was considered. Here, this problem will be examined in the slip-flow regime. A quantitative knowledge of thermophoresis is of great importance in many areas such as optical fiber fabrication, nuclear reactor safety, micro-contamination control, etc. Thermophoresis has been studied extensively, both experimentally [1018] and theoretically [19-29]. In spite of the numerous articles associated with this problem, both the experimental data and the theory remain controversial. The former due to the difficulties in measuring the relatively small effect because of competing phenomena, and the latter because of the approximations employed in the theoretical efforts. None of the previous work, with the exception of Loyalka [29], has attempted to solve this problem in its general form. Loyalka’s work contains the most general numerical analysis that may be used to obtain results for all Knudsen numbers.

219

Chapter 10. The Slip-Flow Regime

1.2

Sphere Drag Ratio (dimensionless)

1.0

0.8

0.6

0.4

0.2

0.0 0.0

5.0

10.0

15.0

20.0

Inverse Knudsen Number (dimensionless)

Figure 10-1. Comparison of the experimental and theoretical values of the drag on a sphere. Theoretical values are given by the solid line. Experimental points are obtained from the interpolation formula of Buckley and Loyalka [8] for DW 0.95 .

Consider the various aspects of this problem within the slip-flow framework. Let a single spherical particle of radius, R , be located in an infinite gas with a temperature gradient T , which is constant at large distances from the particle. All the other conditions have been specified in Section 7.4. For the slip-flow regime, it is shown here that the NavierStokes equations may be applied, with boundary conditions appropriate for Kn 1 , to the quantitative description of the thermal force. The momentum and continuity equations are given by Eqs. (10-11)-(10-13). The temperature distribution, both inside and outside of the particle, is described by the thermal conductivity equation which may be written as:

1 w § 2 wT · 1 w r r 2 wr ¨© wr ¸¹ r 2 sin T wT

wT · § ¨ sin T ¸ 0 . wT ¹ ©

(10-27)

The proper slip-flow boundary conditions for the motion of the gas at the particle surface are defined by the relations:

220

Analytical Methods for Problems of Molecular Transport

vr

0 ;

r

R ,

vT

ª w §v cm OP « r ¨ T ¬ wr © r

(10-28)

· 1 wvr º

1 § wT · ¸ r wT » cTslQ RT ¨ wT ¸ ; r ¹ ¹ ¼ 0 ©

R ,

(10-29)

where cTsl is the thermal-creep coefficient. The most exact analytical expression for cTsl is given by Eq. (8-88). At large distances from the particle we have:

vr

vT

0 .

(10-30)

The appropriate boundary conditions for the thermal conductivity equation have the form:

Ng

wTg wr

Tg Tp

Np

wTp wr

cT ON

; r

wTg wr

R ,

; r

R ,

Tg o T f r cos T ; r o f ,

(10-31)

(10-32)

(10-33)

where cT is the temperature-jump coefficient. The most exact expression for cT is given by Loyalka’s formula, Eq. (8-50). The solutions of Eqs. (1011)-(10-13) and (10-27) can be presented in the form:

vr

§ A B· ¨ 3 r ¸ cos T , ©r ¹

vT

B· § A ¨ 3 ¸ sin T , 2r ¹ © 2r

221

Chapter 10. The Slip-Flow Regime

B cos T , r 2

p

p0 P

Tg

T0 T f r cos T

Tp

T0 C2 r cos T ,

C1 cos T , r 2

where the constants A , B , C1 , and C2 can be found from boundary conditions. By means of Eq. (10-17), the following expression is obtained for the thermal force:

F

12S

P 2 R T f cTsl ] m] T , U T0

(10-34)

where the quantities ] m and ] T are defined by:

]m

1 , 1 3cm Kn

]T

N N c Kn 1 2 N N 2c Kn g

p

g

T

p

.

T

The trustworthiness of the results obtained here may be analyzed by comparing the thermal force values given by Eq. (10-34) with Loyalka’s numerical values [29]. For this comparison, the thermal force expression of Eq. (10-34) must be transformed into the form used by Loyalka. It is convenient to introduce a dimensionless thermal force, FT , by means of the relation:

F

p0 R 2

O T f T0

FT ,

(10-35)

where, FT , may be written as:

FT

75 32

S 2 KncTsl ] m] T ,

(10-36)

222

Analytical Methods for Problems of Molecular Transport

Table 10-3. The reduced thermal force,

Kn

FT .

FT

N p N g 10.0

1

Eq. (10-36) 2.0175 1.7303

1.2463 1.6617

N p N g 100.0

Loyalka [29] 2.2056 1.7965

Eq. (10-36) 1.9936 1.6972

Loyalka [29] 2.1378 1.7099

Table 10-4. The reduced thermal force on NaCl aerosol particles.

FT

FT experimental [15] 2.4300 2.1552 1.9115 1.6935 1.5036 1.3336

Kn 1 1.00 1.25 1.50 1.75 2.00 2.25

theoretical Eq. (10-36) 2.2059 1.9879 1.8006 1.6387 1.4977 1.3742

Kn 1 2.50 2.75 3.00 3.25 3.50 3.75

experimental [15] 1.1828 1.0490 0.9304 0.8252 0.7319 0.6491

theoretical Eq. (10-36) 1.2654 1.1690 1.0832 1.0066 0.9378 0.8759

and molecules are assumed to be rigid spheres. It is important to note that FT depends on the nature of the aerosol particle through the accommodation coefficients, the Knudsen number, and the ratio of the thermal conductivity coefficients of the gas and the particle. Let the accommodation coefficients, DW DT 1 . For this case, the slip and jump coefficients are specified by: cTsl

1.0193

,

cm

1.1006

,

cT

2.0633

.

A comparison of the results obtained from Eq. (10-36) with Loyalka’s numerical results [29] is given in Table 10-3. It should be noted that the results reported here as Loyalka’s were transformed in a simple fashion to conform to the current notation. One can see that the discrepancy between these results is very small. One may further compare the theoretical results with the experimental values for FT that have been reported by Jacobsen and Brock [15]. The experimental results for sodium chloride aerosols in argon can be fitted, for 0 Kn 1 4 , by the expression:

F

T

exp

5 4

§ W · ¸ , © Kn ¹

S exp ¨

(10-37)

where the constant W 0.48 . The theoretical and experimental results are presented in Table 10-4. A graphical comparison of these values is shown in

223

Chapter 10. The Slip-Flow Regime

Reduced Thermal Force (dimensionless)

2.5

2.0

1.5

1.0

0.5 1.0

1.5

2.0

2.5

3.0

3.5

4.0

Inverse Knudsen Number (dimensionless)

Figure 10-2. A comparison of experimental and theoretical reduced thermal forces for NaCl. Theoretical values from Eq. (10-36) are shown as a solid line and experimental values reported in [15] are shown as solid circles.

Fig. 10-2. One can easily see that the agreement of the theory with the experiment is within 10% for the very narrow region of inverse Knudsen number, 0 Kn 1 3 . A substantial discrepancy occurs outside this region, however, and further experimental and theoretical work is clearly necessary before an acceptable understanding of the whole slip-flow region is possible. One contributing factor to the discrepancy between theory and experiment is that the theory does not take into account all of the first-order boundary effects. A strict theory must account for corrections that are proportional to Kn for the thermal-creep. These additional terms occur due to surface curvature and to Barnett’s thermal stresses [30]. The influence of all the corrections that are proportional to Kn on the thermophoresis of aerosol particles has been analyzed in [30]. Unfortunately, the accuracy of the additional terms was not sufficiently discussed and, moreover, a kinetic theory analysis of these phenomena leads to very complicated calculations that are of the same order of difficulty as an appropriate solution for all Knudsen numbers. Consequently, it stands to reason that the most reliable

224

Analytical Methods for Problems of Molecular Transport

theory must ultimately be based on the direct solution of the Boltzmann equation by use of different moment methods.

PROBLEMS 10.1. A sphere suspended in a gas has radius, R , and temperature, Determine the heat flux and the temperature T0 'T 'T T0 . distribution for the surrounding gas if its number density and temperature far from the sphere are n0 and T0 , respectively. Solve this problem for the slipflow regime. Solution: The temperature distribution can be determined from the Laplace equation:

1 w § 2 wT · ¨ r ¸ 0 . r 2 wr © wr ¹ The boundary conditions are:

T

T0 'T cT O

wT wr

and T

T0 ;

r o f .

r R

The solution of this boundary problem may then be expressed as:

T r T0 'T ]

R ; ] r

O· § ¨1 cT R ¸ © ¹

1

.

The heat flux then takes the form:

Q

4S R 2N

wT wr

4S RN'T ] . r R

10.2. To maintain the constant surface temperature, T0 'T 'T T0 , of a sphere (radius R ), there is a heat source of power, Q , located at the center of the sphere. Determine the thermal conductivity coefficient of the surrounding gas if its number density and temperature far from the sphere are n0 and T0 , respectively. Solve this problem for the slip-flow regime. Solution: Using the expression for the heat flux obtained in Problem 10.1, one has that:

225

Chapter 10. The Slip-Flow Regime

N

Q ; ] 4S R'T ]

O· § ¨1 cT R ¸ © ¹

1

.

10.3. One method of measuring the thermal conductivity of a gas is to determine the loss of heat through the gas from a hot sphere of radius, R . For the same conditions used in Problem 10.2, determine a calculation error for the thermal conductivity coefficient of the gas if the temperature-jump at the surface of the sphere is not taken into account. Solution: In the continuum regime, the thermal conductivity coefficient is given by: Q . 4S R'T

N0

Using the result of Problem 10.2, one can obtain:

G

N N0 N

1

1

]

cT

O R

.

10.4. Determine the heat flux and temperature distribution of a gas between two concentric spheres the radii and temperatures of which are R1 , T0 'T and R2 , T0 'T , respectively. Assume 'T T0 . The equilibrium number density of the gas is n0 . Solve this problem for the slip-flow regime. Solution: The temperature distribution is given by:

T r

A B . r

The two constants, A and B , are determined from the boundary conditions:

T R1

T0 'T cT O

wT , and: wr

T R2

T0 'T cT O

wT . wr

The temperature distribution and the thermal flux may be expressed as:

O · R1 § O · º °½ ° 'T ª 2 R1 § ] « T r T0 ®1 ¨1 cT ¸ ¨ 1 cT ¸» ¾ , T0 «¬ r R1 ¹ R2 © R2 ¹ »¼ ¿° © ¯°

226

Analytical Methods for Problems of Molecular Transport

Q

8S R1N'T ] ,

]

ª§ O · R1 § O ·º «¨1 cT ¸ ¨1 cT ¸» R1 ¹ R2 © R2 ¹ »¼ «¬©

1

.

10.5. Determine the heat flux and temperature distribution of a gas between two coaxial cylinders the radii and temperatures of which are R1 , T0 'T and R2 , T0 for the inside and outside cylinders, respectively. Assume 'T T0 . Solve this problem for the slip-flow regime. Solution: The temperature distribution may be found from the following boundary problem:

1 w § wT · r 0 , r wr ¨© wr ¸¹

T R1

T0 'T cT O

T R2 T0 cT O

wT , wr

wT . wr

After some calculations, one can obtain:

ln r R2 cT O R2 ° 'T °½ T r T0 ®1 ¾ , and: T0 ln R2 R1 cT O 1 R1 1 R2 ¿° ¯°

Q

ª §R 2SN'T « ln ¨ 2 ¬« © R1

· § 1 1 ·º ¸ cT O ¨ ¸» ¹ © R1 R2 ¹ ¼»

1

.

10.6. One method of measuring the thermal conductivity of a gas is to determine the loss of heat through the gas from a long hot wire enclosed by a cylinder having a radius much greater than that of the wire. If the thermal power generated per unit length of the wire, Q , and the temperature difference, 'T , are measured during the experiment, determine the thermal

Chapter 10. The Slip-Flow Regime

227

conductivity coefficient of the ambient gas. Assume that the temperature of the external cylinder is T0 and that slip-flow conditions are satisfied during this experiment. Solution: Using the expression for the heat flux from Problem 10.5 and allowing for R r , one can obtain:

N

Q 2S'T

ª §R· Oº «ln ¨ r ¸ cT r » . ¬ © ¹ ¼

10.7. For the conditions given in Problem 10.6, determine the relative error of the predicted values of the thermal conductivity coefficient of the ambient gas if the temperature-jump at the surface of the wire is neglected. Solution: In the continuum regime, the thermal conductivity coefficient is given by:

N0

Q §R· ln ¨ ¸ . 2S'T © r ¹

From this one may generate the following relative error:

G

N N0 N

cT O r . ln R r cT O r

10.8. Two flat parallel round disks of equal radius, R , are separated by a small distance, h R . While the lower disk is stationary, the upper disk moves toward it with a velocity, u . Determine the resistance force on the upper disk for the slip-flow regime. Use the geometry in Fig. 10-3.

228

Analytical Methods for Problems of Molecular Transport

Figure 10-3. The geometry to be used in Problem 10.8 in determining the resistance force on a flat, round disk in the slip-flow regime due to its approach to an identical, parallel, stationary disk.

Solution: The motion of the gas in this problem is axially symmetric such that uI 0 and, therefore, the Navier-Stokes system of equations in cylindrical coordinates may be written as [1]:

wu 1 w r z ru r wr wz

§

P ¨ 'ur ©

P'u z

'f

ur · ¸ r 2 ¹

0 ,

wp , wr

wp , where: wz 1 w § wf · w 2 f r . r wr ¨© wr ¸¹ wz 2

It may readily be shown that u z ~ u and ur ~ R h u and, therefore, that ur u z . For the partial derivatives, one can obtain the following estimates:

wur Ru ~ wr Rh

u , h

wur Ru ~ hh wz

R u , h2

229

Chapter 10. The Slip-Flow Regime

P'u z ~ UV O

u , h2

wp U kT UV 2 ~ ~ . mh h wz Consequently:

P'u z wp wz

~

wu wr h u O 1 , and r ~ 1 . V h wur wz R

Taking into consideration these estimates, one can reduce the Navier-Stokes system to the form:

wu 1 w r z ru r wr wz

P

w 2 ur wz 2

wp wz

0 ,

(P-1)

wp , wr

(P-2)

0 .

(P-3)

The boundary conditions may be expressed as:

z

0 , uz

0 , ur

z

h , uz

u , u r

r

R , p

p0 .

wur , wz

cm

O P P rz

cm O

cm

O P P rz

cm O

wur , wz

From Eqs. (P-2) and (P-3) and the boundary conditions, one can obtain:

230

Analytical Methods for Problems of Molecular Transport

ur

1 dp ª z z h cm O h º¼ 2 P dr ¬

.

Integration of Eq. (P-1) yields: h

u

1 w r dz ru r wr ³0

h3 d § dp · § O· r ¸ ¨ 1 6cm ¸ . ¨ 12P r dr © dr ¹ © h¹

From this relation, the pressure may be expressed in the form:

p

p0

3P u R 2 r 2 . h3 1 6cm O h

The resistance force on the upper disk is then given by:

Fz

3 2

S

P R 4] u h3

where ]

O· § ¨ 1 6cm h ¸ © ¹

1

.

10.9. Determine the velocity distribution and the net mass transport of a gas flowing along a cylindrical tube owing to a pressure gradient, dp dx . The tube radius is R . Solution: The gas velocity has only an x -component which depends only on the radial coordinate, r . The Navier-Stokes system may be written as:

1 d § du · r r dr ¨© dr ¸¹ wp wr

1 wp , P wx

0 .

The slip-flow boundary condition is given by: u R cm O

wu . wr

The solution of this boundary problem can be written as:

231

Chapter 10. The Slip-Flow Regime

u r

º dp 1 dp ª 2 § O· R ¨ 1 2cm ¸ r 2 » , with « dx 4 P dx ¬ © R¹ ¼

const .

The net mass transfer is given by: R

M

2SU ³ rudr 0

SU R 4 dp § O· ¨1 4cm ¸ . R¹ 8P dx ©

(P-4)

10.10. Determine the pressure difference at the ends of a cylindrical tube of radius, R , and length, L , if the net mass transfer through the tube is M . The gas flow is assumed to be isothermal. Solution: The pressure gradient may be expressed in the form (see Problem 10.9):

dp dx

8P kTM , S mR p 4cm p1 O1 R 4

where p1 and O1 are the pressure and mean free path at the entrance of the tube. Integrating this expression, one can obtain:

p1 p2

12 ª º ½° 16P kTML ° » ¾ . p1 1 4cm O1 R ®1 «1 2 4 2 ° «¬ S mR p1 1 4cm O1 R »¼ ° ¯ ¿

10.11. Two vessels containing the same gas at the different temperatures, T1 and T2 , are connected by a long cylindrical tube of radius, R . There exists a pressure difference between the vessels owing to the phenomenon of thermal-creep. Determine this pressure difference and the velocity distribution of the gas in the tube. Solution: The net mass transfer per unit time, ignoring transfer in the Knudsen layer, may be written as (see Problem 10.9):

M

18 S

U R4 § O · dp 1 dT cTslQUS R 2 1 4cm ¸ . ¨ P © R ¹ dx T dx

The mechanical equilibrium condition, M

O · dp § p ¨ 1 4cm ¸ R ¹ dx ©

8k P 2 cTsl dT . mR 2 dx

0 , gives the following relation: (P-5)

232

Analytical Methods for Problems of Molecular Transport

If one takes into consideration that O ~ 1 p and P ~ T , this equation may be transformed to:

O1 · dp § ¨ p 4cm p1 R ¸ dx © ¹

8k P12 cTsl dT T . dx mR 2T1

Integrating this expression, one obtains: p2 p1

]

p1] 1

1 A T

O1 · § ¨ 1 4cm R ¸ © ¹

2 2

1

and A

T12 1 where:

8k P12] 2 cTsl . mT1 p12 R 2

The velocity distribution is then given by:

u

º dp 1 ª 2§ O· 1 dT cTslQ R ¨1 2cm ¸ r 2 » . R¹ dx T dx 4 P «¬ © ¼

Substituting dp dx into this relation from Eq. (P-4), one obtains:

§ 2r 2 · 1 dT u r cTslQ] ¨ 2 1¸ . © R ¹ T dx For r

0 and r

R , the gas velocity has opposite signs.

10.12. For the same conditions used in Problem 10.11, determine the pressure difference if the variation in the mean velocity due to thermal-creep in the Knudsen layer is taken into account. Solution: The mean velocity due to the temperature gradient is given by Eq. (8-87) and may be expressed in the form:

uth r cTslQ ª¬1 \ exp D R r º¼

\

0.3882 DW 0.4354 0.2179 DW

,

1 dT where: T dx

233

Chapter 10. The Slip-Flow Regime

Figure 10-4. The geometry to be used for Problem 10.13 in determining the net mass transfer in a closed capillary loop with different radii and end temperatures.

and D 2.2015 O 1 . Taking into account the mass transport in the Knudsen layer, Eq. (P-5) from Problem 10.11 may be cast in the form: 1

O ·§ 2\ · dp § p ¨ 1 4cm ¸¨1 R ¹© D R ¸¹ dx ©

8k P 2 cTsl dT . mR 2 dx

The pressure difference is given by:

p1] 11

p2 p1

]1

1 A T 1

§ § 2\ · O1 · ¨1 ¨ 4cm ¸ DO ¸¹ R ¹ © ©

2 2

T12 1 where:

1

and A1

8k P12] 12 cTsl . mT1 p12 R 2

10.13. Two capillary tubes having an equal length, L , but different radii, R1 and R2 where R1 R2 , are connected at each end to form a closed loop (see the figure). The ends of the capillary tubes are maintained at the temperatures, T0 and T0 'T where 'T T0 , as shown in Fig. 10-4. Determine the net mass transfer through any cross section of either tube that is due to a circular motion of the gas. The gas parameters, P and U , are assumed to be constant and the isothermal end connections are assumed to make no contribution to the flow. Solution: The solution for the net mass transfer from Problem 10.9 may be presented in the form:

M]i Ri4

SU dp cTsl SP] i 1 dT where: 8P dx Ri2 T0 dx

234

Analytical Methods for Problems of Molecular Transport

Figure 10-5. The geometry to be used in Problem 10.14 in determining the torque on a rotating sphere in the slip-flow regime.

]i

§ O· ¨1 4cm ¸ Ri ¹ ©

1

.

Integrating over all values of x along a closed path, one can obtain:

M

cTsl P

'T S R12 R22 ] 1 R22 ] 2 R12 . T0 L ] 1 R24 ] 2 R14

10.14. A sphere of radius, R , is rotating with an angular velocity, Z , such 12 that Z R 2kT m . Determine the torque, K , on the sphere in the slipflow regime. Use the geometry in Fig. 10-5. Solution: Owing to the axial symmetry of this problem, the mean velocity of the gas has only one component, uI r,T u r,T . Additionally, one also has that p const . The Navier-Stokes system of equations thus reduces to a single equation given by [1]:

235

Chapter 10. The Slip-Flow Regime

1 w § 2 wu · 1 w r r 2 wr ¨© wr ¸¹ r 2 sin T wT

wu · u § ¨ sin T ¸ 2 2 wT ¹ r sin T ©

0 .

The boundary conditions may be expressed as:

u R Z R sin T cm

O P R and u f 0 where: P rI

§ wu u · PrI R P ¨ ¸ © wr r ¹ r

. R

Let u r ,T be expressed in the following form u r,T f r sin T . Then, the function, f r , may be determined from the equation:

d § 2 df · 2f r dr ¨© dr ¸¹

0 .

The solution of the above boundary problem is then:

u r,T

Z R 3] r 2

sin T ,

PrI R 3PZ] sin T ,

dFI where ]

3PZ R 2] sin 2 T dT dI ,

1 3cm O R 1 . 2S

Kx

The net torque, K x , is then:

S

R ³ dI ³ sin T dFI dT 0

8SPZ R 3] .

0

10.15. An infinite cylinder of radius, R , is rotating with an angular velocity, 12 Z , such that Z R 2kT m . Determine the torque per unit length in the slip-flow regime.

236

Analytical Methods for Problems of Molecular Transport

Figure 10-6. The geometry to be used for Problem 10.16 in determining the average speed of a steady gas flow between two, parallel, planar surfaces in the slip-flow regime.

Solution: The mean velocity of the gas possesses only one component, uI u r . Additionally, p const . The velocity distribution may be found from the following boundary problem:

1 w § wu · u r r wr ¨© wr ¸¹ r 2 u R Z R cm

0 ,

O P R and u f 0 , where: P rI

§ wu u · ¸ . PrI R P ¨ © wr r ¹ The solution of this problem may be expressed as:

u

Z R 2] r

and PrI R 2 PZ] where ]

1 2cm

O R

.

The torque is then given by K x 4SPZ R 2] where the direction of the angular velocity vector, Ȧ , is in the direction of the x -axis. 10.16. In the steady flow of a gas between two parallel planar surfaces, if d O 1 (the slip-flow regime) determine the average gas speed, u , owing to a constant pressure gradient, dp dy const , where the velocity is assumed 1-dimensional in the y -direction. Use the geometry in Fig. 10-6. Solution: The gas flow may be described by Eqs. (10-8) and (10-9) which, for this particular case, take the form:

wu y wy

§ w 2u · wp 0, P¨ 2 ¸ , wp wx 0 , and u y u x . © wx ¹ wy

Chapter 10. The Slip-Flow Regime

237

The boundary conditions are given by u d 2 cm O wu wx u d 2 cm O wu wx . This statement of the problem then yields:

and

d 2

u d 1

³

d 2

u x dx 121

d2 § O · dp 1 6cm ¸ . d ¹ dy P ¨©

REFERENCES 1. Landau, L.D. and Lifshitz, E.M., Fluid Mechanics (Pergamon Press, Oxford, 1987). 2. Hidy, G.M. and Brock, J.R., Topics in Aerosol Research, vols. 1-3 (Pergamon Press, Oxford, 1971-1973). 3. Williams, M.M.R. and Loyalka, S.K., Aerosol Science: Theory and Practice, with Special Applications to Nuclear Industry (Pergamon Press, Oxford, 1991). 4. Allen, M.D. and Raabe, O.G., “Re-evaluation of Millikan’s Oil Drop Data for the Motion of Small Particles in Air,” J. Aerosol Sci. 13(6), 537-547 (1982). 5. Cercignani, C., Pagani, C.D., and Bassanini, P., “Flow of Rarefied Gas Past an Axisymmetric Body. II. Case of a Sphere,” Phys. Fluids 11(7), 1399-1403 (1968). 6. Phillips, W.F., “Drag on a Small Sphere Moving Through a Gas,” Phys. Fluids 18(9), 1083-1089 (1975). 7. Lea, K.C. and Loyalka, S.K., “Motion of a Sphere in Rarefied Gas,” Phys. Fluids 25(9), 1550-1557 (1982). 8. Buckley, R.L. and Loyalka, S.K., “Cunningham Correction Factor and Accommodation Coefficient: Interpretation of Millikan’s Data,” J. Aerosol Sci. 20(3), 347-349 (1989). 9. Loyalka, S.K., “Motion of a Sphere in a Gas: Numerical Solution of the Linearized Boltzmann Equation,” Phys. Fluids A 4(5), 1049-1056 (1992). 10. Rosenblatt, P. and La Mer, V.K., “Motion of a Particle in a Temperature Gradient. Thermal Repulsion as a Radiometer Phenomenon,” Phys. Rev. 70(5-6), 385-395 (1946). 11. Saxton, R.L. and Ranz, W.E., “Thermal Force on an Aerosol Particle in a Temperature Gradient,” J. Appl. Phys. 23(8), 917-923 (1952). 12. Schadt, C.F. and Cadle, R.D., “Thermal Forces on Aerosol Particles in a Thermal Precipitator,” J. Colloid Sci. 12, 356-362 (1957). 13. Schadt, C.F. and Cadle, R.D., “Thermal Forces on Aerosol Particles,” J. Phys. Chem. 65(10), 1689-1694 (1961). 14. Schmitt, K.H., “Untersuchungen an Schwebstoffteilchen im Temperaturfeld,” Z. Naturforschg. 14a, 870-881 (1959). 15. Jacobsen, S. and Brock, J.R., The Thermal Force on Spherical Sodium Chloride Aerosols,” J. Colloid Sci. 20(6), 544-556 (1965). 16. Waldmann, L. and Schmitt, K.H., “Thermophoresis and Diffusiophoresis of Aerosols,” in Aerosol Science, edited by Davies, C.N. (Academic Press, London, 1966). 17. Derjaguin, B.V., Storozhilova, A.I., and Rabinovich, Ya.I., “Experimental Verification of the Theory of Thermophoresis of Aerosol Particles,” J. Colloid and Interface Sci. 21(1), 35-58 (1966). 18. Keng, E.Y.H. and Orr, C.Jr., “Thermal Precipitation and Particle Conductivity,” J. Colloid and Interface Sci. 22, 107-116 (1966). 19. Epstein, P.S., “Zur Theorie des Radiometers,” Z. Physik 54(4), 537-563 (1929).

238

Analytical Methods for Problems of Molecular Transport

20. Brock, J.R., “On the Theory of Thermal Forces Acting on Aerosol Particles,” J. Colloid Sci. 17, 768-780 (1962). 21. Dwyer, H.A., “Thirteen-Moment Theory of the Thermal Force on a Spherical Particle,” Phys. Fluids 10(5), 976984 (1967). 22. Ivchenko, I.N. and Yalamov, Yu.I., “The Hydrodynamic Calculation Method of the Thermophoresis of Sufficiently Large Nonvolatile Aerosol Particles,” J. Phys. Chem. (Russia) 45(3), 577-582 (1971) (in Russian). 23. Derjaguin, B.V. and Yalamov, Yu.I., “The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and their Experimental Testing,” in Topics in Current Aerosol Research, vol. 3, part 2, edited by Hidy, G.M. (Pergamon Press, Oxford, 1972). 24. Sone, Y. and Aoki, K., “Negative Thermophoresis: Thermal Stress Slip Flow Around a Spherical Particle in a Rarefied Gas,” in Rarefied Gas Dynamics, vol. 74, part I, edited by Fisher, S.S. (American Institute of Aeronautics and Astronautics, 1981). 25. Yamamoto, K. and Ishihara, Y., “Thermophoresis of a Spherical Particle in a Rarefied Gas of a Transition Regime,” Phys. Fluids 31(12), 3618-3624 (1988). 26. Bakanov, S.P., “Thermophoresis in Gases at Small Knudsen Numbers,” Aerosol Sci. Tech. 15(1), 77-92 (1991). 27. Loyalka, S.K., “Mechanics of Aerosols in Nuclear Reactor Safety: A Review,” Prog. Nucl. Energy 12(1), 1-56 (1983). 28. Loyalka, S.K., “Rarefied Gas Dynamics Problems in Environmental Sciences,” in Rarefied Gas Dynamics, XVI Symposium, edited by Boffi, V. and Cercignani, C. (Teubner, Stuttgart, 1986). 29. Loyalka, S.K., “Thermophoretic Force on a Single Particle-I. Numerical Solution of the Linearized Boltzmann Equation,” J. Aerosol Sci. 23(3), 291-300 (1992). 30. Poddoskin, A.B., Yushkanov, A.A., and Yalamov, Yu.I., “The Thermophoresis Theory for Sufficiently Large Aerosol Particles,” J. Tech. Phys. (Russian) 52(11), 2253-2261 (1982) (in Russian).

Chapter 11 BOUNDARY VALUE PROBLEMS FOR ALL KNUDSEN NUMBERS

1.

THE MOMENT EQUATIONS IN ARBITRARY CURVILINEAR COORDINATES.

The most advanced analysis possible for boundary value transport problems involving arbitrary Knudsen number and for various molecular interaction laws must be founded on the solution of the Boltzmann equation. In the transition regime, the effects of intermolecular collisions become dominant and their influence on the gas behavior is accurately described only by the full Boltzmann form of the collision operator found in the Boltzmann equation. For this reason, different model forms of the collision integral and model equations will not be considered here. There are two kinds of difficulties associated with the solution of the Boltzmann integro-differential equation [1]. The first difficulty is connected with the direct solution of this equation in which the distribution function depends, in the general case, on seven independent variables. It is well known that the difficulty in solving differential equations increases very rapidly with the number of independent variables. The second difficulty is associated with the very complicated non-linear structure of the collision term in the Boltzmann equation. The first difficulty may be overcome by replacing the Boltzmann equation with a system of equations for macroscopic values (moments). Such a moment system involves, in general, functions of only four independent variables and, thus, the difficulty of solution does not increase to the same level as it would with the Boltzmann equation due to this lesser

240

Analytical Methods for Problems of Molecular Transport

number of independent variables. Additionally, for specific transport problems involving stationary or symmetry conditions, the number of independent variables may even be less than four. The second difficulty may be overcome, for some applied transport problems, through the process of linearization. This may be done, however, only if the appropriate linearization conditions are physically satisfied. In order to investigate boundary value transport problems for all Knudsen numbers a moment method is considered. Instead of seeking exact solutions of the Boltzmann equation, this equation may be satisfied in a certain average sense. For boundary value transport problems, one is often not particularly interested in the distribution function itself since only certain lower moments of this function have the practical significance of being connected with the macroscopic values that describe the gas. Since one is mainly interested in mean flow quantities rather than the distribution function itself, the Maxwell integral transport equations, or moment equations, will be taken as the basis of the theory of transport phenomena for arbitrary Knudsen number. The general moment equation can be obtained by multiplying the Boltzmann equation by the function, Q v , of the components of the molecular velocity and integrating over the velocity space. The general form of this equation was derived in [2]. The transport moment equation of the molecular property, Q v , in arbitrary, orthogonal, curvilinear coordinates, has the following form:

w wt

3

³ fQ v dv h h1 h ¦ wwD h h ³ fv Q v dv j k

1 2 3 i 1

3 § 3 F w · ° 1 ³ f ¨ ¦ i Q v ¸ dv ³ f ¦ ® hh i 1° © i 1 m wvi ¹ ¯ i j

wh 1 § 2 whk vi vk i ¨ vk wD k hi hk © wD i

i

i

§ wh j wh vi v j i ¨ v 2j ¨ wD i wD j ©

· ½° w Q v dv ¸¾ ¹ °¿ wvi

· ¸ ¸ ¹

(11-1)

n' Q v ,

where D1 , D 2 , D 3 are the curvilinear coordinates, hi D1 , D 2 ,D 3 denotes the metric coefficient such that dli hi dD i , Fi is the component of the external force acting on a molecule, and the moment of the collision operator has been expressed in standard notation as n'Q( v ) . For the spherical and cylindrical geometries these metric coefficients can be written as:

D1

r

h1 1

D2 T

D3 I ,

h2

h3

r

r sin T ,

Chapter 11. Boundary Value Problems for All Knudsen Numbers

241

and:

D1

r

h1 1

z ,

D2 T

D3

h2

h3 1 .

r

For the stationary problem with radial symmetry, in the absence of external forces, the general moment equation, Eq. (11-1), takes the following form [2,3]:

w 1 d j 1 r ³ fvr Q v dv ³ f ® ¬ªvT2 j 1 vI2 ¼º Q v j wvr r r dr ¯ w ª¬ j 1 cot T vI2 vT vr º¼ Q v wvT

j 1 ª¬vI vr cot T vI vT º¼

½° w Q v ¾ dv wvI °¿

(11-2)

Rn'Q v ,

where, for the cylindrical problem, j 1 , for the spherical problem, j 2 , r r R is a dimensionless radial vector and R is the radius of the sphere or cylinder. If Q v is successively selected to be the collisional invariants, m , mvr , and 12 mv 2 , for which n'Q ( v ) 0 , one can obtain the ordinary continuity equation, and the radial momentum and energy transport equations. These moment equations are usually employed in all boundary value transport problems. Other equations can be obtained by using the arbitrary functions, Q v . It is necessary to note, however, that such a system of moment equations will never be closed as the number of the unknown moments is always greater than the number of moment equations. A common feature of all moment methods is their inclusion of some technique for closing the moment system. Typically, the technique for closing the moment system consists of making a special choice of the distribution functions so that they contain a certain number of unknown quantities that depend upon the various coordinates and time. The number of these unknown quantities must be equal to the number of moment equations and thus is a characteristic of the order of a given moment approximation. Two ways may be distinguished by which to close the moment system. The first is based on the use of polynomial expansions in velocity space for the distribution function. The coefficients in this expansion are unknown functions of time and space but may be determined from the moment system

242

Analytical Methods for Problems of Molecular Transport

if one employs a final number of expansion terms and uses an equal number of moment equations [4-7]. The second method of closing the system of moment equations involves making a special choice for the distribution function that is pertinent to the specific transport problem under consideration [2,3,8-12]. Simple approximate functions may be chosen if one takes into account conditions specific to the problem. However, one must insure that the distribution function chosen has the features normally characteristic of both the freemolecular and continuum regimes. Mott-Smith [13], to describe planar shock waves, was the first to suggest the use of the discontinuous Maxwellian distribution function. This idea was then generalized by Lees and Liu [2,3] for arbitrary curvilinear geometries. At present, the distribution functions proposed by Mott-Smith and Lees are considered to be the most suitable for the solution of boundary value transport problems. The Lees method will be discussed in the next section.

2.

THE TWO-SIDED MAXWELLIAN DISTRIBUTION FUNCTIONS.

As mentioned above, for arbitrary Knudsen number it is most convenient to use a moment method. Any moment method is an integral method. Just as in any other integral method, the distribution function that is employed is not necessarily an exact solution of the original Boltzmann equation. Nevertheless, this function must have features common to both the freemolecular and continuum distributions. It is also desirable that the distribution function be adaptable in such a fashion as to satisfy the boundary conditions. The best distribution function satisfying these conditions for the Maxwellian boundary model is that proposed by Lees and Liu [2,3]. In the Lees method the distribution function is assumed to be twosided Maxwellians which are discontinuous on the surface of the ‘so-called’ ‘cone of influence’ in the velocity space. Consider the arbitrary point, M r , where r is a dimensionless position vector. For each specified point there exists a cone of influence the surface of which is formed by tangents to the particle surface passing through the point, M r . At this point it is necessary to distinguish two regions in the molecular velocity space. In the body coordinates all the outwardly directed molecular velocity vectors lying within the cone of influence are considered to belong to the second region in the velocity space. All other molecules belong to the first velocity set, denoted as region 1. For these velocity regions, the simplest distribution function having a discontinuous structure and being capable of exhibiting a

243

Chapter 11. Boundary Value Problems for All Knudsen Numbers

smooth transition between the free-molecular and continuum regimes is the two-sided Maxwellian distribution given by: 32

f1

ª º m n1 r, t « » ¬« 2S kT1 r, t ¼»

f2

ª º m n2 r, t « » ¬« 2S kT2 r, t ¼»

§ m v u r, t 2 1 exp ¨ ¨ 2kT1 r, t ©

· ¸ , ¸ ¹

(11-3a)

and: 32

§ m v u r, t 2 2 exp ¨ ¨ 2kT2 r, t ©

· ¸ , ¸ ¹

(11-3b)

where ni r, t , Ti r, t , and ui r, t are ten unknown functions of position and time. These ‘so-called’ characteristic densities, temperatures, and mean velocities form the basic set of parameters over which all of the moments of the distribution function and, therefore, all of the macroscopic quantities for a gas, may be calculated. As one can see, the two-sided Maxwellian distribution function is a straightforward generalization of the distribution function for free-molecular flow. Such a choice for the distribution function gives a chance to close the moment system but, unfortunately, it limits the number of moment equations and consequently the accuracy of this analysis. An improvement in the accuracy can be achieved by the use of the ‘so-called’ modified two-stream Maxwellians proposed by Krook [14] and described in detail by Kogan [15]. For various regions of the velocity space these modified two-stream Maxwellians can be represented by:

fi

ª m º ni « » ¬ 2S kTi ¼

32

§ m v u 2 i exp ¨ ¨ 2kTi ©

· ¸ 1 Ak i vk Akl i vk vl , (11-4) ¸ ¹

where, i 1 for the first region and i 2 for the second. The functions ni , i i ui , Ti , Ak , and Akl depend on r and t in the general case. To facilitate the employment of the discontinuous distribution functions given by Eqs. (11-3a), (11-3b), and (11-4), one should attempt to represent these functions by analytical formulas that might be suitable for the entire velocity space [16]. For two specific geometries of interest, this type of representation will be discussed in detail. For the spherical geometry, at each space point, M r , there exists a cone of influence, the surface of which is described by the equation F F 0 , F 0 arcsin R r , where R is the radius of the particle, r Rr , and F is

244

Analytical Methods for Problems of Molecular Transport

Figure 11-1. The cone of influence in cylindrical coordinates.

the angle between vectors v and r (Fig. 7.1). In velocity space, the cone of influence divides the molecules into two groups with distribution functions f1 and f 2 ( f 2 describes the molecules for which vr ! vr , vr v cos F 0 ; f1 describes all other molecules). In the full velocity space the distribution function may be expressed as [16]:

f

1 2

f1 f 2 12 f 2 f1 sign vr vr

.

(11-5)

This is a representation of the distribution function that is very convenient to use in a moment system.

Chapter 11. Boundary Value Problems for All Knudsen Numbers

245

Now, consider the corresponding representation for the cylindrical geometry shown in Fig 11-1. While still traditionally termed a ‘cone of influence,’ the region is actually tent-shaped and is bounded by the two planes passing through the point M r and lying tangent to the surface of the cylinder. The cylinder, which nominally has a dimensional radius of R , is shown in Fig. 11-1 in dimensionless form where r r R and thus the cylinder has a dimensionless radius of unity. Let vT , vr , vz and vU ,T , vz be Cartesian and cylindrical coordinates in velocity space, respectively. Now consider an arbitrary velocity, v vT v r v z . From Fig. 11-1 one can see that v U vT v r with vr vU sin T and vT vU cos T . The velocity component, v z , which is the same in both coordinate systems, is not shown in Fig. 11-1 which is only a two-dimensional representation, but it may be envisioned as projecting normally upward from the plane of the page perpendicular to the components, vT and v r . If one defines vr vU sin D for the special case when v U and v both lie on the surface of the cone of influence, i.e. when T D arccos 1 r , then from Fig. 11-1 one can conclude that all molecules for which vr ! vr belong to the second region of the velocity space in which the angle, T , varies over the domain D d T d S D . For all of the remaining molecules, vr vr and the molecules belong to the first region for which S D T 2S D . Eliminating the angular dependence, vr is given in terms of vU and r by:

vr

vU 1 r 2

12

.

(11-6)

This analysis allows one to write an expression for the distribution function in the full velocity space in the same form as that given by Eq. (11-5) which is a form of the distribution function that allows one to find all of the moments necessary to construct the moment system. The next section contains an analytical method for the calculation of these moments for both the spherical and cylindrical geometries.

3.

MOMENTS OF DISCONTINUOUS DISTRIBUTION FUNCTIONS.

Let value Q v be any function of the molecular velocity, v . Then, the mean value of Q v , being a moment of the distribution function, can be calculated by the formula:

246

Analytical Methods for Problems of Molecular Transport

nQ v

³ Q v f1dv ³ Q v f 2 dv

1

³

2

Q v f1dv

1 2

³

Q v f 2 dv

2

³

Q v f1dv ,

(11-7)

2

where ³ i and ³ 1 2 are interpreted to mean integration over the specified region of the velocity space and the entire velocity space, respectively. For the spherical geometry, integration over velocities lying within the cone of influence proceeds according to the formula:

³

dv

2

f

2S

F0

0

0

0

2 ³ v dv ³ d E ³ sin F d F ,

(11-8)

where F 0 arcsin 1 r and r is the dimensionless radial coordinate of the point, M r . The relationships between spherical and Cartesian coordinates in velocity space are:

vr

v cos F , vT

v sin F sin E , vI

v sin F cos E .

Thus, the same integration for the cylindrical geometry proceeds according to the formula: f

³

2

dv

³ vU dvU 0

S D

f

³

f

dvz

³

dT ,

(11-9)

D

where D arccos 1 r and vr vU sin T , vT vU cos T , and vz vz . For example, one may calculate the gas number density in the fourmoment approach for which the Lees distribution function is given by:

fi

ª º m ni r , t « » «¬ 2S kTi r, t »¼

32

§ mv 2 · exp ¨ ¸¸ ; i 1; 2 , ¨ © 2kTi r, t ¹

(11-10)

Using Eq. (11-7) with Q v 1 , one obtains for the spherical geometry:

Chapter 11. Boundary Value Problems for All Knudsen Numbers

n r, t n1 r, t

n2 r, t f

1 2

0

n1 r, t

S3 2

f

0

0

2S

F0

0

0

³ c1 exp c1 dc1 ³ d E ³ sin F d F 2

2

0

n1 n2 12 n2 n1 x 12

F0

³ d E ³ sin F d F

2 2 ³ c2 exp c2 dc2

S3 2

2S

247

(11-11)

, 12

and ci m 2kTi v . The appropriate expression for where x 1 r 2 the cylindrical geometry is then given by: 1 2

n

where D

4.

n1 n2 12 n2 n1

2D

S

(11-12)

,

arccos 1 r . Other moments can be calculated in the same way.

ANALYTICAL EXPRESSIONS FOR THE BRACKET INTEGRALS.

Let this analysis be confined to a description of linearized transport problems in which the four-moment approach is used. The characteristic densities and temperatures in Eqs. (11-3a) and (11-3b), for these linearized problems, may be written as: ni r, t n0 ª¬1 Q i r, t º¼ ,

(11-13a)

Ti r, t T0 ª¬1 W i r, t º¼ .

(11-13b)

The corrections, Q i and W i , are assumed to satisfy the following conditions Q i 1 and W i 1 . The two-sided Maxwellian distribution function takes the form:

^

where:

f

`

0 f 1 Q i r, t c 2 32 W i r, t ,

fi

0

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

exp c 2 .

(11-14)

248

Analytical Methods for Problems of Molecular Transport

In the full velocity space, for the spherical and cylindrical geometries, this distribution function may be expressed, according to Eq. (11-5), as:

^

0 f 1 12 Q 12 c 2 23 W

f

ª 12 Q 12 c 2 23 W º sign cr cr ¬ ¼

(11-15)

` ,

where Q r Q 2 r Q 1 and W r W 2 r W 1 . If the specific molecular properties, 1, cr , c 2 , c 2 cr , are substituted into Eq. (11-2) instead of Q v , then four-moment equations will be produced. The right-hand-side of the fourth equation will contain the following bracket integrals:

ªsign cr cr , c 2 cr º , ¬ ¼

ªc 2 sign cr cr , c 2 cr º . ¬ ¼

Consider the first of the above bracket integrals when molecules are assumed to act as rigid spheres. For the spherical geometry, this integral has the following analytical form:

12

ªsign cr cr , c 2 cr º ¬ ¼ 2S

S

u³ sin T dT 0

f

³ 0

V 2 § kT0 · S 3 ¨© m ¸¹ 2S

S

d H ³ sin D dD

0

u ³ exp Gr2 dGr

f

f

f

f

³ dE 0

³ exp GT dGT

f

1 2

2

(11-16)

u ³ exp GI2 dGI ³ g 3 exp g 2 dg f

^

0

u 2 Gr g r2 g rc2 GT g r gT g rc gTc GI g r gI g rc gIc § u sign ¨ Gr g r ¨ ©

`

12 1 · · 2 § G g 1 ¨ 2 ¸ ¸¸ . © r ¹ ¹

Here, the relations given in Appendix A are used to specify the components of the relative velocity vector, g , before and after a collision. From Eq. (11-16) one can see that this integral depends only on the radial coordinate, r . Two limiting values of this integral may easily be found. On

249

Chapter 11. Boundary Value Problems for All Knudsen Numbers

the particle surface, when r 1 , the integrand has the usual form encountered in planar transport problems. This integral may be expressed as [16]:

ªsign cr cr , c 2 cr º ¬ ¼r

1

12

2kT0 · V2 . ¸ © m ¹ 2§ 3¨

ªsign cr , c 2 cr º ¬ ¼

In the other limit, when r o f , sign Gr g r G g 1 and, therefore, the integrand is an odd function with respect to the components of the molecular velocity, Gr , GT , GI . Consequently, in this limit one has:

lim ªsign cr cr , c 2 cr º ¼

r of ¬

0 .

For arbitrary values of r , these multiple integrations result in very cumbersome algebraic calculations. After integration with respect to the variables T , H , and g , and after introducing the new variables Gr xg , GT yg , GI zg , and cos D t , this integral becomes:

12

ªsign cr cr , c 2 cr º ¬ ¼

96V 2 § kT0 · ¨ ¸ 2S 2 © m ¹

I1> @ r , 1

(11-17)

where:

I1> @ r 1

1

2S

f

³ dt ³ d E

1

0

F t , E , y, z, x

³

f

f

f

dy ³ dz ³ F t , E , y , z , x dx , f

x0

x t 2 13 yt 1 t 2 cos E zt 1 t 2 sin E

1 x

2

2

y z

2

5

and:

x0

t ª«1 t 2 y 2 z 2 2 y 1 t 2 cos E ¬ 12

2 z 1 t 2 sin E º» ¼

r

2

12

1

.

,

250

Analytical Methods for Problems of Molecular Transport

For the cylindrical geometry the integrand is the same, but it is convenient to perform the first integration over all values of z as the lower limit of the integral with respect to x does not depend on z . In this case, the bracket integral may be written as:

12

96V 2 § kT0 · ¨ ¸ 2S 2 © m ¹

ªsign cr cr , c 2 cr º ¬ ¼

I 2> @ r , 1

(11-18)

with: 1

>[email protected]

I2 r

2S

f

f

f

³ dt ³ d E ³ dy ³ dx ³ F t , E , y, z, x dz

1

0

f

x0c

,

f

and:

x0c

12

t y 1 t 2 cos E r 2 1

.

In the same way, analogous expressions may be obtained for the second bracket integral that is contained in the right-hand-side of the heat flow transport equation. For the spherical geometry this integral may be represented as:

12

240V 2 § kT0 · ¨ ¸ 2S 2 © m ¹

ª c 2 sign cr cr , c 2 cr º ¬ ¼

2 I1> @ r ,

(11-19)

with: 2 I1> @ r

1

2S

³ dt ³ d E

1

0

f

³

f

f

f

dy ³ dz ³ F t , E , y, z , x ) t , E , y, z , x dx , f

x0

and:

) t , E , y, z, x 1 x 2 y 2 z 2 2 xt 2 y 1 t 2 cos E 2 z 1 t 2 sin E

1 x

2

y2 z2

.

251

Chapter 11. Boundary Value Problems for All Knudsen Numbers

For the cylindrical geometry, the appropriate integral is given by:

12

240V 2 § kT0 · ¨ ¸ 2S 2 © m ¹

ª c 2 sign cr cr , c 2 cr º ¬ ¼

I 2> @ r , 2

(11-20)

with: 1

I 2> @ r 2

2S

³ dt

³

1

f

dE

0

³

f

f

f

dy ³ dx ³ F t , E , y, z , x ) t , E , y, z , x dz , x0c

f

The numerical values of the functions, I i> @ r , are necessary to obtain results of practical interest for arbitrary Knudsen numbers. The application of these functions will be more convenient if one introduces some special functions depending only on a radial coordinate by some normalization procedure. Using the values of the bracket integrals for the planar geometry [16], one can introduce the following notation for these functions: j

ªsign cr cr , c 2 cr º ¬ ¼

12

2kT0 · V 2 I i >[email protected] r , ¸ © m ¹

ª c 2 sign cr cr , c 2 cr º ¬ ¼

2§ 3¨

12

2kT0 · 2 > [email protected] ¸ V Ii r , © m ¹

23 14 S §¨

(11-21)

(11-22)

j where the normalized functions, I i > @ r , may be written as:

Ii > @ r

1

72 >[email protected] I r , 2 i

S

and:

2 Ii > @ r

1440 2 I i> @ r , S 8 3S 2

and in which the index, i , assumes the value of i 1 for the spherical geometry and assumes the value i 2 for the cylindrical geometry. Now,

j the functions, I i > @ r , have the common range, [1,0], for values of r from 1 to f . These special functions allow one to represent the solutions to many boundary value transport problems in analytical forms.

252

Analytical Methods for Problems of Molecular Transport

As one can see, the calculational procedure for the bracket integrals presents the main difficulty in the application of Lees’s method. The aforementioned difficulty typically results in the use of various approximate methods which will be considered later in Sections 11.7 and 11.8. Another method for simplifying the analysis is based on a specific form of the moment system which has been generated using some general properties of the Chapman-Enskog solution (see Section 11.10). One method that may be used to alleviate the need for numerical values of the special functions described above has been discussed for cylindrical and spherical geometries in [17,18] and in Appendix B. Since it is impossible to obtain these integrals

j in analytical forms, numerical values of the special functions, I i > @ r , would be very useful for many boundary value transport problems. Some of these numerical values have been given in Tables B-1 and B-2.

5.

BOUNDARY CONDITIONS FOR MOMENT EQUATIONS.

In a general case the microscopic boundary conditions on the body surface can be formulated for reflected molecular distribution function by means of the dispersion operator, A , that may be expressed by the relation:

f v, rS Af v, rS ,

(11-23)

where rS is the position radius vector for a given point on the particle surface. For the arbitrary boundary model the approximating functions given by Eqs. (11-3a), (11-3b) and (11-4) do not coincide with the distribution function of reflected molecules, expressed by Eq. (11-23), for any values of the characteristic parameters that are contained in them. In that case the boundary condition may, on the average, be satisfied only approximately if one uses the integral boundary model which may be written as: ³ v n Q v f dv

vn !0

³ v n Q v A f dv

,

(11-24)

vn ! 0

where n is a unit normal vector to the surface directed into the gas and Q v usually represents the molecular properties used in the moment equations. For various Q v , this relation yields the necessary number of boundary conditions for the moments.

Chapter 11. Boundary Value Problems for All Knudsen Numbers

253

A question arises concerning what molecular properties must be used for a given problem on each of the sections of the boundary. There are some uncertainties connected with this choice. Nevertheless, a general recommendation might be useful to facilitate this choice. Both on inner and outer surfaces it is convenient to use the integral form of the boundary conditions constructed on the integral definition of the accommodation coefficients. For heat and tangential momentum transport problems, the following integral boundary conditions should be used: Qn rS DT ªQn rS Qn rS ¬«

PW rS DW ª PW rS PW rS «¬

º ,

» eq ¼

º .

eq » ¼

(11-25a)

(11-25b)

The second relation is the generalized form of Eq. (6-7) in which ( PW (rS ))eq is the tangential momentum that would be carried away by reflected molecules if the tangential momentum accommodation coefficient were equal to unity. Note that, for stationary surfaces, ( PW (rS ))eq 0 . Another way of constructing boundary conditions [4] may be used for transport problems for which the approximating distribution function given by Eqs. (11-3a), (11-3b) and (11-4) is adapted to the Maxwellian boundary model (for example, the single-sphere heat transport, drag and torque problems). If the two-sided Maxwellian distribution function is adapted to the boundary model, the boundary conditions may be satisfied exactly by choosing certain characteristic parameters. To simplify this analysis, one can use the general expression for the reflected distribution function (Problem 6.5) which can be written in the form:

) c, rS

1 DW ) cc, rS

2G · ½ 2G (11-26a) §G DW 2 c 2 ®DTW w 1 DT ¨ 2 1 ¸ ¾ DW 1 , S ¹¿ S ©S ¯

where ) r is a measure of the perturbation in the distribution function from an absolute Maxwellian, cc c 2n c n , W w is a correction to the temperature at the surface point, rS , and G1 and G 2 are given by:

G1

³

cn 0

) c, rS c n exp c 2 dc ,

(11-26b)

254

Analytical Methods for Problems of Molecular Transport

G2

³

) c, rS c 2 c n exp c 2 dc .

cn 0

(11-26c)

Now, equating terms in Eq. (11-26a) with the same molecular velocity factors, one can obtain the boundary conditions necessary to determine the parameters in the two-sided Maxwellian distribution function associated with a given boundary value problem.

6.

THERMAL CONDUCTION FROM A HEATED SPHERE.

In this section, the main features of the transition between gas-kinetics and gas-dynamics are analyzed by means of the four-moment approach, utilizing the two-sided Maxwellian distribution function to close the moment system. This approach is illustrated by considering, for arbitrary Knudsen number, the steady heat conduction problem from a heated sphere. Consider the following brief statement of this problem [18]. Let a sphere of radius, R , be located in an infinite expanse of a gas which, far from the sphere, has a temperature, T0 , and a number density, n0 . The temperature, T0 'T , of the sphere surface is supposed to be slightly different from that of the surrounding gas 'T T0 . For this assumption the gas is near to the equilibrium state, and, therefore, this problem may be linearized be means of using the small parameters that are proportional to 'T T0 . The problem being considered possesses radial symmetry, and, therefore, the moment equation for transport of any molecular property, Q v , is:

1 w 2 r ³ fvr Q v dv r 2 wr wQ v wQ v 1 ° ³ ® ª¬ vT2 vI2 º¼ ª¬vI2 cot T vT vr º¼ wvr wvT r °¯ wQ v °½ ¬ªvI vr vT vI cot T ¼º ¾ fdv Rn'Q v . wvI °¿

(11-27)

Alternately selecting Q v to be the collisional invariants, 1, vr , and v 2 , for which 'Q v 0 , one obtains the ordinary continuity, radial momentum, and energy equations. Since the primary interest is in the radial heat flux, one also takes Q v vr v 2 which yields a fourth moment equation involving this heat flux [2].

Chapter 11. Boundary Value Problems for All Knudsen Numbers

255

For the four-moment approach the linearized distribution function may be used in the form given by Eq. (11-15). From Eq. (11-27), the following system of moment equations may be derived:

w 2 r nur wr

(11-28a)

0 ,

w 1 Prr 2 Prr PTT PII r wr

w 2 r Qr wr

(11-28b)

0 ,

0 ,

(11-28c)

1 w 2 2 2 1 r v vr fdv ³ v 2 vT2 vI2 fdv ³ 2 r r wr

Rn' v 2 vr .

(11-28d)

All moments of this system can be calculated by means of the method described in Section 11.3. These moments are given by the relations:

n r n0 1 12 Q 12 Q x ,

Pr r

12

§ 2kT0 · ¸ 2 ¨ 2 Sr © m ¹ 1

ur r

Q

12 W ,

p0 1 12 Q 12 Q x 3 12 W 12 W x3 ,

pr

p0 1 12 Q 12 Q x 12 W 12 W x ,

12

Qr r

p0 § 2kT0 · ¨ ¸ S r2 © m ¹

Q

32 W ,

256

Analytical Methods for Problems of Molecular Transport 2 2

³ c cr fdv

5 4

n0

1

1 2

Q 12 Q x3 W W x3 ,

and: 2 2 2 ³ c cT cI f dv

n0

5 2

1

1 2

Q 34 Q x W 32 W x 14 Q x3 12 W x3 ,

12

d 1 Q 2W dr

where x 1 r 2 . The moment system given by Eqs. (11-28) for the parameters of the distribution function can be expressed as:

d Q W dr

0 ,

(11-29b)

0 ,

d 3 Q 2W dr

0 ,

d Q 2W dr

8 5

(11-29a)

(11-29c)

12

R§ m · ¨ ¸ n0 © 2kT0 ¹

n' c 2 cr ,

(11-29d)

where the moment of the collision operator, n'c 2 cr , is determined by:

n' c 2 cr

n02

^ Q 1 2

32 W ªsign cr cr , cr c 2 º ¬ ¼

`

12 W ªc 2 sign cr cr , cr c 2 º . ¬ ¼ Now, consider the boundary conditions for these moment equations. Since the distribution function that has been employed is well suited to the Maxwellian boundary model, the ordinary microscopic boundary condition on the sphere surface can be used to yield:

Q 2 1 c 2 32 W 2 1

1 DW ª¬Q 1 1 c 2 32 W1 1 º¼ DW ª¬Q r c 2 32 W r º¼

.

(11-30a)

Chapter 11. Boundary Value Problems for All Knudsen Numbers

257

This expression contains two unknown quantities, Q r and W r , that may be found from the two additional conditions:

Nr

n0ur

0 ,

DT

Qr Qr , Qrw Qr

(11-30b)

(11-30c)

where the first relationship expresses the fact that the sphere is impenetrable to the gas molecules and the second one is the thermal accommodation condition. At large distances from the sphere the number density and the temperature tend to the equilibrium values that can be written as:

Q f Q f 0 ,

(11-30d)

W f W f 0 .

(11-30e)

After the integration of Eqs. (11-29a)-(11-29d), one can obtain:

Q

12 A3 ,

(11-31a)

W

A3 ,

(11-31b)

Q

A3[ r 2 A2 A4 ,

(11-31c)

W

A3[ r A4 A2 ,

(11-31d)

where A2 , A3 , and A4 A1 0 are the arbitrary constants of integration and a function, [ r , may be expressed as:

[ r

r ª r >[email protected] º

2 c c Kn «D1 ³ I1 r dr D 2 ³ I1 > @ r c dr c» . «¬ 1 »¼ 1 1

(11-31e)

258

Analytical Methods for Problems of Molecular Transport

In this expression, the following notations were introduced:

D1

8 3 S

8 2 , D2 15S

2

,

30S

and the functions I1 > @ r and I1 > @ r are determined by Eqs. (11-21) and (11-22). Using the boundary conditions given by Eqs. (11-30a)-(11-30e) and (1131a)-(11-31e) and performing some simple transformations, one obtains the following expression for the heat flux through the sphere surface:

1

Q Q fm

2

1 , 1 ] Kn 1

(11-32a)

where: 12

n kT § 2kT0 · 'T 4S R 0 0 ¨ , DW DT ¸ T0 S © m ¹ 2

Q fm

(11-32b)

is the free-molecular heat flux and the quantity, ] , is given by:

]

1 2

ª

f

f

º

«¬

1

1

»¼

DW DT «D1 ³ I1 >[email protected] r dr D 2 ³ I1 > [email protected] r dr » ,

(11-33)

in which DW and DT are the accommodation coefficients. This analytical expression gives the exact result for the free-molecular regime when Kn o f . For another limiting regime, when Kn o 0 , Eq. (11-32a) may be written as:

Q

4S RN'T ,

where N is the thermal conductivity coefficient that, for the given approach, can be expressed in the form: 12

N

2 § 2kT0 · ¨ ¸ S© m ¹

f ª f 1 º

2 n0 k O «D1 ³ I1 > @ r dr D 2 ³ I1 > @ r dr » «¬ 1 »¼ 1

1

.

259

Chapter 11. Boundary Value Problems for All Knudsen Numbers

If Kn o 0 , this expression for the thermal conductivity coefficient must tend to that obtained by the same approach for the planar geometry [16]. Using results from [16], one obtains: f

f

1

1

D1 ³ I1 >[email protected] r dr D 2 ³ I1 > [email protected] r dr D1 D 2 . This relationship may be considered a general property of the special functions. Having employed this property, one can express Eq. (11-33) in the form:

]

1 2

DW DT D1 D 2

0.2508 DW DT

(11-34)

.

The results obtained here will be discussed later in detail together with other results that will be derived by the use of the various approximate methods.

7.

METHOD OF THE ‘SMOOTHED’ DISTRIBUTION FUNCTION.

The absence of analytical expressions for the bracket integrals containing discontinuous Lees’ distribution functions prevents one from obtaining solutions for boundary value transport problems, with the exception of cases involving the planar geometry and Maxwellian molecules. Some approximate results can be obtained by utilizing a new procedure proposed in [19] that allows one to overcome these difficulties. A continuous distribution function of the Chapman-Enskog type is assumed to be employed in the collision integral instead of the actual discontinuous function. This auxiliary function may be called a ‘smoothed’ distribution function. The unknown parameters of the ‘smoothed’ distribution function may be determined by equating its basic moments with those obtained from the actual distribution function. Employing this method to solve the problem investigated in Sec. 11.6, one assumes that the ‘smoothed’ distribution function used in the collision operator has the following form:

f

^

` , (11-35)

1 0 f 1 Q r c 2 32 W r 2cr ur r A r cr S3 2 c 2

where Q r and W r are corrections to the number density and the temperature. The quantity, A r , is determined by:

260

Analytical Methods for Problems of Molecular Transport

³ cr c

2

fdc

³ cr c

2

f dc ,

(11-36)

where, f , is given by Eq. (11-15). All unknown parameters in Eq. (11-35) may be expressed through the basic functions Q r r and W r r . This ‘smoothed’ function allows one to obtain analytical results for arbitrary molecular potential interactions. The standard bracket integrals, which are tabulated for many molecular potential models, may be utilized for these analyses. For example, the expression, n 'c 2 cr , from the moment system, Eqs. (11-29d), may be written as:

n 'c 2 cr

1 1 n02 A r ª cr S3 2 c 2 , cr S3 2 c 2 º . ¬ ¼

Using this relation in the right-hand-side of Eq. (11-29d), one can obtain (for rigid-sphere molecules) the following final formula for the heat flux:

Q Q fm

1 , 1 ] Kn 1

(11-37)

where ] DW DT 64 S 1 . 75 As one can see, this formula gives the exact expression for the heat flux in a continuum regime when Kn o 0 . This method allows one to simplify the calculation of the bracket integrals. Moreover, there are no difficulties in generalizing this analysis for arbitrary models of the intermolecular potential.

8.

THE POLYNOMIAL EXPANSION METHOD.

The above mentioned difficulties of calculating the bracket integrals in the analytical form can be overcome by means of the use of a special representation for the function, sign cr cr . In this section a method will be presented to calculate the bracket integrals in the analytical form that consists of an expansion of the discontinuous function, sign cr cr , in velocity space polynomials. First, consider the problem in a spherical geometry. In Cartesian coordinates, this function can be written as:

sign cr c cos F 0 sign §¨ cr cr2 cT2 cI2 1 r 2 ©

12

· . ¸ ¹

(11-38)

It is easy to see that this function depends upon the three velocity coordinates and on the radial coordinate of the position of a molecule.

Chapter 11. Boundary Value Problems for All Knudsen Numbers

261

Owing to this notation, one must utilize a three-dimensional expansion of the velocity components in Hermite polynomials. This expansion was first proposed by Ivchenko [16]. But it is necessary to note that the expansion given in [16] is very difficult to be realized for practical calculations as there are many difficulties connected with the use of general schemes for the three-dimensional expansions. Since the function given by Eq. (11-38) depends only on the sum of the squares of the independent variables, cT and cI , one could hope to find a more simple scheme in which the sum, cT2 cI2 9 2 , is considered as a new independent variable. As 0 d 9 f , Hermite polynomials cannot be employed for this expansion. Therefore, Sonine polynomials are used instead. A polar coordinate system is introduced for the above velocity components by means of the relations cT 9 cos E , cI 9 sin E , and dcT dcI 9 d 9 d E . Integration with respect to these variables is determined by: f f

³³

exp cT2 cI2

f f

F 9 2 dcT dcI

f

2S ³ exp 9 2 F 9 2 9 d9 0

f

S ³ F x exp x dx , 0

and therefore the integrand contains the weight function, x m 1 . This p means that m 0 and the Sonine polynomials, S0 9 2 , must used in the expansion. The orthonormal conditions for this kind of the Sonine polynomials can be expressed as:

f

p q ³ exp x S0 x S0 x dx

G pq .

(11-39)

0

This polynomial can be defined by the expression: n

S0

x p n x ¦ 2 p 0 p n p n

,

where: 0 1 2 S0 x 1 , S0 x 1 x , S0 x 1 2 x 12 x 2 , and:

(11-40)

262

Analytical Methods for Problems of Molecular Transport 3 S0 x 1 3x 32 x 2 16 x 3 .

Taking into account this analysis, one can use the following twodimensional expansion:

sign §¨ cr cr2 cT2 cI2 1 r 2 ©

12

· ¸ ¹

(11-41)

¦¦ Akn r H k cr S0 cT2 cI2 . n

k

n

Since this expansion contains only the odd Hermite polynomials, the coefficients, Akn r , are given by: f

4

Akn r

2k k S

³c 0

2

exp c 2 dc (11-42)

1

n

u ³ H k c t S0 t0

c

2

2

ª1 t º dt , ¬ ¼

12

. where t0 1 r 2 Now, consider the features of this expansion for the cylindrical geometry. As one can see from Eq. (11-6), for this case, the quantity, cr , depends only on the two molecular velocity components, cr and cT , and, therefore, the appropriate expansion is defined by:

sign §¨ cr cr2 cT2 1 r 2 ©

12

· ¸ ¹

¦¦ Aij r H i cr H j cT i

, (11-43)

j

where the coefficients, Aij r , in this expansion can be expressed as: f

Aij r

1 c9 exp c92 dc9 i j 2 ijS ³0

2S S D ° °½ u ® ³ H i cr H j cT dT 2 ³ H i cr H j cT dT ¾ , °¯ 0 °¿ D

where cT c9 cos T and cr c9 sin T . In the planar geometry, one has a simple one-dimensional expansion given by:

263

Chapter 11. Boundary Value Problems for All Knudsen Numbers

sign cx

f

¦ Ai H i cx

(11-44)

,

i 0

where the constant coefficients can be determined by:

Ai

1 i 2 i S

f

³ exp cx H i cx sign cx dcx 2

.

f

This expansion was first proposed by Savkov in [20]. Now, consider some of the features of this method associated with the spherical geometry. If the expansion given by Eq. (11-41) is used in the bracket integrals presented by Eqs. (11-21) and (11-22), one can obtain the

i following analytical expressions for the special functions, I1 > @ r : >@ I1 > @ r 32 Z 1 ¦¦ Akn r D kn ,

1

1

k

2 I1 > @ r

(11-45)

n

12 > [email protected] Z 1 ¦¦ Akn r D kn . 8 3S k n

(11-46)

>@ > @ Here, the constants D kn and D kn are the common bracket integrals: 1

2

>[email protected] D kn

ª H k cr S0 n cT2 cI2 , cr c 2 º , ¬ ¼

(11-47a)

> [email protected] D kn

ª H k cr c 2 S0 n cT2 cI2 , cr c 2 º . ¬ ¼

(11-47b)

Instead of the infinite sums in Eqs. (11-45) and (11-46) one assumes that finite sums can be employed. It is not difficult to determine that the convergence of these series is very rapid. Indeed, if three terms n >[email protected] 0 ; k 1,3,5 are used, one will find that I1 >[email protected] 1 1.0102 and I1 1 1.1198 . The relative errors of such an approach are about 1% and 12% r 1 for the first and second functions respectively. When n 0 ; 1 d k d 7 , the relative errors are less than 1% r 1 for both functions. All of the bracket integrals necessary for this analysis can be found in Table 11-1. For arbitrary r , the other terms for which n z 0 must be considered. To

i obtain some appropriate expressions for the special functions, I1 > @ r , the

264

Analytical Methods for Problems of Molecular Transport

Table 11-1. Expressions for some bracket integrals.

ª cri , cr c 2 º ¬ ¼

i a †

a 2S Z

ª cri c 2 , cr c 2 º ¬ ¼

†

a 2S Z

3

5

7

1

3

5

7

4 5

30 7

287 12

4 3

6

123 4

741 4

12

Z V 2 2kT m

following finite sums, for which n 0 ; k 1,3,5,7 and k 1; n 1, 2,3 , are employed. Moreover, to obtain the more accurate expressions for

2 I1 > @ r , an additional term k 3 ; n 1 has been included in the expansion given by Eq. (11-46). For both expansions, the exact values of the bracket integrals are given in Appendix C and the expression > [email protected] D 31 925 2S Z has been employed. Having substituted the bracket integrals into Eqs. (11-45) and (11-46), one can obtain:

I1 > @ r

53 6 53 4 2 ª¬ 105 r 84 r 163 r 2 º¼ , 280

(11-48)

2 I1 > @ r

12 2 95 4 ª 19 r 6 84 r 17 r 2 º¼ . 21 8 3S ¬ 21

(11-49)

1

These analytical expressions can be readily integrated, after which, one obtains: f

>@ ³ I1 r dr

1

f

0.9779

1

,

> [email protected] ³ I1 r dr 0.9793 .

(11-50)

1

It is necessary to note, however, that the accuracy of the approach proposed here may only be specified by a comparison of these analytical results with those obtained numerically. The deviations of the integrals given in Eq. (1150) from the corresponding numerical values are approximately 2% for both of these integrals. This very close agreement with the numerical results indicates a rapid convergence of the expansions being used. Therefore, one can see that the approach proposed here is a reasonable approximate method for solving boundary value transport problems. For the problem of thermal conduction from a heated sphere, the method of polynomial expansion yields:

Q Q fm

1 , 1 ] Kn 1

(11-51a)

265

Chapter 11. Boundary Value Problems for All Knudsen Numbers Table 11-2. Values of the reduced heat flux ratio,

Q Q fm

DW

DT

1 .

Q Q fm

Kn 1 0.00 0.25 0.50 0.75 1.00 2.00 3.00 4.00 5.00 6.00 8.00 10.0 12.0 14.0 16.0 18.0 20.0

Exact Moment Solution

Smoothed Distribution Function Method

Method of Polynomial Expansion

1.0000 0.9410 0.8886 0.8417 0.7995 0.6660 0.5706 0.4992 0.4437 0.3992 0.3326 0.2851 0.2494 0.2217 0.1995 0.1813 0.1662

1.0000 0.9364 0.8804 0.8308 0.7864 0.6480 0.5510 0.4793 0.4241 0.3803 0.3152 0.2691 0.2348 0.2082 0.1871 0.1698 0.1555

1.0000 0.9422 0.8907 0.8445 0.8029 0.6707 0.5759 0.5045 0.4489 0.4044 0.3374 0.2894 0.2534 0.2254 0.2029 0.1845 0.1692

]

0.2508

]

0.2716

]

0.2455

where, for rigid-sphere molecules:

]

0.2455 DW DT

.

(11-51b)

The reduced heat flux given by Eqs. (11-51a) and (11-51b) deviates only slightly from the solution given by Eq. (11-34). Table 11-2 gives values of the reduced heat flux ratio for DW DT 1 and Fig. 11-2 illustrates these values for the various methods employed here.

9.

SOLUTION OF ONE CLASSIC TRANSPORT PROBLEM.

In this section the accuracy of the various methods previously proposed will be analyzed. A comparison is made between the various approximate analyses and the exact moment solution for a classical problem of heat transport between two parallel plates with the temperature difference being 2'T . Let the temperature of the lower plate x L 2 be T T0 'T and that of the upper one x L 2 be T T0 'T , where T0 is the temperature

266

Analytical Methods for Problems of Molecular Transport

Reduced Heat Flux (dimensionless)

1.0

0.8

0.6

0.4

0.2

0.0 0.0

5.0

10.0

15.0

20.0

Inverse Knudsen Number

Figure 11-2. A comparison of reduced heat flux values, Q Q fm , obtained from various approximate methods assuming (DW D T 1) with the Exact Moment Solution given in Table 11-2. The Exact Moment Solution is represented as a solid line while values obtained using the Smoothed Distribution Function method are shown as crosses and values obtained using the Polynomial Expansion Method are shown as open circles.

at x 0 . The temperature difference is assumed to be small compared with T0 , and therefore, the linearized theory is used in this problem. This problem has been previously formulated and studied by WangChang and Uhlenbeck [21], Jackson [22], Gross and Ziering [23], Bassanini, Cercignani and Pagani [24], Loyalka [25], and Ivchenko [16]. Moreover, the heat conduction characteristics and density distribution were experimentally determined by Teagan and Springer [26]. The solution of this planar problem may be based on the Maxwellian integral transport equations which can be directly derived from the Boltzmann equation by means of the relation:

w vx Q v fdv wx ³

Gf

³ Q v G t

dv ,

(11-52)

Chapter 11. Boundary Value Problems for All Knudsen Numbers

267

where x is the normal coordinate to the plate. The corresponding moment equations can be obtained by setting Q v equal to 1, vx , v 2 , and vx v 2 in Eq. (11-52). These equations can be written as:

d nu 0 , dx

(11-53a)

d Pxx dx

0 ,

(11-53b)

d Qx dx

0 ,

(11-53c)

d vx2 v 2 fdv dx ³

n' v x v 2 .

(11-53d)

In order to close this system of moment equations, the distribution function is chosen to be of the form:

f

^

0 f 1 12 ªQ x c 2 23 W x º ¬ ¼

`

12 ªQ x c 2 23 W x º sign cx . ¬ ¼

(11-54)

All the moments of the distribution function in Eqs. (11-53a)-(11-53d) may be calculated from the expressions given in Sec. 11.6 by setting r 1 . The solution of the moment system then contains four arbitrary constants of integration which can be specified by boundary conditions. For this problem the Maxwellian boundary model is employed in which the tangential accommodation coefficient is assumed to be equal to unity. Utilizing the values of the bracket integrals which are given in Table 111, one obtains the exact analytical solution via a four-moment approach. The heat flux may then be written as [16]:

ª º DT < Kn 1 » Q Q fm «1 ¬ 2 DT ¼

1

,

@ r . Solution: The moment system for this cylindrical boundary value problem may be written as: d rnur 0 , dr d 1 Prr Prr PTT 0 , dr r

d rQr 0 , dr

12

1 d 1 r ³ fcr2 c 2 dv ³ fcT2 c 2 dv r dr r

§ m · R1 ¨ ¸ © 2kT0 ¹

n 'cr c 2 .

For the distribution function given in Eq. (11-5), the moments connected with this system may be calculated by means of the technique described in Section 11.3. For example, the number density, mean velocity, and heat flux are given by:

2D · § n r n0 ¨ 1 12 Q 12 Q , S ¸¹ ©

12

1 § 2kT0 · ¨ ¸ r S © m ¹

ur

1 2

Qr

p0 § 2kT0 · ¨ ¸ r S © m ¹

12

Q

Q

12 W ,

32 W ,

280

Analytical Methods for Problems of Molecular Transport

where D arccos r 1 . After some transformations, the moment system takes the same form as that given by Eqs. (11-29a)-(11-29d). Therefore, one can use the general solution described by Eqs. (11-31a)-(11-31d). The function, [ r , is given by: r r >1@ ½°

2 c c Kn ®D1 ³ I 2 r dr D 2 ³ I 2> @ r c dr c¾ . °¯ 1 °¿ 1 1 °

[ r

The boundary conditions may be written as n 1 n0 and:

DT Qrreq Qr# for r 1 and r

Qr

z

R2 . R1

The equilibrium number densities associated with Qrreq may be found from the impenetrability condition. This statement gives: 12

§ 2kT0 · 2 S R1 p0 ¨ ¸ © m ¹

Q

] z

1 2

DW DT 'T , 1 T0 1 z 1 DW DT Kn 1] z

°

z

z

°½

°¯

1

1

°¿

DW DT ®D1 ³ I 2 >1@ r dr D 2 ³ I 2 > 2@ r dr ¾ ,

where Q is the heat flux per unit length of the inner cylinder. 11.4. Using the Chapman-Enskog solution, Ar , as a molecular property in the moment system, determine the heat flux between two coaxial cylinders the radii and temperatures of which are R1 , T0 and R2 , T0 'T for the inner and outer cylinders, respectively. Assume 'T T0 . The number density of the gas is n0 at r R1 . Solution: The moment system for this problem is: d 1 Q 2W dr

d Q W dr

0 ,

0 ,

Chapter 11. Boundary Value Problems for All Knudsen Numbers

d 3 Q 2W dr

281

0 ,

12

d W dr

8 5

R1n0 § m · ¨ ¸ a1 © 2kT0 ¹

> Ar , ) @

,

where Ar is the Chapman-Enskog solution for thermal conductivity and ) is the correction to the distribution function. The basic function, [ r , is discussed in Section 11.6 and, in this case, may be written as:

[ r

8 5

R1 1 2 S ln r . a1

It is very important to note that this relation is applicable for arbitrary models of the intermolecular potential. For rigid-sphere molecules this expression becomes [ r 128 S 1 Kn 1 ln r . The heat flux per unit 75 length of the inner cylinder is then given by: 12

Q

§ 2kT0 · 2 S R1 ¨ ¸ © m ¹

p0

DW DT 'T , T0 1 z 1 1 DW DT Kn 1] z

where:

] z

64 75

S 1DW DT ln z and z

R2 . R1

11.5. Determine the general expression for the torque on one of two concentric spheres if the inner sphere of radius, R1 , is slowly rotating with an angular speed, Z , such that Z R1 (2kT m)1 2 and the outer sphere of radius, R2 , is stationary. Assume that molecules are rigid spheres. From this general expression, derive the torque for the specific limiting case when h R2 R1 R1 . Solution: To solve this problem, one can use the same distribution function as was obtained in Section 11.11. The corrections, G r r , are given by:

G r r ª A2 A1 43 ] 1 r 3 º and G r 2 S A1r . ¬ ¼

282

Analytical Methods for Problems of Molecular Transport

The constants of integration, A1 and A2 , may be specified from two boundary conditions at the sphere surfaces. It is convenient to use the integral form of the boundary conditions that are expressed by:

PrI

DW ª Pr#I PrrI

¬«

º

» eq ¼

;

r 1, z .

These relations yield:

G 1

1 2

ª§ 2m ·1 2

º

DW «¨ ¸ Z R1 G 1 G 1 » , kT ¹ ¬«© ¼»

z 4G z DW ª 12 G z G z z 4G z º . ¬ ¼ The torque on the inner sphere is found to be:

ª «1 DW « ¬

Kx

K x

K x

23 DW SUVR14Z

8 15

1

S Kn

1

z

3

1 z º » z 4 1 DW » ¼

1

where:

z4 , z 4 1 DW

12

V 8kT S m is the mean speed of the gas molecules, Kn 1 R1 O , z R2 R1 , and the x -axis is in the direction of the vector, Ȧ . In the limiting case when h R2 R1 R1 , the torque is given by: Kx

ª DW 8 1 h º S K1 x «1 » 5 O¼ ¬ 2 DW

K1 x

DW 2 SUVR14Z . 3 2 DW

1

with:

283

Chapter 11. Boundary Value Problems for All Knudsen Numbers

11.6. Determine the general expression for the torque on one of two coaxial cylinders if the inner cylinder of radius, R1 , is slowly rotating with an angular speed, Z , such that Z R1 ( 2kT m )1 2 and the outer cylinder of radius, R2 , is stationary. Assume that molecules are rigid spheres. From this general expression, derive the torque for the specific limiting case when h R2 R1 R1 . Solution: Using Q v cT and Q v cr cT B c 2 as the molecular properties in Eq. (11-2), one can obtain:

1 d 1 rncr cT ncr cT r dr r

1 d rncr2 cT B c 2 r dr

0 ,

1r nc B c 1r nc c B c

3

2

2 r T

T

12

§ m · R1 ¨ ¸ © 2kT ¹

2

n' cr cT B c 2 .

where r r R1 . For closing this moment system the two-sided Maxwellian 0 distribution function may be chosen in the form f i f 1 2cT Gi r . This choice yields:

n

ncr cT

2 Sr

ncr2 cT B c 2

ncT3 B c 2

3

G r ,

x ª º nb1 « 14 G r 12 S 1D G r 12 S 1G r 2r 2 1 » , r ¬ ¼

D x ª º nb1 « 34 G r 32 G r 12 S 1G r 2r 2 3 » , r S ¬ ¼

where D arccos r 1 system is found to be:

and x

12

1 r 2

. The solution of the moment

G r 2 S A1r , and: G r

A2 r A1r ª 2] 1 r 2 º , where: ¬ ¼

284

Analytical Methods for Problems of Molecular Transport 12

]

R1 § m · ¨ ¸ b1 © 2kT ¹

.

The constants, A1 and A2 , may be specified from the boundary conditions which are given by: PrT 1 DW ª PrT 1 PrT ¬«

º and P z D ª P z P rT W « rT rT ¬

» eq ¼

º.

» eq ¼

The torque per unit length of the inner cylinder may be expressed as:

Kx

ª z2 1 z º

« 1 4 1 » K x 1 DW 5 S Kn 3 « z 1 DW » ¬ ¼

K x

12 DW SUVR13Z

1

, where:

z3 , z 3 1 DW

12

V 8kT S m is the mean speed of the gas molecules, Kn 1 R1 O , z R2 R1 , and the x -axis is in the direction of the vector, Ȧ . In the limiting case when h R2 R1 R1 , the torque is given by: Kx

K1 x

K1 x

ª DW 8 1 h º S «1 » 5 O¼ ¬ 2 DW

1

,

DW 1 SUVR13Z . 2 DW 2

11.7. Generalize the sphere torque problem described in Section 11.11 for arbitrary models of the intermolecular potential. Derive the reduced torque for the Lennard-Jones (6-12) potential. Consider two different gases (N2 and CO2) when T 293 K. The quantity, ] , depending upon the model of the Solution: intermolecular potential, may be expressed in the form:

]

>] @RS V

2

: 2,2 å T ,

Chapter 11. Boundary Value Problems for All Knudsen Numbers

285

where the first-order Chapman-Enskog approximation for b1 is utilized, V V V RS , and T kT H is the reduced temperature ( V and H k are parameters of the potential function). In terms of the : -integral, the torque on the sphere is given by:

ª «1 DW ¬

Kn 1

Kx

K x

K x

23 DW SU vR 4Z ,

8 15

S

V

2

:

2,2 å

º T » ¼

1

,

where the Knudsen number is calculated for rigid-sphere molecules. For the specific condition of T 293 K, the reduced temperatures and : -integrals of the gases under consideration are: TN 2

TCO 2

293 91.5

293 190

3.2022 ;

: 2,2 å 3.2022 1.0218 ,

1.5491 ;

: 2,2 å 1.5491 1.2988 .

The reduced torque for the Lennard-Jones (6-12) potential can be expressed in the form:

Kx K x

ª1 DW< Kn 1 º ¬ ¼

1

,

where < N 2 0.1672 and < CO 2 0.1642 . For comparison, it should be noted that < RS 0.1698 for rigid-sphere molecules. 11.8. Determine the domains of applicability of the free-molecular and continuum formulas for the heat flux between two coaxial cylinders. Assume that the limiting formulas have to be correct to within about 5%. Consider the case when DW DT 1 . Solution: The heat flux expression derived in Problem 11.4 and the limiting expressions derived from it should be combined in the following manner for the free-molecular and continuum regimes:

Q 1 Q 1 . and 1 Q fm 1 Kn ] z Qc 1 Kn ª] z º 1 ¬ ¼

286

Analytical Methods for Problems of Molecular Transport

For ] z 64 S 1DW DT ln z , these free-molecular and continuum formulas 75 are valid to within about 5% if the following inequalities are satisfied:

Kn t 5.43249 ln z and Kn d 0.01358 ln z where Kn O R1 and z R2 R1 . 11.9. Determine the domains of applicability of the free-molecular and continuum formulas for the heat flux between two concentric spheres. Assume that the limiting formulas have to be correct to within about 5%. Consider the case when DW DT 1 . Solution: The heat flux expression derived in Problem 11.2 and the limiting expressions derived from it should be combined in the following manner for the free-molecular and continuum regimes:

Q 1 Q 1 and . Q fm 1 Kn 1] z Qc 1 Kn ª] z º 1 ¬ ¼

S 1DW DT z 1 z , these free-molecular and continuum For ] z 64 75 formulas are valid to within about 5% if the following inequalities are satisfied:

Kn t 5.43249 1 z 1 and Kn d 0.01358 1 z 1

where Kn O R1 and z R2 R1 .

REFERENCES 1. Kogan, M.N., “Recent Developments in the Kinetic Theory of Gases,” in Rarefied Gas Dynamics 6th Symposium 1, 1-39, Academic Press Inc., N.Y. (1969). 2. Lees, L., “Kinetic Theory Description of Rarefied Gas Flow,” J. Soc. Indust. Appl. Math. 13(1), 278-311 (1965). 3. Lees, L. and Liu, C.Y., “Kinetic Theory Description of Conductive Heat Transfer from a Fine Wire,” Phys. Fluids 5(10), 1137-1148 (1962). 4. Grad, H., “On the Kinetic Theory of Rarefied Gases,” Comm. Pure and Appl. Math. 2(4), 311-407 (1949). 5. Grad, H., “Principles of the Kinetic Theory of Gases,” in Handbuch der Physik, vol. 12, ed. Flügge, S. (Springer Verlag, Berlin, 1958). 6. Grad, H., “Asymptotic Theory of the Boltzmann Equation,” Phys. Fluids 6, 147 (1963). 7. Kumar, K., “Polynomial Expansions in Kinetic Theory of Gases,” Ann. Phys. 37(2), 113141 (1966).

Chapter 11. Boundary Value Problems for All Knudsen Numbers

287

8. Muckenfuss, C., “Some Aspects of Shock Structure According the Bimodal Model,” Phys. Fluids 5(11), 1325-1336 (1962). 9. Salwen, H., Grosch, C.E. and Ziering, S., “Extension of the Mott-Smith Method for a Onedimensional Shock Wave,” Phys. Fluids 7(2), 180-189 (1964). 10. Hurlbut, F.C., “Note on Conductive Heat Transfer from a Fine Wire,” Phys. Fluids 7(6),904-906 (1964). 11. Liu, C.Y. and Sigimura, T., “Rarefied Gas Flow Over a Sphere at Low Mach Numbers,” in Rarefied Gas Dynamics 6th Symposium 1, 789-794, Academic Press Inc., N.Y. (1969). 12. Phillips, W.F., “Drag on a Small Sphere Moving Through a Gas,” Phys. Fluids 18(9), 1089-1093 (1975). 13. Mott-Smith, H.M., “The Solution of the Boltzmann Equation for a Shock Wave,” Phys. Rev. 82(6), 885-892 (1951). 14. Krook, M., “Continuum Equations in Dynamics of Rarefied Gases,” J. Fluid. Mech. 6(4), 523-541 (1959). 15. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, N.Y., 1969). 16. Ivchenko, I.N., “Generalization of the Lees Method in Boundary Problems of Transfer,” J. Colloid and Interface Sci. 135(1), 16-19 (1990). 17. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “A Solution of the Problem of Heat Transfer Between Two Cylinders for Arbitrary Knudsen Number,” High Temperature 31(4), 776-783 (1993). 18. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “A Method for Solving Linearized Problems of the Transfer Theory for Spherical Geometry at Arbitrary Knudsen Numbers,” Fluid Dynamics 29(6), 888-893 (1994). 19. Ivchenko, I.N., “Evaporation (Condensation) Theory of Spherical Particles with All Knudsen Numbers,” J. Colloid and Interface Sci. 120(1), 1-7 (1987). 20. Savkov, S.A., The Slip Boundary Conditions of the Non-Uniform Binary Gas Mixture and an Application of Them in Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 21 Wang-Chang, C.S. and Uhlenbeck, G.E., “Heat Transport Between Parallel Plates,” University of Michigan Project M999 (1953). 22. Jackson, E.A., Boundary Value Problems in Kinetic Theory of Gases (Ph.D. Dissertation, Syracuse University, 1958). 23. Gross, E.P. and Ziering, S., “Heat Flow Between Parallel Plates,” Phys. Fluids 2(6), 701712 (1959). 24. Bassanini, P., Cercignani, C., and Pagani, C.D., “Comparison of Kinetic Theory Analyses of Heat Transfer Between Parallel Plates,” Inter. J. Heat Mass Transfer 10, 447 (1967). 25. Loyalka, S.K., “Linearized Couette Flow and Heat Transfer Between Two Parallel Plates,” in Rarefied Gas Dynamics 6th Symposium 1, 195, Academic Press Inc., N.Y. (1969). 26. Teagan, W.P. and Springer, G.S., “Heat-Transfer and Density-Distribution Measurements Between Parallel Plates in the Transition Regime,” Phys. Fluids 11(3), 497-506 (1968). 27. Halbritter, L., “Torque on a Rotating Ellipsoid in a Rarefied Gas,” Z. Naturforsh. 29a, 1717-1722 (1974). 28. Smirnov, L.P. and Chekalov, V.V., “Slow Rotation of a Sphere in a Bounded Rarefied Gas Volume,” Izv. AN SSSR, M. Zh. G. (4), 117-124 (1978) 29. Chandrasekhar, S., Hydrodynamic and Hydromagnetic Stability (Dover, New York, 1961). 30. Landau, L.D. and Lifshitz, E.M., Mechanics (Pergamon Press, New York, 1960). 31. Landau, L.D. and Lifshitz, E.M., Fluid Mechanics (Addison-Wesley, Reading, Massachusetts, 1960). 32. Loyalka, S.K., “Motion of a Sphere in a Gas: Numerical Solution of the Linearized Boltzmann Equation,” Phys. Fluids A 4(5), 1049-1056 (1992).

Chapter 12 BOUNDARY SLIP PHENOMENA IN A BINARY GAS MIXTURE

1.

THE FIRST-ORDER CHAPMAN-ENSKOG APPROXIMATION FOR A BINARY GAS MIXTURE.

The non-uniform state of a binary gas mixture is described by the distribution functions:

fi where Ci

0 1 f i r , t 1 < i Ci , r , t ; i 1, 2 ,

mi

0 f i r , t

12

2kT

(12-1)

0 Vi and fi r , t is given by:

§ · mi ni r , t ¨ ¨ 2S kT r , t ¸¸ © ¹

32

§ m v u r , t 2 i i ¨ exp ¨ 2kT r , t ©

· ¸. ¸ ¹

(12-2)

The corrections, < i Ci , r , t , may be expressed in terms of the first-order Chapman-Enskog solutions [1], A i , Di , and Bi , in the following manner: 1

< i1 Ci , r , t Ai1 d12

x1

w ln T wr

n1n2 m2 m1 nU

Di d12 2Bi : 1

ln p ,

1

wu , wr

(12-3)

(12-4)

290

Analytical Methods for Problems of Molecular Transport

where ni is the number density of the i -th constituent, mi is the molecular mass, p p1 p2 is the hydrostatic pressure, xi ni n , n n1 n2 , and U n1m1 n2 m2 . In Eq. (12-4), an external force field is assumed to be absent. The first-order Chapman-Enskog solutions for the thermal conductivity, 1 1 A1 and A2 , are given by [1]:

A1 1

1

A2

a1S3 2 C12 C1 1

1

a1S3 2 C22

12

kT D12 § m1 · § 2m2 · C1 ¸ , x1 ¨© 2kT ¸¹ ¨© m ¹

(12-5)

12

k D § m · § 2m · C2 T 12 ¨ 2 ¸ ¨ 1 C2 ¸ , x2 © 2kT ¹ © m ¹

(12-6)

where a1 and a1 are the transport coefficients, kT DT D12 is the thermal diffusion ratio, D12 is the diffusion coefficient, and m x1m1 x2 m2 . In the 1 same manner, the first-order Chapman-Enskog solutions for diffusion, D1 1 and D2 , are given by [1]: 12

D1

D12 § m1 · § 2m2 · C1 ¸ , x1 ¨© 2kT ¸¹ ¨© m ¹

D2

D12 § m2 · § 2m1 · C2 ¸ . ¨ ¸ ¨ x2 © 2kT ¹ © m ¹

1

1

(12-7)

12

(12-8)

and the first-order Chapman-Enskog solutions for viscosity, B1 and B2 , are given in terms of their respective components by [1]: 1

B1ij

b1 C1i C1 j 13 C12G ij ,

B2 ij

b1 C2i C2 j 13 C22G ij .

1

1

1

(12-9)

(12-10)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

2.

291

THE TRANSPORT COEFFICIENTS FOR A BINARY GAS MIXTURE.

The first-order Chapman-Enskog solutions for the unknown transport coefficients a1 , a1 , d 0 , b1 , and b1 , are derived from the following algebraic systems of equations [1]: 12

154

a11a1 a11a1 D 1

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

,

(12-11)

12

a00 d 0

G0

3 2

,

(12-12)

12

2kT · n ¨ ¸ © m0 ¹

b11b1 b11b1

b11b1 b11b1

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

154

a11a1 a11a1 D1

1 §

5 2

E 1 5 2

E1

,

(12-13)

n2 , n2

(12-14)

n1 , n2

(12-15)

The thermal conductivity, diffusion, and viscosity coefficients are then found to be [1]:

N

12 ª § 2kT ·1 2 º § 2kT · nk « x1 ¨ ¸ a1 x2 ¨ ¸ a1 » , «¬ © m1 ¹ »¼ © m2 ¹ 5 4

(12-16)

12

D12

1 xx 2 1 2

§ 2kT · ¨ ¸ © m0 ¹

d0 ,

(12-17)

and:

P

p x1b1 x2b1 ; p

nkT .

(12-18)

292

Analytical Methods for Problems of Molecular Transport

If one introduces the following standard notations of Chapman and Cowling [1]:

A

C

1 5

:12 2,2

1,3

: 1,2 , B 5:12 15 121,1 , 1,1 :12 :12

1

2 5

:121,2 1,1

:12

1 8

, and E

kT

, 1,1 M 1M 2 :12

then the known algebraic coefficients in Eqs. (12-11)-(12-15), a11 , a11 , a11 , a11 , a00 , b11 , b11 , b11 , and b11 , may be expressed in the following manner: a00

x1 x2

a11

x12

5 2

kT , E

(12-19)

kT

5 x1 x2

> P1 @1

u ª 14 6 M 12 ¬

a11

x22

5 2

kT

> P2 @1

kT 1 M1 E

5M 22

5 x1 x2

(12-20a)

M 22 B

2M 1M 2 Aº , ¼

kT 1 M2 E

(12-20b)

u ª 14 6 M 22 5M 12 M12 B 2M 1M 2 Aº , ¬ ¼

a11

b11

a11

x12

5 2

5 x1 x2

kT

> P1 @1

kT M1 M 2 E

kT 1 M1 E

5 x1 x2

12

2 3

ª¬ 114 B 2 Aº¼ ,

M1 M 2 A ,

(12-21)

(12-22a)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

b11

x22

b11

b11

5 2

kT

> P2 @1

5 x1 x2

5 x1 x2

kT E

kT 1 M2 E

2 3

M 2 M1 A ,

293 (12-22b)

23 A .

(12-23)

> P1 @1

is given by Eq. (5-62) with the substitutions m m1 and The coefficients a11 and b11 have been derived from the coefficients a11 and b11 by changing the indices associated with the gas constituents; > P1 @1 o > P2 @1 , x1 o x2 , x2 o x1 , M 1 o M 2 , and M 2 o M1 . Then, the transport coefficients may be presented in the form: where

V

V1 .

12

a1

52

m1 § 2kT · ¨ ¸ k © m1 ¹

>N1 @1 p > ' a @1

(12-24)

^

`

u x1Q1 x2 ª¬ P1M1M 2 11 4 B 8 A 2 P1 P2 º¼ ,

12

a1

m2 § 2kT · ¨ ¸ k © m2 ¹ 2 5

>N 2 @1 p > ' a @1

(12-25)

^

`

u x2Q2 x1 ª¬ P2 M 1M 2 11 4 B 8 A 2 P1 P2 º¼ ,

12

§ 2kT · ¸ © m0 ¹

E nx1 x2 1 , kT

d0

3 2¨

b1

1 ª x1 R1 x2 p > ' b @1 ¬«

b1

(12-26)

1 2

1 E > P2 @1 23 A º , ¼»

1 ª x2 R2 x1 p > ' b @1 «¬

1 2

1 E > P1 @1 32 A º . »¼

Here, the following notations have been incorporated:

(12-27)

(12-28)

294

Analytical Methods for Problems of Molecular Transport

> ' a @1

x12Q1 x22Q2 x1 x2 Q12 ,

> 'b @1

x12 R1 > P1 @1 x22 R2 > P2 @1 x1 x2 R12 ,

(12-30)

(12-31)

(12-32)

1

(12-29)

1

Q1

P1 6M 22 5M 12 4 M12 B 8M 1 M 2 A ,

Q2

P2 6 M12 5M 22 4 M 22 B 8M 1 M 2 A ,

Q12

3 M1 M 2

P1

M1 E > P1 @1 , P2

2

5 4 B 4M1 M 2 A 11 4 B 2P1P2 ,

1

1

M 2 E > P2 @1 ,

(12-33)

(12-34)

R1

2 3

M1 M 2 A ,

(12-35)

R2

2 3

M 2 M1 A ,

(12-36)

R12

3.

1 2

1

1

1

E > P1 @1 > P2 @1 34 A EM1M 2 .

(12-37)

THE SECOND-ORDER CHAPMAN-ENSKOG APPROXIMATION FOR A BINARY GAS MIXTURE.

The second-order Chapman-Enskog solutions for the functions, A i , Di , and Bi , can be expressed as:

Ai

1 2 ar1Ci ª S3 2 Ci2 ar 2 S3 2 Ci2 º kT Di , ¬ ¼

(12-38)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

295

12

Di

ª º D12 § mi · i 1 m j 1 zi S3 2 Ci2 » , ¨ ¸ Ci « 1 2 xi © 2kT0 ¹ m ¬ ¼

Bi

Ci Ci Bi Ci2

(12-39)

and: $

$

,

br1 Ci Ci 1 br*2 S5 2 Ci2 1

(12-40) 12

where ar*2 ar2 ar1 , br*2 br2 br1 , zi xi 2kT mi d r1 D12 , the upper and lower signs, r , refer to the first and second constituents of the gas mixture, respectively, and i z j . The transport coefficients of the thermal conductivity, diffusion and viscosity, ar1 , ar2 , d r1 , d 0 , br1 , and br2 for the second-order Chapman-Enskog approximation can be calculated from the three linear algebraic systems of equations given by: 2

¦ a p a pq

D q ; p, q z 0 ,

(12-41)

E q ; p, q z 0 ,

(12-42)

Gq ,

(12-43)

p 2

2

¦ bp bpq

p 2

and: 1

¦ d p a pq

p 1

where D r1 , E r1 , and G 0 are given by Eqs. (12-11)-(12-15) in Section 12-2 while the remaining coefficients, D r2 E r2 G r1 0 . The coefficient, a00 , of the algebraic system for diffusion, Eq. (12-43), is specified by Eq. (12-19) while the other coefficients in this system, a01 a10 and a0 1a10 , are expressible in the form: a01

x1 x2

a01

x1 x2

5 2

kT CM 2 M 11 2 , E

(12-44)

kT CM 1M 21 2 . E

(12-45)

and: 5 2

296

Analytical Methods for Problems of Molecular Transport

All three of the systems given in Eqs. (12-41)-(12-43) involve the coefficients, a11 , a11 , b11 , and b11 which are given by Eqs. (12-20)-(1223). For p 1, 2 and q r2 , explicit analytical expressions for a pq and bpq may be found if one uses the analytical representations of the appropriate bracket integrals given in [2,3]. These coefficients, a pq and bpq , are then found to be [2,3]:

: 2,3 · kT § ¨ 7 2 1 2,2 ¸ > P1 @1 ¨© :1 ¸¹ 1,2 M kT ª 35 2 2 2 2 :12 x1 x2 2 4 5 M M « 16 12M 1 5M 2 21 1 2 8 M 1 E «¬ :121,1 x12 85

a12

§ : 1,3 : 1,4 M 22 ¨ 194 121,1 12 121,1 ¨ : :12 12 ©

· § : 2,2 : 2,3 ¸ M 1M 2 ¨ 7 121,1 2 121,1 ¸ ¨ : :12 12 ¹ ©

·º ¸» , ¸» ¹¼

: 2,3 · kT § ¨ 7 2 2 2,2 ¸ > P2 @1 ¨© : 2 ¸¹ 1,2 M kT ª 35 2 2 2 2 :12 x1 x2 1 4 5 M M « 16 12M 2 5M 1 21 2 1 8 M 2 E «¬ :121,1

a1 2

x22 85

§ : 1,3 : 1,4 M 12 ¨ 194 121,1 12 121,1 ¨ : :12 12 ©

x1 x2 M 23 2 M11 2

a12

1,4 1 :12

2

:121,1

7

:12 2,2 :121,1

kT E

kT E 2

· § :12 2,2 :12 2,3 7 2 M M ¸ 1 2¨ ¸ ¨ : 1,1 :121,1 12 ¹ ©

1,2 (1,3) § 189 :12 19 :12 ¨ 595 8 4 ¨ 16 :121,1 :121,1 © : 1,4 : 2,2 : 2,3 · 12 121,1 7 121,1 2 121,1 ¸ , :12 :12 :12 ¸¹

x1 x2 M13 2 M 21 2

a1 2

(12-46a)

1,2 (1,3) § 595 189 :12 19 :12 ¨ 16 8 4 ¨ :121,1 :121,1 ©

:12 2,3 ·

¸,

:121,1 ¸¹

(12-46b)

·º ¸» , ¸» ¹¼

(12-47a)

(12-47b)

297

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

x12 85

a22

x1 x2

: 2,3 : 2,4 kT § 77 ¨ 4 7 1 2,2 1 2,2 > P1 @1 ¨© :1 :1

kT ª 35 4 2 2 4 « 64 40 M1 168M1 M 2 35M 2 M1E ¬

78 M 22 84M 12 35M 22

1,2 12 1,1 12

:121,4

x22 85

x1 x2

72 M 14

1,1 12

:

2,2

:12 2,3

2,4 3,3 º 3 :12 2 2 :12 2 M M 2 M M », 1 2 1 2 :121,1 :121,1 :121,1 »¼

: 2,3 : 2,4 kT § 77 ¨ 4 7 2 2,2 2 2,2 > P2 @1 ¨© :2 :2

· ¸ ¸ ¹

kT ª 35 4 2 2 4 « 40 M 2 168M 1 M 2 35M 1 M 2 E ¬ 64

:

: 1,5

78 M 12 84M 22 35M 12 1 M 12 8

(12-48a)

:121,3

14 M 24 121,1 72 M 1M 2 4 M 12 7 M 22 121,1 :12 :121,1 :12

14 M 1M 23

a2 2

::

18 M 22 108M 12 133M 22 72 M 24

· ¸ ¸ ¹

108M 22

:121,2

:

133M 12

:121,4

1,1 12

(12-48b)

:121,3

:

: 1,5

1,1 12

:

2,2

14 M 14 121,1 72 M 1M 2 4 M 22 7 M12 121,1 1,1 :12 :12 :12

14 M 2 M 13

:12 2,3

2,4 3,3 º 3 :12 2 2 :12 2 M M 2 M M », 2 1 1 2 :121,1 :121,1 :121,1 »¼

298

Analytical Methods for Problems of Molecular Transport

x1 x2 M 13 2 M 23 2

a2 2

241 8

14

b12

b1 2

:121,3

kT E

: 1,4

1,2 § 833 :12 ¨ 8505 8 ¨ 64 :121,1 ©

: 1,5

: 2,2

72 121,1 14 121,1 772 121,1 :121,1 :12 :12 :12

(12-49)

:12 2,3

:121,1 2,4 :12 3,3 · 2 2 ¸, :121,1 :121,1 :121,1 ¸¹

: 2,3 · kT § kT M 2 ¨ 7 2 1 2,2 ¸ x1 x2 23 ¨ ¸ E M1 > P1 @1 © :1 ¹ § : (1,2) : (2,2) : (2,3) u ¨ 352 M 1 7 M 1 121,1 214 M 2 121,1 23 M 2 121,1 ¨ :12 :12 :12 ©

x12 85

x22 85

: 2,3 kT § ¨ 7 2 2 2,2 > P2 @1 ¨© :2

· kT M 1 ¸ x1 x2 23 ¸ E M2 ¹

§ : (1,2) u ¨ 352 M 2 7 M 2 121,1 ¨ :12 ©

21 M1 4

§ : 1,2 ¨ 352 7 121,1 ¨ :12 ©

21 4

: kT § 35 M1 ¨ 2 7 121,1 E ¨© :12

21 4

b1 2

x1 x2 23 M 2

b12

x1 x2 23

kT E

· ¸, ¸ ¹

1,2

:12(2,2) :121,1

32 M 1

:12(2,3) ·

¸,

2,3 · 3 :12 ¸, 2 :121,1 :121,1 ¸¹

:121,1

3 2

(12-50b)

:121,1 ¸¹

:12 2,2

:12 2,2

(12-50a)

:12 2,3 ·

¸,

:121,1 ¸¹

(12-51a)

(12-51b)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

: 2,3 : 2,4 kT § 301 ¨ 12 7 1 2,2 1 2,2 > P1 @1 ¨© :1 :1

x12 85

b22

· ¸ ¸ ¹

kT ª 1 2 2 « 4 M1 140M 1 245M 2 M1E ¬ § : 1,2 : 1,3 · M 1M 22 ¨ 49 121,1 8 121,1 ¸ ¨ :12 :12 ¸¹ © x1 x2

2 3

18 M 2 154 M 12 147 M 22 21 M 23 2

b2 2

:12 2,2

:

1,1 12

2,4 3,3 º 3 :12 2 :12 3 M M M 3 », 2 1 2 2 :121,1 »¼ :121,1 :121,1

: 2 2,3 : 2 2,4 kT § 301 7 ¨ > P2 @1 ¨© 12 : 2 2,2 : 2 2,2

kT ª 1 2 2 « M 2 140 M 2 245M1 M2E ¬ 4 § : 1,2 : 1,3 · M 2 M 12 ¨ 49 121,1 8 121,1 ¸ ¨ :12 :12 ¸¹ ©

1 8

2 3

M 1 154 M 22

212 M 13

(12-52a)

:12 2,3

x22 85

x1 x2

147 M 12

· ¸ ¸ ¹

(12-52b)

:12 2,2

:

1,1 12

:12 2,3

2,4 3,3 º 2 :12 3 :12 3 M M M 3 », 1 2 1 2 :121,1 »¼ :121,1 :121,1

§ :121,2 :121,3 49 8 ¨ 385 ¨ 4 :121,1 :121,1 © :12 2,2 21 :12 2,3 3 :12 2,4 :12 3,3 · 3 301 ¸. 8 2 2 :121,1 :121,1 :121,1 :121,1 ¸¹

b2 2

299

x1 x2 23 M 1M 2

kT E

(12-53)

300

Analytical Methods for Problems of Molecular Transport

The coefficients, a pq and bpq , when p, q 0 , have been specified from the expressions for when p, q ! 0 by changing the indices connected with the gas constituents in the manner shown previously, i.e. > P1 @1 o > P2 @1 , x1 o x2 , x2 o x1 , M 1 o M 2 , and M 2 o M 1 . The same procedure applied to the expressions for a1 2 and b1 2 yield expressions for a12 and b12 . The remaining coefficients, a pq and bpq , can be found from the symmetry conditions, a pq aqp and bpq bqp . All of the coefficients described above depend upon the : -integrals, many of which have been previously tabulated in [2]. In Appendix F, a full set of these : -integrals, sufficient for use with the second-order approximation, have been included. Among these are values for the 1,4 1,5 3,3 previously unreported, : , : , and : . All of the : -integral values included in Appendix F have been fully recomputed and an annotated Mathematica® program for use in further calculations has been provided.

4.

ANALYTICAL METHODS OF SOLUTION FOR PLANAR BOUNDARY VALUE PROBLEMS INVOLVING BINARY GAS MIXTURES.

A semi-infinite expanse of a binary gas mixture is considered which is bounded by a planar surface lying in the y z plane and located at x 0 . Far from the surface, the gas mixture is maintained at a constant mean-mass velocity gradient normal to the surface, h wu wx f , and constant tangential gradients of the partial concentrations, d12 wx1 wy f , and temperature (in the y -direction), g w ln T wy f . The total pressure is assumed to be constant. For these conditions in a binary gas mixture, the mean-mass velocity far from the surface is expressible in the following asymptotic form: u x

u0 , xh

(12-54)

where u0 is a constant velocity that arises owing to the influence of the surface on the gas mixture. The asymptotic behavior of a binary gas mixture may only be found via solution of the Boltzmann equations with appropriate boundary conditions at the wall. For this geometry, the distribution function for the i -th constituent may be expressed as:

fi r

m º 0 ª iy < i ci , r ) ir ci , x » , f i «1 i xhv ¬ kT ¼

(12-55)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

301

12

where ci mi 2kT v i is the dimensionless velocity, < i ci , r is the Chapman-Enskog solution, ) ir ci , x is a function that allows for the influence of the wall, and the signs, and , refer to the reflected and incident molecules, respectively. The functions, ) ir ci , x , can be found from the following linearized Boltzmann equations: 0 w) i vix f i wx

ni2 I i ) i ni n j I ij ) i ) j ,

(12-56)

where i 1, 2 with i z j , and I i and I ij are the standard notations [1]. These linearized collision integrals are given by

ni2 I i ) i

0 0 ³³ fi vi fi v ª¬) i vi ) i v

) i vci ) i vc º¼ giD i T , gi d : dv ,

ni n j I ij ) i ) j

³³ f f ª¬) 0

i

0

j

i

) j

) ic ) cj º¼ gijD ij T , gij d : dv j .

(12-57)

(12-58)

Using the Maxwellian boundary model, Eq. (6-4), the boundary conditions for the functions, ) ir ci , x , can be written in the form:

) i ci ,0 D iW ª¬ Aiy g Diy d12 º¼ 2 2 D iW Bixy h 1 D iW ) i cix , ciy , ciz ,0 .

(12-59)

Let the scalar product be introduced by: 2

M1 ci , x ,M2 ci , x ¦ ³ M1 fi0 M2 dvi .

(12-60)

i 1

Then, using both the commutative property of the standard bracket integrals and the conservation of momentum present during molecular encounters, the scalar product of Eqs. (12-56) with mi viy gives:

302

Analytical Methods for Problems of Molecular Transport

w n 2 ) i , mi viy mi vix viy ,) i ci , x n 2 mi viy ,) i wx n12 ª¬)1 , m1v1 y º¼ n1 n2 ª¬)1 ) 2 , m1v1 y m2 v2 y º¼ 1 12

n22 ª¬) 2 , m2 v2 y º¼

^

2

`

^

` (12-61)

0.

where the bracket integral notations of Eq. (12-61) are defined in [1] as:

n 2 ^ F , G` n12 > F1 , G1 @1 n1n2 > F1 F2 , G1 G2 @12 n22 > F2 , G2 @2 ,

> Fi , Gi @i ³ Gi Ii Fi dvi , > F1 G2 , H1 K 2 @12 ³ F1I12 H1 K 2 dv1 ³ G2 I 21 H1 K 2 dv 2 , with F , G , Fi , Gi , H i , and Ki representing arbitrary functions of the molecular velocities of the subscripted constituents, and I i and I ij having been previously defined in Eqs. (12-57) and (12-58), respectively. In the dimensionless variables, ci , this moment solution may be written in the form:

c c

) i ci , x const .

ix iy ,

(12-62)

If one takes into account the asymptotic behaviors of the functions, 12 ) i ci , x 2 mi 2kT u0 ciy , then this function in the two-moment approach may be chosen in the form:

) ir ci , x M i1 2 a r x ciy .

(12-63)

This form of the correction to the distribution function means that Eq. (1262) may be expressed in the form:

c c

) i ci , x 0 ,

ix iy ,

(12-64)

which is the exact analytical solution of Eq. (12-56). The Maxwellian method [4] involves the use of the analytical solution expressed by Eq. (12-64) at the surface where x 0 , i.e.:

303

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

c c

) i ci ,0 0 ,

ix iy ,

(12-65)

where ) i ci ,0 K cix ) i ci ,0 K cix ) i ci ,0 and K cix is the Heaviside step function that is defined as:

1 ; x ! 0, ¯0 ; x 0.

K x ®

For this method, the boundary conditions are satisfied only in an integral sense and not in an explicit sense. To get some useful results using this method, additional assumptions must be made. The Maxwellian analysis is constructed around the assumption that, in Eq. (12-63), a r f a 0 a . Then, the constant, a , may be calculated if one uses the analytical solution given by Eq. (12-65). The Maxwellian method may be classified as a onemoment approach. Another analytical method, first proposed by Loyalka [5], involves using the two general moment solutions of the Boltzmann equations, Eq. (12-56). The first of these moment solutions, given by Eq. (12-65), is the same as that used in the Maxwellian method. The second moment solution may be derived from Eqs. (12-56) if one uses a property of the Chapman-Enskog solution for viscosity. For the particular planar geometry being considered here, the Chapman-Enskog corrections may be found from the following expressions: 0 f1 c1x c1 y

0 f 2 c2 x c2 y

n12 I1 B1xy n1n2 I12 B1xy B2 xy ,

(12-66)

n22 I 2 B2 xy n1n2 I 21 B1xy B2 xy .

(12-67)

Now, the scalar product of Eq. (12-56) with Bixy , taken in accordance with Eq. (12-60), yields:

w vix Bixy ,) i ci , x wx

^ ` n ^) c , x , B ` ³ ) c , x ª n I B n n I B B º dv ¬ ¼ ³ ) c , x ª n I B n n I B B º dv . ¬ ¼ 1

1

2

2

n 2 Bixy ,) i ci , x 2 1 1

2 2 2

1 xy

2 xy

1 2 12 1

2 21

2

i

1 xy

1 xy

i

2 xy

2 xy

ixy

1

2

304

Analytical Methods for Problems of Molecular Transport

Here, the commutative property of the bracket integrals has been used. From this, taking into account Eqs. (12-64), (12-66) and (12-67), one obtains:

w vix Bixy ,) i ci , x wx

) c , x , c c i

i

ix iy

0,

which yields:

M

1 2 2 cix ciy Bi i

c ,) c , x 2 i

i

i

const .

(12-68)

The Loyalka method is a generalization of the Maxwellian method and may be classified as a two-moment approach because two-moment solutions to the Boltzmann equations are used in its formulation. Specifically, for the Loyalka method, the two solutions used are the following exact moment solutions:

c c

) i ci , x 0 ,

ix iy ,

M

1 2 2 cix ciy Bi i

(12-69)

c , ª¬) c ,0 ) c , f º¼ 2 i

i

i

i

i

0.

(12-70)

These two analytical solutions give useful results if one takes into account two assumptions concerning the corrections to the distribution functions. Specifically, one must introduce two constants, a r f a and a 0 aw , where a z aw , which may be calculated from the two analytical solutions given in Eqs. (12-69) and (12-70).

5.

THE SLIP COEFFICIENTS FOR A BINARY GAS MIXTURE.

The slip phenomena in gas mixtures are of fundamental significance when it comes to the specification of appropriate boundary conditions for flows in the slip-flow regime. In this regime, the boundary condition for the mean mass velocity component tangential to a body surface may be presented in the form [6-13]:

uW rS cm

OP w ln T * Q* PnW cTsl cDsl D12 d12 W , P wrW

(12-71)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

305

where rS is the radius vector of a given point on the body surface, n indicates a direction normal to the surface and points into the gas at rS while W indicates a direction tangential to the body surface at the point rS . Also, * cm , cTsl , and cDsl are the velocity-slip, thermal-creep, and diffusion-slip coefficients, respectively and the following notations have been used: 12

P § 2kT · ¨ ¸ S p© m ¹

OP

8 5

Q*

P ON , U OP

ON

64 75

,

12

N § m · ¨ ¸ T, S p © 2kT ¹

in which OP and ON are the mean free paths calculated from the viscosity and thermal conductivity coefficients, P and N , respectively, and PnW is the pressure tensor component. The first term in Eq. (12-71) yields a correction to continuum expressions which is proportional to the Knudsen number, Kn , and which, when Kn o 0 , may be neglected. However, the other terms in Eq. (12-71), which are associated with the thermal-creep and diffusion-slip phenomena, may be significant even if Kn o 0 because the influence of these phenomena depends only upon the value of the tangential component of the external non-uniformities in the gas mixture. In the slip-flow regime, for the bulk flow of the gas mixture, the NavierStokes equations can be used to develop the same description as the Boltzmann equations if one introduces the appropriate slip boundary condition given by Eq. (12-71). For this regime, the use of the NavierStokes equations together with the slip boundary condition is preferable to the use of the full Boltzmann equations because it allows one to reduce the number of independent variables from ten (for the Boltzmann equations for a binary gas mixture) to four (in the general case with the Navier-Stokes equations). For stationary boundary value problems and for some specific geometries the number of these independent variables may be additionally reduced. It should be noted that the slip conditions themselves must be determined by a solution of the planar boundary value problem involving the Boltzmann equations. These slip coefficients can then be used with suitable accuracy to formulate transport problems for arbitrary geometries.

306

Analytical Methods for Problems of Molecular Transport

The slip-flow phenomena are basic to understanding cross-effects in gas mixtures (where they can be the most important factors), the motions of aerosol particles suspended in non-uniform media, and the flow of gas mixtures through capillary tubes, all of which have numerous technological applications. The slip effects are also of importance in studies of many natural phenomena such as aerosol mechanics and flow in porous media, which also play a significant role in numerous other technological applications such as chemical and physical vapor deposition, nanofabrication, and most low-pressure applications. In some cases, such as physical vapor transport experiments under micro-gravity conditions, for example, these phenomena may become the dominant transport factors due to the production of side-wall gas creep which may suppress natural convection [14-20]. In spite of considerable work in this field [21-63], there is no completely accurate analysis employing realistic models of the intermolecular interactions that would have reliable accuracy in all applications. The half-range moment method [34,35,51] described in Chapter 8, was previously believed to be able to yield accurate results for slip problems. Unfortunately, this method involves a very complicated analysis (even in low-moment approaches) and the accuracy of the method is not predictable. The relatively simple Loyalka’s method described in this chapter results in a simpler analysis and good accuracy [5]. To calculate the slip coefficients for binary gas mixtures, Loyalka’s method is among the most useful. First, it yields relatively simple expressions and, second, it has no difficulties associated with the use of general intermolecular and gas-surface interaction laws. Moreover, the use of the two general conservation laws as exact moment solutions provides a suitable accuracy to the results. Next in this section, Loyalka’s method is used to derive analytical expressions for the slip coefficients of a binary gas mixture. This analysis employs the second-order Chapman-Enskog solutions for the transport parameters. One uses the same statement of the planar boundary value problem as was described in Section 12.4. The velocity, u0 , given by Eq. (12-54) is presented in the form:

u0

cm OP h cDsl D12 d12 cTslQ * g .

(12-72)

Additionally, this asymptotic velocity may be calculated by simple integration, i.e.:

U u0

m v

12 r i iy , M i a

f ciy ,

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

307

where a r f a const . This integration yields: 12

u0

§ 2kT · ¸ © m0 ¹

1 2¨

(12-73)

a.

Within the framework of the Loyalka method, one has two unknown constants, a r f a and a 0 aw , which can be found from the two exact moment solutions given in Eqs. (12-69) and (12-70). Substitution of the boundary condition given by Eq. (12-59) into Eqs. (12-69) and (12-70) yields the analytical expressions for the constants, a and aw . Then, Eqs. (12-72) and (12-73) yield the following expressions for the slip coefficients:

cm

cDsl

5 8

pM 1 2S

P

2

ª

i 1

¬

4b

º

¦ 2 D iW xi br1 « K1 S Mr11 2 K 2Z1i » ,

2

ª

§

i 1

¬«

©

¦ 2D iW « M i1 2 K1 ¨ 1

i 1

mj

· 14 zi ¸ m ¹

§ ·º i 1 m j br1 K 2 ¨ 1 12 ziZ2i ¸ » , m © ¹ ¼»

* cTsl

3 a*

ª 2 « ¦ D iW xi ar1 «¬ i 1

1 2

K1H1i br1M i1 2 K 2H 2i

where:

x1M 1 x2 M 2 , a*

x1a1M 11 2 x2 a1M 21 2 ,

2

Z1i 1 br*2 174 br*2 , Z2i 1 72 br*2 ,

H1i 1 14 ar*2 , and H 2i 1 72 br*2 72 ar*2br*2 .

(12-75)

12 º § m · kT D12 ¨ 0 ¸ cDsl » , © 2kT ¹ »¼

M

(12-74)

¼

i

(12-76)

308

Analytical Methods for Problems of Molecular Transport

Here, also, the following notations have been used:

K1

2 D1W x1b1 2 D 2W x2b1 K

K2

1 4

D1W x1M 11 2 D 2W x2 M 21 2

2

,

x1b1 x2b1 1 .

(12-77)

(12-78)

Equations (12-74)-(12-78) are suitable for the computation of results corresponding to arbitrary models of the intermolecular interaction provided that appropriate values of the transport coefficients, ar1 , br1 , etc, are available. Now, consider the limiting case of a simple gas. The limiting formulas may be derived from Eqs. (12-74)-(12-78) if one makes the substitutions: x1 1 , m1 m2 m , V 1 V 2 V , and D1W D 2W DW . The factors K1 and K 2 are then given by:

K1

1 4

2 2 DW DW1 and K 2

1 1 b . 4 1

(12-79)

The limiting expressions for the slip coefficients are the same as those given by Eqs. (9-62) and (9-63) and the diffusion-slip coefficient is found to be cDsl 0 .

6.

DISCUSSION OF THE SLIP COEFFICIENT RESULTS.

The expressions for the slip coefficients that have been obtained in the preceding sections of this chapter are in simple algebraic forms readily usable in computations. Such computations are straightforward provided that values of the coefficients, ar1 , br1 , etc., are available. These values depend upon the intermolecular interaction (potential) model that is used to compute the : -integrals. While many of the necessary : -integrals are tabulated in the classical texts [2,3], there are several additional : -integrals that have been newly computed for inclusion in this text. The various : integrals, both old and new, are given in Appendix F. Tables 12-1 and 12-2 contain values of the transport coefficients, P , N ,

and D12 , and of the slip coefficients, cm , cDsl , cTsl , and cTsl , that have been computed for a selection of binary gas mixtures using different intermolecular potential models. In Table 12-1 the rigid-sphere potential

309

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

Table 12-1. Values of the transport coefficients ( P , N , and D12 ) and of the slip coefficients

( cm , cDsl , cTsl , and cTsl ) for a selection of binary gas mixtures obtained using the rigidsphere potential model and the first- and second-order Chapman-Enskog approximations ( p =1 atm, T0 =293 K). The units of the viscosity, P , are gm cmí1 secí1, the units of the thermal conductivity, N , are cal cmí1 secí1 Kí1, and the units of the diffusion coefficient, D12 , are cm2 secí1. The slip coefficients are all dimensionless. The thermal conductivity coefficients have been calculated under the assumption that the molecules are effectively monatomic. For all of these calculated values, the assumption has been made that the tangential momentum accommodation coefficient for each gas species is equal to unity, i.e. D1W = D 2W =1. The column labeled ‘Order’ refers to the order of the Chapman-Enskog approximation that was used in calculations.

P

N

( u 105) 4.5642 4.6736

D12

cm

cDsl

cTsl

cTsl

1 2

( u 104) 1.6874 1.7125

0.6379 0.6857

1.1189 1.0914

-0.1775 -0.1920

1.1102 0.9786

1.1200 0.9961

N2-Ar

1 2

1.9262 1.9545

4.1837 4.2788

0.1659 0.1690

1.1206 1.0936

0.1170 0.1265

1.1120 0.9825

1.1029 0.9822

x1 =0.5

N2-O2

1 2

1.8202 1.8471

4.5041 4.6063

0.1752 0.1783

1.1167 1.0893

0.0363 0.0395

1.1235 0.9924

1.1211 0.9979

N2-CO2 x1 =0.5

1 2

1.5254 1.5493

3.1895 3.2666

0.1269 0.1296

1.1253 1.0982

0.2211 0.2374

1.0976 0.9625

1.1106 0.9821

Mixture

Order

N2-H2

x1 =0.99 x1 =0.5

model was used while in Table 12-2 the Lennard-Jones (6-12) potential model was used. For all mixtures and models the results obtained using the first-order Chapman-Enskog approximation are compared with those obtained using the second-order Chapman-Enskog approximation. In all of the cases shown in Tables 12-1 and 12-2 molecules have been assumed to be diffusely reflected at the relevant surfaces such that D1W D 2W 1 . Table 123 summarizes values of the parameters that are associated with each of the intermolecular potential models used. A variety of additional models are also available in the literature for which results have not been computed here but which may also be of interest to the reader [64]. This source contains tabulated values of the transport collision integrals for a variety of modern potential models and provides substantial information about the various intermolecular collision parameters included in these models. Some approximate methods using the various alternative potential models described in this reference, and which may be used to calculate the various transport coefficients for the second- and thirdorder Chapman-Enskog approximations, may be found in the literature [6569] but, since these analyses generally are not complete, they tend to be of limited use. It is clear from Tables 12-1 and 12-2 that there exists a significant dependence of the slip coefficients upon the intermolecular potential (within the context of the Loyalka method) even when the calculations are limited to

310

Analytical Methods for Problems of Molecular Transport

Table 12-2. Values of the transport coefficients ( P , N , and D12 ) and of the slip

coefficients ( cm , cDsl , cTsl , and cTsl ) for a selection of binary gas mixtures obtained using the Lennard-Jones (6-12) potential model and the first- and second-order Chapman-Enskog approximations ( p =1 atm, T0 =293 K). The units of the viscosity, P , are gm cmí1 secí1, the units of the thermal conductivity, N , are cal cmí1 secí1 Kí1, and the units of the diffusion coefficient, D12 , are cm2 secí1. The thermal conductivity coefficients have been calculated under the assumption that the molecules are effectively monatomic. The slip coefficients are all dimensionless. For all of these calculated values, the assumption has been made that the tangential momentum accommodation coefficient for each gas species is equal to unity, i.e. D1W = D 2W =1. The column labeled ‘Order’ refers to the order of the Chapman-Enskog approximation that was used in calculations. All of the necessary : -integrals have been calculated numerically using the appropriate reduced temperatures and without using the typical interpolation procedure.

P

N

( u 105) 4.7246 4.7576

D12

cm

cDsl

cTsl

cTsl

1 2

( u 104) 1.7375 1.7446

0.7361 0.7564

1.1189 1.1021

-0.1896 -0.1959

1.1072 1.0342

1.1228 1.0518

N2-Ar

1 2

1.9872 1.9932

4.3334 4.3543

0.1870 0.1877

1.1208 1.1064

0.1307 0.1331

1.1138 1.0514

1.1091 1.0489

x1 =0.5

N2-O2

1 2

1.8825 1.8893

4.6598 4.6858

0.1982 0.1990

1.1168 1.1009

0.0380 0.0390

1.1238 1.0548

1.1216 1.0549

N2-CO2 x1 =0.5

1 2

1.6006 1.6037

3.3829 3.3940

0.1477 0.1481

1.1252 1.1141

0.2589 0.2621

1.1051 1.0560

1.1303 1.0815

Mixture

Order

N2-H2

x1 =0.99

x1 =0.5

the first-order Chapman-Enskog solutions whereas, in a simple gas, such a dependence is observed only for calculations employing the second- or higher-order Chapman-Enskog solutions. As is evident from Eqs. (12-74)(12-76), the thermal-creep and diffusion-slip coefficients are significantly more sensitive to the order of the Chapman-Enskog approximation than the velocity-slip coefficient. This is undoubtedly due to the dependence of the thermal-creep and diffusion-slip coefficients on cross-effects; a supposition that is confirmed by direct calculation of the slip coefficients for a selection of gas mixtures. For example, with the rigid-sphere model for a mixture of N2 and O2, the relative differences between the first-order Chapman-Enskog derived values with respect to the second-order Chapman-Enskog derived values are 13.5 % for cT* sl and 8.1 % for cD sl . For the same mixture and the rigid-sphere model the relative difference observed for cm is only 2.5 %. The same situation occurs in a simple gas as well where the relative differences observed for cT* sl and cm are 13 % and 2.5 %, respectively. Of course, for a simple gas, one must remember that there is no diffusion-slip effect to observe. The relative differences of the slip coefficients that one observes when using the generally more realistic Lennard-Jones (6-12) potential model are somewhat less than those observed for the rigid-sphere model but the same

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

311

Table 12-3. Relevant parameters for two of the most commonly used intermolecular potential models; the rigid-sphere model [1] and the Lennard-Jones (6-12) potential model [2]. Gas He H2 N2 Ar O2 CO2

Rigid-sphere model V (Å) 2.193 2.745 3.784 3.659 3.636 4.643

Lennard-Jones (6-12) potential model V (Å) H k (K) 2.576 10.22 2.94 35.65 3.72 85.65 3.418 124.0 3.487 100.5 3.95 201.5

trend with respect to the sensitivity is observed. In a mixture of N2 and H2, * the relative differences determined for cTsl and cDsl (with respect to the second-order approximation values) are 6.7 % and 3.2 %, respectively while the relative difference determined for cm is only 1.5 %; again indicating that the velocity-slip coefficient is not as sensitive to the order of the approximation used. Further, if one compares the relative differences between the values of these coefficients using the different potential models with the same order * approximation for the same mixture of N2 and CO2 one finds that cTsl and cDsl vary from the Lennard-Jones (6-12) model to the rigid-sphere model by 9.1 % and 9.4 %, respectively; clearly showing that the more realistic model likely produces substantially more realistic results. From this discussion it is clear that whenever possible it is desirable to use the higher-order approximations and most realistic potential models as significant improvements in the theory of the slip phenomena may be realized under these circumstances. One slip phenomenon of interest is the thermal transpiration effect which is observed in capillary tubes. Table 12-4 compares the experimental and theoretical values of the slip factor, csl cTsl cDsl kT D12 Q , which governs this effect [48]. The experimental values given in Table 12-4 are slightly different from those reported in [48] owing to the use of the secondorder Chapman-Enskog approximations in determining the viscosity and velocity-slip coefficients which are used to calculate the slip factor for experimentally obtained pressure difference data. The statement of the problem for the thermal transpiration effect and the method solution are given in Problem 12.5. The specific values of the tangential momentum accommodation coefficients, D1W and D 2W , shown in Table 12-4 were chosen so as to produce the best possible agreement between the measured and theoretical slip factors for the selected binary gas mixtures. Table 12-4 shows that excellent agreement between theory and experiment is obtained if these values are assumed. This further demonstrates the substantial improvement in the accuracy of theoretically calculated slip phenomena

312

Analytical Methods for Problems of Molecular Transport

Table 12-4. The tangential momentum accommodation coefficients, D iW , in a binary gas mixture chosen for calculation of the slip coefficients and slip factors. These specific D iW were chosen so as to produce the best agreement between the measured and theoretical slip factors for the selected binary gas mixtures. Gas mixture

D1W

D 2W

>csl @exp

>csl @th

Error (%) †

He:Ar (4:1) 0.8 0.8 1.1863 1.1708 He:Ar (2:1) 0.8 0.8 1.1874 1.1559 He:Ar (1:1) 0.8 0.8 1.1043 1.1146 He:Ar (1:2) 0.8 0.8 1.0903 1.0714 He:CO2 (2:1) 0.8 0.8 1.3846 1.3141 0.8 0.8 1.2997 1.2720 He:CO2 (1:1) Ar:CO2 (2:1) 0.8 0.8 0.9593 1.0098 Ar:CO2 (1:1) 0.8 0.8 0.9573 1.0143 † Relative error has been calculated with respect to the experimental values of i.e. csl csl csl u 100% .

> @exp > @th > @exp

1.3 2.7 0.9 1.7 5.1 2.1 5.3 6.0 the slip,

values that can be obtained when using higher-order Chapman-Enskog approximations; in this case the second-order approximation. The above discussion has shown that there are strong dependencies of * cTsl and cDsl on the potential model used and on the order of the ChapmanEnskog approximation used. However, cm does not show the same degree of sensitivity as the other slip coefficients. There exists a direct way to assess the accuracy of the analysis reported here by comparing the values of the slip coefficients that it predicts with the very accurate numerical values reported by Takata and Aoki in [62]. The relative deviation of the values calculated from Eqs. (12-75) and (12-76) from the values reported in [62] is less than 2 % over a wide range of the molecular mass ratio. From this, one may conclude that the overall degree of accuracy of the slip coefficients reported here is, almost certainly, less than 2 % as well.

PROBLEMS 12.1. Using the first-order Chapman-Enskog approximation, determine the diffusion-slip coefficient for a binary gas mixture by means of the Maxwell method. Use a planar geometry in which the pressure and temperature are assumed to be constant. Solution: The distribution function for the i -th constituent may be expressed in the form:

fi

r

fi

0

12 ª º · D12 § mi · § i 1 m j d12 ¸ M i1 2 a r x ciy » , «1 ciy ¨ 1 2 ¨ ¸ xi © 2kT ¹ © m «¬ »¼ ¹

313

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

where d12 wx1 wy f , and x , and y are the coordinates normal and tangential to the planar surface, respectively. The Maxwell method is constructed on the assumptions that a r f a and a 0 a f a . The constant, a , is found from the moment solution given by Eq. (12-65):

c c

) i ci ,0 0 .

ix iy ,

The boundary conditions (one for each constituent) are given by:

)

i

ci ,0

12 ª i 1 m j D12 § mi · º 2D iW d12 « 1 » ciy 1 D iW M i1 2 aciy , ¨ ¸ m xi © 2kT ¹ »¼ «¬

where D iW is the tangential momentum accommodation coefficient for the i th constituent. The constant, a , can then be found by simple integration. Thus, Eq. (12-73) yields: cDsl

D

1W

m2 D 2W m1 K 0 ,

where:

K0

m1m2

m D1W x1 m1 D 2W x2 m2

.

12.2. For a non-uniform binary gas mixture, determine the mean velocities of the constituent gases and the mean mass velocity of the mixture. Consider the specific case in which there is no net number flow (transport) of molecules and in which the pressure and temperature of the gas mixture are assumed to be uniform. Solution: The distribution function of the i -th constituent is given by:

fi

m · 0 § f i ¨1 i u v i Di d12 ¸ , © kT ¹

where u is the mean mass velocity of the mixture and in which Di and d12 are expressed as: 12

Di

ª º D12 § mi · i 1 m j 1 zi S3 2 Ci2 » and d12 ¨ ¸ Ci « 1 2 xi © 2kT0 ¹ m ¬ ¼

n 1n1 .

314

Analytical Methods for Problems of Molecular Transport

The mean mass velocity of the i -th constituent can be calculated by integrating in the following manner:

viE

m · 0 § ni1 ³ viE fi dv i ni1 ³ f i ¨1 i uD viD DiD d12 D ¸ viE dv i © kT ¹ i 1 m j D12 uE 1 d12 E . m xi

This yields:

v1

u

m2 D12 d12 , m x1

(P-1)

v2

u

m1 D12 d12 . m x2

(P-2)

and:

For the specific case under consideration in which there is no net number flow of molecules, the velocities, v1 and v 2 , may be found from the equations:

v1 v 2

D12 n n1n2 n1 ,

n1 v1 n2 v 2

0,

(P-3)

(P-4)

which yield:

v1

D12 n11n1 and v 2

D12 n21n1 .

From these, the mean mass velocity is then determined to be:

u

U1 v1 U 2 v 2 U1 U 2

m2 m1

U

D12n1 .

(P-5)

12.3. For a non-uniform binary gas mixture, determine the number flux vector (current) of the first constituent assuming that the first constituent is

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

315

diffusing through the second and that there is no net number flow (current) of the second constituent. Assume the pressure and temperature of the gas mixture to be uniform. Solution: From the statement of the problem, one has that v 2 0 . Since the gas mixture conditions in this problem are similar to those in Problem 12.2, one then has, from Eq. (P-2), that:

u

m1 D12 d12 . m x2

Using this in Eq. (P-1) yields: n1 v1

n n2 D12n1 .

12.4. For a non-uniform binary gas mixture, determine the mean velocities of the constituent gases. In this more general problem, assume that the partial densities of the constituents, and the pressure and temperature of the gas mixture are all non-uniform. Solution: The distribution function of the i -th constituent is given by:

fi

m wu · 0 § f i ¨1 i u v i Ai ln T Di d12 2Bi : ¸ , wr ¹ © kT

(P-6)

where d12 is specified by Eq. (12-4). The same integration as in Problem 12.2 yields:

v1

u

m2 D12 d12 kT ln T , m x1

(P-7)

v2

u

m1 D12 d12 kT ln T , m x2

(P-8)

and:

where kT is the thermal diffusion ratio and u is the mean mass velocity which is unknown until the exact conditions governing the diffusion are specified. Note: Terms containing Sonine polynomials which occur in the functions, A i and Di , should vanish after the integration procedure and integrations over the function Bi should be identically zero.

316

Analytical Methods for Problems of Molecular Transport

12.5. For a non-uniform binary gas mixture, determine the relationship between the mean velocity of the gas mixture and its mean mass velocity. Solution: Let vi be the mean velocity of the i -th constituent of the gas mixture in the laboratory frame of reference. This velocity may be expressed in the form:

vi

w Wi

u Ui ,

(P-9)

where w and u are the mean and mean mass velocities of the gas mixture, respectively, and Wi and Ui are the diffusion velocities of the i -th constituent relative to two different frames of reference moving with the velocities, w and u , respectively. If one takes into account Eqs. (P-7) and (P-8), the relative diffusion velocities, Wi and Ui , may be found from the two following algebraic systems of equations:

W1 W2

v1 v 2

n1W1 n2 W2

U1 U 2

D12 d12 kT ln T , x1 x2

0,

v1 v 2

U1U1 U 2 U 2

D12 d12 kT ln T , x1 x2

0.

Upon solving these systems for Wi and Ui , the results may be substituted into Eq. (P-9) to yield:

u

w

m1 m2 D12 d12 kT ln T . m

(P-10)

12.6. Calculate the diffusion coefficient for a H2-N2 binary gas mixture at Use the first-order Chapman-Enskog T 293 K and p 1 atm. approximation. Assume that the molecules act as rigid spheres. Solution: Using Eqs. (12-13), (12-17) and (12-19) one obtains: D12 1

3 16

kT

. 1,1 M 1M 2 nm0 :12

The : -integral for rigid-sphere molecules is given by [3]:

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture 12

§ kT · * ¸ © 2S m ¹

:12 l , r SV 122 ¨

1 2

l ª 1 1 º 1 « », r 1 1 2 l 1 » «¬ ¼

317 (P-11)

where m* m1m2 m1 m2 . Then, one can obtain the following analytical expression for the diffusion coefficient. 1

D12

12

3 8

kT § kT · ¨ ¸ pV 122 © 2S m* ¹

(P-12)

.

For

the specific mixture under consideration, one may use 3.2645 u 108 cm from [1] and, from [70], p 1.01325 u 106 dyne cmí2, mH 1.00794 gm molí1, mN 14.00674 gm molí1, and k 1.38066 u 1016 erg Kí1 and N A 6.02214 u 1023 molí1 for Boltzmann’s constant and Avogadro’s number, respectively. Then, Eq. (P-12) yields 1 D12 0.6379 cm2 secí1.

V 12

12.7. For the second-order Chapman-Enskog approximation, obtain an analytical expression for the diffusion coefficient that relates it to the expression for the diffusion coefficient obtained in Problem 12.6 using the first-order Chapman-Enskog approximation. Solution: The transport coefficient, d 0 , is calculated from the following algebraic system of equations:

d 1a11 d 0 a01 d1a11

0, 12

2kT · n ¨ ¸ , © m0 ¹

d 1a10 d 0 a00 d1a10

3 2

d 1a11 d 0 a01 d1a11

0,

1 §

where the coefficients, a00 , a11 , a11 , and a11 a11 are given in Section 12.2 while the coefficients, a01 a10 and a01 a10 , are given by Eqs. (1244) and (12-45), respectively. Then, taking into account Eqs. (12-13) and (12-17), one can represent the diffusion coefficient obtained using the second-order approximation in the form: D12 2

D12 1

1 , 1 '

318

Analytical Methods for Problems of Molecular Transport

where the following notation has been introduced:

a0 1 '

a a10 a01 11 a10 a11 a a a00 11 11 a11 a11

a10 a11

a11 a10

2 a01 a11 a021a11 2a01a10 a11

a00 a11a11 a121

.

12.8. For the first-order Chapman-Enskog approximation, derive an analytical expression for the viscosity coefficient for a binary gas mixture. Solution: The viscosity coefficient is given by P p (x1b1 x2b1 ) . The transport coefficients, b1 and b1 , are specified from the following algebraic system of equations found in Section 12.2 (Eqs. (12-14) and (12-15)):

b11b1 b11b1

E 1

5 2

n2 , and b11b1 b11b1 n2

E1

5 2

n1 , n2

The solution of these equations yields:

> P @1

5 2

kT

x12b11 2 x1 x2b11 x22b11 . b11b11 b121

12.9. For the first-order Chapman-Enskog approximation, derive an analytical expression for the thermal conductivity coefficient for a binary gas mixture. Solution: The thermal conductivity coefficient is given by Eq. (12-16):

N

12 ª § 2kT ·1 2 º § 2kT · 54 nk « x1 ¨ a x ¸ 1 2¨ ¸ a1 » , «¬ © m1 ¹ »¼ © m2 ¹

The transport coefficients, a1 and a1 , are specified from the algebraic system of equations found in Section 12.2 (Eqs. (12-11) and (12-12)): 12

a11a1 a11a1 D 1

154

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

, and:

319

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture 12

a11a1 a11a1 D1

154

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

,

The solution of these equations yields:

>N @1

2

75 8

k T

x12 m11a11 2 x1 x2 m1m2

1 2

a11 x22 m21a11

a11a11 a121

.

12.10. For the second-order Chapman-Enskog approximation, derive an analytical expression for the viscosity coefficient for a binary gas mixture. Solution: The viscosity coefficient is given by P p (x1b1 x2b1 ) . The transport coefficients, b1 and b1 , are specified from the following algebraic system of equations:

b2 2 b2 b1 2b1 b1 2b1 b2 2b2

0,

5 2

b21b2 b11b1 b11b1 b21b2

n2 , n2

n1 , n2

b21b2 b11b1 b11b1 b21b2

5 2

b22 b2 b12b1 b12b1 b22b2

0.

The solution of this system of equations is given by:

b1

' b1

> 'b @2

and b1

' b1

> 'b @2

,

where the following notations have been introduced:

> 'b @2

b2 2 b21 b21 b22

b1 2 b11 b11 b12

b1 2 b11 b11 b12

b2 2 b21 , b21 b22

320

Analytical Methods for Problems of Molecular Transport 5 2

' b1

n1 n 'cb1 52 22 'ccb1 , 2 n n n1 n 'cb1 52 22 'ccb1 , 2 n n

' b1

52

'cb1

b2 2 b21 b22

b1 2 b11 b12

b2 2 b21 , 'ccb1 b22

'cb1

b2 2 b21 b22

b1 2 b11 b12

b2 2 b21 , and 'ccb1 b22

b2 2 b21 b22

b1 2 b11 b12 b2 2 b21 b22

b2 2 b21 , b22 b1 2 b11 b12

b2 2 b21 . b22

Using these notations, the viscosity coefficient is found to be:

> P @2

5 2

kT

x12 'cb1 x1 x2 'ccb1 'cb1 x22 'ccb1

> 'b @2

.

12.11. For the second-order Chapman-Enskog approximation, derive an analytical expression for the thermal conductivity coefficient for a binary gas mixture. Solution: The thermal conductivity coefficient is given by Eq. (12-16):

N

12 ª § 2kT ·1 2 º § 2kT · « nk x1 ¨ ¸ a1 x2 ¨ ¸ a1 » , «¬ © m1 ¹ »¼ © m2 ¹ 5 4

(12-16)

The transport coefficients, a1 and a1 , are specified from the following algebraic system of equations:

a2 2 a2 a1 2 a1 a1 2 a1 a2 2 a2

0,

a21a2 a11a1 a11a1 a21a2

154

12

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

,

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture 12

a21a2 a11a1 a11a1 a21a2

154

a22 a2 a12 a1 a12 a1 a22 a2

0.

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

,

The solution of this system of equations is given by:

a1

' a1

> ' a @2

' a1

and a1

,

> ' a @2

where the following notations have been introduced:

a2 2 a21 a21 a22

> ' a @2

a1 2 a11 a11 a12

a1 2 a11 a11 a12

a2 2 a21 , a21 a22

12

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

12

'ca1 154

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

15 4

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

' a1

154

' a1

15 4

'ca1

a2 2 a21 a22

a1 2 a11 a12

a2 2 a21 , 'cca1 a22

'ca1

a2 2 a21 a22

a1 2 a11 a12

a2 2 a21 , and 'cca1 a22

12

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

'cca1 ,

12

'ca1

a2 2 a21 a22

'cca1 ,

a1 2 a11 a12

a2 2 a21 a22

a2 2 a21 , a22

a1 2 a11 a12

a2 2 a21 . a22

321

322

Analytical Methods for Problems of Molecular Transport

Table 12-5. A comparison of diffusion-slip coefficient values. Values determined numerically [62] are compared with corresponding values determined using the Maxwell method for selected values of the mass ratio, Z , for the specific case when D1W D 2W 1 and x1 x2 12 .

Z m2 m1

2

cDsl , Eq. (P-13) cDsl , numerical G (%)

0.3235

0.5333

0.5694

0.5974

0.1637

0.2514

0.2634

0.2619

97.6

112

116

128

4

5

10

G (%) = cDsl ,num cDsl , Max cDsl ,num u 100 . Using these notations, the thermal conductivity coefficient is found to be:

>N @2

2

75 8

k T

x12 m11'ca1 x1 x2 m1m2

1 2

'cc 'c x m

> ' a @2

a1

a1

2 2

1 cc 2 ' a1

.

12.12. Evaluate the accuracy of the Maxwell method for determining the diffusion-slip coefficient by comparing it with the numerical results [62] for different values of the mass ratio, Z m2 m1 , Consider the specific case when D1W D 2W 1 and x1 x2 12 . Solution: For the specific case under consideration, the diffusion-slip coefficient given in Problem 12.1 may be expressed in the form:

cDsl

4Z Z 1

Z

2

1 Z 1

.

(P-13)

The numerical values and values computed from Eq. (P-13) can then be assembled in tabular form for ease of comparison. This has been done and is given in Table 12-5. As one can see from Table 12-5, with relative errors on the order of 100%, the Maxwell method does not appear, at least for the specific case under consideration in this problem, to be an effective technique for determination of the diffusion-slip coefficient. However, the current case is quite limited and one should not make the mistake of inferring that the accuracy of the Maxwell method overall is represented by the current problem. Rather, one should understand from the current problem only that the Maxwell method can lead to unacceptably inaccurate results in some circumstances.

12.13. Using the first-order Chapman-Enskog approximation and the Maxwell method, determine the slip-flow coefficient for a binary gas mixture. Consider the case in which the temperature and partial pressures of the gas constituents are uniform.

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

323

Solution: The distribution function of the i -th constituent in the gas mixture is given by:

fi r

m º 0 ª iy 2cix ciy br1h ) ir ci , x » , f i «1 i xhv ¬ kT ¼

where one should use b1 if i 1 and b1 if i model yields:

2 . The Maxwellian boundary

) i ci ,0 2 2 D iW cix ciy br1h 1 D iW ) i cix , ciy , ciz ,0 . The correction to the distribution function, ) i ci , x , can be chosen in the form ) ir ci , x M i1 2 a r x ciy . The Maxwell method is constructed on the assumption that a r f a 0 a . Then this constant can be calculated from the exact analytical solution to the Boltzmann equations given by:

c c

) ir ci ,0

ix iy ,

0.

Taking into account Eqs. (12-72) and (12-73), one can then obtain:

cm

pM 1 2

P

5 8

S

2 D1W x1b1 2 D 2W x2b1 . D1W x1M11 2 D 2W x2 M 21 2

12.14. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at the same constant temperature, T c T cc T , and are filled at the beginning t 0 with a binary gas mixture having slightly different mixture fractions in each vessel such that x1c 0 x1 0 and x2c 0 x2 0 in the first vessel while in the second vessel x1cc 0 x1 0 G x1 0 and x2cc 0 x2 0 G x1 0 . When the capillary is opened a pressure difference, G p pcc pc , will develop as a function of time. Determine the time dependence of the relative density difference G x1 t x1cc t x1c t and the time dependence of the pressure difference, G p t . Also, determine the maximum value of the pressure difference that develops. Solution: The basic equations of balance for the number of molecules of each constituent gas are given by:

324

Analytical Methods for Problems of Molecular Transport

V c

dnic dt

J iz and V cc

dnicc dt

J iz ,

where J iz S R 2 ni viz is the net molecular current (referred to as the molecular number flux in some texts) of the i -th constituent through the capillary. If one takes into account Eq. (P-9) from Problem 12.5, the basic equations can be written in the form:

V c

V cc

dnic dt

S R 2 ni wz Wiz ,

dnicc S R 2 ni wz Wiz , dt

(P-14)

(P-15)

where wz is the mean velocity of the gas mixture and Wiz is the diffusion velocity of the i -th constituent relative to a frame of reference moving with velocity, w . For the number density difference, G ni t nicc t nic t , one can obtain:

S R2

d G ni dt

V

ni wz Wiz ,

(P-16)

where V V cV cc V c V cc . The number density difference may then be expressed in the form:

G ni

· § pxi · § xi ¨ G p G xi ¸ n . ¸ © kT ¹ © p ¹

G¨

(P-17)

Then, Eq. (P-16) yields: · d § x1 ¨ G p G x1 ¸ dt © p ¹

S R2

· d § x2 ¨ G p G x1 ¸ dt © p ¹

S R2

V

V

x1 wz W1z ,

(P-18)

x2 wz W2 z .

(P-19)

The average values introduced here are slightly different from those that would exist in the equilibrium state because the pressure difference that

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

325

develops is much less than the initial pressure. The mean molecular velocity, w , is related to the mean mass velocity, u , by (see Problem 12.5): § m m1 · w u¨ 2 ¸ D12d12 , © m ¹

(P-20)

and the diffusion velocities are then defined by:

n1W1

n2 W2

nD12d12 ,

(P-21)

where the vector, d12 , is specified in Eq. (12-4). In the particular case under consideration in this problem, the z -component of this vector can be expressed in the form:

d12 z

§ § m m1 · G p · L1 ¨ G x1 x1 x2 ¨ 2 ¸. ¸ © m ¹ p ¹ ©

(P-22)

Note that in the slip-flow regime, the mean mass velocity in the capillary may be specified from the boundary value problem of:

u 0 and P 2u p with:

u z R cm OP

wu z cDsl D12 d12 z and p r , L p r ,0 G p . wr

The solution of this boundary value problem is found to be:

uz r

1 wp ª R 2 1 4cm Kn r 2 º cDsl D12 d12 . z ¬ ¼ wz 4P

from which one can obtain: R

uz

2 u z r rdr R 2 ³0

R 2 1 4cm Kn 8P L

G p cDsl D12 d12 z .

(P-23)

Now, taking into account Eqs. (P-20)-(P-23), Eqs. (P-18) and (P-19) can be written in the form:

326

Analytical Methods for Problems of Molecular Transport

d G x1 dt

Z1xG x1 Z1 pG p ,

(P-24)

d Gp dt

Z xG x1 Z pG p ,

(P-25)

where the following notations have been employed:

§ m m1 · D12 , a ax1 x2 ¨ 2 ¸ © m ¹ p

S R2

Z1x

aD12 , Z1 p

Zx

apc12 D12 , c12

Zp

ª pR 2 1 4cm Kn º § m m1 · c12 x1 x2 ¨ 2 a« D ». 12 ¸ 8P © m ¹ «¬ »¼

VL

§ m m1 · cDsl ¨ 2 ¸ , and © m ¹

The characteristic equation for this homogeneous linear system of differential equations is the quadratic equation:

Z1x s Z1 p Z x Z p s

0,

which has the following roots:

s1,2

1 2

ª « Z Z r Z p p 1x « «¬

2 º · 4Z xZ1 p » . ¸ 2 » ¸ Z p ¹ »¼

§ Z1x ¨1 ¨ Zp ©

For the different terms in this expression one can make the following determinations regarding the orders of various quantities, specifically: 12

§ 2kT ·

P UV O , D12 V O , V ¨ ¸ , and: © m ¹

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

327

Z p R 2 nkT R 2 2 Kn 2 1 . 2 2 Z1x UV O O The last of these implies that:

Z Z1x Z 1 , x 1 , and 1 p 1 . Zp Zp Zp Now, taking into account these determinations of order, it follows that s1 Z1x and s2 Z p . Thus, since both of the roots of the characteristic equation have been shown to be negative, it further follows that each term in the complete solution to the homogeneous system of equations will decay exponentially to zero as t o f . The solution of the system is found to be:

G x1 t G x1 0 exp Z1x t and:

G p t K exp Z1x t exp Z p t where:

K

8P c12 D12 § Z1x G x1 0 2 ¨1 R 1 4cm Kn ¨© Z p

1

· ¸ . ¸ ¹

The maximum pressure difference is found to be:

'p max where c12

G x1 0

8P c12 D12 § Z1x § Z p · · ln ¨ ¨1 ¸¸ . R 1 4cm Kn ©¨ Z p © Z1x ¹ ¹¸ 2

cDsl (m2 m1 )/m .

12.15. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O (which corresponds to the slip-flow regime) where O is the mean free path of the gas molecules. The two vessels are maintained at different temperatures, T c and T cc with T c ! T cc . Assume that the two vessels initially t 0 contain identical gas mixtures, each with the same pressure and mole ratios, i.e. and x1c (0) x1cc(0) q (1 q)

328

Analytical Methods for Problems of Molecular Transport

x2c (0) x2cc (0) q (1 q) , where q is the initial pressure ratio of the gas constituents. When the capillary is opened a pressure difference, G p t pcc pc , will develop as a function of time. Determine the steadystate pressure difference that will develop in this system, 'pst G p f . Solution: For this problem one can use the same basic system of differential equations that was employed in Problem 12.14 in which the mean velocity, wz , and the diffusion velocities, Wiz , must be modified by the inclusion of additional terms to account for the non-uniformity of the temperature in the system. This basic system of differential equations is given by: · d § x1 ¨ G p G x1 ¸ dt © p ¹

S R2

· d § x2 ¨ G p G x1 ¸ dt © p ¹

S R2

V

V

x1 wz W1z ,

x2 wz W2 z .

where G p t pcc pc and G x1 x1cc x1c . The diffusion velocities, Wiz , are given by:

n1W1z

GT º ª nD12 « d12 z kT , where: TL »¼ ¬

n2W2 z

d12 z

§ § m m1 · G p · L1 ¨ G x1 x1 x2 ¨ 2 ¸, ¸ © m ¹ p ¹ ©

and in which kT is the thermal diffusion ratio. The mean velocity, wz , can be expressed in the form:

wz

R 2 1 4cm Kn 8P L

G p cDsl D12 d12 z

GT º § m m1 · ª Q D12 « d12 z kT , 2 TL ¨© m ¸¹ TL »¼ ¬

* * cTsl

GT

(P-26)

* where cm , cTsl , and cDsl are the velocity-slip, thermal-creep, and diffusionslip coefficients, respectively. Then, the basic system of equations can be written in the form:

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

329

d G x1 dt

Z1xG x1 Z1 pG p Z1T G T ,

(P-27)

d Gp dt

Z xG x1 Z pG p ZT G T ,

(P-28)

where the following notations have been employed:

§ m m1 · D12 , a ax1 x2 ¨ 2 ¸ © m ¹ p

S R2

Z1x

aD12 , Z1 p

Z1T

Zp

ª pR 2 1 4cm Kn º § m m1 · c12 x1 x2 ¨ 2 a« D », 12 ¸ 8P © m ¹ «¬ »¼

ZT

ap T * T c12 Q , and c12 T

akT D12 , Zx T

apc12 D12 , c12

VL

§ m m1 · cDsl ¨ 2 ¸, © m ¹

§ m m1 · kT D12 * ¨ 2 cTsl ¸ * © m ¹ Q

The last terms of Eqs. (P-27) and (P-28) can be regarded as the driving forces which will determine the ultimate steady-state (or stationary state) of this process. The characteristic equation of the homogeneous system has two negative roots as was shown in Problem 12.14. This again indicates that the solutions to the homogeneous system decay exponentially to zero as t o f . However, the steady-state solution for the non-homogeneous system, where the contribution of the particular solutions dominates the behavior of the system, can be specified from the following equations:

Z1xG x1 f Z1 pG p f Z1T G T ,

Z xG x1 f Z pG p f ZT G T .

330

Analytical Methods for Problems of Molecular Transport

G p f , is then found to be:

The steady-state pressure difference, 'pst

'pst

Z § ZZ G T T ¨ 1 x 1T Z p © Z1xZT

· § Z xZ1 p ¸ ¨¨1 ¹ © Z1xZ p

1

· ¸ . ¸ ¹

An evaluation of the two terms found in the expression for Z p shows that the first term is the dominant term, i.e.: 1

R2 p R 2 nkT R 2 nkT R2 § m2 m1 · 2 1 , ¨ ¸ 2 2 8P c12 D12 x1 x2 © m ¹ P D12 mnV O O where the following estimations of order have been used: 12

§ 2kT ·

P mnV O , D12 V O , and V ¨ ¸ . © m ¹ This means that:

Zp

ap

R 2 1 4cm Kn 8P

.

Further, the order of the ratio, Z xZ1 p Z1xZ p , is determined to be:

Z xZ1 p D12 P V 2 O 2 nm O 2 2 1 . Z1xZ p pR 2 nkTR 2 R Now, neglecting the small terms, one can obtain:

'pst

8PQ *csl , T R 2 1 4cm Kn

GT

where the following notation has been introduced:

csl

c12 c12 T

kT D12

Q

*

* cTsl cDsl

kT D12

Q*

.

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

331

REFERENCES 1. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 2. Hirschfelder, J.O., Curtiss, C.F., and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954). 3. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 4. Maxwell, J.C., in The Scientific Papers of J.C. Maxwell, Vol. 1, ed. by Niven, W.D. (Dover, Mineola, NY, 2003). 5. Loyalka, S.K., “Approximate Method in the Kinetic Theory,” Phys. Fluids 14(11), 22912294 (1971). 6. Kennard, E.H., Kinetic Theory of Gases (McGraw-Hill, New York, 1938). 7. Williams, M.M.R. and Loyalka, S.K., Aerosol Science, Theory and Practice (Pergamon Press, New York, 1991). 8. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, NY, 1969). 9. Cercignani, C., Theory and Application of the Boltzmann Equation (Scottish Academic Press, Edinburg, UK, 1975). 10. Hidy, G.M. and Brock, J.R., The Dynamics of Aerocolloidal Systems, Vol. 1 (Pergamon Press, New York, 1970). 11. Derjaguin, B.V. and Yalamov, Yu.I., “The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and their Experimental Testing,” in Topics in Current Aerosol Research, vol. 3, part 2, edited by Hidy, G.M. (Pergamon Press, Oxford, 1972). 12. Loeb, L.B., The Kinetic Theory of Gases, 3rd edition (Dover, New York, 1961). 13. Present, R.D., Kinetic Theory of Gases (McGraw-Hill, New York, 1958). 14. Rosner, D.E., “Side-wall Gas ‘Creep’ and ‘Thermal Stress Convection’ in Microgravity Experiments on Film Growth by Vapor Transport,” Phys. Fluids A 1(11), 1761-1763 (1989). 15. Gupta, R.N., Scott, C.D., and Moss, J.N., Slip-Boundary Equations for Multicomponent Non-equilibrium Air Flow (NASA TP2455, Nov. 1985). 16. Hess, D.W. and Jensen, K.F., Microelectronic-Processing Chemical Engineering Aspects (American Chemical Society, Washington DC, 1989). 17. Wolf, S. and Tauber, R.N., Silicon Processing for the VLSI ERA, Vol. 1. Process Technology (Lattice Press, Sunset Beach, CA, 1986). 18. Ikegawa, M. and Kobayashi, J., “Deposition Profile Simulation Using the Direct Simulation Monte Carlo Method,” J. Electrochemical Soc. 136(10), 2982-2986 (1989). 19. Coronell, D.G. and Jensen, K.F., “Analysis of Transition Regime Flows in Low-pressure Chemical Vapor Deposition Reactors Using the Direct Monte Carlo Method,” J. Electrochemical Soc. 139(8), 2264-2273 (1992). 20. Coronell, D.G. and Jensen, K.F., “Simulation of Rarefied Gas Transport and Profile Evolution in Nonplanar Substrate Chemical Vapor Deposition,” J. Electrochemical Soc. 141(9), 2545-2551 (1994). 21. Kramers, H.A. and Kistemaker, J., “On the Slip of a Diffusing Gas Mixture Along a Wall,” Physica (Amsterdam) 10(8), 699-713 (1943). 22. Kucherov, R.Ya. and Richenglas, L.E., “Slip and Temperature Jump at the Boundary of a Gaseous Mixture,” Zh. Eks. Teor. Fis. (Russian) 36(6), 1758-1761 (1959). 23. Kucherov, R.Ya., “Diffusion Slip and Convective Diffusion of a Gas in Capillaries,” Zh. Tekh. Fiz. (Russian) 27(9), 2158-2161 (1957).

332

Analytical Methods for Problems of Molecular Transport

24. Schmitt, K.H., “Untersuchungen an Schwebstoffteilchen in Diffundierendem Wasserdampf,” Z. Naturforschg. 16a, 144-149 (1961). 25. Brock, J.R., “The State of a Binary Gas Mixture Near a Catalytic Surface,” J. Catalysis 2, 248-250 (1963). 26. Brock, J.R., “Forces on Aerosols in Gas Mixtures,” J. Colloid Sci. 18(6), 498-501 (1963). 27. Zhdanov, V.M., “The Theory of Slip on the Boundary of a Gaseous Mixture,” Zh. Tekh. Fiz. (Russian) 37(1), 192-197 (1967). 28. Lang, H. and Eger, K., “Gas Diffusion in Narrow Capillaries,” Z. Physik Chem. 68, 130148 (1969). 29. Lang, H. and Müller, W.J.C., “Slip Effects in Mixtures of Monoatomic Gases for General Surface Accommodation,” Z. Naturforschg. 30a, 885-867 (1975). 30. Schmitt, K.H. and Waldmann, L., “Untersuchungen an Schwebstoffteilchen in Diffundierenden Gasen,” Z. Naturforschg. 15a(10), 843-851 (1960). 31. Waldmann, L. and Schmitt, K.H., “Über das bei der Gasdiffusion Auftretende Druckgefälle,” Z. Naturforschg. 16a, 1343-1354 (1961). 32. Waldmann, L., “On the Motion of Spherical Particles in Nonhomogeneous Gases,” in Rarefied Gas Dynamics, ed. Talbot, L. (Academic Press, New York, 1961), pp. 323-344. 33. Waldmann, L. and Schmitt, K.H., “Thermophoresis and Diffusiophoresis of Aerosols,” Chapter VI in Aerosol Science, ed. Davies, C.N. (Academic Press, New York, 1966), pp.137-162. 34. Breton, J.P., “Interdiffusion of Gases Through Porous Media – Effect of Molecular Interactions,” Phys. Fluids 12(10), 2019-2026 (1969). 35. Breton, J.P., “The Diffusion Equation in Discontinuous Systems,” Physica 50, 365-379 (1970). 36. Yalamov, Yu.I., Ivchenko, I.N., and Derjaguin, B.V., “Calculation of the Surface Diffusion Slip Velocity of a Gas Mixture,” in Rarefied Gas Dynamics, Vol. 1, eds. Trilling, L. and Wachman, H.Y. (Academic Press, New York, 1969), pp. 295-300. 37. Ivchenko, I.N. and Yalamov, Yu.I., “Diffusion Slip of a Binary Gas Mixture,” Mekh. Zhid. G. (Russian) 4, 22-26 (1971). 38. Loyalka, S.K. and Ferziger, J.H., “Model Dependence of the Slip Coefficient,” Phys. Fluids 10(8), 1833-1839 (1967). 39. Loyalka, S.K. and Ferziger, J.H., “Model Dependence of the Temperature Slip Coefficient,” Phys. Fluids 11(8), 1668-1671 (1968). 40. Loyalka, S.K., “Momentum and Temperature Slip Coefficient with Arbitrary Accommodation at the Surface,” J. Chem. Phys. 48(12), 5432-5436 (1968). 41. Loyalka, S.K., “Velocity Slip Coefficient and the Diffusion Slip Velocity for a Multicomponent Gas Mixture,” Phys. Fluids 14(12), 2529-2604 (1971). 42. Loyalka, S.K., “Kinetic Theory of Thermal Transpiration and Mechanocaloric Effect,” J. Chem. Phys. 55(9), 4497-4503 (1971). 43. Lang, H. and Loyalka, S.K., “Diffusion Slip Velocity. Theory and Experiment,” Z. Naturforschg. 27a, 1307-1319 (1972). 44. Loyalka, S.K., “Temperature Jump in a Gas Mixture,” Phys. Fluids 17(5), 897-899 (1974). 45. Loyalka, S.K., “Kinetic Theory of Thermal Transpiration and Mechanocaloric Effect. II,” J. Chem. Phys. 63(9), 4054-4060 (1975). 46. Loyalka, S.K. and Storvick, T.S., “Kinetic Theory of Thermal Transpiration and Mechanocaloric Effect. III,” J. Chem. Phys. 71(1), 339-350 (1979). 47. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “Slip Coefficients for Binary Gas Mixtures,” JVST A 15(4), 2375-2381 (1997).

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

333

48. Huang, C.M., Tompson, R.V., Ghosh, T.K., Ivchenko, I.N., and Loyalka, S.K., “Measurements of Thermal Creep in Binary Gas Mixtures,” Phys. Fluids 11(6), 16621671 (1999). 49. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “Boundary Slip Phenomena in a Binary Gas Mixture,” ZAMP 53(1), 58-72 (2002). 50. Yalamov, Yu.I., Shkanov, Yu. and Savkov, S.A., “Boundary Conditions of Sliding of a Binary Gas Mixture and Their Use in the Dynamics of Aerosols. 1. Flow of a Gas Mixture Along a Solid Flat Wall,” Inzh. Fizich. Zh. (Russian) 66(4), 421-426 (1994). 51. Savkov, S.A., Slip Boundary Conditions for Non-Uniform Binary Gas Mixtures and their Application to Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 52. Volkov, I.V. and Galkin, V.S., “Analysis of the Slip Coefficients and Temperature Jump in a Binary Gas Mixture,” Mekh. Zhid. G. (Russian) 6, 152-159 (1990) 53. Siewert, C.E. and Sharipov, F., “Model Equations in Rarefied Gas Dynamics: ViscousSlip and Thermal-Slip Coefficients,” Phys. Fluids 14(12), 4123-4129 (2002). 54. Siewert, C.E., “Viscous-Slip, Thermal-Slip, and Temperature-Jump Coefficients as Defined by the Linearized Boltzmann Equation and the Cercignani-Lampis Boundary Condition,” Phys. Fluids 15(6), 1696-1701 (2003). 55. Siewert, C.E., “The Linearized Boltzmann Equation: A Concise and Accurate Solution of the Temperature-Jump Problem,” J. Quantitative Spectroscopy and Radiative Transfer 77(4), 417-432 (2003). 56. Takata, A. and Aoki, K., “Two Surface Problems of a Multicomponent Mixture of Vapors and Non-condensable Gases in the Continuum Limit in the Light of Kinetic Theory,” Phys. Fluids 11(9), 2743-2756 (1999). 57. Sone, Y., Takata, A., and Golse, F, “Notes on the Boundary Conditions for FluidDynamics Equations on the Interface of a Gas and Its Condensed Phase,” Phys. Fluids 13(1), 324-334 (2001). 58. Takata, A. and Aoki, K., “The Ghost Effect in the Continuum Limit for a Vapor-Gas Mixture Around Condensed Phases: Asymptotic Analysis of the Boltzmann Equation,” Transport Theory and Stat. Phys. 30(2&3), 205-237 (2001). 59. Aoki, K., Bardos, C., and Takata, A., “Knudsen Layer for Gas Mixtures,” J. Statistical Phys. 112(3&4), 625-655 (2003). 60. Takata, A. and Aoki, K., “The Ghost Effect in the Continuum Limit for a Vapor-Gas Mixture Around Condensed Phases: Asymptotic Analysis of the Boltzmann Equation,” Transport Theory and Stat. Phys. 31(3), 289-290 (2002). 61. Takata, A., Yasuda, S., Kosuge, S., and Aoki, K., “Numerical Analysis of Thermal-Slip and Diffusion-Slip Flows of a Binary Mixture of Hard-Sphere Molecular Gases,” Phys. Fluids 15(12), 3745-3766 (2003). 62. Takata, A. and Aoki, K., “Numerical Analysis of the Shear Flow of a Binary Mixture of Hard-Sphere Gases Over a Plane Wall,” Phys. Fluids 16(6), 1989-2003 (2004). 63. Takata, A., “Kinetic Theory Analysis of the Two-Surface Problem of a Vapor-Vapor Mixture in the Continuum Limit,” Phys. Fluids 16(7), 2182-2198 (2004). 64. Maitland, G.C., Rigby, M., Smith, E.B., and Wakeham, W.A., Intermolecular Forces: Their Origin and Determination (Oxford University Press, New York, 1981, reprinted and corrected edition, 1987). 65. Mason, E.A., “Transport Properties of Gases Obeying a Modified Buckingham (exp-six) Potential,” J. Chem. Phys. 22(2), 169-186 (1954). 66. Mason, E.A., “Higher Approximations for the Transport Properties of Binary Gases Mixtures. I. General Formulas,” J. Chem. Phys. 27(1), 75-84 (1957).

334

Analytical Methods for Problems of Molecular Transport

67. Mason, E.A., “Higher Approximations for the Transport Properties of Binary Gases Mixtures. II. Applications,” J. Chem. Phys. 27(3), 782-790 (1957). 68. Mason, E.A. and Saxena, S.C., “Approximate Formula for the Thermal Conductivity of Gas Mixtures,” Phys. Fluids 1(5), 361-369 (1958). 69. Mason, E.A. and Monchick, L., “Heat Conductivity of Polyatomic and Polar Gases,” J. Chem. Phys. 36(6), 1622-1639 (1962). 70.CRC Handbook of Chemistry and Physics, 71 edition, ed. Lide, D.R. (CRC Press, Boca Raton, 1990).

Appendix A BRACKET INTEGRALS FOR THE PLANAR GEOMETRY

1.

BRACKET INTEGRALS INVOLVING TWO SONINE POLYNOMIALS. In this section, bracket integrals of the form: ars

ªcS r c 2 , cS s c 2 º , 32 ¬ 32 ¼

are discussed. Consider, for example, the bracket integral a11 that is contained in the first-order Chapman-Enskog solution for thermal conductivity. This integral can be written as:

a11

ªcS 1 c 2 , cS 1 c 2 º 3 ª cx S 1 c 2 , cx S 1 c 2 º 32 32 ¬ 32 ¼ ¬ 32 ¼

^³ D T , g d : ³ ³ gf f ª¬c S c S c cc S cc cc S cc º dv ` dv . ¼

3n 2 ³ cx S3 2 c 2 1

c1 x

1 32

2 1

0

0

1

1 x 32

2

1 1x 3 2

1 x 32

2

(A-1)

2

1

1

Using the momentum conservation equation, one obtains:

a11

3 ª¬ cx c 2 , cx c 2 º¼ .

(A-2)

336

Analytical Methods for Problems of Molecular Transport

The differential scattering cross section for rigid-sphere molecules can be expressed in the form:

D T , g d :

bdbd H

b

db dT d H dT

1 4

V 2 sin T dT d H ,

(A-3)

where V is the diameter of a molecule. and Let the variables of integration be changed from v and v1 to G g v1 v (the center of mass and relative molecular velocities, respectively) by the relations:

1 g , v G 2

1 g , G 2

v1

1 2

(A-4a)

m v 2 v12

(A-4b)

m G 2 14 g 2 ,

(A-4c)

and: dvdv1

dg , dG

(A-4d)

and g are dimensional variables. where G After changing the variables of integration, one can make the following substitutions: 12

G

§ m· ¨ kT ¸ G , © ¹

g

1§ 2¨

c

12

m· ¸ g , © kT ¹

1 2

(A-5a)

G g

,

(A-5b)

(A-5c)

337

Appendix A. Bracket Integrals for the Planar Geometry

c1

dG

1 2

G g

§ kT · ¨ ¸ © m¹

,

(A-5d)

dG ,

(A-5e)

32

and:

§ kT · dg 8 ¨ ¸ © m¹

32

(A-5f)

dg .

It is convenient to introduce the spherical coordinates g ,D , E for g . The relationships between these spherical coordinates and the coordinates in Cartesian velocity space, g x , g y , g z , are given by:

gx

g cos D ,

g sin D cos E ,

gy

g sin D sin E .

gz

Substituting the dimensionless variables into a11 , and neglecting all vanishing integrals, one can obtain:

a11

3 4

12S

§ kT · ¸ © m¹

V 2S 3 ¨

f

2S

³ sin T dT 0

u ³ exp G y2 dG y f

f

³

f

³ 0

f

dH

³ exp Gx dGx 2

f

f

exp Gz2 dGz ³ g 3 exp g 2 dg 0

2S

S

u³ sin D dD 0

(A-6)

³ dE 0

^

u Gx4 Gx2G y2 Gx2Gz2 Gx2 g 2 2Gx2 g x2

g

2 x

g cx2

`

2G y2 g x g y g x g y g cx g cy 2Gz2 g x g z g x g z g cx g cz , where g cx , g cy , and g cz are given by: g cx

g ª¬ cos D cos T sin D sin T cos H º¼ ,

g cy

g ª¬sin D cos E cos T cos D cos E sin T cos H sin E sin T sin H º¼ ,

(A-7a)

(A-7b)

338

Analytical Methods for Problems of Molecular Transport

g cz

g ª¬sin D sin E cos T cos D sin E sin T cos H . cos E sin T sin H º¼

(A-7c)

After integration over all values of G and neglecting vanishing integrals one can express the bracket integral, a11 , in the following form:

a11

3 4

2

V S

3 2

12f

§ kT · ¨ m¸ © ¹

S

u ³ sin D dD 0

³g

3

0

2S

2S

0

0

S

exp g 2 dg ³ sin T dT

³ d H ³ d E ^ 54 g

2

0

12 g 4

u ª¬ cos 2 D sin 2 T sin 2 D sin 2 T cos 2 H º¼

(A-8)

`

g 4 cos 2 D sin 2 T 8 3

12f

§ kT · ¸ © m¹

SV 2 ¨

7 2 ³ g exp g dg 0

12

§ kT · 8 SV 2 ¨ ¸ © m¹

.

Finally, in summary, the bracket integral, a11 , is given by: 12

ªcS 1 c 2 , cS 1 c 2 º 8 SV 2 § kT · , (A-9) ¨ m¸ 32 ¬ 32 ¼ © ¹ where one must remember that molecules have been assumed to be rigid spheres. a11

2.

BRACKET INTEGRALS CONTAINING SEVERAL COMPONENTS OF MOLECULAR VELOCITY.

In this section, the bracket integral associated with the first-order Chapman-Enskog solution for viscosity:

b11c

ª¬cx c y , cx c y º¼ ,

is considered. Using the dimensionless variables from Section A.1 and omitting vanishing integrals, one can obtain the following expression for this bracket integral:

339

Appendix A. Bracket Integrals for the Planar Geometry 12

b11c

1 4

ª¬cx c y , cx c y º¼ 2S

S

u³ sin T dT 0

2S

u ³ dE 0

f

³ 0

f

³ exp

§ kT · ¸ © m¹

V 2S 3 ¨ f 0

f

dG ³ exp

Gx2

0

G y2

x

f

S

d H ³ g 3 exp g 2 dg ³ sin D dD (A-10)

dG

y

f

^

u ³ exp Gz2 dGz g x g y g x g y g cx g cy f

` .

After integration over the components of G , and using the relations for g cx and g cy given by Eqs. (A-7a) and (A-7b), one can obtain:

b11c

1 4

12S

§ kT · ¸ © m¹

V 2S 3 2 ¨ f

2S

³ sin T dT

³ dH

0

0

S

u³ g 3 exp g 2 dg ³ sin D dD 0

2S

0

³ dE g

4

(A-11)

0

^

`

u sin 2 D cos 2 D cos 2 E sin T ª¬1 cos 2 H º¼ . The remaining simple integrations then yield: 12

b11c

¬ªcx c y , cx c y ¼º

4 5

§ kT ·

V2 S ¨ ¸ © m¹

.

(A-12)

Molecules have been assumed to be rigid spheres. Analytical expressions for other bracket integrals contained in higher-order Chapman-Enskog solutions and a more general calculation scheme for arbitrary intermolecular potentials may be found in [1-3].

3.

BRACKET INTEGRALS CONTAINING TWO DISCONTINUOUS FUNCTIONS.

In this section a method of calculating bracket integrals that contain the discontinuous sign functions is considered. As an example, the following bracket integral will be evaluated for the case of rigid-sphere molecules:

340

Analytical Methods for Problems of Molecular Transport

ª¬ c y sign cx , c y sign cx º¼

^

1 0 f 0 dv1 n 2 ³ c y sign cx ³ D T , g d : ³ gf

(A-13)

u ª c y sign cx c1 y sign c1 x ¬

`

ccy sign ccx c1c y sign c1cx º dv . ¼ This integral may be represented as the sum of four terms: 1 2 3 4 I I I I .

I

(A-14)

The first term does not contain the discontinuous functions and, therefore, may be calculated by the standard method described in Section A.1. Consider the second integral in detail. The integrand does not contain any collision parameters and, therefore, ³ D T , g d : can be calculated independently of the other integrations yielding:

³ D T , g d :

SV 2 .

Now, let the variables of integration be changed from v, v1 to , g v v . Then, after introducing the dimensionless variables G and G 1 g , this integral becomes:

I

2

2

V S

2

2S

f

u ³ dE 0

f

12f

§ kT · ¨ ¸ © m¹

S

3 2 ³ g exp g dg ³ sin D dD 0

f

0

³ exp Gx dGx ³ exp Gy dGy 2

f

2

(A-15)

f

^

u ³ exp Gz2 dGz G y2 g 2 sin 2 D cos 2 E f

u sign ª¬Gx g cos D º¼ ª¬Gx g cos D º¼

` .

After integration over all values of G y , Gz , and E this expression becomes:

341

Appendix A. Bracket Integrals for the Planar Geometry 12 f

§ kT · ¸ © m¹

2 I

V2¨ f

S

3 2 ³ g exp g dg ³ sin D dD 0

0

u ³ exp Gx2 dGx 1 g 2 sin 2 D

(A-16)

f

u sign ª¬Gx g cos D º¼ ª¬Gx g cos D º¼ . Note that the method of calculation used in these integrations has been previously described in [4,5]. Here, this method has been generalized by the introduction of appropriate recurrence relations. Using the new variables of integration, x cos D and z Gx g , one can reduce Eq. (A-16) to the form: 2 I

12 1

§ kT · ¸ © m¹

V2¨

f

³ dx

³

1

f

f

dz ³ g 4 exp g 2 ¬ª1 z 2 ¼º dg 0

u ª1 g 2 1 x 2 º sign > z x @> z x @ ¬ ¼ 12 1

§ kT · V ¨ ¸ © m¹ 2

(A-17)

f

³ dx ³ dz sign > z x @> z x@

1

f

u 83 S ª 1 z 2 «¬

5 2

52 1 x 2 1 z 2

7 2

º . »¼

It is convenient to use the following notations:

J m 1 2 1

1

f

2³ dx ³ 0

2

J m 1 2

1

sign > z x @> z x @

1 z 2

f

2

f

2³ 1 x dx ³ 0

m 1 2

sign > z x @> z x @

1 z 2

f

(A-18a)

dz ,

m 1 2

dz .

(A-18b)

After transformation, these integrals may be written as:

J m 1 2 1

1 x ° f ½° m 1 2 m 1 2 2 ³ dx ® ³ 1 z 2 dz 2 ³ 1 z 2 dz ¾ , x 0 ¯° f ¿°

(A-19a)

342

Analytical Methods for Problems of Molecular Transport

J m 1 2 and J m 1 2 . 1

Table A-1. Values of the integrals,

2

J m 1 2

J m 1 2

1

m

2

4 3 2

3 2 2 3 2 2 15 11 2 58 45 2

1

J 3 2

4 3 2 2

J 3 2

16 3

2

J 5 2

4 3

J 5 2

8 9

1

1

J 7 2 1

3

1

1

J m 1 2 2

2

28 23 2 8 7 2

2 15

J 9 2

4

2

J 7 2 2

8 45

J 9 2

4 105

2

13 35

2³ 1 x 2 dx 0

(A-19b)

f x m 1 2 m 1 2 ° °½ u® ³ 1 z2 dz 2 ³ 1 z 2 dz ¾ . x ¯° f ¿°

Then, using the following relationship [6]:

³

dz

1 z 2

z m 1 2

2 m 1 1 z 2

m 1 2

2m 2 2m 1

³

dz

1 z 2

m 1 2

, (A-20)

recurrence formulas for m t 2 may be obtained:

8

1 § ¨ 1 m 3 2 2 m 1 2 m 3 © 2

· 2m 2 1 J m 1 2 , ¸ ¹ 2m 1

J m 1 2

J m 1 2

8 § 25 2 m 2 25 2 m 1 · 2 m 2 2 J . ¨ ¸ 2m 1 © 2m 3 2m 5 ¹ 2m 1 m 1 2

1

2

(A-21a)

(A-21b)

Table A-1 gives the expressions for these integrals for several values of m . Using the integrals from Table A-1, one obtains: 2 I

12

3 8

§ kT · ¸ © m¹

SV 2 ¨

J 1 52

12

§ kT · SV 2 ¨ ¸ © m¹

1 6

5 2

J 7 2 2

3 2 2 .

(A-22)

343

Appendix A. Bracket Integrals for the Planar Geometry 3 Now, consider the third integral, I , which may be written as:

3 I

12f

§ kT · 14 V 2S 3 ¨ ¸ © m¹ 2S

f

u ³ dE 0

f

S

3 2 ³ g exp g dg ³ sin D dD 0

f

0

³ exp Gx dGx ³ exp Gy dGy 2

f

2

f

S

u ³ exp Gz2 dGz ³ sin T dT f

0

2S

³ dH 0

u ª¬G y g sin D cos E º¼ u ª¬G y g sin D cos E cos T (A-23) cos D cos E sin T cos H sin E sin T sin H º¼ sign Gx g cos D

u sign Gx g ª¬cos D cos T sin D sin T cos H º¼ . First, the variable of integration should be changed from T to T . Then, after integration with respect to E , G y , and Gz , the integrand may be written as:

^

`

S 2 1 g 2 ª sin 2 D cos T sin D cos D sin T cos H º ¬

¼

u sign Gx g cos D

u sign Gx g ª¬cos D cos T sin D sin T cos H º¼ . Here, it is convenient to introduce new collision angles to simplify integration over the collision parameters. Let the variables of integration be changed from g ,T , H to g c, D c, H c . The relationships between these variables are shown in Fig. A-1. From the spherical triangle, the following expressions may be obtained:

cos T cos D cos D c sin D sin D c cos H c ,

(A-24a)

cos D c cos D cos T sin D sin T cos H .

(A-24b)

Neglecting vanishing integrals, one obtains:

344

Analytical Methods for Problems of Molecular Transport

Figure A-1. The relationship between various collision angles.

I

3

12f

kT · V S ¨ ¸ © m¹ 1 4

1 §

2

f

³g 0

3

S

exp g 2 dg ³ sin D dD

S

S

0

0

0

u ³ exp Gx2 dGx ³ sin D c dD c³ d H c f

(A-25)

u sign ª¬Gx g cos D º¼ ª¬Gx g cos D c º¼ . At this point, changing the variables of integration to ( z Gx g , x cos (D ) , y cos (D c) ) and then performing the integrations with respect to g and H c , Eq. (A-25) becomes:

I

3

12 1

§ kT · SV ¨ ¸ © m¹ 2

163 f

u ³ dz 1 z

2

1

³ dx ³ dy

1

5 2

1

(A-26)

sign > z x @> z y @ .

f

For other bracket integrals in planar boundary value transport problems the integrand may contain the following expressions:

345

Appendix A. Bracket Integrals for the Planar Geometry

1 z 2

m 1 2

sign > z x @> z y @ ,

and:

1 x 1 z 2

2

m 1 2

sign > z x @> z y @ .

For these terms, let the following notations be introduced:

I m 1 2 1

1 1

I m 1 2 2

1

f

³ dx ³ dy ³

1

1

sign > z x @> z y @

1 z 2

f

1

f

1

f

³ 1 x dx ³ dy ³ 2

1

m 1 2

sign > z x @> z y @

1 z 2

(A-27a)

dz ,

m 1 2

dz .

(A-27b)

To facilitate calculations of these integrals it is convenient to obtain recurrence formulas that may be derived by using the relation given by Eq. (A-20). 1 First, consider the integral, I m 1 2 , which may be represented in the form:

I m 1 2 1

x f sign > z x @> z y @ ° dx dz ³ ® ³ dy ³ m 1 2 1 ° 1 f 1 z2 ¯ 1

½ sign > z x @> z y @ ° ³ dy ³ dz ¾ . m 1 2 x f ° 1 z2 ¿ 1

f

In the first term y d x and therefore:

1 ; f z y , x z f , sign > z x @> z y @ ® yzx . ¯ 1 ; In the second term y t x and therefore:

1 ; f z x , y z f , sign > z x @> z y @ ® xz y . ¯ 1 ;

(A-28)

346

Analytical Methods for Problems of Molecular Transport

1 Using these representations for sign > z x @> z y @ , the integral I m 1 2 may be expressed in the following form:

1

I m 1 2

1 f x x dz dz ° dx dy dy 2 ® ³ ³ ³ ³ ³ m 1 2 m 1 2 2 1 1 y 1 z ° 1 f 1 z 2 ¯ 1

½ dz ° . 2 ³ dy ³ m 1 2 ¾ 2 x x 1 z ° ¿

(A-29)

y

1

Now, by means of Eq. (A-20), the following recurrence formula may be obtained:

2m 2 1 211 2 m I m 1 2 2m 1 2m 1 2m 3

I m 1 2 1

1

8

dx

2m 1 2m 3 ³1 1 x 2 m3 2

(A-30)

;

mt2 .

1 From this general expression the basic integral, I 3 2 , becomes:

I 3 2 1

1

8 1 2 4³

dx

1

12

1 x 2

.

(A-31)

In the same way, the following recurrence formula may be obtained for 2 I m 1 2 :

I m 1 2 2

2m 2 2 211 2 m I m 1 2 2m 1 3 2m 1 2m 3

1 x dx 2m 1 2m 3 ³ 1 x 1 x dx ; m t 2 , 4 ³ 2m 1 1 x 1

2

1

1

with the basic integral:

2

1

4 2m 2

2

m 1 2

2

m 3 2

(A-32)

347

Appendix A. Bracket Integrals for the Planar Geometry

2

I3 2

16 3

1 x dx 2 4³ 1 x 2

1

1

2

1

.

12

(A-33)

Expressions for several of these integrals are given in Table A-2. Using these table integrals yields: 3 I

12

§ kT · 163 SV 2 ¨ ¸ © m¹

I 5 2 1

12

§ kT · 12 S V 2 ¨ ¸ © m¹

2 2 .

(A-34)

The integral given by Eq. (A-23) does not alter if the variable of integration, T , is changed from T to T S . This integral then becomes 4 equal to I , and: 3 I

4 I .

1 The integral, I , after simple integration may be written as:

I

1

12

§ kT · SV ¨ ¸ © m¹ 2

7 3

(A-35)

.

Thus, for ª¬ c y sign cx , c y sign cx º¼ , the final expression is given by:

I

1 2 3 I I 2 I

12

§ kT · 16 S V 2 ¨ ¸ © m¹

1 8 2 .

(A-36)

It is necessary to note that this bracket integral is different from that used in the Gross-Ziering theory. The relationship between these integrals is: 12

>) ,< @

§ m · >) ,< @ nS 3 2 ¨ ¸ © 2kT ¹

(A-37)

,

where >) ,< @ is the Gross-Ziering notation. obtains:

ª¬ c y sign cx , c y sign cx º¼

1 12

1 8 2 SO

1

Using this notation, one

.

This expression is identical to that obtained previously in [7].

(A-38)

348

Analytical Methods for Problems of Molecular Transport

Table A-2. Values of the integrals,

1 2 I m 1 2 and I m 1 2 .

I m 1 2

I m 1 2 1

m

I 3 2 1

1

8 1 2 1

2

4 ³ 1 x 2 1

I 3 2 2

1 2

8 3

I5 2

2

2 2

1 2 1

4 ³ 1 x 2

dx

1

I 5 2 2

1

16 3

8 9

I 7 2 1

I 9 2 1

4

4.

4 15 2 105

16 9 2

192 115 2

2

3 2

2

4 3 2 1

43 ³ 1 x 2 1

3

1 2

1 x dx

I 7 2 2

4 45

I 9 2

2 105

2

1 x dx

32 17 2 128 73 2

BRACKET INTEGRALS CONTAINING ONE DISCONTINUOUS FUNCTION.

In the case of the planar geometry for stationary transport problems, a moment system can be obtained by means of the Maxwellian integral equation given by Eq. (11.52). For these problems it is convenient to employ the discontinuous distribution functions which might satisfy the boundary conditions at the wall surface. In the full velocity space these distribution functions for the linearized transport problems can be presented by the following general expression:

f

^

0 f 1 12 ª¬) 2 c, x )1 c, x º¼

`

12 ª¬) 2 c, x )1 c, x º¼ sign cx ,

(A-39)

where ) 2 c, x and )1 c, x are the corrections to the distribution function for which cx ! 0 and cx 0 , respectively, and x is a normal coordinate to the wall. If this distribution function is employed for closing the full velocity space moment system, the right-hand-sides of moment equations contain both ordinary bracket integrals and others which can be expressed in the form:

ª¬< c sign cx , Q c º¼ ,

(A-40)

349

Appendix A. Bracket Integrals for the Planar Geometry

where < c and Q c are functions of the molecular velocity components. Consider, as an example, evaluation of the bracket integral defined by:

I

ªsign cx , c 2 cx º . ¬ ¼

(A-41)

This integral occurs in the planar heat transport problem (see Section 11.9) and can be obtained from Eq. (11.16) by substitution of r 1 and cr 0 . Integrating with respect to the variables, T , H , and g , and introducing new variables of integration via the following relationships:

Gx

xg , G y

yg , Gz

zg ,

12 1

f

cos D t ,

one can obtain:

I

§ 2kT0 · 96V 2S 1 ¨ ¸ © m ¹

f

f

³ dt ³ dx ³ dy ³ F t , x, y, z dz ,

1

t

f

(A-42)

f

where:

F

x t 2 13 1 x 2 y 2 z 2

5

,

and molecules have been assumed to be rigid spheres. In order to facilitate the integration procedure in Eq. (A-42) with respect to variables y and z , it is convenient to introduce polar coordinates by means of the relationships:

y

U cos \ , z U sin \ .

(A-43)

Then, after integrations with respect to \ and U , this integral becomes:

I

12 1

§ 2kT0 · 24V ¨ ¸ © m ¹ 2

³ t

1

2

f

13 dt ³ 1 x 2 t

4

xdx .

(A-44)

The integrations with respect to x and t offer no difficulty. On integrating over all values of these variables one can obtain: 12

I

§ 2kT0 · 2 23 ¨ ¸ V . © m ¹

The same value of this integral has been obtained in [8].

(A-45)

350

Analytical Methods for Problems of Molecular Transport

REFERENCES 1. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 2. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 3. Hirschfelder, J.O., Curtiss, C.F., and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954). 4. Rolduguin, V.I., Application of the Non-Equilibrium Thermodynamics Method in Boundary Problems of the Kinetic Theory of Gases (M.S. Thesis, Moscow, 1979). 5. Savkov, S.V., The Slip Boundary Conditions of Non-Uniform Binary Gas Mixture and Application of them in Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 6. "Mathematical Tables," in Handbook of Chemistry and Physics, 71st edition, edited by Lide, D.R. (CRC Press, Boston, 1990), #283, p.A-38. 7. Derjaguin, B.V. and Yalamov, Yu.I., "The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and Their Experimental Testing," In Topics in Current Aerosol Research, vol. 3, part 2, edited by Hidy, G.M. (Pergamon Press, Oxford, 1972). 8. Ivchenko, I.N., "Generalization of the Lees Method in Boundary Problems of Transfer," J. Colloid and Interface Sci. 135(1), 16-19 (1990).

Appendix B BRACKET INTEGRALS FOR CURVILINEAR GEOMETRIES

1.

THE SPECIAL FUNCTION OF THE FIRST KIND FOR THE SPHERICAL GEOMETRY.

Here, the special functions introduced in Section 11.4 will be

1 investigated in detail. For spherical transport problems, the function I1 > @ r is determined by Eq. (11-21) and, taking into account Eqs. (11-17) and (1118), the following expression may be written: 1

2S

I1 > @ r 72S 2 ³ dt ³ d E

1

1

0

f

³

f

f

f

dy ³ dz ³ F t , E , y, z , x dx , f

(B-1)

x0

where the notations introduced in Eq. (11-17) have been used. It is very difficult to use Eq. (B-1) to obtain numerical values for

1 I1 > @ r , however some analytical simplification is readily achievable. The integration with respect to x offers no difficulty if the following recurrence formula is used [1]:

³

dx

a x 2

ª

m 1

1 x 2ma a x 2

m

2m 1 dx 2ma ³ a x 2

m

(B-2)

º

r r 1 2m « x m 1 dx » ¦ ³ 2 « r m m «¬ 2a r 1 4a mr 2r a x 2 4a a x 2 »¼»

.

352

Analytical Methods for Problems of Molecular Transport

To integrate with respect to y and z , the variables of integration are changed from y and z to y c and z c which are defined by: 12

yc

y 1 t2

zc

z 1 t2

12

cos E ,

(B-3a)

sin E .

(B-3b)

After this substitution it is convenient to introduce the following polar coordinates:

U cos M c ,

yc

zc

U sin M c .

Then, the substitution, M M c E , permits one to integrate over all values of E . Having performed this transformation and integration, one obtains the following analytical expression: I1 > @ r

1

1

2S

f

1

0

0

144

S

>1@ >1@ >1@ >1@ >1@ >1@ ³ dt ³ dM ³ U f1,1 \ 1,1 f1,2\ 1,2 f1,3\ 1,3 d U .

(B-4)

>@ Here, the quantities f1,i> @ and \ 1,i may be written as: 1

f1,1> @ 1

>@ f1,2 1

t 2 13 ,

(B-5a)

ªt 1 t 2 1 2 cos M U 2 t 1 t 2 U º r 2 1 1 2 «¬ »¼ 12 t 2 ª 1 t 2 1 t 2 cos M U º , «¬ »¼

>@ f1,3 1

>1@ \ 1,1

1

t 1 t2

1 8

12

a x 2 0

,

(B-5b)

cos M U t 1 t 2 ,

4

(B-5c)

(B-5d)

353

Appendix B. Bracket Integrals for Curvilinear Geometries

>i @

Table B-1. Numerical values of the functions, I1 r . Note that for i 1, 2 , the apparent values of these functions (evaluated numerically) appear to be the same although the analytical formulations of these functions appear to be different. As of this time these functions have not been proven to be identical.

r

i I > @ r

r

i I > @ r

r

i I > @ r

1.00 1.10 1.20 1.30 1.40

9.982Eí1 8.267Eí1 6.944Eí1 5.917Eí1 5.093Eí1

1.50 1.75 2.00 2.50 3.00

4.440Eí1 3.261Eí1 2.489Eí1 1.591Eí1 1.116Eí1

3.50 4.00 5.00 6.00 10.0

8.143Eí2 6.052Eí2 3.982Eí2 2.791Eí2 1.049Eí2

>1@ \ 1,2

35 128

a 4 a x02

>1@ \ 1,3

35 128

7 48

a

a 9 2

2

1

35 3 a a x02 192

a x02

3

1 8

a

1

x0

t U r2 1

12

12

2

a

U 2 2U 1 t 2

x02

S arctan x0 a 1 2

1 2

a

4

(B-5e)

,

,

(B-5f)

(B-5g)

cos M 2 t 2 ,

.

(B-5h)

The numerical values of the functions, I1 > @ r and I1 > @ r (which will be discussed in Section B.2), are given in Table B-1 and the dependence of

i I1 > @ r on the radial coordinate is shown in Fig. B-1. A numerical analysis

i shows that the special functions, I1 > @ r , are essentially equal to each other for i 1, 2 with the relative deviations of the numerical values being less than 1% although equality has not been definitively established.

1

2.

2

THE SPECIAL FUNCTION OF THE SECOND KIND FOR THE SPHERICAL GEOMETRY.

The second special function related to spherical boundary value transport problems can be expressed in the form: 1

2 I1 > @ r 1440S 2 8 3S

1

2S

³ dt ³ d E

1

0

f

³

f

f

f

dy ³ dz ³ F) dx , f

x0

(B-6)

354

Analytical Methods for Problems of Molecular Transport

1.0

Normalized Special Function Value

0.8

0.6

0.4

0.2

0.0 0

2

4

6

8

10

Dimensionless Radial Coordinate

Figure B-1. Dependence of

I1 > @ r on the radial coordinate, r .

i

where the notations of Section 11.4 are used. By means of the transformations employed in Section B.1, this integral can be expressed in the form: 1 1 1

2 I1 > @ r 20 8 3S I1 > @ r 2880S 1 8 3S 1

2S

f

1

0

0

^

`

2 > 2@ > 2@ > 2@ > 2@ > 2@ u ³ dt ³ dM ³ U d U f1,1> @\ 1,1 f1,2 \ 1,2 f1,3 \ 1,3 ,

(B-7)

> @ may be written as: where the quantities f1,i> @ and \ 1,i 2

f1,1> @

15 1 t 2

> @ f1,2

1 t 5

2

2

t

2

2

12

2t

2

13 ª 1 t 2 «¬

12

cos M U º , »¼

13 95 t 1 t 2 a 1

u ª 1 t 2 cos 2 M U 2 2 1 t 2 «¬

12

cos M U º , »¼

(B-8a)

(B-8b)

355

Appendix B. Bracket Integrals for Curvilinear Geometries > @ f1,3 2

2t U 2 ª 1 t 2 ¬ 5

> 2@ \ 1,1

a x

> 2@ \ 1,2

>@ >@ x0\ 1,2 \ 1,3 ,

> 2@ \ 1,3

1 10

2 0

12

2t cos M U cos M t º¼ ,

2t 2t 4 133 t 2 53 4t 1 t 2 2

2

2

4 3

(B-8c)

1 3

,

1

(B-8d)

1

(B-8e)

x0 a 1 a x02

5

.

(B-8f)

Here, the quantities a and x0 are the same as those given in Eqs. (B-5g) and (B-5h), respectively. Numerical values of this special function are given in Table B-1.

3.

THE SPECIAL FUNCTION OF THE FIRST KIND FOR THE CYLINDRICAL GEOMETRY.

This function is defined by Eqs. (11-18) and (11-21). Since the lower limit of integration with respect to x is independent of the variable, z , it is convenient to perform the first integration over all values of z and then to integrate with respect to x . Having performed these integrations, one can obtain a triple integral which can be numerically evaluated. To integrate over all values of y , first change the variable of integration to y c y (1 t 2 )1 2 cos E and then break this integral into two parts by means of the relationship: f

³

dy c

f

0

f

³

dy c ³ dy c .

f

0

After these very simple transformations, one can obtain: 1

2S

f

1

0

0

^

>1@ >1@

S 1 ³ dt ³ d E ³ dy I 2> @ r 315 16

1

>1@ >1@

(B-9) >1@

>1@

>1@

>1@

`

u f 2,1\ 2,1 f 2,2\ 2,2 g 2,1) 2,1 g 2,2) 2,2 ,

356

Analytical Methods for Problems of Molecular Transport

where the following notations are introduced:

t

>@ f 2,1

>@ g 2,1

>@ f 2,2

t 1 t 2

>@ g 2,2

t 1 t 2

>1@ \ 2,1

c x

1

1

1

>1@ \ 2,2

1

1 7

13 ,

cos E ª 1 t 2 «¬

12

cos E ª 1 t 2 «¬

7 2

x0 c x02

>1@ ) 2,2

d x 2 0

1 35

12

cos E y º , »¼

(B-10b)

12

cos E y º , »¼

(B-10c)

,

(B-10d)

1 2

6c

>1@ ) 2,1

(B-10a)

12

2 0

1 35

2

7 2

x0 d x02

2

ª16c 4 8c 3 c x 2 0 «¬

c x02

2

5c

1

1

c x02

3

º . »¼

,

(B-10f)

1 2

6d

(B-10e)

2

ª16d 4 8d 3 d x 2 0 «¬

d x02

2

5d

1

1

d x02

3

º . »¼

(B-10g)

357

Appendix B. Bracket Integrals for Curvilinear Geometries

>i @

Table B-2. Numerical values of the functions, I 2 r . Note that again, as in Table B-1, the values of these functions appear to be the same but that analytical equality has not been proven. Here, the values for the different functions have been given separately in order to demonstrate their similarity.

r

I 2> @ r

2 I 2> @ r

r

I 2> @ r

2 I 2> @ r

1.00 1.10 1.20 1.30 1.40 1.50 1.75 2.00 2.50 3.00 3.50 4.00 5.00 6.00 7.00

0.9999 0.9091 0.8333 0.7692 0.7142 0.6666 0.5714 0.4999 0.3999 0.3333 0.2857 0.2499 0.2000 0.1666 0.1426

1.0000 0.9090 0.8333 0.7691 0.7142 0.6666 0.5714 0.5000 0.4000 0.3333 0.2857 0.2500 0.1999 0.1666 0.1430

8.00 9.00 10.00 12.00 15.00 17.00 20.00 25.00 30.00 50.00 60.00 70.00 80.00 90.00 100.00

0.1249 0.1111 0.1000 0.0833 0.0665 0.0588 0.0500 0.0399 0.0334 0.0200 0.0165 0.0151 0.0123 0.0112 0.0110

0.1250 0.1110 0.0999 0.0833 0.0667 0.0588 0.0499 0.0400 0.0332 0.0199 0.0167 0.0132 0.0123 0.0110 0.0090

1

1

In Eqs. (B-10a)-(B-10g) the quantities x0 , c , and d are defined as:

x0

12

t y r2 1

c 1 ª y 1 t2 «¬

,

12

(B-11a)

2

cos E º , »¼

(B-11b)

and:

d 1 ª 1 t2 «¬

12

2

cos E y º . »¼

(B-11c)

Numerical values of the functions, I 2> @ r and I 2> @ r , are given in

1 Table B-2 and the dependence of I 2> @ r on the radial coordinate is shown z 1 in Fig. B-2. Moreover, the numerical values of ³ I 2> @ r dr for various 1 values of z are given in Table B-3.

1

2

358

Analytical Methods for Problems of Molecular Transport

Normalized Special Function Value

1

0.8

0.6

0.4

0.2

0 1

10

100

Dimensionless Radial Coordinate

Figure B-2. Dependence of

4.

I 2> @ r on the radial coordinate, r .

1

THE SPECIAL FUNCTION OF THE SECOND KIND FOR THE CYLINDRICAL GEOMETRY.

The special function, I 2> @ r , is determined by Eqs. (11-19) and (11.22). Performing the same transformations as in Section B.3, one can obtain:

2

1 1 1

2 I 2> @ r 20 8 3S I 2> @ r 315 S 1 8 3S 4 1

2S

f

(B-12)

u ³ dt ³ d E ³ dy 1

0

0

^

`

> 2@ > 2@ > 2@ > 2@ > 2 @ > 2@ > 2 @ > 2@ u f 2,1 \ 2,1 f 2,2 \ 2,2 g 2,1 ) 2,1 g 2,2 ) 2,2 ,

where the following functions are employed:

359

Appendix B. Bracket Integrals for Curvilinear Geometries Table B-3. Numerical values of the integrals, \ i

z

z ³1 I 2 >i@ r dr .

z

\1

\2

z

\1

\2

1.00 1.05 1.10 1.20 1.30 1.40 1.50

0.0000 0.0488 0.0935 0.1823 0.2624 0.3365 0.4055

0.0000 0.0488 0.0935 0.1823 0.2624 0.3365 0.4055

1.75 2.00 3.00 4.00 5.00 7.00 10.00

0.5596 0.6931 1.0984 1.3860 1.6090 1.9453 2.3019

0.5596 0.6931 1.0983 1.3860 1.6090 1.9454 2.3021

> @ f 2,1 2

ª t ® t 2 13 1 t 2 «sin 2 E ¬ ¯

8c 1 §¨ 1 t 2 ©

> @ f 2,2 2

12

1 t 2t 2

2

12

2

º °½ cos E y ·¸ cos 2 E » ¾ , ¹ ¼ ¿°

13 §¨ 1 t 2 ©

x0 c 1t 1 t 2 §¨ 1 t 2 ©

12

12

(B-13a)

cos E y ·¸ cos E ¹ 2

cos E y ·¸ cos 2 E ¹

(B-13b)

x0t t 2 13 ,

> @ g 2,1 2

ª t ® t 2 13 1 t 2 «sin 2 E ¬ ¯

1 §

8d ¨ 1 t ©

> @ g 2,2 2

12

1 t 2t 2

2

2

12

2 º °½ cos E y ·¸ cos 2 E » ¾ , ¹ ¼ ¿°

13 §¨ 1 t 2 ©

x0 d 1t 1 t 2 §¨ 1 t 2 ©

x0t t 2 13 ,

12

12

(B-13c)

cos E y ·¸ cos E ¹ 2

cos E y ·¸ cos 2 E ¹

(B-13d)

360

Analytical Methods for Problems of Molecular Transport > 2@ \ 2,1

>@ \ 2,2 ,

> 2@ \ 2,2

c x

> 2@ ) 2,1

>@ ) 2,2 ,

> 2@ ) 2,2

d x

1

2 0

(B-13e) 1

c x

>1@ \ 2,1

2 0

9 2

,

(B-13f)

1

2 0

(B-13g) 1

>1@ ) 2,1

d x 2 0

9 2

.

(B-13h)

The quantities shown in these formulas are identical to those given in Eqs.

2 (B-10a)-(B-11c). Numerical values of I 2> @ r are given in Table B-2 and z > 2@ numerical values of ³ I 2 r dr are given in Table B-3. 1

5.

APPROXIMATE EXPRESSIONS FOR THE SPECIAL FUNCTIONS.

Using the general properties of the Boltzmann equation [2], one can obtain very simple approximate expressions for the bracket integrals considered in Sections B.1-B.4. These approximate expressions will be derived in this section by employing the first-order approximation to the Chapman-Enskog solution given in Eq. (5-39). In order to use the first-order approximation of the Chapman-Enskog solution, the bracket integrals must be transformed into the more convenient form:

ª c k sign cr cr , c 2 cr º ¬ ¼

1 ªc k sign cr cr , cr S3 2 c 2 º ¬ ¼

³ c k sign cr cr I cr S3 2 c 2 dv , k 1

(B-14)

0, 2 .

Now, using Eq. (5-39), these bracket integrals can be approximated by:

ª c k sign cr cr , c 2 cr º ¬ ¼ 1 1 0 v S c 2 f c k sign cr cr dv , 2 1 ³ r 3 2 n a1

1 where a1 is given by Eq. (5-40).

(B-15)

Appendix B. Bracket Integrals for Curvilinear Geometries

361

It is very important to note that the integrals in the right-hand-side of Eq. (B-15) are ordinary integrals in the velocity space. The values of these integrals depends on the problem geometry. For the spherical geometry, after integration, one can obtain the following expressions: I1 > @ r

4 5

2 I1 > @ r

64 5

1

2r 2 ,

(B-16)

1

2 8 3S r 2 .

(B-17)

The appropriate expressions for the cylindrical geometry can be written as: I 2> @ r

4 5

2 I 2> @ r

64 5

1

2r 1 ,

(B-18)

1

2 8 3S r 1 .

(B-19)

The accuracy of this analysis can be easily determined by means of a comparison with the appropriate numerical values. The relative errors for

1

2 I i > @ r and I i > @ r , for r 1 , are 13% and 4%, respectively.

REFERENCES 1. "Mathematical Tables," in Handbook of Chemistry and Physics, 71st edition, edited by Lide, D.R. (CRC Press, Boston, 1990), #67, p.A-24. 2. Loyalka, S.K., "Approximate Method in the Kinetic Theory," Phys. Fluids 14(11), 22912294 (1971).

Appendix C BRACKET INTEGRALS FOR POLYNOMIAL EXPANSION METHOD

1.

CALCULATION OF THE BRACKET INTEGRALS OF THE FIRST KIND.

To complete the calculations of the polynomial expansion method in the four-moment approach for the spherical geometry, one must know values of the following bracket integrals:

D k>1n@

ª H k cr S0 n cT2 cI2 , cr c 2 º , ¬ ¼

D k> 2n@

ª c 2 H k cr S0 n cT2 cI2 , cr c 2 º , ¬ ¼

where H k cr and S0 cT2 cI2 are, respectively, Hermite and Sonine polynomials of the molecular velocity components. First, consider the 1 method of calculating the bracket integrals, D k> n@ , when molecules are assumed to be rigid spheres. To obtain general analytical expressions for these integrals, one must use the generating functions for the Hermite and Sonine polynomials, H x, t and S0 x, s , which are given by: n

H x, t exp t 2 2 xt , and:

x cr ,

(C-1)

364

Analytical Methods for Problems of Molecular Transport

1 § xs · exp ¨ ¸ , 1 s © 1 s ¹

S 0 x, s

x cT2 cI2 .

(C-2)

Then, the Hermite and Sonine polynomials may be expressed as:

Hk x

wk H x, t wt k t

(C-3)

, 0

and: n S0 x

1 wn S 0 x, s n ws n s

.

(C-4)

0

Having used Eqs. (C-3) and (C-4), one can obtain: >i @

D kn

2 1 w k w n exp t [ >i @ t , s n wt k ws n 1 s

,

(C-5)

t ,s 0

where:

ª

§ ©

s

·¹

º

[ >i@ t , s «c 2 i 1 exp ¨ 2cr t cT2 cI2 ¸ , cr c 2 » . 1 s ¬

(C-6)

¼

The bracket integral of Eq. (C-6) can be calculated in the same manner as the original one but it has some special properties that can be taken into account to facilitate the calculations. The analysis is simplified if the following new variables of integration are introduced:

t

,

(C-7)

Grc

Gr

GTc

GT

s g sin D cos E , 2s

(C-8)

GIc

GI

s g sin D sin E , 2s

(C-9)

2

and:

Appendix C. Bracket Integrals for Polynomial Expansion Method

365

where Gr , GT , GI are the Cartesian coordinates of the dimensionless velocity, G , and g ,D , E are the spherical coordinates of the dimensionless velocity, g . Then, having performed six integrations with respect to the variables, T , H , E , Grc , GTc , and GIc , and having employed Leibnitz’ formula [1] for the k -th derivative with respect to t , one can obtain: >1@ D kn

12

16 3

§ 2kT · ¸ © m ¹

S 1 2V 2 ¨

1 wn 1 2 s ³ sin D dD n n ws 0 S

f § s ª º· sin 2 D » ¸ u³ g 3 exp ¨ g 2 «1 ¬ 2s ¼¹ © 0

u< k> @ s,D , g dg 1

s 0

(C-10)

,

where:

< k>1@ s,D , g

3 2

F k 0, g

s g 3 f 2 D 2s

1 k 1 k F 0, g g 2 f1 D , 2

F t , g exp 12 t 2 2 gt cos D ,

f1 D

3 2

cos 2 D 12 ,

and:

f 2 D sin 2 D cos D . k To calculate the k -th derivative, F 0, g , one should use the recurrence relation that is given by:

F k 1 0, g 2 g cos D F k 0, g k F

k 1

0, g

;

k t0 ,

(C-11)

366

Analytical Methods for Problems of Molecular Transport

Table C-1. Analytical values of the bracket integral, method.

>1@ , D1n

for the polynomial expansion

12

D1>1n@

anV 2 2S 2kT m

n

1

2

3

4

an

16 15

16 105

4 315

8 3465

with:

F 0 0, g 1 .

(C-12)

>@ For the particular case when the index, k 1 , the bracket integrals, D1n , may be expressed in the form: 1

D1>1n@

12

§ 2kT · 163 2S V 2 ¨ ¸ © m ¹

1

1 1 M n x dx , n ³0

(C-13)

where:

Mn1 x

3 2

9 1 x x 4nQ

n n 1 n2 x 2 12 4Q 3 4nQ 3 n n 1 Q 3 2

2

4

n 1

4n n 1 Q 4

n n 1 n 2 Q 4

n 3

n 2

.

Here, the following notations are introduced:

Q 3 m

Q 4 m

3 4 5 3 m 1 23 m

4 5 6 4 m 1 24 m

x 2m ;

x2m ;

m t1 ,

m t1 .

>@ Specific values of D1n are given in Table C-1. 1

(C-14)

(C-15)

367

Appendix C. Bracket Integrals for Polynomial Expansion Method

2.

ANALYTICAL EXPRESSIONS FOR THE BRACKET INTEGRALS OF THE SECOND KIND.

The method described in the preceding section may be used to calculate 2 the integrals, D k> n@ , as well. After some fairly complicated algebraic transformations, one can obtain: 12

> 2@ D kn

§ kT · 2SV 2 ¨ ¸ © m¹

8 3

f

1 wn 1 2 s ³ sin D dD n n ws 0 S

§ s ª º· 2 u ³ g exp ¨ g 2 «1 sin 2 D » ¸< k> @ s,D , g dg ¬ 2s ¼¹ © 0 s

(C-16)

3

, 0

where:

°

< k> 2@ s,D , g F k 0, g ® f1 D g 3 cos D °¯

2 ª º½ · s § 3s 2 4s 2 3 3 13s 22 s 8 » ° ¨1 ¸ D f 2 D « 32 g 5 sin g ¾ 2 2 ¸ 4 2s¨ « »° 2 s 2 s © ¹ ¬ ¼¿

1 2

kF

k 1

§ 3s 2 4s ·º ª 10 7 s 2 2 ¸ » f1 D D g g 4 ¨1 sin 2 ¨ ¸ 2 s » s 2 « ¯° ¬ © ¹¼

0, g ®°« 12

½° 3s k 2 g 4 cos D f 2 D ¾ k k 1 F 0, g 2 s °¿

s ° °½ u ® g 3 cos D f1 D 34 g 3 f 2 D ¾ 2 s ¯° ¿° 1 k 3 12 k k 1 k 2 g 2 F 0, g f1 D , 2

f1 D

3 2

cos 2 D 12 , and f 2 D sin 2 D cos D .

> @ Here, for the specific case of k 1 , the integrals, D1n , may be expressed in the form: 2

368

Analytical Methods for Problems of Molecular Transport

Table C-2. Analytical values of the bracket integral, method.

> 2@ , D1n

for the polynomial expansion

12

D1>n2@ bnV 2 2S 2kT m n

1

2

3

4

bn

163

46 15

38 105

19 630

> 2@

D1n

12

4 3

§ 2kT · 2S V ¨ ¸ © m ¹ 2

1

1 2 M n x dx , n ³0

where:

Mn 2 x

^88 96 x Q n 168 116 x Q n n 1 110 36 x Q n n 1 n 2 24 x Q ` 3 1 x x ^96Q n 552 48 x Q n n 1 816 12 x Q n n 1 n 2 474 72 x Q n n 1 n 2 n 3 96 27 x Q ` ,

3 2

x 2 12

n

2

n 1

2

4

4

n2

2

4

n 3

2

4

2

n

2

n 1

2

5

5

n2

2

5

n 3

2

5

n4

2

5

and:

Q 5 m

5 6 7 5 m 1 25 m

x2m ;

m t1 .

> @ Some values of D1n are given in Table C-2. 2

REFERENCES 1. Bronshtein, I.N. and Semendyayev, K.A., A Guidebook to Mathematics (Springer Verlag, New York, N.Y., 1973).

Appendix D THE VARIATIONAL PRINCIPLE FOR PLANAR PROBLEMS

1.

SOME DEFINITIONS AND PROPERTIES FOR INTEGRAL OPERATORS.

In the Kinetic Theory of Gases, the following form of an integral operator, Q , is frequently used:

Q) c, r

³ K c, cc ) cc, r dcc ,

(D-1)

where the integration extends over all values of the molecular velocity. For planar transport problems, the appropriate operator is defined by:

Q) c, x

³ K c, cc ) cc, x dcc .

(D-2)

To begin with, a fundamental definition for this integral operator is introduced. The adjoint operator, Q , in the space defined by the scalar product given by Eqs. (9-16) and (9-17) is specified by:

ª¬< , Q) º¼

ª ) , Q < º . ¬ ¼

(D-3)

The operator, Q , is termed a self-adjoint operator if:

Q

Q .

(D-4)

370

Analytical Methods for Problems of Molecular Transport

For a self-adjoint operator, Eq. (D-3) becomes: ª¬< , Q) º¼

ª¬) , Q< º¼ .

(D-5)

It has been proven in [1,2] that the condition for the operator, Q , to be selfadjoint is that:

K c, cc

K cc, c .

(D-6)

Another way of stating Eq. (D-6) is to say that the function, K c, cc , in Eqs. (D-1) and (D-2) is symmetric in the variables, c and cc . From the symmetry of the kernel, K c, cc , in Eq. (9-14), one can easily conclude that the operator, H , is self-adjoint. In the formulation of the variational

principle, the adjoint operator, H L , where L is defined by Eq. (9-13), is

employed to facilitate certain transformations. The adjoint operator, H L , may be derived from the following relationships:

ª¬< , H L) º¼

ª¬ H < , L) º¼

ª ) , L H < º ¬ ¼

ª ) , L H < º , ¬ ¼

(D-7)

which yield:

H L 2.

L H

L H .

(D-8)

THE VARIATIONAL PRINCIPLE.

In the linearized planar transport problems studied to date, it has generally been the case that the main quantity of physical interest can be expressed as:

I

ª ) , p º ¬ ¼

ª p , ) º , ¬ ¼

(D-9)

where the following notation has been introduced:

)

H) .

(D-10)

Appendix D. The Variational Principle for Planar Problems

371

and ) is the correction to the distribution function which satisfies the following integral form of the Boltzmann equation:

) c, x L) c, x p c, x .

(D-11)

Applying the operator, H , to Eq. (D-11), one obtains:

) c, x H L) c, x p c, x L) c, x p c, x ,

(D-12)

where L H LH 1 and H 1 is the inverse of H . Now, a functional may be constructed of the form:

F ) ,)

ª ) , p º ª ) ,) L) p º , ¬ ¼ ¬ ¼

(D-13)

where ) is a trial function. If this trial function is equal to the real correction to the distribution function, ) , then the integral, I , is found from:

F ) ,) .

I

(D-14)

Now, consider a trial function of the form:

) ) G) .

(D-15)

For this trial function, the functional of Eq. (D-13) becomes:

F ) G) ,) G)

ª p ¬

, ) º O ªG) º , ¼ ¬ ¼

ª p , ) º ª G) , ) L) p º ¬ ¼ ¬ ¼

(D-16)

2

which implies that:

GF

2 O ªG) º . ¬ ¼

(D-17)

Since the first variation is equal to zero, the functional is stationary when ) c, x ) c, x and ) c, x ) c, x . Now, the basic integral of

372

Analytical Methods for Problems of Molecular Transport

interest, I , can be expressed in terms of the stationary value of the functional by the following simple expression:

I

F ) ,)

Fst ) ,) ,

(D-18)

and, hence, to find an approximate value for I , one need only employ a simple trial function satisfying the stationary condition [3-10]. It should be noted in this context that a non-self-adjoint operator such as the Boltzmann operator cannot lead to an extremum (minimum) but, rather, must yield a saddle point because it has been proven in [4] that the second variation, G 2 F , may have either sign.

REFERENCES 1. Morse, P.M. and Feshbach, H., Methods of Theoretical Physics (McGraw-Hill, NY, 1953). 2. Courant, R. and Hilbert, D., Methods of Mathematical Physics, Vol. 1 (Interscience, NY, 1953). 3. Kahan, T., Rideau, G., and Roussopolous, P., Les Methodés d'Approximation Variationelles Dans la Theorié des Collisions Atomiques et Dans la Physique des Piles Nucléaires (Gauthier-Villars, Paris, 1956). 4. Pomraning, G.C. and Clark, M.Jr., "The Variational Method Applied to the Monoenergetic Boltzmann Equation. Part I," Nucl. Sci. Engineering 16, 147-154 (1963). 5. Cercignani, C. and Pagani, C.D., "Variational Approach to Boundary-Value Problems in Kinetic Theory," Phys. Fluids 9(6), 1167-1173 (1966). 6. Loyalka, S.K., "Momentum and Temperature-Slip Coefficients with Arbitrary Accommodation at the Surface," J. Chem. Phys. 48(12), 5432-5436 (1968). 7. Lang, H., "Ein Variationsprinzip für die Linealisierte, Stationäre Boltzmann-Gleichung der Kinetischen Gastheorie," Acta Mechanica 5, 163-188 (1968). 8. Loyalka, S.K., "Linearized Couette Flow and Heat Transfer Between Two Parallel Plates," in Rarefied Gas Dynamics, edited by Trilling, L. and Wachman, H.Y. (Academic Press, New York) Proceedings of the 6th International Symposium on Rarefied Gas Dynamics (1969). pp. 195-203. 9. Loyalka, S.K., "Slip Problems for a Simple Gas," Z. Naturforsh. 26a(6), 964-972 (1971). 10. Loyalka, S.K., "Slip in the Thermal Creep Flow," Phys. Fluids 14(1), 21-24 (1971).

Appendix E SOME DEFINITE INTEGRALS

1.

SOME FREQUENTLY ENCOUNTERED INTEGRALS. One commonly encountered integral is of the form (see Table E-1): f

in

³x

n

exp x 2 dx .

0

For even and odd values of n , respectively, this may be rewritten as: f

³x

2n

0

f

³x 0

exp E x 2 dx

2 n 1

2n 1 2n 3 5 3 1

S

2n 1

E 2 n 1

exp E x 2 dx

1 2

n

E n 1

;

nt0 .

(E-1)

;

n t1 ,

(E-2)

374

Analytical Methods for Problems of Molecular Transport

Table E-1. Values of the commonly encountered n

0 1 2

in

5

in

6

10 945 64

in

2.

1 4

15 16

1

n

2

1 2

S

n

in integrals.

1

S

60

3 8

8 105 32

3 12 10395 128

4

1 2

S 7

11

S

3

S 9

S

12

13

S

360

SOME INTEGRALS ENCOUNTERED IN BOUNDARY PROBLEMS. 2 ³ cx exp c dc

1 2

S ,

(E-3)

exp c 2 dc S ,

(E-4)

³ cx c

2

2 2 ³ cx exp c dc

1 4

S3 2 ,

(E-5)

2 2 ³ cx c y exp c dc

1 4

S ,

(E-6)

2 2 2 ³ cx c y exp c dc

1 8

S3 2 ,

(E-7)

5 8

S3 2 ,

(E-8)

1 4

2 2

³ cx c

exp c 2 dc

³ cx c

2

32 exp c 2 dc

S ,

(E-9)

375

Appendix E. Some Definite Integrals 2 2 2 ³ cx c c 32 exp c dc

3 2

S ,

(E-10)

S3 2 ,

(E-11)

³ cx c 2

2

23 exp c 2 dc

³ cx c

2

52 exp c 2 dc 14 S ,

1 4

(E-12)

2 2 2 ³ cx c c 52 exp c dc

1 2

S ,

(E-13)

³ cx c 2

2

52 exp c 2 dc 0 ,

(E-14)

2 1 2 2 ³ cx c y S3 2 c exp c dc

18 S ,

(E-15)

2 2 1 2 2 ³ cx c y S3 2 c exp c dc

18 S 3 2 ,

(E-16)

S3 2 c 2 exp c 2 dc 85 S 3 2 ,

(E-17)

2 2

³ cx c

1

3 1 2 2 ³ cx S3 2 c exp c dc

14 S ,

(E-18)

1 2 2 ³ cx ª¬ S3 2 c º¼

2

5 8

13 8

exp c 2 dc

S3 2 ,

(E-19)

2

1 2 3 2 ³ cx ª¬ S3 2 c º¼ exp c dc

S ,

(E-20)

376

Analytical Methods for Problems of Molecular Transport 3 2 2 2 ³ cx S3 2 c exp c dc

161 S ,

(E-21)

2 2 3 1 2 2 ³ cx S3 2 c S3 2 c exp c dc

23 32 S ,

(E-22)

2 2 2 ³ cx 2 c exp c dc

18 S 3 2 ,

(E-23)

S ,

(E-24)

2 2 2 ³ cx c 2 c exp c dc

1 2 2 2 2 ³ cx 2 c S3 2 c exp c dc

5 8

2 2 2 2 2 ³ cx 2 c S3 2 c exp c dc

0 .

S3 2 ,

(E-25)

(E-26)

Here, the following notation has been used:

f

³ dc

3.

³ dcx 0

f

³

f

f

dc y

³ dcz

.

f

SOME INTEGRALS CONNECTED WITH THE SECOND-ORDER CHAPMAN-ENSKOG SOLUTION. 1 2 2 2 2 2 ³ cx ª¬ S3 2 c a2 S3 2 c º¼ exp c dc

1 4

S 1 34 a2 2 ,

(E-27)

377

Appendix E. Some Definite Integrals

³ cx c

2

7 8

S 1

³ cx c

2

1 2 2 32 ª S3 2 c 2 a2 S3 2 c 2 º exp c 2 dc ¬ ¼ 13 28

2

a2

,

ª S 1 c 2 a2 2 S 2 c 2 º exp c 2 dc 32 ¬ 32 ¼

1 2

1 4

2

S 1 a2

,

1 2 2 2 2 2 2 ³ cx c y ª¬ S3 2 c a2 S3 2 c º¼ exp c dc

18

2 2

³ cx c

1 4

2

S 1 a2

,

ª S 1 c 2 a2 2 S 2 c 2 º exp c 2 dc 5 S 3 2 , 32 8 ¬ 32 ¼

2 2 2 2 º 3 ª 1 2 c S c a S c exp c 2 dc 2 x 32 ³ ¬ 32 ¼

§ 23 2 S ¨1 26 a2 ©

433 208

2 ª a2 2 º ¸· , ¬ ¼ ¹

1 2 2 2 2 2 1 2 2 2 2 ³ cx c y ª¬ S3 2 c a2 S3 2 c º¼ ª¬1 b2 S5 2 c º¼ exp c dc

1 8

S

32

1

7 2 b 2 2

7 2

2 2

a2 b2

2 2 1 2 º 3 2ª c c 1 b S c exp c 2 dc x y 2 5 2 ³ ¬ ¼

(E-30)

(E-31)

,

(E-32)

(E-33)

1 4

(E-29)

13 8

(E-28)

2 ª º S «1 b2 2 174 b2 2 » , ¬ ¼

(E-34)

378

Analytical Methods for Problems of Molecular Transport 2 1

3 2

³ cx c y ª¬1 b2

4.

S5 2 c 2 º exp c 2 dc ¼

1 4

S 1 12 b2 2 .

(E-35)

SOME INTEGRALS CONNECTED WITH NONLINEAR TRANSPORT PROBLEMS. f

³ exp cx dcx 2

1 2

S 1 2 1 erf u x ,

(E-36)

(E-37)

ux

f

³ cx exp cx dcx 2

1 2

exp u x2 ,

1 2

ªu x exp u x2 1 S 1 2 1 erf u x º , 2 ¬ ¼

1 2

exp u x2 1 u x2 ,

ux

f

³ cx exp cx dcx 2

2

ux

f

³ cx exp cx dcx 3

2

ux

where: erf x

2

S1 2

x

2 ³ exp t dt 0

is the error function.

,

(E-38)

(E-39)

Appendix F OMEGA-INTEGRALS FOR SECOND-ORDER APPROXIMATION

All of the coefficients associated with the Chapman-Enskog algebraic systems of equations given in Eqs. (12-11)-(12-13) for the first-order Chapman-Enskog approximation and given in Eqs. (12-41)-(12-43) for the second-order Chapman-Enskog approximation are expressible in terms of the : -integrals which are dependent upon the specific model of the intermolecular potential that one is using. For a simple gas, these integrals have been described in Section 5.9. For gas mixtures, however, the : integrals must be slightly generalized. In this case, the : -integrals are defined by:

:12 l , r

ª :12 l ,r º : l , r å , ¬ ¼ r .s.

where [:12(l ,r ) ]r .s. is the : -integral for rigid-sphere molecules which is defined in [1] as: 12

l ,r º

ª: ¬ 12 ¼ r .s.

§ kT · ¨ * ¸ © 2S m12 ¹

r 1 ª«1 1 1 l º» SV 2 , 12 2 « 2 l 1 » ¬ ¼

where: * m12

m1m2 and V 12 m1 m2

1 2

V 1 V 2 .

380

Analytical Methods for Problems of Molecular Transport

The functions, : , are known as the reduced : -integrals (called the reduced collision integrals in some texts) and are given in terms of the reduced temperature, T * , by [2]: l ,r å

: l ,r å T *

ª r 1 T * ¬«

f r 2 1

º ¼»

³Q 0

l å

§ E* · E * exp ¨ * ¸ E * © T ¹

r 1

dE * ,

in which the following notations have been used:

Q

l å

E *

ª 1 1 l º » 2 «1 2 l 1 » «¬ ¼

f

³ 1 cos F b db l

*

*

,

0

·¸

* * b*2 M r * ¨ S 2b ³ 1 *2 ¨ r E* r0* ©

f§

F

1

¸ ¹

1 2

dr * , r *2

where r * r V 12 is the reduced intermolecular distance, b* b V 12 is the reduced impact parameter, M * M H is the reduced intermolecular potential * 2 energy, T * kT H is the reduced temperature, E * 12 m12 g H is the reduced relative kinetic energy, H is the energy parameter of the potential model being used, and r0* is determined from the largest, positive, real root of the expression:

·¸

§ b*2 M * r0* ¨1 ¨ r0*2 E* ©

¸ ¹

0.

In Table F-1 below, we have presented values of the : å -integrals computed for selected values of the reduced temperature, T * , using the Lennard-Jones (6-12) potential model. The values tabulated here are those needed for the first- and second-order Chapman-Enskog approximations. All listed values have been independently recalculated from those tabulated in [3] using a program developed by Maitland. [2] and subsequently revised by us to include increased numerical quadratures in the variables, b* and E* . Additionally, we note that a simple, concise, and very transparent program can be constructed in Mathematica® which will also permit good graphical display of the results at several stages in the calculations including

381

Appendix F. Omega-Integrals for Second-Order Approximation l

plots of the scattering angle, F (b * , E * ) , the functions, Q ( E * ) , and the l ,r resulting : -integrals, : (T * ) . We have constructed this Mathematica® program and have included it below as Table F-2 for the convenience of the reader. Given with this program is relevant output which includes a complete alternate set of : -integral values calculated using the Lennard-Jones (6-12) potential model which are needed for the first- and second-order Chapman-Enskog approximations. These values are analogous to those given in Table F-1, which were calculated with a modified version of an earlier Fortran program, and are shown computed for the same set of values of the reduced temperature, T * , that were specified in the earlier Fortran program. We note that while much of the nomenclature in the Mathematica® program is self-explanatory, the program has been annotated with a number of comments to assist the reader. We are suggesting that the reader may use this Mathematica® program to explore results for different intermolecular potential models. The reader may also wish to explore ways in which the program may be improved with respect to its accuracy and speed. The major time-consuming parts of this program are the computation of the scattering angle, F (b* , E * ) , and, in particular, the determination of r0* (b* , E * ) . For potential functions where non-polynomial equations are involved, one might wish to explore the use of the Mathematica® command ‘FindRoot’ (with a suitable starting value r0* | b* ).

382

Analytical Methods for Problems of Molecular Transport

Table F-1(a). Values of some : -integrals computed for selected values of the reduced * temperature, T , using the Lennard-Jones (6-12) potential model. The values tabulated here are the complete set of : -integrals needed for the first- and second-order ChapmanEnskog approximations. All listed values have been independently recalculated from those tabulated in [3] using a Fortran program developed by [2] and subsequently modified by us to give improved accuracy. Note that the last column of the table on page 541 of [2] appears to 2,5 3,3 be mislabeled. Instead of being labeled : it should be labeled : .

T*

: 1,1

0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00 2.10 2.20 2.30 2.40 2.50 2.60

2.6505 2.4699 2.3157 2.1826 2.0671 1.9661 1.8775 1.7992 1.7298 1.6679 1.6126 1.5628 1.5178 1.4771 1.4400 1.4062 1.3752 1.3468 1.3206 1.2964 1.2740 1.2531 1.2337 1.2157 1.1988 1.1829 1.1681 1.1541 1.1409 1.1285 1.1167 1.1056 1.0950 1.0850 1.0755 1.0577 1.0416 1.0268 1.0132 1.0007 0.9891

: 1,2

2.2586 2.0816 1.9345 1.8113 1.7077 1.6198 1.5448 1.4803 1.4244 1.3756 1.3328 1.2949 1.2613 1.2312 1.2042 1.1799 1.1578 1.1376 1.1192 1.1024 1.0868 1.0724 1.0591 1.0467 1.0352 1.0244 1.0143 1.0048 0.9959 0.9874 0.9795 0.9719 0.9648 0.9580 0.9515 0.9394 0.9284 0.9183 0.9090 0.9003 0.8923

: 1,3

1.9674 1.7998 1.6659 1.5578 1.4697 1.3971 1.3365 1.2855 1.2419 1.2045 1.1720 1.1435 1.1184 1.0961 1.0761 1.0582 1.0419 1.0271 1.0136 1.0012 0.9897 0.9791 0.9693 0.9601 0.9516 0.9435 0.9360 0.9288 0.9221 0.9158 0.9097 0.9040 0.8985 0.8933 0.8883 0.8790 0.8704 0.8625 0.8551 0.8482 0.8418

: 1,4

1.7426 1.5908 1.4740 1.3828 1.3103 1.2516 1.2034 1.1632 1.1292 1.1001 1.0749 1.0528 1.0333 1.0160 1.0005 0.9865 0.9737 0.9621 0.9515 0.9416 0.9325 0.9240 0.9161 0.9087 0.9018 0.8952 0.8891 0.8832 0.8777 0.8724 0.8674 0.8626 0.8580 0.8536 0.8494 0.8415 0.8341 0.8273 0.8209 0.8149 0.8092

: 1,5

1.5692 1.4363 1.3372 1.2614 1.2021 1.1546 1.1157 1.0834 1.0560 1.0325 1.0121 0.9942 0.9784 0.9642 0.9514 0.9398 0.9292 0.9195 0.9105 0.9022 0.8945 0.8873 0.8805 0.8741 0.8681 0.8624 0.8570 0.8519 0.8470 0.8424 0.8380 0.8337 0.8296 0.8257 0.8219 0.8148 0.8082 0.8019 0.7961 0.7906 0.7854

: 2,2

2.8438 2.6796 2.5337 2.4029 2.2854 2.1798 2.0849 1.9995 1.9226 1.8533 1.7907 1.7341 1.6826 1.6358 1.5932 1.5542 1.5184 1.4856 1.4553 1.4274 1.4015 1.3776 1.3553 1.3346 1.3152 1.2971 1.2802 1.2643 1.2494 1.2353 1.2221 1.2095 1.1977 1.1865 1.1758 1.1561 1.1383 1.1220 1.1071 1.0934 1.0809

383

Appendix F. Omega-Integrals for Second-Order Approximation

T*

: 1,1

2.70 2.80 2.90 3.00 3.10 3.20 3.30 3.40 3.50 3.60 3.70 3.80 3.90 4.00 4.10 4.20 4.30 4.40 4.50 4.60 4.70 4.80 4.90 5.00 6.00 7.00 8.00 9.00 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100 200 300 400

0.9783 0.9683 0.9589 0.9501 0.9418 0.9340 0.9267 0.9197 0.9131 0.9069 0.9009 0.8953 0.8898 0.8847 0.8797 0.8749 0.8704 0.8660 0.8618 0.8577 0.8538 0.8500 0.8464 0.8429 0.8129 0.7898 0.7712 0.7556 0.7423 0.6641 0.6235 0.5963 0.5760 0.5599 0.5465 0.5352 0.5254 0.5168 0.4630 0.4339 0.4142

: 1,2

0.8848 0.8777 0.8711 0.8649 0.8590 0.8534 0.8481 0.8430 0.8382 0.8336 0.8292 0.8250 0.8210 0.8171 0.8134 0.8098 0.8063 0.8030 0.7997 0.7966 0.7936 0.7906 0.7878 0.7850 0.7610 0.7419 0.7261 0.7126 0.7008 0.6295 0.5913 0.5655 0.5461 0.5308 0.5181 0.5073 0.4979 0.4897 0.4383 0.4106 0.3919

: 1,3

0.8357 0.8300 0.8245 0.8194 0.8145 0.8099 0.8054 0.8012 0.7971 0.7932 0.7895 0.7859 0.7824 0.7791 0.7758 0.7727 0.7697 0.7667 0.7639 0.7611 0.7584 0.7558 0.7533 0.7508 0.7292 0.7117 0.6970 0.6844 0.6733 0.6052 0.5684 0.5435 0.5248 0.5099 0.4977 0.4872 0.4782 0.4702 0.4207 0.3940 0.3761

: 1,4

0.8039 0.7988 0.7940 0.7894 0.7850 0.7808 0.7768 0.7730 0.7693 0.7657 0.7623 0.7590 0.7558 0.7527 0.7497 0.7468 0.7440 0.7413 0.7386 0.7361 0.7336 0.7311 0.7288 0.7264 0.7060 0.6894 0.6753 0.6632 0.6526 0.5865 0.5507 0.5265 0.5083 0.4938 0.4819 0.4718 0.4630 0.4552 0.4071 0.3812 0.3641

: 1,5

0.7804 0.7757 0.7713 0.7670 0.7629 0.7590 0.7552 0.7516 0.7481 0.7448 0.7416 0.7384 0.7354 0.7325 0.7296 0.7269 0.7242 0.7216 0.7191 0.7166 0.7142 0.7119 0.7096 0.7074 0.6877 0.6716 0.6580 0.6462 0.6359 0.5714 0.5364 0.5127 0.4949 0.4808 0.4691 0.4592 0.4506 0.4431 0.3962 0.3710 0.3545

: 2,2

1.0692 1.0584 1.0483 1.0389 1.0301 1.0218 1.0141 1.0067 0.9998 0.9932 0.9870 0.9810 0.9754 0.9700 0.9649 0.9599 0.9552 0.9507 0.9463 0.9422 0.9381 0.9343 0.9305 0.9269 0.8963 0.8728 0.8539 0.8381 0.8245 0.7437 0.7008 0.6717 0.6498 0.6323 0.6178 0.6055 0.5947 0.5852 0.5256 0.4931 0.4710

384

Analytical Methods for Problems of Molecular Transport

Table F-1(b).

T*

: 2,3

0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 2.90 3.00 3.10 3.20

2.5812 2.4096 2.2577 2.1241 2.0070 1.9046 1.8150 1.7364 1.6674 1.6065 1.5526 1.5048 1.4620 1.4238 1.3893 1.3582 1.3301 1.3044 1.2810 1.2596 1.2399 1.2218 1.2050 1.1895 1.1750 1.1616 1.1490 1.1373 1.1262 1.1159 1.1061 1.0969 1.0882 1.0800 1.0722 1.0577 1.0446 1.0326 1.0216 1.0115 1.0022 0.9935 0.9854 0.9778 0.9707 0.9640 0.9577

: 2,4

2.3626 2.1835 2.0298 1.8990 1.7881 1.6940 1.6138 1.5452 1.4861 1.4349 1.3903 1.3512 1.3166 1.2860 1.2586 1.2340 1.2119 1.1919 1.1736 1.1570 1.1417 1.1276 1.1147 1.1026 1.0915 1.0810 1.0713 1.0622 1.0536 1.0456 1.0380 1.0308 1.0240 1.0175 1.0114 0.9999 0.9895 0.9799 0.9711 0.9629 0.9553 0.9482 0.9416 0.9353 0.9294 0.9238 0.9185

: 2,5

2.1704 1.9898 1.8407 1.7185 1.6181 1.5350 1.4658 1.4075 1.3581 1.3157 1.2791 1.2471 1.2191 1.1944 1.1723 1.1526 1.1348 1.1186 1.1040 1.0906 1.0783 1.0669 1.0564 1.0466 1.0376 1.0291 1.0211 1.0136 1.0066 0.9999 0.9936 0.9876 0.9820 0.9765 0.9714 0.9618 0.9529 0.9447 0.9372 0.9301 0.9235 0.9173 0.9114 0.9059 0.9007 0.8957 0.8910

: 2,6

2.0009 1.8256 1.6868 1.5764 1.4880 1.4162 1.3572 1.3081 1.2667 1.2314 1.2011 1.1747 1.1515 1.1310 1.1128 1.0964 1.0816 1.0682 1.0560 1.0447 1.0344 1.0248 1.0159 1.0077 0.9999 0.9927 0.9858 0.9794 0.9733 0.9675 0.9621 0.9569 0.9519 0.9471 0.9426 0.9341 0.9262 0.9189 0.9121 0.9057 0.8997 0.8941 0.8887 0.8837 0.8788 0.8742 0.8699

: 3,3

2.3989 2.2249 2.0799 1.9572 1.8523 1.7618 1.6833 1.6147 1.5544 1.5011 1.4538 1.4115 1.3737 1.3396 1.3089 1.2809 1.2555 1.2323 1.2110 1.1914 1.1733 1.1567 1.1412 1.1268 1.1134 1.1008 1.0891 1.0781 1.0678 1.0580 1.0488 1.0401 1.0319 1.0241 1.0167 1.0029 0.9904 0.9790 0.9684 0.9587 0.9497 0.9414 0.9336 0.9263 0.9194 0.9130 0.9069

: 4,4 2.5712 2.3939 2.2395 2.1052 1.9885 1.8871 1.7987 1.7214 1.6537 1.5940 1.5412 1.4943 1.4525 1.4150 1.3814 1.3510 1.3234 1.2984 1.2755 1.2546 1.2354 1.2176 1.2013 1.1861 1.1721 1.1589 1.1467 1.1353 1.1245 1.1145 1.1050 1.0960 1.0876 1.0796 1.0720 1.0579 1.0452 1.0335 1.0229 1.0131 1.0040 0.9956 0.9877 0.9804 0.9735 0.9670 0.9609

385

Appendix F. Omega-Integrals for Second-Order Approximation

T*

: 2,3

3.30 3.40 3.50 3.60 3.70 3.80 3.90 4.00 4.10 4.20 4.30 4.40 4.50 4.60 4.70 4.80 4.90 5.00 6.00 7.00 8.00 9.00 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100 200 300 400

0.9518 0.9461 0.9407 0.9356 0.9308 0.9261 0.9216 0.9174 0.9133 0.9093 0.9056 0.9019 0.8984 0.8950 0.8917 0.8885 0.8854 0.8824 0.8566 0.8363 0.8195 0.8052 0.7927 0.7165 0.6751 0.6469 0.6256 0.6086 0.5945 0.5825 0.5720 0.5628 0.5050 0.4735 0.4521

: 2,4

0.9135 0.9087 0.9041 0.8997 0.8955 0.8915 0.8876 0.8839 0.8803 0.8769 0.8736 0.8703 0.8672 0.8642 0.8612 0.8584 0.8556 0.8529 0.8295 0.8106 0.7948 0.7813 0.7694 0.6956 0.6552 0.6276 0.6068 0.5901 0.5764 0.5646 0.5544 0.5454 0.4891 0.4584 0.4373

: 2,5

0.8865 0.8821 0.8780 0.8740 0.8702 0.8666 0.8630 0.8596 0.8563 0.8531 0.8500 0.8470 0.8441 0.8413 0.8386 0.8359 0.8333 0.8308 0.8086 0.7906 0.7754 0.7623 0.7508 0.6786 0.6389 0.6118 0.5914 0.5751 0.5616 0.5501 0.5401 0.5312 0.4761 0.4461 0.4250

: 2,6

0.8657 0.8616 0.8578 0.8540 0.8505 0.8470 0.8437 0.8404 0.8373 0.8343 0.8314 0.8285 0.8258 0.8231 0.8205 0.8179 0.8154 0.8130 0.7916 0.7742 0.7594 0.7465 0.7353 0.6643 0.6253 0.5986 0.5785 0.5625 0.5492 0.5379 0.5280 0.5194 0.4653 0.4357 0.4144

: 3,3

0.9011 0.8956 0.8904 0.8855 0.8808 0.8763 0.8720 0.8678 0.8639 0.8600 0.8564 0.8528 0.8494 0.8461 0.8429 0.8398 0.8368 0.8339 0.8090 0.7893 0.7731 0.7593 0.7474 0.6743 0.6349 0.6081 0.5879 0.5718 0.5585 0.5471 0.5372 0.5285 0.4740 0.4444 0.4245

: 4,4 0.9551 0.9496 0.9444 0.9395 0.9348 0.9302 0.9259 0.9218 0.9178 0.9140 0.9103 0.9068 0.9033 0.9000 0.8968 0.8937 0.8907 0.8878 0.8627 0.8428 0.8264 0.8124 0.8001 0.7247 0.6834 0.6551 0.6338 0.6167 0.6025 0.5904 0.5799 0.5706 0.5122 0.4804 0.4585

386

Analytical Methods for Problems of Molecular Transport

Table F-2. A Mathematica® program that may be used to compute values of the : * integrals for selected values of the reduced temperature, T . This program is currently configured to use the Lennard-Jones (6-12) potential model and the same set of reduced temperatures reported in Table F-1, but can be readily adapted to other potential models and alternative sets of reduced temperatures.

Evaluating Omega-Integrals: INI, SKL, RVT 2005-2006 We have used the Lennard-Jones (6-12) potential model as an example. The results of this program compare well with those in: 1. Hirschfelder, J.O., Curtiss, C.F. and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954). In the program we refer to this reference as HCB. Take special note of pages 557, 558, and 1132 in this text. 2. Maitland, G.C., Rigby, M., Smith, E.B. and Wakeham, W.A., Intermolecular Forces (Oxford University Press, Oxford, 1981). In the program we refer to this reference as MRSW. Take special note of pages 538-541 in this text. The program consists of three parts: 1. Calculation of the scattering angle, chi (F), and construction of an interpolation approximation to it. 2. Calculation of the Q-integrals using chi, and again interpolation approximations to these. 3. Calculation of the omega-integrals. In the following, we have not focused on high accuracy. Our purpose has been to indicate how a simple program for calculating the omega-integrals can be constructed. We encourage the reader to explore the program. We have not included several comparative tests that were carried out during construction of the program. We thank our students Ryan M. Meyer, Zebadiah Smith, and Earl Lynn Tipton for reviewing and helping us with improvements in parts of the program. Needs["Graphics`Graphics`"]; Needs["Graphics`MultipleListPlot`"]; Needs["Graphics`Legend`"]; Off[NIntegrate::ncvb]; Off[NIntegrate::slwcon];

Appendix F. Omega-Integrals for Second-Order Approximation

387

Off[General::spell]; Off[General::spell1]; (* Define the potential: In the following we have chosen the Lennard-Jones (6-12) potential. *) Clear[ustar, rstar]; ustar[rstar_]:=4(rstar^-12 - rstar^-6); (* Construct the chi function, specialized form for the Lennard-Jones (6-12) potential. In the calculation of chi we have used a standard transformation, pstar=1/(rstar*2), which reduces the order of the polynomial in the integrand, and also changes the interval of integration to a finite interval. For calculating pstar0 we have use the Matrix Eigenvalue method of finding zeros of a polynomial. One can also use the NSolve or FindRoot functions depending upon the potential. We have not accounted for the orbiting condition in the scattering angle chi explicitly; we have smoothed over it through the choice of a coarse computational grid on bstar and estar. A more careful computation would be needed for precise work. For using the NSolve function, one can have a construct of the type: pstar0[bstar_,estar_]:=Min[Select[pstar/.NSolve[(1 – bstar^2 * pstar - 4(pstar^6 * pstar^3)/estar)m0,pstar]//N, (Im[#]m0&&Re[#]>0) &]]; *) Clear[pstar0, bstar, estar, chi, pstar00]; pstar0[bstar_,estar_]:=Min[Select[Eigenvalues[{{0, 0, 0, 0, 0, estar/4}, {1, 0, 0, 0, 0, -estar/4 * bstar^2}, {0, 1, 0, 0, 0, 0}, {0, 0, 1, 0, 0, 1}, {0, 0, 0, 1, 0, 0}, {0, 0, 0, 0, 1, 0}}]//N, (Im[#]m0&&Re[#]>0) &]]; chi[bstar_,estar_]:=Re[Pi – bstar * With[{pstar00=pstar0[bstar,estar]}, NIntegrate[Evaluate[1/Sqrt[pstar(1 - bstar^2 * pstar - 4(pstar^6 - pstar^3)/estar)]], {pstar, 0, pstar00}, MaxRecursiono30, AccuracyGoalo8, SingularityDeptho10]]]; (* We avoid the almost 0 imaginary part that drifts in because of numerics. *) (* Construct and test chiapprox which is based on interpolation. For higher precision or higher range on tstar, the reduced temperature, the limits and the grids below would need to be modified. *)

388

Analytical Methods for Problems of Molecular Transport

Clear[bstarlowlim, bstaruplim, estarlowlim, estaruplim, bstarlist, estarlist, reschi, chiapprox]; bstarlowlim=10^-5; bstaruplim=5; estarlowlim=10^-5; estaruplim=1200; bstarlist=bstarlowlim + Range[0.0,bstaruplim,0.025]; estarlist=estarlowlim + Range[0.0,estaruplim,0.025]; reschi=Outer[chi,bstarlist,estarlist]; (* The above evaluates chi on bstar, estar grid. *) chiapprox=ListInterpolation[reschi,{{First[bstarlist], Last[bstarlist]}, {First[estarlist], Last[estarlist]}}]; (* The above constructs the interpolation function for chi. *) Print[chiapprox]; InterpolatingFunction[{{0.00001,5.00001},{0.00001,1200.}},]

(* Plot (1-Cos[chi[bstar,estar]]), compare with HCB see page 557. *) Clear[chiapproxplotsHCB]; chiapproxplotsHCB=Map[Plot[1.0 Cos[Re[chiapprox[Sqrt[bstarsq], #]]], {bstarsq, o1000, PlotRangeo{{10^-10, 6}, 10^-10, 6}, PlotPointso {0, 2}}, Tickso{{0, 1, 2, 3, 4, 5, 6}, {0, 0.5, 1.0, 1.5, 2.0}}, FrameoTrue, RotateLabeloTrue, FrameLabelo{"b*2", "1-CosF"}, oIdentity] &, {0.4, 0.8, 1.0, 10.0}]; DisplayFunctiono Print["Plots of 1-CosFapprox vs b*2 for E*={0.4,0.8,1.0 and 10.0}", "\n "]; Show[GraphicsArray[Partition[chiapproxplotsHCB, 4]], DisplayFunctiono$DisplayFunction];

389

6

2 1.75 1.5 1.25 1 0.75 0.5 0.25

1Cos¢

1Cos¢

Appendix F. Omega-Integrals for Second-Order Approximation

1

2

3 4 b 2

5

6

2 1.75 1.5 1.25 1 0.75 0.5 0.25 1

2

3 4 b 2

5

6

(* For the construction of qstar[[L]][estar], we first evaluate the integrals on estar grid for given L (b in the textbook notation), and then construct an interpolation function for each L. We call the interpolation functions qstarapprox2 which are elements of the list qstarapproxlist. Lhigh is the user assigned integer, dictated by the orders of omega-integrals needed. *) Clear[qstarapprox1, qstarapprox2, qstarapproxlist, Lhigh]; qstarapprox1[L_,estar_]:=2(1-((1+(-1)^L)/(2(1+L))))^-1 * NIntegrate[(1 - Cos[chiapprox[bstar,estar]]^L) * bstar, {bstar, First[bstarlist], Last[bstarlist]}, MaxRecursion5, AccuracyGoal4]; qstarapprox2[L_]:=Interpolation[Map[{#, qstarapprox1[L,#]} &,estarlist]]; Lhigh=4; qstarapproxlist=Map[qstarapprox2[#] &,Range[1,Lhigh]]; Print[qstarapproxlist]; {InterpolatingFunction[{{0.00001,1200.}},],InterpolatingFunc tion[{{0.00001,1200.}},],InterpolatingFunction[{{0.00001,120 0.}},],InterpolatingFunction[{{0.00001,1200.}},]}

(* Tabulate and Plot qstar (L, estar). page 558. *)

Compare with HCB

Print[TableForm[Table[Map[(Flatten[{N[#], Table[qstarapproxlist[[L]][#], {L, Lhigh}]}]) &, {10^-4, 10^-3, 10^-2, 10^-1, 1, 2, 3, 4, 5, 10, 100, 1000}]], TableHeadings{None, {" E* ", " Q*(1, E*)", "Q*(2, E*)", "Q*(3, E*)", " Q*(4, E*)"}}]]; Clear[qstarplot1]; qstarplot1=Plot[Evaluate[Chop[Table[ qstarapproxlist[[L]][estar], {L, 1, Lhigh}], 10^-8]], {estar, 0.1, 4}, PlotPoints1000, PlotRangeAll, Ticks{{0, 1, 2, 3, 4}, Automatic}, FrameTrue, RotateLabelTrue, FrameLabel{" E* ", " Q*(b, E*)"}, AxesOrigin{0, 0}, PlotStyleMap[AbsoluteDashing[{5, #^2.5}] &,{1, 2, 3, 4}], PlotLegendMap[ToString, N[{1, 2, 3, 4}]], LegendPosition{1, -0.3}, LegendSize{0.8, 0.8}, LegendShadowNone, LegendLabelStyleForm["b", FontSize14]];

390

Analytical Methods for Problems of Molecular Transport

E 0.0001 0.001 0.01 0.1 1. 2. 3. 4. 5. 10. 100. 1000.

Q +1,E / 34.8393 33.1847 19.8657 5.38281 2.40659 1.46589 1.18881 1.06699 0.997701 0.855383 0.594234 0.41265

Q +2,E / 27.3578 26.2679 17.4243 6.12706 2.80343 2.14097 1.61842 1.37593 1.24426 1.01136 0.7044 0.495646

Q +3,E / 29.8057 28.5725 18.5817 6.23423 2.81556 1.94271 1.52617 1.30831 1.18465 0.95951 0.662703 0.465145

Q +4,E / 26.2607 25.2987 17.4597 6.57795 3.02483 2.37249 1.84504 1.54584 1.37286 1.06798 0.73363 0.519093

Q +b,E /

6

b

5

1.

4

2. 3.

3 4. 2

1 1

2 E

3

4

(* Compare again with HCB page 558, who have given a LogLinear Plot of Q*(1,E*), Q*(2,E*), and Q*(4,E*). *) Clear[qstarplot2]; qstarplot2=LogLinearPlot[Evaluate[Chop[Map[ qstarapproxlist[[#]][estar] &, {1, 2, 4}], 10^-8]], {estar, 0.1, estaruplim}, PlotPoints5000, PlotRangeAll, * FrameTrue, RotateLabelTrue, FrameLabel{" E ", " Q*(b, E*)"}, AxesOrigin{0.1, 0}, AspectRatio1.2, PlotStyleMap[AbsoluteDashing[{5, #}] &,{1, 5, 15}], Epilog{Text[" Q*(b, E*) for b = 1,2,4", {1.0, 5.0}], Text[" b = 1", {0.1, 2.0}], Text[" b = 2", {0.5, 1.5}], Text[" b = 4", {1.5, 1.0}]}];

391

Appendix F. Omega-Integrals for Second-Order Approximation

6

b

Q + ,E / f o r

5

b

1,2, 4

Q +b,E /

4

3

b

2

1

b

2

b

1

0.1

1

10 E

4

1 00

1000

(* Evaluate the omega-integrals using qstarapprox2[[L]][estar]. Compare with HCB pages 11261127 and MRSW page 541. We have set AccuracyGoal high to get reasonably accurate results for small Tstar. We have also evaluated the tail part of the integral approximately as per MRSW. *) Clear[omegastar]; Off[NIntegrate::ploss]; omegastar[L_,s_,tstar_]:=(NIntegrate[Chop[ qstarapproxlist[[L]][estar]] * Exp[-estar/tstar] * estar^(s+1), {estar, First[estarlist], Last[estarlist]}, MinRecursion10, MaxRecursion30, AccuracyGoal8]/(Factorial[(s+1)] * tstar^(s+2)) + Chop[qstarapproxlist[[L]][Last[estarlist]]] * NIntegrate[Exp[-estar/tstar] * estar^(s+1), {estar, Last[estarlist], Infinity}]/(Factorial[(s+1)] * tstar^(s+2))); Print["Table of Omega Integrals ", "\n "]; Do[Print["b = ", L];

392

Analytical Methods for Problems of Molecular Transport

Print[TableForm[Chop[Map[{#, omegastar[L, 1, #], omegastar[L, 2, #], omegastar[L, 3, #], omegastar[L, 4, #], omegastar[L, 5, #], omegastar[L, 6, #]} &, {0.1, 0.2, 0.3, 0.5, 0.7, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 50.0, 100.0}]], TableHeadings{None, {"T* ", "r=1 ", "r=2 ", "r=3 ", "r=4 ", "r=5 ", "r=6 "}}]], {L, 1, Lhigh}]; Table of Omega Integrals b =

1

T 0.1 0.2 0.3 0.5 0.7 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 50. 100.

b = T 0.1 0.2 0.3 0.5 0.7 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 50. 100.

r1 4.03052 3.14477 2.6577 2.06944 1.73095 1.44051 1.07552 0.950098 0.884639 0.842821 0.812858 0.789791 0.771157 0.755583 0.742235 0.575969 0.516794

r2 3.57351 2.74038 2.26061 1.70836 1.42466 1.20433 0.951483 0.864833 0.817083 0.78494 0.760923 0.741824 0.726007 0.712527 0.700797 0.546111 0.48969

r3 3.25852 2.44475 1.96826 1.47 1.24203 1.07614 0.888318 0.819374 0.779017 0.75077 0.729126 0.711621 0.696951 0.684343 0.673301 0.524777 0.470324

r4 3.01939 2.20647 1.74315 1.31043 1.12922 1.00046 0.849385 0.78935 0.752666 0.726388 0.705976 0.689324 0.675289 0.663177 0.65254 0.50829 0.455435

r5 2.82695 2.00606 1.5696 1.20216 1.05596 0.951348 0.821901 0.766952 0.732417 0.70733 0.687689 0.67159 0.657978 0.646207 0.635852 0.494927 0.443509

r6 2.6648 1.83615 1.43693 1.12668 1.0054 0.916617 0.800763 0.749051 0.715933 0.691661 0.672565 0.656868 0.643571 0.632057 0.621921 0.483736 0.433762

r1 4.56849 3.63283 3.1838 2.64092 2.27737 1.91273 1.37254 1.17317 1.07194 1.01019 0.967913 0.936631 0.912183 0.892297 0.875624 0.683549 0.616226

r2 4.05719 3.24077 2.83 2.2811 1.92119 1.59293 1.1759 1.0389 0.969964 0.926819 0.896217 0.872757 0.853822 0.837976 0.82437 0.649791 0.585166

r3 3.719 2.98294 2.57354 2.00572 1.66732 1.38953 1.07222 0.97065 0.917304 0.882326 0.85655 0.836199 0.819399 0.805097 0.79265 0.625553 0.562809

r4 3.47506 2.78939 2.35902 1.78801 1.48643 1.2588 1.01133 0.929309 0.883842 0.852839 0.829375 0.810506 0.79473 0.781177 0.769302 0.606706 0.545531

r5 3.2892 2.62689 2.16912 1.61841 1.35839 1.17244 0.971338 0.900594 0.859519 0.830714 0.808538 0.79051 0.775326 0.762215 0.750686 0.591349 0.531643

r6 3.14222 2.47914 2.00066 1.48839 1.26698 1.11283 0.942531 0.878744 0.840357 0.812914 0.791551 0.774064 0.759271 0.746461 0.735172 0.578434 0.520264

2

Appendix F. Omega-Integrals for Second-Order Approximation

b =

3

T 0.1 0.2 0.3 0.5 0.7 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 50. 100.

b = T 0.1 0.2 0.3 0.5 0.7 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 50. 100.

393

r1 4.6529 3.64653 3.12014 2.50274 2.13222 1.7864 1.29264 1.10901 1.01433 0.955952 0.915696 0.885775 0.862323 0.843213 0.827177 0.643025 0.579091

r2 4.12477 3.19922 2.70053 2.11889 1.78412 1.49032 1.11282 0.984202 0.918331 0.876791 0.847243 0.824573 0.806281 0.790986 0.777862 0.610888 0.54966

r3 3.76508 2.88425 2.40109 1.8529 1.5546 1.30887 1.01663 0.919351 0.867753 0.833868 0.808919 0.789249 0.773037 0.759258 0.747282 0.587878 0.528516

r4 3.49582 2.63981 2.17056 1.65819 1.39593 1.19164 0.95883 0.879455 0.835342 0.805309 0.782632 0.764437 0.749251 0.736227 0.724831 0.570016 0.512193

r5 3.28207 2.43999 1.98643 1.51165 1.28309 1.1125 0.920299 0.851593 0.811756 0.783903 0.762519 0.745173 0.73059 0.718018 0.706977 0.555479 0.499083

r6 3.10483 2.27233 1.83608 1.39949 1.20089 1.05667 0.892353 0.830375 0.793201 0.766717 0.746158 0.729366 0.715185 0.702922 0.692127 0.543265 0.488348

r1 4.9391 3.92571 3.43691 2.86481 2.48993 2.10675 1.50388 1.26756 1.14572 1.07167 1.02153 0.984951 0.956765 0.934145 0.915416 0.71244 0.643407

r2 4.38923 3.49768 3.05538 2.49314 2.12111 1.76422 1.27437 1.10635 1.023 0.972207 0.937115 0.910811 0.889963 0.872772 0.858177 0.677832 0.611512

r3 4.02342 3.215 2.78624 2.21284 1.85296 1.53592 1.14761 1.02299 0.960207 0.920578 0.892193 0.870234 0.852368 0.83732 0.824323 0.65302 0.588504

r4 3.75748 3.00503 2.56878 1.98837 1.65362 1.38128 1.07189 0.973386 0.921683 0.887715 0.862615 0.842741 0.826292 0.812258 0.800019 0.633697 0.570684

r5 3.55326 2.83367 2.38056 1.80744 1.5054 1.27473 1.02261 0.940009 0.894636 0.863818 0.840527 0.821803 0.806141 0.792677 0.780869 0.617927 0.556337

r6 3.39054 2.68347 2.21396 1.6625 1.3948 1.19933 0.987894 0.915372 0.873884 0.844987 0.822804 0.804792 0.789623 0.776523 0.764997 0.604644 0.544568

4

REFERENCES 1. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 2. Maitland, G.C., Rigby, M., Smith, E.B., and Wakeham, W.A., Intermolecular Forces: Their Origin and Determination (Oxford University Press, New York, 1981, reprinted and corrected edition, 1987). 3. Hirschfelder, J.O., Curtiss, C.F., and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954).

Author Index

In this author index, bold page entries signify a citation in the references. Unbolded page entries signify a citation by name in the text. –A– Allen, M.D.; 237 Anselm, A.I.; 37 Aoki, K.; 238, 312, 333

–B– Bakanov, S.P.; 178, 238 Barantsev, R.G.; 89 Bardos, C.; 333 Barrett, J.C.; 89 Bassanini, P.; 237, 266, 287 Becker, M.; 89 Bird, G.A.; 28, 139 Bird, R.B.; 73, 331, 350, 393 Boffi, V.; 238 Bogoliubov, N.N.; 37 Breton, J.P.; 332 Brock, J.R.; 89, 140, 222, 237, 238, 331, 332 Bronshtein, I.N.; 368 Brush, S.G.; 51 Buckley, R.L.; 217, 219, 237

–C– Cadle, R.D.; 237 Campbell, P.H.; 89 Cercignani, C.; 51, 89, 237, 238, 266, 287, 331, 372 Chambré, P.L.; 139

Chandrasekhar, S.; 287 Chapman, S.; 14, 28, 51, 67, 73, 292, 331, 350 Chekalov, V.V.; 272, 287 Cheng, R.K.; 140 Cipolla, J.W.Jr.; 209 Clark, M.Jr.; 372 Coronell, D.G.; 331 Courant, R.; 372 Cowling, T.G.; 14, 28, 51, 67, 73, 292, 331, 350 Curtiss, C.F.; 73, 331, 350, 393

–D– Davies, C.N.; 237, 332 de Boer, J.; 37 Derjaguin, B.V.; 89, 178, 237, 238, 331, 332, 350 Dirac, P.A.M.; 89 Dwyer, H.A.; 238

–E– Eger, K.; 332 Epstein, P.S.; 237

–F– Ferziger, J.H.; 14, 37, 51, 73, 208, 331, 332, 350, 393 Feshbach, H.; 209, 372 Fiebig, M.; 89

396

Analytical Methods for Problems of Molecular Transport

Fisher, S.S.; 238 Flügge, S.; 286 Ford, G.W.; 51, 73 Fuchs, N.A.; 140

–G– Galkin, V.S.; 333 Ghosh, T.K.; 333 Golse, F.; 333 Goodman, F.O.; 89 Gora, E.K.; 37 Grad, H.; 286 Grosch, C.E.; 287 Gross, E.P.; 89, 155, 159, 178, 266, 268, 287 Gupta, R.N.; 331

–H– Halbritter, L.; 272, 287 Heineman, M.; 139 Hess, D.W.; 331 Hickey, K.A.; 188, 208, 209 Hidy, G.M.; 89, 179, 237, 238, 331, 350 Hilbert, D.; 372 Hirschfelder, J.O.; 73, 331, 350, 393 Huang, C.M.; 333 Huang, K.; 14, 51 Hurlbut, F.C.; 287

–I– Ikegawa, M.; 331 Ishihara, Y.; 238 Ivchenko, I.N.; 83, 84, 89, 90, 106, 139, 140, 178, 189, 196, 208, 209, 238, 261, 266, 287, 332, 333, 350

–J– Jackson, E.A.; 89, 178, 266, 287 Jacobsen, S.; 222, 237 Jeans, J.; 28 Jensen, K.F.; 331 Jew, H.; 89

–K– Kahan, T.; 372 Kaper, H.G.; 14, 37, 51, 73, 331, 350, 393 Keller, J.B.; 88 Kelly, G.E.; 89 Keng, E.Y.H.; 237 Kennard, E.H.; 51, 139, 331 Kistemaker, J.; 331 Klein, M.J.; 51 Knudsen, M.; 51

Kobayashi, J.; 331 Kogan, M.N.; 14, 73, 89, 139, 208, 243, 286, 287, 331 Kosuge, S.; 333 Kramers, H.A.; 331 Krook, M.; 243, 287 Kucherov, R.Ya.; 331 Kumar, K.; 286 Kušþer, I.; 89

–L– La Mer, V.K.; 237 Landau, L.D.; 14, 28, 37, 90, 237, 287 Lang, H.; 89, 179, 332, 372 Lea, K.C.; 237 Lees, L.; 242, 286 Levich, V.G.; 14 Lide, D.R.; 334, 350, 361 Lifshitz, E.M.; 14, 28, 37, 90, 237, 287 Liu, C.Y.; 242, 286, 287 Liu, V.C.; 89 Loeb, L.B.; 51, 331 Loyalka, S.K.; 83, 84, 89, 90, 149, 178, 179, 183, 188, 189, 196, 197, 198, 199, 208, 209, 217, 218, 219, 221, 222, 237, 238, 266, 273, 287, 303, 331, 332, 333, 361, 372

–M– Maitland, G.C.; 73, 333, 380, 393 Malinauskas, A.P.; 140 Mason, B.J.; 140 Mason, E.A.; 140, 333, 334 Maxwell, J.C.; 141, 178, 331 Millikan, R.A.; 79, 89 Monchick, L.; 334 Morse, P.M.; 209, 372 Moss, J.N.; 331 Mott-Smith, H.M.; 242, 287 Muckenfuss, C.; 287 Müller, W.J.C.; 332 Muntz, E.P.; 89

–N– Niven, W.D.; 331

–O– Orr, C.Jr.; 237

–P– Pagani, C.D.; 237, 266, 287, 372 Pang, S.C.; 89 Patterson, G.N.; 139 Phillips, W.F.; 237, 287

397

Author Index Pitaevskii, L.P.; 37 Poddoskin, A.B.; 238 Pomraning, G.C.; 372 Porodnov, B.T.; 159, 178 Present, R.D.; 331

–R– Raabe, O.G.; 237 Rabinovich, Ya.I.; 237 Ranz, W.E.; 237 Richenglas, L.E.; 331 Rideau, G.; 372 Rigby, M.A.; 73, 333, 393 Rolduguin, V.I.; 90, 178, 350 Rose, M.H.; 88 Rosenblatt, P.; 237 Rosner, D.E.; 331 Roussopolous, P.; 372

–S– Salwen, H.; 287 Savkov, S.A.; 178, 263, 287, 333 Savkov, S.V.; 90, 350 Saxena, S.C.; 334 Saxton, R.L.; 237 Schaaf, S.A.; 139 Schadt, C.F.; 237 Schefer, R.W.; 140 Schmitt, K.H.; 140, 237, 332 Scott, C.D.; 331 Semendyayev, K.A.; 368 Sengers, J.V.; 89 Sharipov, F.; 333 Shizgal, B.; 89 Shkanov, Yu.; 333 Siewert, C.E.; 333 Sigimura, T.; 287 Smirnov, L.P.; 272, 287 Smith, E.B.; 73, 333, 393 Smoluchowski, M.; 140 Sone, Y.; 238, 333 Springer, G.S.; 266, 287

Storozhilova, A.I.; 237 Storvick, T.S.; 332 Suetin, P.E.; 159, 178 Szymanski, Zd.; 139

–T– Takata, A.; 312, 333 Talbot, L.; 106, 140, 332 Tauber, R.N.; 331 Teagan, W.P.; 266, 287 Tompson, R.V.; 83, 84, 90, 189, 196, 208, 209, 287, 332, 333 Trilling, L.; 332, 372

–U– Uhlenbeck, G.E.; 37, 51, 73, 139, 155, 178, 266, 287

–V– Volkov, I.V.; 333

–W– Wachman, H.Y.; 89, 332, 372 Wakeham, W.A.; 73, 333, 393 Waldmann, L.; 97, 140, 237, 332 Wang-Chang, C.S.; 139, 140, 155, 178, 266, 287 Weaver, D.P.; 89 Williams, M.M.R.; 140, 237, 331 Willis, D.R.; 140 Wolf, S.; 331

–Y– Yalamov, Yu.I.; 89, 106, 140, 178, 238, 331, 332, 333, 350 Yamamoto, K.; 238 Yasuda, S.; 333 Yushkanov, A.A.; 238

–Z– Zhdanov, V.M.; 332 Ziering, S.; 89, 139, 155, 159, 178, 266, 268, 287

Subject Index

–A– accommodation coefficients; 77 additive quantities; 1 adjoint operator; 369 aerosol particle; 211 external parameters; 212 anti-symmetric unit tensor; 276 apse-line; 25 arbitrary intermolecular potential, first-order approximation; 67 second-order approximation; 68 auxiliary parameter; 92 auxiliary system of ordinary differential equations; 92 average gas speed; 236 Avogadro’s number; 317 azimuthal symmetry in the sphere torque problem; 273

–B– binary encounters; 31 binary gas mixtures; 289 average molecular mass; 290 Chapman-Enskog solutions, first-order, diffusion; 290 thermal conductivity; 290 viscosity; 290 second-order, diffusion; 294 thermal conductivity; 294 viscosity; 294

collision interval; 43 diffusion coefficient; 290 diffusion coefficient for a H2-N2 mixture; 316 diffusion coefficients for different order approximations; 317 diffusion-slip coefficient by Maxwell method; 312 distribution functions; 289 flow through a capillary, constant temperature with unequal initial mixtures; 323 temperature gradient with equal initial mixtures; 328 hydrostatic pressure; 290 mass density; 290 mean free path; 42 mean mass velocity; 313 mean velocities of constituents; 313, 315 mole fractions of constituents; 290 molecular masses of constituents; 290 number densities of constituents; 290 number flux vector (current); 314 planar problems; 300, 306 diffusion-slip coefficient; 312 limiting case of a simple gas; 308 Loyalka method; 303 Maxwell method; 302 slip coefficients; 307

400

Analytical Methods for Problems of Molecular Transport

relationship between mean velocity and mean mass velocity; 316 slip coefficient; 304 slip-flow coefficient using the firstorder Chapman-Enskog approximation; 322 tangential momentum accommodation coefficients; 312 thermal conductivity coefficien,t first-order Chapman-Enskog approximation; 318 thermal conductivity coefficient, second-order Chapman-Enskog approximation; 320 thermal diffusion ratio; 290 transport coefficients; 291, 309, 310 two vessels connected by a slot; 47 uniform steady-state; 42 viscosity coefficient, first-order Chapman-Enskog approximation; 318 second-order Chapman-Enskog approximation; 319 Boltzmann collision integral; 16, 34 Boltzmann constant; 10, 317 Boltzmann equation; 15 condition of applicability; 35 difficulties in solution; 239 dimensionless form; 53 infinite, unsteady, non-uniform gas; 53 integral form; 182 Boltzmann H-theorem; 39 boundary conditions, condensable gas; 79 non-condensable gas; 77 ordinary moment form; 168 reflected molecules of the same liquid vapor; 86 small Knudsen number; 82 solid impenetrable wall; 86 boundary dispersion kernel; See dispersion kernel boundary values; 77 bracket integrals, analytical expressions; 247 arbitrary intermolecular potentials; 67 commutative property; 151 curvilinear geometry; 351

cylindrical geometry, special function approximations; 361 special function of the first kind; 355 special function of the second kind; 358 independent of intermolecular potential; 271 linearized problems; 248 method of calculation; 159 notations; 302 of Gross-Ziering; 347 planar geometry; 251 first-order Chapman-Enskog solution for viscosity; 338 with one discontinuous sign function; 348 with Sonine polynomials; 335 with two discontinuous sign functions; 339 special functions, approximate expressions; 360 spherical geometry, polynomial expansion method; 363, 367 special function approximations; 361 special function of the first kind; 351

–C– capillary flow between vessels, isothermal gas mixture; 130 isothermal simple gas; 130 non-isothermal gas mixture; 127 non-isothermal simple gas; 129 Cauchy problem; 109 center of mass; 336 central force field; 25 Chapman-Enskog correction to the distribution function; 142 Chapman-Enskog method; 91 Chapman-Enskog solutions, diffusion in a gas mixture; 290 first-order; 289 polynomial expansion, thermal conductivity; 71 viscosity; 72 second coefficients; 70

401

Subject Index second-order; 68, 294 thermal conductivity; 71 viscosity; 72 thermal conductivity in a gas mixture; 290 viscosity in a gas mixture; 290 Chapman-Enskog theory in the limit of a simple gas; 60 characteristic, density; 243 in linearized problems; 247 dimension; 75 mean velocity; 243 system; 92 temperature; 243 in linearized problems; 247 characteristics as molecular trajectories; 92 classical mechanics; 1 Clausius-Clapeyron equation; 87, 88 closure of moment systems; 19 coaxial cylinders, heat flux between; 133, 226 temperature distribution between; 226 collision interval; 43 collision operator; 34, 71 equal to zero; 71 self-adjointness; 151 collisional invariants; 241 heat conduction from a sphere; 254 communicable energy; 9 concentric spheres, heat flux between in a binary gas mixture; 225 heat flux between in a monatomic gas; 134 temperature distribution between in a binary gas mixture; 225 condensation coefficient; 79 condensation on a spherical droplet; 106 cone of influence; 94, 242 cylindrical geometry; 244, 245 sphere; 93 spherical geometry; 243 conservation of energy; 71 conservation of momentum; 71 conservation of tangential momentum; 146 continuity equation; 241

continuum equations; 19 in the slip-flow regime; 211 continuum medium; 18 continuum regime; 76 coordinates in velocity space, cylindrical vs. Cartesian; 246 spherical vs. Cartesian; 246 correction terms to the distribution, integral equations for first- and second-orders; 55 Couette flow, Gross-Ziering method; 171 Loyalka method; 171 Maxwell method; 169 Cunningham slip correction factor; 217 current density from a metal surface with a work function; 49 curvilinear coordinates; 239, 240 curvilinear geometry; 271

–D– de Broglie wavelength; 12 definite integrals; 373 in boundary problems; 374 in non-linear transport problems; 378 in the second-order Chapman-Enskog solution; 376 degree of rarefaction; 75 degrees of freedom; 212 density; 2 density of a drop; 108 diameter of a molecule; 336 differential scattering cross section for rigid-sphere molecules; 336 diffusely reflected molecules; 46 diffusion coefficient, binary gas mixture; 291 different order approximations; 317 H2-N2 gas mixture; 316 diffusion velocities; 316 diffusion-pressure effect; 131 diffusion-slip coefficient, accuracy of Maxwell method; 322 binary gas mixture; 305, 312 dimensionless radial vector; 241 dimensionless thermal force; 221 NaCl aerosols in Ar; 222 Dirac delta function; 81, 162 discontinuous Maxwellian distribution function,

402

Analytical Methods for Problems of Molecular Transport

curvilinear geometries; 242 planar geometry; 242 disk in a rarefied gas, thermophoretic force on; 120 dispersion kernel; 77, 80, 81, 85 dispersion operator; 252 distribution function; 3, 11 Chapman-Enskog correction; 142 cylindrical coordinates; 44 discontinuous in velocity space; 94, 142 first-order approximation; 55 free-molecular in the full velocity space; 94 full corrections to; 81 in a linear expansion; 54 integral equation for the second-order correction; 56 kinetic energy of translation in a Maxwellian gas; 45 Maxwellian; 41 moment of; 4 of incident molecules; 78 of reflected molecules; 78 rebounding molecules; 87 rebounding molecules with evaporation/ condensation; 88 second-order approximation; 56 second-order conditions; 56 slip-flow problem; 142 surface of discontinuity; 94 thermal-creep problem; 142 drag force, in the free-molecular regime; 217 on a sphere in a monatomic gas; 112 drag ratio; 217

–E– effective molecular diameter; 67 effusion; 47 electron gas, flux density; 48 energy accommodation coefficient; 79, 87, 88 energy conservation; 21 energy parameter of a potential model; 380 energy transport equation; 241 equation of state for a perfect gas; 10 error function; 378 evaporation coefficient; 79

expansion in powers of a small parameter; 53 expansion into a vacuum, from a spherical surface; 138 from an infinite planar surface; 137 exponential integrals; 188 external force; 15 external non-uniformity of a gas; 193

–F– first variation; 371 first-order Chapman-Enskog approximation, slip-flow coefficient for a binary gas mixture; 322 thermal conductivity coefficient for a binary gas mixture; 318 viscosity coefficient for a binary gas mixture; 318 flow between two parallel surfaces; 236 flow in a cylindrical tube; 230, 231 flow through a capillary, binary gas mixtures, driving forces; 329 pressure difference; 323 relative density; 323 steady-state pressure difference; 328 diffusion velocities; 325 mean mass velocity; 325 net mass transfer in a closed loop; 233 flux density vector; 6 flux of the energy; 79 flux of the number of molecules; 79 force, particle in a uniform flow; 96 round disk in a uniform flow; 116 fourth moment equation for heat conduction from a sphere; 254 free-molecular regime; 76, 91 full-range moment equations, apparent viscosity coefficient; 172 Couette flow; 172 mean velocity; 172 pressure tensor component; 172 two-moment approach; 172 functional of a trial function; 371

–G– gamma function; 62 gas flow,

403

Subject Index along a plane wall; 82 number density around a sphere in the free-molecular regime; 100 number density at large distance from a small slot; 123 number density at the exit of a round hole; 122 past a sphere in the free-molecular regime; 99 generating functions, Hermite polynomials; 363 Sonine polynomials; 363 geometry of an encounter; 21 Gross-Ziering bracket integrals and their relationship to Chapman-Enskog bracket integrals; 159 Gross-Ziering method; 155 apparent viscosity coefficient; 171 Couette flow; 171 mean velocity profile; 171 pressure tensor component; 171 growth rate of a drop; 109

–H– half-range moment method; 141, 155 four-moment approximation; 155 mean velocity of the gas; 161 polynomials in velocity space; 156 thermal-creep problem; 166 velocity-slip problem; 155 half-space integral notation; 376 heat conduction between two parallel plates; 265 ‘smoothed’ distribution function method; 268 half-range moment method; 268 polynomial expansion method; 268 heat conduction from a sphere; 254, 271 free-molecular heat flux; 258 polynomial expansion method; 264 surface heat flux; 258 heat flux, between coaxial cylinders; 226 Chapman-Enskog solution as a molecular property; 280 domain of applicability of continuum formula; 285 domain of applicability of freemolecular formula; 285 four-moment approach; 279

between concentric spheres; 225 ‘smoothed’ distribution function method; 278 domain of applicability of continuum formula; 286 domain of applicability of freemolecular formula; 286 four-moment approach; 277 from a sphere; 224 per unit length, inner cylinder; 280, 281 Heaviside step function; 303 Hermite polynomials; 261 generating function; 363 Hilbert space; 145 hydrostatic pressure; 8

–I– impenetrable surface condition; 7 infinite gas in an unsteady, non-uniform state; 53 integral boundary conditions for heat and tangential momentum transport problems; 253 integral boundary model; 252 integral operators in Kinetic Theory; 369 planar problems; 369 integral properties of a gas; 181 integral theorem; 57 intermolecular potential models, Lennard-Jones (6-12); 67 parameters of; 67, 311 invariant quantities; 18 isothermal-creep coefficient; 83, 164 isothermal-slip coefficient in the sphere drag problem; 216

–J– Jacobian; 42, 111

–K– kinematic viscosity; 84, 146, 213 kinetic energy; 9 peculiar motion; 9 volume element; 9 Kinetic Theory of Gases; 5, 11 regimes of; 76 Knudsen iteration; 76 Knudsen layer; 181 boundary conditions for NavierStokes equations; 82 Knudsen number; 75

404

Analytical Methods for Problems of Molecular Transport

–L– Lagrange system; 92, 110 latent heat of evaporation/condensation; 87, 88 Legendre polynomials; 104 Leibnitz’ formula; 365 Lennard-Jones (6-12) potential; 202 linearized boundary condition; 96 linearized transport problems; 181 Liouville theorem; 15, 34 local Maxwellian distribution function; 55 Loyalka method; 149 analytical formulas, pressure jump, monatomic condensable gas over liquid; 176 monatomic non-condensable gas; 174 temperature-jump, monatomic condensable gas over liquid; 176 monatomic non-condensable gas; 174 apparent viscosity coefficient; 171 boundary-jump effects; 175 Couette flow; 171 generalization of the Maxwell method; 150 mean velocity profile; 171 pressure tensor component; 171 slip coefficients; 151 slip velocities; 151 temperature-jump problem; 152

–M– macroscopic quantities; 11 free-molecular regime; 98 mass density; 2 mass of the Earth’s atmosphere; 50 Maxwell integral transport equations; 240 heat conduction between plates; 266 Maxwell method; 141 apparent thermal conductivity coefficient; 173 apparent viscosity coefficient; 169 Couette flow; 169 form of the distribution function; 145 main deficiency; 149 mean velocity far from the wall; 146 mean velocity profile; 169

pressure tensor component; 169 slip factors; 147 slip-flow problem; 142 temperature distribution between parallel plates; 173 temperature-jump problem; 147 thermal-creep problem; 142 Maxwellian boundary model; 77, 183, 274 dispersion kernel; 85 heat conduction between two parallel plates; 267 heat conduction from a sphere; 256 number density of fictitious gas; 78 reflected distribution function; 85 temperature of fictitious gas; 78 Maxwellian distribution function; 41 as a first approximation; 55 local; 55 Maxwellian gas, distribution function of kinetic energy of translation; 45 number of molecules impinging upon a unit surface element in a given angular range; 45 number of molecules impinging upon a unit surface element in a given velocity range; 45 Maxwellian velocity distribution; 40 mean distance between collisions; 43 mean free path; 36, 43, 65 from thermal conductivity coefficient; 305 from viscosity coefficient; 305 in N2 at STP; 44 Knudsen number; 75 other representation; 146 simple gas; 44 two-dimensional, rigid-sphere gas; 37 mean heat velocity; 13 mean time between collisions; 43 mean value of a molecular property; 110 mean velocity; 4, 11, 167 equilibrium gas; 46 fictitious value; 82 molecular beam; 46 of a gas near a moving wall; 82 profile; 200 real; 82

405

Subject Index thermal-creep in a tube; 232 uniform flow past a sphere; 113 variational method; 187 mechanical equilibrium condition, thermal-creep in a cyindrical tube; 231 methods of closure, polynomial expansions in velocity space; 241 special choice for the distribution function; 242 metric coefficients; 240 cylindrical geometry; 240 spherical geometry; 240 microscopic boundary conditions; 252 mobile operator; 18 model of diffuse reflection; 79 modified two-stream Maxwellians; 243 molecular encounters; 19 summational invariants of; 55 molecular flux, in a given angular range; 45 in a given velocity range; 45 molecular interaction; 19 molecular property; 4, 16 conserved during encounters; 18 fluxes; 5, 11 mean value of; 4, 98 rate of variation; 34 moment equations; 16, 19 features of boundary conditions; 162 moment of the collision integral; 34 moment of the collision operator; 256 moments of discontinuous distribution functions; 245 momentum conservation; 20 monatomic gas; 18 mutual potential energy; 25

–N– Navier-Stokes equations; 76 as moment equations in the Maxwellian analysis; 143 near free-molecular regime; 76 net flux; 6 net mass transport, in a cylindrical tube; 230 thermal-creep in a cyindrical tube; 231 Newton’s equations of motion; 27 non-dimensionalization; 53 non-stationary gas flows; 109

normal coordinate; 182 number density; 2, 3, 11 saturated vapor; 107 simple gas in a rotating cylinder; 50 number flow of molecules; 6 numerical quadratures; 380

–O– omega-integrals, first-order approximation; 379 with arbitrary intermolecular potential; 67 gas mixtures; 379 reduced; 68, 380 rigid-sphere molecules; 68, 379 second-order approximation; 379 tables of; 382

–P– parallel plates, frictional force on due to horizontal motion; 125 gas flow between; 126 gas temperature between; 131 parameter of non-uniformity; 54 parameters of an encounter; 22 peculiar velocity, mean value of; 4 perfect gas; 10 phase space; 4 phase volume element; 27 planar boundary transport problems; 181 Planck constant; 13 plate in a steady gas flow, drag when angled to flow; 118 drag when parallel to flow; 117 point-mass hypothesis; 12 polar coordinates; 261, 349, 352 polynomial expansion method; 260 cylindrical geometry; 262 planar geometry; 262 spherical geometry; 260 potential flow field for a sphere; 114 pressure; 6, 11 pressure difference due to thermal-creep in a tube; 231 pressure distribution; 8 pressure tensor; 6 general expression; 66

–R– radial momentum equation; 241 rarefied gases; 1

406

Analytical Methods for Problems of Molecular Transport

reduced heat flux; 266 experiment vs. theory; 270 ratio; 265, 269 reduced impact parameter; 380 reduced intermolecular distance; 380 reduced intermolecular potential energy; 380 reduced omega-integrals; 380 reduced relative kinetic energy; 380 reduced temperature; 68, 202, 285, 380 reduced torque on a rotating sphere; 277 reflected distribution function, condensible gas; 80 correction to; 81 reflection, diffuse; 77 operator; 184 specular; 77 relative fluctuation of an additive quantity; 12 relative molecular velocity; 336 relative motion of two molecules; 25 resistance force on a moving disk; 227 Reynolds number; 213 Richardson thermionic emission formula; 49 rigid-sphere gas, model; 19 molecular mean free path; 36 temperature independence; 203 two-dimensional; 37 root mean square velocity of a molecular beam; 46 rotating gas, equivalence to an external force field; 50 number density distribution; 50 round disk in a gas, force on from different temperature sides; 119 force on in a uniform flow; 116 number density behind in a flow; 121, 124

–S– saturated vapor density at a surface; 80 scalar product; 185, 301 scattering cross section; 29 differential; 29 integral; 30

rigid spheres; 30 total; 30 scattering solid angle; 31 second variation; 372 second-order Chapman-Enskog solutions; 306 thermal conductivity coefficient for a binary gas mixture; 320 viscosity coefficient for a binary gas mixture; 319 second-order correction, as a summational collision invariant; 56 homogeneous term; 56 non-homogeneous term; 57 self-adjoint operator; 369 sign function; 156 simple gas, in a rotating cylinder; 50 steady-state number density variation in a force field; 49 two vessels connected by a slot; 47 uniform steady-state; 39, 41 slip coefficient; 165, 201 binary gas mixture; 322 isothermal; 83 Lennard-Jones (6-12) potential; 202 second-order Maxwell method; 203 variational method; 187 slip factor; 311 experiment vs. theory; 312 slip problem geometry; 142 slip velocity; 82 slip-flow regime; 76, 211 average gas speed between two parallel surfaces; 236 boundary conditions; 219 calculation error in thermal conductivity coefficient; 225 drag force on parallel round disks; 227 flow in a cylindrical tube, isothermal pressure difference; 231 net mass transport; 230 velocity distribution; 230 flow in capillary tubes; 233 heat flux, between coaxial cylinders; 226 between concentric spheres; 225 from a sphere; 224

407

Subject Index temperature distribution, around a sphere; 224 between coaxial cylinders; 226 between concentric spheres; 225 thermal conductivity coefficient, around a heated sphere; 224 calculation error in heated wire in a cylinder method; 227 heated wire in a cylinder method; 227 thermal-creep in a cylindrical tube; 231, 232 torque, on a sphere; 234 on an infinite cylinder; 235 slip-flow results; 188 half-range moment method; 189 ILT boundary model; 189 Maxwellian boundary model; 189 Loyalka method; 189 variational method; 189 smooth molecules; 20 smoothed distribution function method; 259 collision operator; 259 corrections to density and temperature; 259 Sonine polynomials; 62, 261 generating function; 363 orthonormal conditions; 261 special functions; 252 cylindrical geometry; 361 spherical geometry; 263 specific heat per unit mass at constant volume in a monatomic gas; 68 speed ratio; 117 sphere drag problem; 95, 211, 214 dimensional analysis; 213 experiment vs. theory; 218, 219 external parameters; 212 satisfied conditions; 212 sphere in a monatomic gas, drag force; 112 heat flux; 115 temperature distribution; 115 temperature of; 111 spherical coordinates; 337 standard collision operator; 56, 145 standard temperature and pressure; 1

stationary functional; 371 STP; 1 summational invariants; 55 supersaturation; 107 surface element of a sphere; 276 surface tension coefficient; 108 system of moment equations, method of closure; 241

–T– tangential momentum, accommodation coefficient; 78, 117, 312 conservation of; 146 temperature; 9, 11 of a gas near a wall; 84 of a particle; 104 of a sphere in a monatomic gas; 111 of a surface; 79 real; 84 temperature defect at the wall, Loyalka method; 206 Maxwell method, first-order; 205 second-order; 205 temperature distribution, around a sphere; 224 between coaxial cylinders; 226 between concentric spheres; 225 inside a particle; 104 temperature-jump; 83, 84, 202, 220 Lennard-Jones (6-12) potential; 202 Loyalka method and general boundary conditions; 206 Maxwell method and general boundary conditions; 204 Maxwell method, second-order; 204 thermal conduction equation, boundary conditions; 220 first-order Chapman-Enskog solution; 62 rigid-sphere molecules; 63 thermal conductivity; 212 binary gas mixture; 291, 318, 320 first-order approximation; 68 first-order Chapman-Enskog theory; 147 heat conduction from a sphere; 258 heat loss from a hot sphere method; 224, 225

408

Analytical Methods for Problems of Molecular Transport

heated wire in a cylinder method; 226 calculation error; 227 monotomic gas of rigid spheres; 66 of a particle; 104 second-order approximation; 69 rigid-sphere molecules; 71 thermal diffusion ratio; 315, 328 thermal flux; 11 thermal flux vector; 10 thermal force; 106 dimensionless; 221 one component; 103 thermal force problem; 211, 218 dimensional analysis; 213 experiment vs. theory; 222, 223 external parameters; 212 geometry; 102 satisfied conditions; 212 third moment equation; 213 thermal transpiration effect; 311 thermal-creep; 84, 220 binary gas mixture; 305 geometry; 85 in a cylindrical tube, mean velocity variation; 232 pressure difference; 231 influence of boundary models; 168 Lennard-Jones (6-12) potential; 202 second-order Maxwell method; 203 velocity; 84 thermal-creep results; 196 mean velocity profile half-range moment method; 199 thermal-creep coefficient, half-range moment method, ILT boundary model; 196 integral boundary model; 196 Loyalka method; 196 variational method; 196 velocity defect, half-range moment method, ILT boundary model; 196 integral boundary model; 196 Loyalka method; 196 variational method; 196 thermal-slip; See thermal-creep thermophoresis; 101, 218 corrections to; 223 time duration of an encounter; 35 torque problem; 272

torque, on a rotating body, flat disk; 135 infinite cylinder; 136 sphere; 137 on a sphere; 234, 273 arbitrary potential model; 284 in CO2; 284 in N2; 284 Lennard-Jones (6-12) potential; 284 Maxwellian boundary model; 274 moment system; 274 net; 276 on an infinite cylinder; 235 on two coaxial cylinders; 283 on two concentric spheres; 281 per unit length; 136 coaxial cylinders; 284 slip-flow regime; 235 transport coefficients; 70 functions for calculating second-order; 70 functions for Lennard-Jones (6-12) potential; 70 monatomic gas; 68 transport equation; 17 trial function; 371 two vessels connected by a slot, gas mixture density ratio; 47 kinetic equilibrium; 47 two vessels connected by a tube, pressure difference due to thermalcreep; 231, 232 velocity distribution in the tube; 231 two-dimensional gas; 28 two-sided Maxwellian distribution function; 242 in linearized problems; 247

–U– uniform flow past a sphere, mean velocity distribution; 113 potential flow field; 114 pressure; 114 temperature distribution; 114 uniform steady-state of a gas; 41

–V– variational method; 181 planar geometry; 181

409

Subject Index slip-flow problem; 183 thermal-creep coefficient; 195 thermal-creep problem; 192 velocity defect; 195 variational parameter; 186 variational principle; 370 velocity defect; 164, 168 at the wall; 190 influence of boundary models; 168 Maxwell method, slip-flow problem; 204 thermal-creep problem; 204 slip-flow problem, integral representation of the distribution function; 207, 208 thermal-creep problem, integral representation of the distribution function; 208 Loyalka method; 208 variational method; 187 velocity distribution, due to thermal-creep; 231 in a cylindrical tube; 230 Maxwellian; 40 velocity,

actual; 3 mean; 2 peculiar; 3 relative; 3 velocity-slip coefficient for a binary gas mixture; 305 viscosity; 212 binary gas mixture; 291, 318, 319 first-order approximation; 68 first-order Chapman-Enskog theory; 147 monotomic gas of rigid spheres; 66 second-order approximation; 69 rigid-sphere molecules; 73 viscosity equation, first-order Chapman-Enskog solution; 64 rigid-sphere molecules; 65 volume element; 4

–W– work function; 49

–Z– zone of influence for a cylinder; 133

FLUID MECHANICS AND ITS APPLICATIONS

Volume 83 Series Editor: R. MOREAU

MADYLAM Ecole Nationale Supérieure d'Hydraulique de Grenoble Boîte Postale 95 38402 Saint Martin d'Hères Cedex, France

Aims and Scope of the Series The purpose of this series is to focus on subjects in which fluid mechanics plays a fundamental role. As well as the more traditional applications of aeronautics, hydraulics, heat and mass transfer etc., books will be published dealing with topics which are currently in a state of rapid development, such as turbulence, suspensions and multiphase fluids, super and hypersonic flows and numerical modeling techniques. It is a widely held view that it is the interdisciplinary subjects that will receive intense scientific attention, bringing them to the forefront of technological advancement. Fluids have the ability to transport matter and its properties as well as to transmit force, therefore fluid mechanics is a subject that is particularly open to cross fertilization with other sciences and disciplines of engineering. The subject of fluid mechanics will be highly relevant in domains such as chemical, metallurgical, biological and ecological engineering. This series is particularly open to such new multidisciplinary domains. The median level of presentation is the first year graduate student. Some texts are monographs defining the current state of a field; others are accessible to final year undergraduates; but essentially the emphasis is on readability and clarity.

For a list of related mechanics titles, see final pages.

Analytical Methods for Problems of Molecular Transport by

I.N. IVCHENKO University of Missouri-Columbia, Columbia, Missouri, U.S.A.

S.K. LOYALKA University of Missouri-Columbia, Columbia, Missouri, U.S.A.

and

R.V. TOMPSON, JR. University of Missouri-Columbia, Columbia, Missouri, U.S.A.

A C.I.P. Catalogue record for this book is available from the Library of Congress.

ISBN-10 ISBN-13 ISBN-10 ISBN-13

1-4020-5864-0 (HB) 978-1-4020-5864-6 (HB) 1-4020-5865-9 (e-book) 978-1-4020-5865-3 (e-book)

Published by Springer, P.O. Box 17, 3300 AA Dordrecht, The Netherlands. www.springer.com

Printed on acid-free paper

All Rights Reserved © 2007 Springer No part of this work may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work.

Dedications

This book is dedicated to the memory of my son Yaroslav.

I. N. Ivchenko This book is dedicated to the memory of Joel H. Ferziger, Heinz Lang, and Lloyd B. Thomas who inspired me to work in this area.

S. K. Loyalka This book is dedicated to my family; my parents, my siblings, my nephews and nieces, and especially to my wife Elena and my son Arseniy.

R. V. Tompson, Jr.

Special Dedication

This book is especially dedicated to the memory of our very dear friend and colleague

Igor Nikolaevich Ivchenko without whom this book would not have been possible.

S. K. Loyalka R. V. Tompson, Jr.

Contents

Table of Tables

xiii

Table of Figures

xv

Preface

xvii

Acknowledgments

xxiii

Chapter 1. The General Description of a Rarefied Gas 1. Some Introductory Remarks. 2. Density and Mean Motion. 3. The Distribution Function of Molecular Velocities. 4. Mean Values of Functions of Molecular Velocities. 5. Transport of Molecular Properties. 6. The Pressure Tensor. 7. The Hydrostatic Pressure. 8. The Amount of Heat. 9. The Kinetic Temperature. 10. The Equation of State for a Perfect Gas. 11. The Thermal Flux Vector. 12. Summary. Problems References

Chapter 2. The Boltzmann Equation 1. 2. 3. 4. 5. 6.

Derivation of the Boltzmann Equation. The Moment Equations. Another Form of the Moment Equations. The Equations for a Continuum Medium. Molecular Encounters. The Relative Motion of Two Molecules.

1 1 2 3 4 4 6 8 9 9 10 10 11 12 14

15 15 16 18 18 19 25

x

Analytical Methods for Problems of Molecular Transport Problems References

Chapter 3. The Collision Operator

27 28

29

1. The Differential and Total Scattering Cross Sections. 2. The Statistics of Molecular Encounters. 3. The Transformation of Some Integrals. Problems References

29 31 35 35 37

Chapter 4. The Uniform Steady-State of a Gas

39

1. The Boltzmann H-Theorem. 2. The Maxwellian Velocity Distribution. 3. The Mean Free Path of a Molecule. Problems References

Chapter 5. The Non-Uniform State for a Simple Gas 1. Expansion in Powers of a Small Parameter. 2. The First Approximation. 3. A General Formal Solution for the Second Correction. 4. The Transformation of the Non-Homogeneous Term. 5. The Second Approximation. 6. The First-Order Chapman-Enskog Solution for Thermal Conduction. 7. The First-Order Chapman-Enskog Solution for Viscosity. 8. The Thermal Conductivity and Viscosity Coefficients. 9. The First-Order Approximation for Arbitrary Intermolecular Potential. 10. The Second-Order Approximation for Arbitrary Intermolecular Potential. Problems References

Chapter 6. Regimes of Rarefied Gas Flows 1. The Knudsen Number. 2. A General Analysis of the Different Gas Flow Regimes. 3. The Boundary Conditions. 4. The Boundary Dispersion Kernel. 5. Features of the Boundary Conditions for Small Knudsen number. Problems References

Chapter 7. The Free-Molecular Regime 1. The Free-Molecular Distribution Function. 2. The Force on a Particle in a Uniform Gas Flow. 3. Calculation of Macroscopic Values in the Free-Molecular Regime. 4. Thermophoresis of Particles in the Free-Molecular Regime. 5. Condensation on a Spherical Droplet. 6. Non-Stationary Gas Flows. Problems References

39 40 42 44 51

53 53 55 56 57 59 62 64 66 67 68 70 73

75 75 76 77 80 82 85 88

91 91 96 98 101 106 109 111 139

Contents

Chapter 8. Methods of Solution of Planar Problems 1. Maxwell’s Method. 2. Loyalka's Method. 3. The Half-Range Moment Method. 4. Features of the Boundary Conditions for the Moment Equations. 5. Solution of the Thermal-Creep Problem by the Half-Range Moment Method. 6. Influence of the Boundary Models on the Thermal-Creep Coefficient. Problems References

Chapter 9. The Variational Method for the Planar Geometry 1. 2. 3. 4. 5. 6.

Another Form of the Boltzmann Equation. The Variational Technique for the Slip-Flow Problem. Discussion of the Slip-Flow Results. The Variational Solution for the Thermal-Creep Problem. Discussion of the Thermal-Creep Results. Slip-Flow and Temperature-jump Coefficients for the Lennard-Jones (6-12) Potential Model. Problems References

Chapter 10. The Slip-Flow Regime 1. Basic Equations. 2. The Spherical Drag Problem. 3. The Thermal Force Problem. Problems References

Chapter 11. Boundary Value Problems for All Knudsen Numbers 1. The Moment Equations in Arbitrary Curvilinear Coordinates. 2. The Two-Sided Maxwellian Distribution Functions. 3. Moments of Discontinuous Distribution Functions. 4. Analytical Expressions for the Bracket Integrals. 5. Boundary Conditions for Moment Equations. 6. Thermal Conduction from a Heated Sphere. 7. Method of the ‘Smoothed’ Distribution Function. 8. The Polynomial Expansion Method. 9. Solution of One Classic Transport Problem. 10. A Simplification of Moment Systems for Curvilinear Problems. 11. The Torque Problem. Problems References

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture 1. 2. 3. 4. 5. 6.

The First-Order Chapman-Enskog Approximation for a Binary Gas Mixture. The Transport Coefficients for a Binary Gas Mixture. The Second-Order Chapman-Enskog Approximation for a Binary Gas Mixture. Analytical Methods of Solution for Planar Boundary Value Problems Involving Binary Gas Mixtures. The Slip Coefficients for a Binary Gas Mixture. Discussion of the Slip Coefficient Results.

xi 141 141 149 155 162 165 168 169 178

181 181 183 188 192 196 200 203 208

211 211 214 218 224 237

239 239 242 245 247 252 254 259 260 265 270 272 277 286

289 289 291 294 300 304 308

xii

Analytical Methods for Problems of Molecular Transport Problems References

Appendix 1. Bracket Integrals for the Planar Geometry 1. Bracket Integrals Involving Two Sonine Polynomials. 2. Bracket Integrals Containing Several Components of Molecular Velocity. 3. Bracket Integrals Containing Two Discontinuous Functions. 4. Bracket Integrals Containing One Discontinuous Function. References

Appendix 2. Bracket Integrals for Curvilinear Geometries 1. The Special Function of the First Kind for the Spherical Geometry. 2. The Special Function of the Second Kind for the Spherical Geometry. 3. The Special Function of the First Kind for the Cylindrical Geometry. 4. The Special Function of the Second Kind for the Cylindrical Geometry. 5. Approximate Expressions for the Special Functions. References

Appendix 3. Bracket Integrals for Polynomial Expansion Method 1. Calculation of the Bracket Integrals of the First Kind. 2. Analytical Expressions for the Bracket Integrals of the Second Kind. References

Appendix 4. The Variational Principle for Planar Problems 1. Some Definitions and Properties for Integral Operators. 2. The Variational Principle. References

Appendix 5. Some Definite Integrals 1. 2. 3. 4.

Some Frequently Encountered Integrals. Some Integrals Encountered in Boundary Problems. Some Integrals Connected with the Second-Order Chapman-Enskog Solution. Some Integrals Connected with Non-Linear Transport Problems.

Appendix 6. Omega-Integrals for Second-Order Approximation

312 331

335 335 338 339 348 350

351 351 353 355 358 360 361

363 363 367 368

369 369 370 372

373 373 374 376 378

379

References

393

Author Index

395

Subject Index

399

Table of Tables

Table 2-1. Transformation of the various derivatives to new variables. 5-1. Functions for calculating the second-order transport coefficients. 8-1. Values of the slip-flow coefficient. 9-1. The slip-flow coefficient. 9-2. The slip-flow coefficient. 9-3. The velocity defect at the wall. 9-4. The velocity defect at the wall. 9-5. The dimensionless velocity defect. 9-6. The mean velocity. 9-7. The thermal-creep coefficient. 9-8. The thermal-creep coefficient. 9-9. The velocity defect at the wall. 9-10. The velocity defect at the wall. 9-11. The dimensionless velocity defect. 9-12. Numerical values of the velocity defect. 9-13. Velocity defect data obtained during the solution of Problem 9.8. 10-1. The isothermal-slip coefficient for different values of the tangential momentum accommodation coefficient. 10-2. Comparison of experimental and theoretical values of drag on a sphere. 10-3. The reduced thermal force. 10-4. The reduced thermal force on NaCl aerosol particles. 11-1. Expressions for some bracket integrals. 11-2. Values of the reduced heat flux ratio. 11-3. Values of the reduced heat flux ratio. 11-4. Analytical and numerical values of reduced torque on a rotating sphere.

page 17 70 165 190 190 191 191 192 192 197 197 198 198 199 199 207 217 218 222 222 264 265 269 277

xiv

Analytical Methods for Problems of Molecular Transport

12-1. Values of the transport coefficients and of the slip coefficients for a selection of binary gas mixtures obtained using the rigid-sphere potential model and the firstand second-order Chapman-Enskog approximations. 12-2. Values of the transport coefficients and of the slip coefficients for a selection of binary gas mixtures obtained using the Lennard-Jones (6-12) potential model and the first- and second-order Chapman-Enskog approximations. 12-3. Relevant parameters for two of the most commonly used intermolecular potential models; the rigid-sphere model and the Lennard-Jones (6-12) potential model. 12-4. The tangential momentum accommodation coefficients in a binary gas mixture chosen for calculation of the slip coefficients and slip factors. 12-5. A comparison of diffusion-slip coefficient values. A-1. Values of the J-integrals for use in evaluating bracket integrals containing two discontinuous functions. A-2. Values of the I-integrals for use in evaluating bracket integrals for planar problems. B-1. Numerical values of the spherical special functions of the first- and second-kind. B-2. Numerical values of the cylindrical special functions of the first- and second-kind. B-3. Numerical values of the integrals of the cylindrical special functions of the firstand second-kind. C-1. Values of the bracket integrals of the first-kind for the polynomial expansion method. C-2. Analytical values of the bracket integrals of the second-kind for the polynomial expansion method. E-1. Values of the commonly encountered i-integrals. F-1. Values of some reduced omega-integrals. F-2. A Mathematica® program to compute the reduced omega-integrals for selected values of the reduced temperature.

309

310 311 312 322 342 348 353 357 359 366 368 374 382 386

Table of Figures

Figure 1-1. Passage of molecules across an arbitrary surface element. 1-2. Components of the pressure tensor. 2-1. Geometry of a generic encounter. 2-2. Parameters of a generic encounter. 2-3. The spherical coordinates of the relative velocity. 3-1. The scattering solid angle. 3-2. Encounters of rigid-sphere molecules. 4-1. The geometry to be used in Problem 4.8 in determining the rate at which molecules reflected from one differential surface element cross another. 6-1. The variation of the mean velocity of a gas near a moving wall. 6-2. The variation of the gas temperature near a wall. 6-3. The thermal-creep geometry. 7-1. The cone of influence for a sphere. 7-2. The geometry of the sphere drag problem. 7-3. The geometry of the thermal force problem. 7-4. The geometry to be used for Problem 7.7 in determining the force on a round disk perpendicular to a uniform gas flow. 7-5. The geometry to be used for Problem 7.8 in determining the drag on a plate parallel to a steady, free-molecular gas flow. 7-6. The geometry to be used for Problem 7.9 in determining the drag on a plate oriented at an angle to a steady, free-molecular gas flow. 7-7. The geometry to be used for Problem 7.10 in determining the force on a round disk in a rarefied gas when the two sides of the disk have different temperatures. 7-8. The geometry to be used for Problem 7.11 in determining the thermophoretic force on a thin disk located in a rarefied gas having a constant temperature gradient.

page 5 7 21 22 23 31 32 47 82 84 85 93 95 102 116 117 118 120 120

xvi

Analytical Methods for Problems of Molecular Transport

7-9. The geometry to be used for Problem 7.12 in determining the number density behind a round disk perpendicular to a free-molecular gas flow in which the mean gas velocity is much less than the thermal molecular velocity. 7-10. The geometry to be used for Problem 7.13 in determining the number density of a free-molecular flow exiting through a round hole. 7-11. The geometry to be used for Problem 7.14 in determining the number density of an effusing gas at large distances from a small slot. 7-12. The geometry to be used for Problem 7.16 in determining the frictional force on a plate due to the horizontal motion of a second, parallel plate. 7-13. The geometry to be used in Problem 7.22 in determining the gas temperature between two parallel plates in the free-molecular regime. 7-14. The geometry to be used for Problem 7.25 in determining the torque on a flat, rotating disk in the free-molecular regime. 8-1. The geometry of slip problems. 8-2. The geometry to be used in Problem 8.1 in determining the mean velocity profile, the pressure tensor component, and the apparent viscosity for Couette flow. 8-3. The geometry to be used in Problem 8.5 in determining the temperature distribution and apparent thermal conductivity coefficient between parallel plates having different temperatures. 9-1. The dimensionless, mean velocity profile. 9-2. The dependence of the slip coefficients on reduced temperature for the LennardJones (6-12) potential model. 9-3. The dependence of the temperature-jump coefficient on reduced temperature for the Lennard-Jones (6-12) potential model. 10-1. Comparison of the experimental and theoretical values of the drag on a sphere. 10-2. A comparison of experimental and theoretical reduced thermal forces for NaCl. 10-3. The geometry to be used in Problem 10.8 in determining the resistance force on a flat, round disk in the slip-flow regime due to its approach to an identical, parallel, stationary disk. 10-4. The geometry to be used for Problem 10.13 in determining the net mass transfer in a closed capillary loop with different radii and end temperatures. 10-5. The geometry to be used in Problem 10.14 in determining the torque on a rotating sphere in the slip-flow regime. 10-6. The geometry to be used for Problem 10.16 in determining the average speed of a steady gas flow between two, parallel, planar surfaces in the slip-flow regime. 11-1. The cone of influence in cylindrical coordinates. 11-2. A comparison of reduced heat flux values 11-3. A comparison between reduced heat flux data for Argon and analytical results. A-1. The relationship between various collision angles. B-1. Dependence of the spherical special functions of the first- and second-kind on the radial coordinate. B-2. Dependence of the cylindrical special functions of the first- and second-kind on the radial coordinate.

122 123 124 126 132 135 142 170

174 200 201 202 219 223

228 233 234 236 244 266 270 344 354 358

Preface

The transport of a given species (atoms, molecules, neutrons, photons, etc.), either through its own kind or through some other host medium, is a problem of considerable interest. Practical applications may be found in many technologically and environmentally relevant areas such as the transport of neutrons in a nuclear power reactor or in a nuclear weapon, the transport of ions and electrons in plasma, the transport of photons which constitutes radiative heat transfer in various industrial, environmental and space applications, the transport of atoms or molecules of one species either through itself or as one component of a multi-component gas mixture, and the interactions of such gas mixtures with various solid and liquid surfaces such as one might find associated with capillary tubes, aerosol particles, interstellar dust grains, etc.. These application areas are obviously quite broad and it is readily apparent that there are, indeed, few scientific activities that do not require some level of understanding of transport processes. One of the most important and influential texts in the area of transport theory has been The Mathematical Theory of Non-Uniform Gases by Sidney Chapman and T.G. Cowling that was first printed in 1939. This book, along with several other more recent texts (Hirschfelder, J.O., Curtiss, C.F. and Bird, R.B., Molecular Theory of Gases and Liquids, John Wiley and Sons, NY, 1954; Kogan, M.N., Rarefied Gas Dynamics, Plenum Press, NY, 1969; Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases, North-Holland, Amsterdam-London, 1972; Cercignani, C., Theory and Application of the Boltzmann Equation, Scottish Academic Press, Edinburg, UK, 1975.), have provided students and researchers in the area of transport theory with an excellent basis for pursuing in-depth research in the area and have made possible some limited, albeit very useful,

xviii

Analytical Methods for Problems of Molecular Transport

applications of the theory. These texts have focused primarily on the development and simplification of the Boltzmann equations and on the derivation of general expressions for the coefficients associated with the transport phenomena of viscosity, thermal conduction, and diffusion in essentially unbounded gases. In recent years, however, it has been recognized that not only do intermolecular collisions within the bulk of the gas play an important role in determining the various transport coefficients but that interactions at boundaries are also significant. Hence, a simultaneous understanding of both is required in order to obtain reasonably complete descriptions of, and ultimately solutions to, any given transport problem. The relative importance of a boundary in a given transport problem is generally specified by Knudsen number, Kn O d , where O represents the mean free path of the species of interest in the host medium while d is some characteristic dimension of the system. Thus, in transport problems in general, there exist several somewhat arbitrarily defined regimes of interest. These cover the range from Kn 1 where a given boundary exerts a strong influence on the transport processes to Kn 1 where a given boundary exerts an essentially negligible influence on the transport processes. Accounts of the effects of boundaries on transport problems have been discussed in a variety of texts on subjects such as Neutron Transport Theory, Radiative Heat Transfer, and the Kinetic Theory of Gases. These, and the basic transport theory books mentioned above, are generally written at a level suitable only for specialists in the field or very advanced students. Thus, even though many have tried to use them as teaching texts, their application in this capacity has remained somewhat constrained with a clear need continuing to exist for a more student-friendly text having a clearer emphasis on basic problem solving. At the University of MissouriColumbia, we became clearly aware of this need in the Winter of 1992 when one of us (INI), as a part of teaching obligations associated with his status as a visiting professor to Columbia from Russia under the Fulbright exchange program, taught a class on the subject of transport theory. Based on this experience, we felt that the time had arrived for a new text that would combine a concise, but basically complete, introduction to the field of transport theory with a fairly tight focus on a few recently successful analytical solution techniques. Further, we recognized that any successful text of this type would, of necessity, include a good selection of easily applicable, representative problems that were either fully solved or which included sufficient basic guidance to assist the problem solver in reaching the correct solution. In this context, we have been strongly influenced by the organization and style of the excellent series of texts by Landau and Lifshitz.

Preface

xix

The subject material in the book divides itself quite naturally into two parts: the non-equilibrium properties of an infinite expanse of gas and those of a bounded gas. The first part (Chapters 1-5) contains the basis of the Kinetic Theory of Gases developed along the traditional lines of Chapman and Cowling. Some simplifications have been made during this development but only to facilitate the reader’s understanding of the basic principles of the kinetic theory. For example, the derivation of the Boltzmann collision operator is substantially simpler than that described in the above mentioned texts. The analysis presented in these chapters is based on the framework of the scattering cross section being the basic quantity characteristic of molecular encounters. In description of the ChapmanEnskog method, we focus our attention on the first-order approximation for rigid-sphere molecules to facilitate one’s understanding of this method. Nevertheless, the second-order Chapman-Enskog solutions along with tabulated values of all quantities necessary for their complete evaluation, are also provided for the Lennard-Jones (6-12) intermolecular potential as this development is potentially very important for practical applications. The presentation of this material here is in the most complete form currently available. The second part of the book (Chapters 6-12) presents focused descriptions of a limited selection of analytical methods that may be used to solve boundary value problems of transport with reasonable accuracy. The typical features found in the statements of boundary value problems for various gas flow regimes are discussed in Chapter 6. Within the framework of the traditional Maxwellian boundary model, new expressions for the reflected distribution functions at surfaces are presented which are then adapted to numerous applied problems of a practical nature. The physical features associated with statements of the boundary conditions for condensable and non-condensable gases are discussed and appropriate definitions for the basic accommodation coefficients are described in detail. Chapter 7 contains a kinetic theory treatment of gas transport problems for extremely high Knudsen numbers. The theory presented here starts with a Boltzmann equation that takes its simplest form due to the absence of the collision operator. Under stationary conditions, the constancy of the distribution function along molecular trajectories results in its discontinuity in velocity space; an effect that may be described in terms of the classic ‘cone of influence.’ For the spherical geometry, a new analytical representation of the distribution function in the full velocity space is given. This representation is a generalization of that previously proposed for the planar geometry. A number of basic problems are considered which show, in detail, the features of a mathematical solution technique that is presented

xx

Analytical Methods for Problems of Molecular Transport

here, in monographic form, for the first time and which otherwise could only be found by perusing the original journal articles in which it was developed. The next two chapters, Chapters 8 and 9, present selected methods of solution of planar transport problems. Detailed analytical solutions of some applied problems are given for the special case when molecules can be approximated as rigid spheres. In our description of selected methods, we have concentrated on various moment approaches in which we have emphasized the importance of using conservation laws to construct exact moment solutions. In Chapter 8, our development of the various moment approaches starts with the classic Maxwellian method and proceeds on to a generalized form of the Maxwellian method, which we term the Loyalka method after the one of us (SKL) who developed it. A key feature of this latter method that we emphasize is the opportunity that it represents to construct an accurate theory of slip phenomena for arbitrary models of the intermolecular potential. We follow these with a description of the characteristic features of the half-range moment method which, in combination with the mathematical solution technique described in Appendix A, is presented here in monographic form for the first time. Also in Chapter 8 we discuss in detail various boundary models and introduce a new model that allows for the incorporation of two conservation laws. We continue our discussion of planar transport problems in Chapter 9 with a discussion of the variational method which is followed by a discussion of the accuracy of all the various solution methods described for planar transport problems. Features of the integral representation of the distribution function and of the construction of basic functionals for slip problems are discussed in detail. We touch briefly on a new approximate method (Prob. 9.8) that may be used to describe gas behavior in the Knudsen layer. This comparative analysis of various approximate methods and estimates of their accuracy are compiled here in monographic form for the first time allowing the reader to make easy comparisons between the various methods. Chapter 10 contains straightforward applications of the slip-flow theory to various classical problems of interest. Derivations of the basic equations and statements of the boundary conditions in the slip-flow regime are discussed as well. While the subject treated here may be found in various places in the literature, several gaps in the subject have been bridged in order to make possible a systematic description of the various linearized problems. Chapter 11 contains original material that has been worked out only recently and which, in some cases, is still under development (such as the sphere drag problem). In many cases, methods and results for arbitrary Knudsen number and different geometries are described for the first time in such a complete manner. For instance, analytical expressions for the bracket

Preface

xxi

integrals associated with each transport problem as well as numerical and various approximate methods of calculation of these bracket integrals for different curvilinear geometries are original and have not been reported previously. The authors were the first to suggest the use of the ChapmanEnskog solutions as the molecular properties used to construct moment systems. This approach yields a very simple procedure that may be used to calculate bracket integrals for arbitrary intermolecular potentials. The advantage of this method is shown in analyses of some classical spherical transport problems such as the torque and thermal conduction problems. In Chapter 12 the analytical methods presented earlier for the Maxwell and Loyalka methods are again used to solve planar boundary value problems involving the various boundary slip phenomena. Here, however, the emphasis is on binary gas mixtures where the preceding chapters have focused only on simple one-component gases. The relevant boundary slip phenomena are discussed in detail and the accuracy of the various methods is evaluated further in the context of gas mixtures. This chapter also contains all of the material necessary to completely specify the transport coefficients for a binary gas mixture for both the first- and second-order Chapman-Enskog approximations. In addition to the material in Chapters 1-12, a number of appendices have been included to facilitate the ease of the reader in finding and using such information as is necessary to effectively make use of the analytical methods described in this book. Appendices A-C contain descriptions of analytical methods needed to evaluate the necessary bracket integrals associated with the various transport problems in both the planar and curvilinear geometries. Of particular importance, the methods of calculation of bracket integrals containing the discontinuous distribution function have all been collected together in one monograph for the first time. Appendix D provides additional information about the variational principle for use in planar transport problems. Here, the reader can find greater detail regarding the general principles behind the construction of the necessary variational functionals needed for specific boundary value transport problems. Appendix E contains an extensive listing of definite integrals that are most frequently encountered in boundary value transport problems and Appendix F contains tabulated numerical values of the : -integrals that are encountered when using the Lennard-Jones (6-12) potential model. This set of tabulated : -integrals is the complete set that it is necessary to use when employing either the first-order or the second-order Chapman-Enskog approximations. The problems following each chapter may be useful to both students and instructors. While all of these problems have been solved, sometimes only an outline of the solution is given and the reader must supply the details.

xxii

Analytical Methods for Problems of Molecular Transport

The authors recommend that students of the subject seek to obtain their own solutions to these problems in order to gain familiarity with the techniques. In addition to basic experience, this will also help students of the subject to understand some of the subtleties of the analytical methods described. While some of the problems presented at the ends of the chapters only require the reader to utilize basic information that may be readily obtained in the necessary form directly from the book, other problems will require the reader to engage in some significant extension and generalization of the basic material presented in the book. This book is designed to serve a dual function. It is intended that it be capable of serving as a teaching instrument, either in a classroom environment or independently, for the study of basic analytical methods and mathematical techniques that may be used in the Kinetic Theory of Gases. It is primarily suitable for use in graduate level physics and engineering courses on the Kinetic Theory of Gases. This book should also prove to be useful as a reference for scientists and engineers working in the fields of Rarefied Gas Dynamics and Aerosol Mechanics. In addition, the material in this book may prove to be of interest to individuals working in the areas of Physical Chemistry, Chemical Engineering, or any other applied discipline in which gas-surface interactions can be expected to play a significant role. INI SKL RVT

Acknowledgments

The authors gratefully acknowledge the invaluable assistance that they have received from many sources including: Dr. Robert L. Buckley for extensive help in typing and proof reading. Dr. James L. Griffin for Mathematica® assistance in some calculations. Professor Mikhail N. Kogan for helpful discussions relating to arbitrary Knudsen number in Chapter 11. Professor Yuri I. Yalamov for helpful discussions relating to boundary conditions for the slip-flow regime in Chapters 8 and 10. Dr. David Gabis for discussions relating to the spinning sphere problem at the end of Chapter 10. Dr. Perapong Tekasakul for help with figures and equations. Mr. Earl L. Tipton for extensive help with Adobe Illustrator® and for proof reading the final version. Mr. Ryan Meyer, Mr. Zeb Smith, and Mr. Earl L. Tipton for contributions made in preparing the omega-integral program of Appendix F. The Council for the International Exchange of Scholars (CIES) for the Fulbright grant that supported INI during his first visit to Columbia Missouri, 1/15/92-7/4/92 and which sponsored two subsequent visits, 8/25/92-1/15/93 and 7/29/93-2/18/94. The U.S. National Aeronautics and Space Administration (NASA) for funding from grant NAG-3-1420 that was used to support INI during his second and third Fulbright visits to Columbia as well as during his fourth visit, 7/01/94-3/15/95. This manuscript was initially prepared on a Macintosh® IIci using System 6.0.5. The text was initially formulated using Microsoft Word® (ver.

xxiv

Analytical Methods for Problems of Molecular Transport

4.0). Equations were initially prepared using MathType® (ver. 2.03), plots using Cricket Graph® (ver. 1.3.2), and figures using MacDraft® (ver. 1.2b). The manuscript was subsequently converted to Microsoft Word® (ver. 2002) employing MathType® (ver. 5.2a) on a Dell Dimension® 2400 configured with Microsoft Windows® XP (ver. 2002 including Service Pack 2). Plots were converted to Microsoft Excel® (ver. 2002) and figures were converted to Adobe Illustrator® CS (ver. 11.0.0). Omega-integral calculations in Appendix F were performed using Compact Visual Fortran® 77 (ver. 6.4) and Mathematica® (ver. 5.2) on a number of different machines.

Chapter 1 THE GENERAL DESCRIPTION OF A RAREFIED GAS

1.

SOME INTRODUCTORY REMARKS.

Systems consisting of what are usually called rarefied gases will be considered in this book. Such systems contain large numbers of molecules; specifically, about 2.7 u 1019 cm–3 for a gas at standard temperature and pressure (STP) defined to be 0 °C (273.15 K) and 1.0 atm (760 torr). How may the behavior of such huge collections of particles be described? If one were to try to apply the methods of classical mechanics, one would have to construct and solve a system of equations of motion containing as many equations and initial conditions as there are numbers of interacting particles. Obviously, such a problem could not be solved today even with the help of the most advanced computers. On this basis, one might naively suspect that the greater the number of particles, the more difficult the problem. This is not strictly true, however, as it turns out that useful results can be obtained by applying statistical descriptions to such systems. The state of a gas can be analyzed by employing statistical laws which allow one to determine average values for the different macroscopic quantities that characterize the behavior of the gas. It has been proven in statistical mechanics [1-3] that the relative fluctuations of additive quantities (i.e. quantities whose values for the body as a whole are equal to the sum of the values for its separate parts) are proportional to N 1 2 , where N is the number of molecules of the gas. In accordance with this theorem, the additive quantities are really equal to their average values to an extremely high degree of accuracy and, therefore, deviations of the actual quantities

2

Analytical Methods for Problems of Molecular Transport

from the average values do not have a practical influence on the trustworthiness of the statistical description. The mean values mentioned above may be found by using a probability framework in which the main interest is in the statistical distribution function for the molecules. It is very important to note that the extreme accuracy of this type of probabilistic analysis, due to the relatively small deviations of the macroscopic quantities from their mean values, is much greater than the accuracy of actual experimental measurements and, hence, small deviations of the macroscopic quantities from their mean values are typically neglected.

2.

DENSITY AND MEAN MOTION.

First, some macroscopic quantities for a ‘simple’ gas composed entirely of identical molecules [4-6] will be determined. Let the mass of any molecule be m and let dr denote a small volume element surrounding the point, r . This volume element is assumed to be large enough to contain a great number of molecules while still possessing dimensions small compared to the scale of variation of the macroscopic quantities of interest. Let the mass contained in dr be averaged over a time interval, dt , which is long compared to the average time needed for a molecule to traverse dr yet short compared to the scale of the time variations in the macroscopic properties of the gas. Then, the average value of the mass contained by dr will be proportional only to its volume and will not depend upon its shape. This mass will be denoted by U dr , where U r ,t may be termed the mass density (or typically just ‘density’) of the gas at r ,t . The number density of the gas, n r , t , is identified by analogy. These densities are connected by the relationship:

U r , t mn r , t , where m is the mass of a single gas molecule. If the velocity of a molecule is denoted by v , then the mean velocity of a gas at r ,t may be denoted by u r ,t which is defined by the vector equation:

ndr u ¦ v

,

3

Chapter 1. The General Description of a Rarefied Gas

where the summation over v on the right-hand-side extends over all of the molecules in dr , and both ndr and ¦ v are averaged over the time interval, dt . The translational motion of an individual molecule in dr may be specified either by its ‘actual’ velocity, v , relative to some standard frame of reference, or by its velocity, V c , relative to axes moving with a velocity, uc , so that:

Vc

v uc .

If uc u , then the quantity, V molecule.

3.

v u , is called the peculiar velocity of the

THE DISTRIBUTION FUNCTION OF MOLECULAR VELOCITIES.

Within a statistical framework, the state of a rarefied gas is described by the distribution function, f v, r , t , so that the quantity, fdvdr , gives the probable number of molecules which, at time, t , are situated around r in the volume element, dr , and have velocities near v lying in the range, dv . It is also important to note that fdvdr is the number of molecules in dvdr averaged over the time interval, dt . The distribution function may be specified as the probability density of points in the phase space (i.e. the sixdimensional space of r and v ) of a molecule. The description of a gas by its distribution function is excessively detailed. For practical purposes this amount of detail is not always needed. For instance, the gas as a continuum may be characterized by various macroscopic quantities which may be experimentally measured. These quantities may be considered to be moments of the distribution function and these moment relationships will be discussed in detail later. It is clear that the number density may be obtained by integrating the distribution function throughout the entire velocity space. This integration may be performed over either the actual or peculiar velocities of the molecules since the distribution of velocity points is unaffected if the origin in the velocity space is shifted to the point, u . The same integration concept also applies with respect to all similar integrals over the distribution function (moments of the distribution function). Hence, the number density may be expressed in the form:

n r , t

³ f v, r , t dv ³ f V u, r , t dV

.

(1-1)

4

Analytical Methods for Problems of Molecular Transport

4.

MEAN VALUES OF FUNCTIONS OF MOLECULAR VELOCITIES.

Let I v, r ,t be any function of the molecular velocity, v , position, r , and time, t . This function will be called a molecular property. The mean value of I v, r ,t in the volume element, dr , is defined by the relation:

ndr 1 ¦ I ,

I

where ¦ I denotes the time average during dt of the sum of the values of I v, r , t for all the molecules in dr and can be expressed in terms of the distribution function, f v, r , t . For this purpose, consider the set of molecules in the volume element of the phase space, dvdr . Molecules of this set all have the same velocity and thus their contribution to the sum, ¦ I , is I fdvdr . The net contribution of all molecules in the volume element, dr , may then be expressed in the form:

¦I

dr ³ I fdv .

Hence I , the mean value of any molecular property, I v, r ,t , may be represented by a moment of the distribution function, i.e.:

I

n 1 ³ I fdv .

(1-2)

In particular, the mean velocity of a gas is given by:

u

n 1 ³ vfdv .

(1-3)

From Eq. (1-3), it then follows directly that the mean value of the peculiar molecular velocity is equal to zero, i.e.:

V

5.

v u

v u 0 .

TRANSPORT OF MOLECULAR PROPERTIES.

Consider the passage of molecules, as shown in Fig. 1-1, across a small element of surface, dS , moving in the gas with arbitrary velocity, uc . Let n be a unit vector positioned normal to the surface element and pointing in

5

Chapter 1. The General Description of a Rarefied Gas

Figure 1-1. Passage of molecules across an arbitrary surface element,

dS .

the direction from the negative side to the positive side. The velocity, V c , of a molecule relative to dS is equal to v uc , or V u uc . In the Kinetic Theory of Gases, a standard analysis is usually employed to calculate the fluxes of the various molecular properties. First, the contribution to the flux across the surface element, dS , is determined for each separate velocity group of molecules. Then, the net flux is calculated by integrating over all velocity groups. Consider the molecules of one velocity group in dr whose peculiar velocities lie in the range, dV . Let a molecule cross the element dS in a time dt which is so short that the possibility of molecular encounters may be ignored. In this case, the molecule must lie somewhere inside the region having dS as the base and having a length and direction determined by V c dt . Thus, if dr denotes the volume of this region, the number of molecules crossing dS during dt is fdVdr . From the above description, it follows that dr r V c cos T dSdt where T is the angle between V c and n and the sign, plus or minus ( r ), is chosen so as to make the expression for dr positive. Thus, the flux of I V for

6

Analytical Methods for Problems of Molecular Transport

this molecular group is expressed, both in magnitude and in sign, by the following relation:

I V Vnc f V dVdSdt .

(1-4a)

The net flux is then the integral over all of the velocity groups and is written as:

dSdt ³ I V Vnc f V dV .

(1-4b)

The rate of flow of property I per unit area is expressed in the form:

nVncI V .

(1-4c)

This represents the n -component of the corresponding vector quantity: n V cI V .

(1-5)

Substituting the definition of V c into Eq. (1-5), one can obtain: n V cI V n VI V n u uc I V .

(1-6)

The vector n VI V is typically termed the flux density vector for the molecular property, I V , while the second term on the right in Eq. (1-6) is connected with the number flow of molecules across dS ; each of which possesses the mean value of the property, I V .

6.

THE PRESSURE TENSOR.

Let I V be equal to some component of the molecular momentum, mv . At the boundary of some containment vessel, every molecule that rebounds from the surface imparts momentum to it. These impacts simulate a continuous force on the surface equal to the rate at which momentum is imparted to the surface. The force per unit area on the surface is called the pressure which is a vector that is not necessarily normal to the surface. Let an element of the surface, dS , move with a velocity, uc , and let the direction pointing into the gas from the surface be taken as negative. The pressure on this boundary surface element, Pn , is equal to the total rate of

7

Chapter 1. The General Description of a Rarefied Gas

Figure 1-2. Components of the pressure tensor. Here, the point, the center of the cube face nearest to the observer.

M , is physically located at

flow of the molecular momentum, mnv , per unit area. Using Eq. (1-4c), one obtains [4]:

Pn

(1-7)

UVncv .

If the surface is impenetrable to the gas molecules, then the mean velocity relative to the wall has no component normal to the wall. This may be expressed as: (1-8)

Vnc 0 . Using this condition, one can perform the following transformations:

Vncv Vnc V u

ª¬n V u uc º¼ V

n V V

,

after which the following expression for Pn may be obtained:

Pn

UVn V n U VV n P .

(1-9)

8

Analytical Methods for Problems of Molecular Transport

Here, P is a symmetrical tensor defined by:

P

U VV

V2 V V V V ½ x y x z ° x ° ° ° 2 U ®VyVx Vy VyVz ¾ . ° 2 ° °¯VzVx VzVy Vz °¿

(1-10)

It is important to note that this tensor depends only upon the distribution of the peculiar velocities. The same analysis may be used to define the pressure distribution at any point, M , within the gas. Let the surface element, dS , containing M share the mean motion of the gas so that uc u, V c V 0 and, therefore, Vnc 0 . Since the basic condition for deriving Eq. (1-9) is satisfied in this particular case, the pressure distribution across the surface element, dS , is defined by the same pressure tensor, P , given by Eq. (1-10). The physical sense of each pressure tensor component is illustrated in Fig. 1-2.

7.

THE HYDROSTATIC PRESSURE.

The sum of the normal pressures at the point, M , across the three planes parallel to the coordinate planes, is given by:

Pxx Pyy Pzz

UV 2 .

The mean of the normal pressures across these planes is called the hydrostatic pressure, p , or the pressure at M . Immediately from this definition, one can obtain:

p

1 3

Pik G ik ,

(1-11)

where G i k is the unit symmetrical tensor whose components are given by:

Gi k

1 ; i k , ® ¯0 ; i z k .

If Pik 0 when i z k and the diagonal elements are equal, the pressure, p , may be expressed in the form:

9

Chapter 1. The General Description of a Rarefied Gas

p

Pxx

Pyy

Pzz ,

(1-12a)

or:

Pik

pG ik .

(1-12b)

In this case, the pressure vector may be written as Pn i Pik nk pni and, therefore, the pressure on any surface element through the point, M , is normal to the surface and is independent of the orientation of the surface.

8.

THE AMOUNT OF HEAT.

The kinetic energy of translational motion of molecules in the volume element, dr , at time, t , is:

ndr 12 mv 2 , which may be expressed in the form:

dr 12 U u 2 V 2 .

(1-13)

Here, the first term is the kinetic energy associated with the motion of the volume element which has a net mass. The second term is the kinetic energy associated with the peculiar motions of the molecules in the volume element and which is one component of the thermal energy of the gas in the volume element (and is the only component of the thermal energy for monatomic gases). This energy is communicable between molecules at encounters. It should be noted that there are also other forms of communicable energy that occur in polyatomic gases. In this case, the total heat energy, E , of a molecule is the sum of its peculiar kinetic energy and any of these other forms of communicable energy that may be present.

9.

THE KINETIC TEMPERATURE.

In the Kinetic Theory of Gases, a temperature, T , is defined directly by the relationship:

10

Analytical Methods for Problems of Molecular Transport 1 2

mV 2

3 2

kT ,

(1-14)

where k 1.38066 u 1016 erg K–1 is Boltzmann’s constant. It is important to note that the temperature definition given by Eq. (1-14) is more general than that employed in Thermodynamics and Statistical Mechanics where only equilibrium states are considered. Eq. (1-14) can be used for non-steadystates of a gas as well.

10.

THE EQUATION OF STATE FOR A PERFECT GAS.

From the definitions of the hydrostatic pressure and the temperature given in Eqs. (1-11) and (1-14), one can obtain the following relationship between the pressure and the temperature:

p

nkT .

(1-15)

This equation is called the equation of state for a perfect gas.

11.

THE THERMAL FLUX VECTOR.

The thermal flux vector, Q , is an important quantity that expresses the rate of flow of heat energy across a unit surface element. This vector may be obtained from Eq. (1-6) by substituting the heat energy of a molecule (the energy which is communicable between molecules at encounters) in place of the generic molecular property, I V . After this substitution, one obtains:

Q

nEV .

(1-16)

If a molecule possesses only kinetic energy, then this heat flux vector is defined by:

Q

n 12 mV 2 V .

(1-17)

Chapter 1. The General Description of a Rarefied Gas

12.

11

SUMMARY.

Rarefied gases, since they are systems containing a huge number of molecules, can only be realistically described within a statistical framework. The probabilistic basis of this description is not readily apparent in many practical transport problems, however, because the accuracy of the results predicted by the Kinetic Theory of Gases is much greater than the realistically achievable accuracy in actual experimental measurements and because the relative deviations of the various macroscopic quantities from their mean values are extremely small. Within this framework, the state of a rarefied gas is described by the distribution function, f v, r , t , so that the quantity, fdvdr , gives the probable number of molecules at time, t , which are located in the spatial region at r in dr , and which have velocities of v in the velocity space interval, dv . The description of a gas by the distribution function is excessively detailed while the gas, as a continuum, may be characterized by certain macroscopic quantities which are relatively few in number and which can be experimentally measured. These macroscopic quantities, and any other molecular property fluxes of interest, may be calculated as moments of the distribution function. The most significant of these macroscopic values for a monatomic gas include the number density, the mean velocity, the temperature, the pressure, and the thermal flux, and are given by:

n

³ fdv

u

n 1 ³ vfdv ,

3 2

(1-19)

2

n 1 ³ 12 m v u fdv ,

(1-20)

m ³ vi ui vk uk fdv ,

(1-21)

kT

Pik

(1-18)

,

and:

Q

1 2

2

m ³ v u v u fdv .

(1-22)

12

Analytical Methods for Problems of Molecular Transport

Thus, a complete knowledge of the distribution function can be used to fully describe the behavior of a gas.

PROBLEMS 1.1. Prove that the relative fluctuation of an additive quantity is proportional to N 1 2 , where N is the number of molecules of a gas. Solution: Let L be an additive quantity for a gas that contains N molecules. The mean value of this quantity may be written as:

L

N

¦ Li

,

i 1

and, therefore, this value is proportional to N . The relative fluctuation of L is given by:

G

L1

¦ 'L i

i

2

.

The radicand may be expressed in the form:

§ · ¨ ¦ 'Li ¸ © i ¹

2 2

¦ 'Li ¦¦ 'Li 'Lk i

.

i k izk

The later term in this relation is equal to zero since separate molecules are statistically independent and, for one molecule, 'Li 0 . Hence, this sum containing N terms is proportional to N and one can obtain the following:

G ~ N 1 N

N 1 2 .

1.2. Derive the condition of applicability of the point-mass hypothesis for a gas molecule. Solution: A molecule in a gas may be considered as a point particle when the de Broglie wavelength of the molecule is much less than the mean distance between the gas molecules. This relationship may be expressed as:

13

Chapter 1. The General Description of a Rarefied Gas 13

2S ! §V · ¨ ¸ mv ©N¹

n 1 3 ,

where ! is Planck’s constant ( ! 1.054 u 1027 erg sec) and v is the mean heat velocity of a molecule. The following expression may be written:

2S ! 2mkT

n1 3 1 .

1.3. Determine the temperature for which the de Broglie wavelength of a molecule is the same order of magnitude as the mean distance between gas molecules. Consider the case of hydrogen having a number density of n 2.69 u 1019 cm–3. Solution:

2S 2 ! 2 n 2 3 ; T0 mk

T0

0.43 K.

1.4. Let the behavior of a mechanical system be described by exact equations of motion ( N is the number of molecules, V is the volume, and N !! 1 ). How might a statistical description be introduced for such a system? How might the mean values and the distribution function be specified? What accuracy does the statistical description have? Solution: Consider the state of a subsystem which is situated in a small volume element, dr , surrounding the point, r , during a time interval, dt . This interval is assumed to satisfy the following relations:

dr 1 3 v

dt

V1 3 , v

where v is the mean velocity of the particles. Let the time interval, dt , be divided into k smaller subintervals of width, dW , and the whole velocity space be divided into m smaller velocity ranges, 'v j 'v jx 'v jy 'v jz , so that ¦ mj 1 'v j is equal to the product of the maximum values of the velocity components. If the equations of motion are known, it is easy to calculate the number of molecules at time, t i dW , which have velocities lying in the range, 'v j . Let this quantity be denoted by nij n r , v j , t idW . The number density and the mean value of any molecular property may be defined, respectively, by:

14

Analytical Methods for Problems of Molecular Transport

n

1 dr

m

k

¦¦

nij

j 1i 1

k

, and nI

m

k

¦¦ j 1i 1

I v j nij k

.

The distribution function may be written as:

f

1 dr'v j

k

nij

¦k

,

i 1

The relative fluctuations of additive quantities for such a system are given by G ~ N 1 2 .

REFERENCES 1. Landau, L.D. and Lifshitz, E.M., Statistical Physics (Pergamon, London; Addison Wesley, Reading, Mass, 1958). 2. Levich, V.G., Theoretical Physics: An Advanced Text (North-Holland, Amsterdam, 1970). 3. Huang, K., Statistical Mechanics (John Wiley and Sons, New York, 1963). 4. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 5. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 6. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, New York, 1969).

Chapter 2 THE BOLTZMANN EQUATION

1.

DERIVATION OF THE BOLTZMANN EQUATION.

It has been established previously that knowledge of the distribution function gives all the necessary information for a gas. To obtain the basic equation for the distribution function, consider a balance of the number of molecules that are located in the element, dvdr , of the six-dimensional phase space for the time interval, dt . Consider a gas in which each molecule is subject to an external force, mF , that is a function of r and t , but not a function of v . For the time interval, dt , the position of this phase element will change from v, r to v Fdt , r vdt . It is important to note that the magnitude of this volume element is unaltered in accordance with the Liouville theorem of classical mechanics [1]. There are f v, r , t dvdr molecules which, at the time, t , are situated in the volume element, dr , and have velocities in the velocity range, dv . These molecules are assumed to belong to the first molecular set. After the time interval, dt , there is another set in which the number of molecules is given by:

f v Fdt , r vdt , t dt dvdr . The number of molecules in the second set will differ from that in the first set owing to molecular encounters. The net gain of molecules to the second set is proportional to dvdrdt and will be denoted by G f G t dvdrdt . Consequently, one can obtain:

16

Analytical Methods for Problems of Molecular Transport

ª¬ f v Fdt , r vdt , t dt f v, r , t º¼ dvdr

G f

G t dvdrdt .

(2-1)

Now, if dt tends to zero, it is easy to obtain the Boltzmann equation which may be written in the form [2]:

Df

Gf . Gt

(2-2a)

Here, the following notation is introduced:

Df

wf wf wf vi Fi , wt wxi wvi

(2-2b)

where xi are components of the position vector, r . The quantity G f G t dvdr is the rate of change of the number of molecules in the phase element, dvdr , owing to molecular encounters.

2.

THE MOMENT EQUATIONS.

Let I I v, r ,t be any molecular property. After multiplying the Boltzmann equation by I dv and integrating throughout the velocity space, one can obtain the following integral relation:

³ I Dfdv

n'I ,

(2-3)

where:

n'I

Gf

³I G t

dv .

In Eq. (2-3), all the integrals are assumed to be convergent and, moreover, products such as I f tend to zero if v o f . This equation is the general form of a moment equation in which n'I is a moment of the Boltzmann collision integral. The various terms on the left-hand-side of Eq. (2-3) may be transformed by means of the following relationships:

17

Chapter 2. The Boltzmann Equation

Table 2-1. Transformation of the various derivatives to new variables.

v, r , t

V, r , t

wI wv

wI wV

wI wt

wI wI wui wt wVi wt

vi

wf

wI wxi

vi

w wI I fdv ³ f dv ³ wt wt

³ I wt dv wf

³ I vi wxi dv

w wI nI n , wt wt

w wI I vi fdv ³ fvi dv ³ wxi wxi

wf

³ I wvi dv ³ ³ I f

vi f vi f

wI wI wuk vi wxi wVk wxi

dv j dvk ³ f

w wI , nI vi nvi wxi wxi

wI dv wvi

n

wI . wvi

In the last of the above relationships, the double integral vanishes because I f |vvii ff 0 in accordance with the notation made above. Substituting these expressions into Eq. (2-3), one can obtain [2]:

w nI wt

w wI wI ° wI °½ nI vi n ® vi Fi 'I ¾ . wxi wxi wvi °¯ wt °¿

(2-4)

This equation is usually called the moment equation or the transport equation of the molecular property, I .

18

3.

Analytical Methods for Problems of Molecular Transport

ANOTHER FORM OF THE MOMENT EQUATIONS.

The basic moment relationship given in Eq. (2-4) is suitable for the laboratory frame of reference relative to which the gas has a mean velocity, u r , t . To transform this equation into the variables, V , r , t , where V v u r ,t , one must consider that I V, r ,t depends upon r and t , both explicitly and implicitly through its dependence on V . Table 2-1 expresses the transformation of the various derivatives in Eq. (2-4). Substituting the derivatives from Table 2-1 into Eq. (2-4), one can obtain [2]:

wu D w nI nI i nI Vi Dt wxi wxi

° DI Dui · wI wI § wI wuk ½° n ® Vi ¨ Fi Vi ¾ n' I , ¸ wxi © Dt ¹ wVi wVk wxi °¿ ¯° Dt

(2-5)

where D Dt w wt ui w wxi is the ‘mobile operator’ or time derivative with respect to the frame of reference moving with velocity, u . By analogy, the Boltzmann equation for these variables may be expressed in the form:

Dui · wf wu wf § wf Df Vi k Vi Fi wxi ¨© Dt Dt ¸¹ wVi wVk wxi

4.

Gf . Gt

(2-6)

THE EQUATIONS FOR A CONTINUUM MEDIUM.

Some important moment equations can be obtained from Eq. (2-5) by using the molecular properties, I , which are conserved during encounters. It is very important to note that these equations may be derived without knowledge of the collision operator because 'I 0 for such molecular properties. The invariant quantities of molecular encounters in a monatomic gas are the number of molecules, momentum, and kinetic energy because the mutual potential energy of two molecules equals zero both before and after an encounter. Let the following notations be introduced with respect to these invariant quantities:

\ 1 1 , \ 2

3 mV , \

1 2

mV 2 .

Chapter 2. The Boltzmann Equation

19

The statement n'\ 0 (for i 1, 2,3 , respectively) expresses the principles of conservation of number (of molecules), momentum, and energy i during encounters. Substituting I \ into Eq. (2-5), one can obtain: i

wu Dn n i Dt wxi

0 ,

(2-7)

wPik Dui · § U ¨ Fi 0 , wxk Dt ¸¹ ©

DT Dt

2 3nk

§ wu wQ ¨ Pij i i ¨ wx j wxi ©

(2-8)

· ¸ . ¸ ¹

(2-9)

These equations are usually called the continuum equations. This system is not closed as it contains 13 unknown functions that cannot be specified from the five equations. This lack of closure is a general characteristic of all moment systems. Since the distribution function is unknown, the moment transport equations may be considered to be the most general relations between the time and space derivatives of the macroscopic quantities because this system cannot be solved without definite suppositions or hypotheses about the distribution function which forms the basis of any moment method. The distribution function being used to close a moment system must satisfy certain general conditions. It must describe the characteristic features of the gas flow and must contain as many unknown parameters as the number of moment equations being employed. Then, all moments will depend on these parameters and the moment system becomes closed.

5.

MOLECULAR ENCOUNTERS.

Knowledge of the nature of the molecular interaction during encounters is necessary to derive the expression for the Boltzmann collision integral. At present, there is not a theory that describes the details of molecular encounters exactly. Therefore, a definite model of the molecular interaction has to be introduced to complete this description. One of the simplest molecular models is that of a smooth, rigid, and perfectly elastic sphere; commonly referred to as the rigid-sphere model for

20

Analytical Methods for Problems of Molecular Transport

brevity. The force between two molecules at a collision appears only if these molecules are in close contact with each other. This model is approximate since the real force varies continuously with the distance between the molecules. This fact might be better interpreted if a molecule was considered as a point-center of force depending on the nature of the interacting molecules and on their separation distance. Molecules represented as either rigid spheres or as point-centers of force may be termed smooth molecules since their internal energy is not altered during encounters. In this book, only smooth molecules are considered and the analysis of encounters follows that described in [2]. Let the masses of the two molecules undergoing an encounter be m1 and m2 . The interaction force is directed along the line joining their centers and it depends only upon the distance between molecules. The velocities before and after the encounter are denoted by v1 , v 2 and v1c , vc2 , respectively. Some general relations between these velocities may be found without consideration of the details of the molecular encounter. To start, one introduces the following notations:

m0

m1 m2 ,

M1

m1 m0 ,

M2

m2 m0 .

Owing to the momentum conservation during encounters, the center of mass of the two molecules moves with a constant velocity, G , that is given by:

m0G

m1 v1 m2 v 2

m1 v1c m2 vc2 .

(2-10)

Let g 21 and gc21 denote the relative velocities of the two molecules before and after an encounter:

g 21

v 2 v1 ,

gc21

vc2 v1c .

(2-11)

which have the following magnitudes:

g 21

g ,

gc21

gc .

From Eqs. (2-10) and (2-11), one can obtain:

v1

G M 2 g 21 ,

v1c

G M 2 gc21 ,

(2-12a)

v2

G M 1g 21 ,

vc2

G M 1gc21 .

(2-12b)

21

Chapter 2. The Boltzmann Equation

Figure 2-1. Geometry of a generic encounter.

The mutual potential energy of the two molecules equals zero both before and after an encounter and, hence, the energy conservation equation may be written in the following very simple form: 1 2

m v

2 1 1

m2 v22

m vc 1 2

1 1

2

m2 v2c2 .

Using this relation and Eqs. (2-12a) and (2-12b), one is able to determine that the relative velocity during an encounter is changed only in direction and not in magnitude, i.e.:

g

gc .

Consideration of momentum and energy conservation laws alone does not suffice to determine the direction of gc21 which depends not only upon the initial velocities but also upon two geometric variables characterizing the spatial orientation of the two interacting molecules under consideration.

22

Analytical Methods for Problems of Molecular Transport

Figure 2-2. Parameters of a generic encounter.

Consider the motion of the center B of the second molecule relative to the center A of the first. This relative motion can formally be considered as the motion of an imaginary particle with a reduced mass, P m1m2 m0 , in a central force field and, therefore, it will be confined to the plane through A [1] (the curve LMN in Fig. 2-1 is a trajectory of the second molecule). The asymptotes of this trajectory, PO and OQ , are in the directions of g 21 and gc21 . Let PcA be a line parallel to PO so that the direction of PcA is fixed by g 21 . The orientation of the plane LMN is specified by the angle H between this plane and another plane containing both PcA and the arbitrary fixed axis, Ax , as shown in Fig. 2-2. Let the plane RAO1 ( O1 is the point of intersection of the line PO with the plane RAO1 ) be perpendicular to PcA and AR be the line of intersection of the plane RAO1 with the plane through A which contains both PcA and the axis, Ax . Let the polar coordinates of O1 in this plane be b and H . These geometric variables complete the specification of the encounter.

Chapter 2. The Boltzmann Equation

23

Figure 2-3. The spherical coordinates of the relative velocity.

The angle, T , through which g 21 is deflected, depends, in general, on the magnitude of g and on the parameter, b . To derive the geometrical connection between g 21 and gc21 , this angle is assumed to be known. Let the spherical coordinates of the vector g 21 g be g ,D , E . Then, the Cartesian and spherical coordinates shown in Fig. 2-3 are related by the following:

gx

g cos D ,

(2-13a)

gy

g sin D cos E ,

(2-13b)

gz

g sin D sin E .

(2-13c)

24

Analytical Methods for Problems of Molecular Transport

Let H be the angle between the two planes, POQ and POR , as shown in Fig. 2-3. The vectors, g and gc , and g and i , lie in the planes, POQ and POR , respectively, hence one can obtain [3]:

cos H

gc u g i u g

gc u g i u g

g cx g x cos T g sin D sin T

(2-14)

.

To simplify this vector expression, the following relation is employed:

a u b c u b a c b 2 a b c b

.

Let Z be the angle between the two planes, POR and POy , shown in Fig. 2-3. Then, one can obtain:

cos Z

i u g ju g i u g ju g

cos D cos E 1 sin 2 D cos 2 E

,

(2-15)

and:

cos Z H

j u g g u gc j u g g u gc

g c g sin D cos E cos T y

sin T 1 sin 2 D cos 2 E

(2-16)

.

Using Eqs. (2-14)-(2-16) and analogous relations for g cz , one can express the components of the relative velocity by: g cx

g ª¬ cos D cos T sin D sin T cos H º¼ ,

g cy

g ª¬sin D cos E cos T cos D cos E sin T cos H sin E sin T sin H º¼ ,

(2-17a)

(2-17b)

Chapter 2. The Boltzmann Equation

g cz

25

g ª¬sin D sin E cos T cos D sin E sin T cos H

(2-17c)

cos E sin T sin H º¼ . If the angle, T , is known, these relations together with Eqs. (2-12a) and (212b) permit one to determine the velocity components after an encounter. Some other relations for the molecular velocities before and after an encounter may be derived from Eqs. (12-12a) and (12-12b). The trajectory, LMN , of the second molecule relative to the first molecule is symmetrical about the line, OA (usually known as the apse-line), shown in Fig. 2-1. The vectors, g 21 and gc21 , differ by twice the component of g 21 in the direction of k . This difference is expressed as:

g 21 gc21

2 g 21 k k .

Using this relation and Eqs. (2-12a) and (2-12b), one can obtain:

v1c v1

2 M 2 g 21 k k

vc2 v 2

2 M1 g 21 k k

2 M 2 gc21 k k ,

(2-18a)

2 M 1 gc21 k k .

(2-18b)

These formulas form another set of relations between molecular velocities during an encounter.

6.

THE RELATIVE MOTION OF TWO MOLECULES.

The relative motion of two molecules is equivalent to the motion of one particle with the reduced mass, P m1m2 m0 , moving in a central force field, F r wU wr r r , where U r is the mutual potential energy of the two molecules [1]. This problem can be solved in a general form by employing the energy and angular momentum conservation laws which, in polar coordinates r ,\ with the origin at the point, A , as shown in Fig. 21, may be written as [1,4]:

E

1 2

P r 2 r 2\ 2 U r ,

(2-19)

26

Analytical Methods for Problems of Molecular Transport

M

P r 2\ ,

(2-20)

where E is the energy before an encounter and M is the constant angular momentum component which is normal to the motion plane. Using Eq. (219), one can obtain:

M2 ª¬ E U r º¼ 2 2 . P P r 2

r

(2-21)

The polar angle, \ , can be found from Eq. (2-20) which may be expressed in the form:

M dt . P r 2

d\

From this relation and Eq. (2-21), after integrating with respect to r , one can obtain the following expression for \ as a function of r :

\

M dr r 2

³

M2 2P ª¬ E U r º¼ 2 r

After substituting E

b³ r0

P g 2 and M

dr r 2

f

\0

1 2

const .

b 2 2U r 1 2 r Pg2

(2-22)

P bg , Eq. (2-22) becomes:

.

(2-23)

Here, the lower limit of integration, r0 , is defined as:

r02 b 2

2r02U r0

Pg2

.

and the scattering angle, T , can be found from:

T

S 2\ 0 .

(2-24)

27

Chapter 2. The Boltzmann Equation

PROBLEMS 2.1. Prove that, if molecules move in accordance with Newton’s equations of motion, the phase volume element is not altered for the time interval, dt . Solution: For the time interval, dt , the position of the phase element, d * dvdr , is changed from v, r to vc v Fdt , r c r vdt . Then:

d* c

w vc, r c d* w v, r

w vc, r c w v, r c d* w v, r c w v, r

w v , v , v

w x c, y c, zc w x , y , z

w vcx , vcy , vzc x

y

z

r c const

d*

d* .

v const

The external force, F , is assumed to be independent of v .

2.2. Prove that the relative motion of two interacting molecules with mutual potential energy depending only on the distance between the molecules may be considered as the motion of a single particle in a central force field. Solution: First, introduce the vector: r r2 r1 e r r . The equations of motion of these molecules may be written in terms of this vector as m1 r1 f r e r , and m2 r2 f r e r , where f r wU r wr and U r is the mutual potential energy. From these equations, it is easy to find the r f r e r . This equation is the equation of motion of a relationship, P single particle of mass, P m1m2 m0 , in a central force field. 2.3. Prove that the relative motion of two interacting molecules is planar in nature. Solution: The differential equation of the relative motion of two molecules is given by: P g f r e r , where P m1m2 m0 , e r r r , r r2 r1 , g v 2 v1 , f r wU r wr and U r is the mutual potential energy. The vector product of this equation and the position vector, r , may be expressed as P r u g 0 or dM dt 0 , where M r u P g . This shows that M const and, consequently, that the position vector lies in the plane which is perpendicular to the constant vector, M . 2.4. Determine the distribution function for a gas in which the magnitude of the velocity vector is the same for all of the molecules, in which all directions of velocity are equally probable, and where the number density of the gas is n . Solution: The number of molecules situated in a phase element, dvdr , may be expressed as: fdvdr n\ v G v v0 dvdr , where \ v is a

28

Analytical Methods for Problems of Molecular Transport

function that depends only on the magnitude of the molecular velocity and G v v0 is the Dirac delta function. The function, \ v , may be found from the condition:

n ³\ v G v v0 v 2 sin T dvdT dI

³ fdv

n .

Then, it is easy to obtain the following form for the distribution function:

f

n G v v0 . 4S v 2

2.5. Using the conditions given in Problem 2.4, determine the distribution function for a two-dimensional gas (in which all molecular velocity vectors are parallel to the same plane) with a number density, n . Answer:

f

n 2S v

G v v0 .

REFERENCES 1. Landau, L.D. and Lifshitz, E.M., Mechanics (Pergamon, Oxford, 1960). 2. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 3. Jeans, J., Dynamical Theory of Gases (Cambridge University Press, Cambridge, U.K., 1990). 4. Bird, G.A., Molecular Gas Dynamics (Clarendon Press, Oxford, 1976).

Chapter 3 THE COLLISION OPERATOR

1.

THE DIFFERENTIAL AND TOTAL SCATTERING CROSS SECTIONS.

One of the most important concepts relating to the statistics of molecular collisions is that of the scattering cross section which is the basic quantity characteristic of molecular encounters. Consider a uniform flux of molecules that have a constant velocity, g , towards a spherically symmetric force center, O . The parameters of importance to the scattering events are shown in Fig. 3-1. Let N g be the g -component of the uniform molecular number flux and d : sin T dT d H be the solid angle in which molecules are deflected after encounters. The differential scattering cross section, D T , g , is defined by [1]:

D T , g d :

dN , Ng

(3-1)

where dN is the number of molecules deflected per unit time by the central force field in d : . Since the incident molecules scattered into d : must primarily pass through that portion of the annular differential element represented by dS bdbd H , dN may be expressed in the form:

dN

N g bdbd H .

From Eqs. (3-1) and (3-2), one can obtain:

(3-2)

30

Analytical Methods for Problems of Molecular Transport

D T , g

b db . sin T dT

(3-3)

The absolute value of db dT must be included in this relation since the signs of db and dT are usually opposite. The integral or total scattering cross section is then defined as:

D tot g 2S ³ D T , g sin T dT .

(3-4)

The differential scattering cross section has a very simple physical sense. The quantity, D T , g d : bdbd H , is that part of the annular differential area through which the molecular trajectory at infinity must pass in order for a molecule to be scattered into the solid angle, d : , after an encounter. The total scattering cross section, D tot g , is the area of the circular region through which the incident molecular trajectory must pass for any scattering to occur. The range of scattering that would contribute to this total cross section would encompass all of the available scattering solid angle 4S and any incident trajectory that were to lie outside of this circular region would be deemed to have not been scattered at all. Now, for example, consider molecules which are rigid spheres with diameters, V 1 and V 2 . The relation between the collision parameters shown in Fig. 3-2 may be expressed in the form:

b V 12 sin \ ,

(3-5)

where V 12 12 V 1 V 2 and \ 12 S T . Since the relative motion of two molecules is equivalent to motion in a central force field, the differential scattering cross section specified by Eq. (3-3) is given by:

D T , g

1 4

V 122 .

(3-6)

The total scattering cross section may be written as:

D tot g SV 122 .

(3-7)

It is very important to note that this molecular interaction model is unique in that D T , g and D tot g are independent of the relative velocity of the two molecules, g . This significant property of this model allows one to facilitate many calculations. For other models of intermolecular

Chapter 3. The Collision Operator

31

Figure 3-1. The scattering solid angle.

interaction that involve forces acting across infinite distances, the integral given by Eq. (3-4) is divergent since molecules are scattered even if the impact parameter, b , tends to infinity. Consequently, in this case, definite assumptions must be made regarding a cutoff of this parameter for which, when b ! bmax , no encounter occurs.

2.

THE STATISTICS OF MOLECULAR ENCOUNTERS.

Consider encounters between two different molecules with masses, m1 and m2 , situated in the volume element, dr . To derive the collision operator, one should calculate the rate of change of the number of m1 molecules having velocities in the range, dv1 , due to collisions with all molecules of the second kind. First, encounters between the two molecular sets, dv1 and dv 2 , in the volume element, dr , must be analyzed. Then, this analysis must be generalized for all molecules of the second kind. Encounters in which more than two molecules take part are assumed to be negligible in number compared with binary encounters and, moreover, molecules of the two velocity sets are assumed to be distributed at random without any correlation between velocity and position in the neighborhood of the point, r . These two assumptions are fundamental to the classical

32

Analytical Methods for Problems of Molecular Transport

Figure 3-2. Encounters of rigid-sphere molecules.

kinetic theory. It is necessary to note that the physical sense of the second assumption cannot be understood within the framework of the one-particle distribution function and, therefore, it may be considered to be a hypothesis the applicability of which may only be confirmed by using a more general theory [2-4] or by comparison with experiment. Examine the scattering of molecules of the second set by a molecule of the first in dr . The molecular number flux in the direction of the relative velocity, may be expressed in the form:

gf 2 v 2 , r , t dv 2 .

(3-8)

Then, in accordance with Eq. (3-1), the number of molecules of the second set deflected into the solid angle, d : , during the time interval, dt , is given by:

gD12 T , g f 2 v 2 , r , t d : dv 2 dt ,

(3-9)

where the time interval, dt , is assumed to be short compared to the scale of the time variation of the macroscopic properties of the gas, but large compared to the duration of an encounter. The appropriate number of v 2 -

Chapter 3. The Collision Operator

33

molecules reflected by all v1 -molecules in dr and dv1 may be written as [5]:

gD12 T , g f1 f 2 d : dv1dv 2 drdt .

(3-10)

In Eq. (3-10) it is assumed that the number of v1 , v 2 -collisions is proportional to f1 f 2 such that any potential correlation between velocity and position of molecules in dr has been ignored. Moreover, this expression holds true only if binary encounters are exclusively considered. The total number of encounters of the first set molecules with all molecules of the second kind in dr during dt may be found by integration over all values of : and v 2 . Then, one can obtain:

N12 dv1drdt

ª ³ ³ f1 f 2 gD12 T , g d : dv 2 º dv1drdt . ¬ ¼

(3-11)

Such encounters change the velocity of the first set molecules and thus the molecules are lost to the first set. As a result, N12 may be interpreted as the rate of loss of molecules from the first set per unit volume and velocity interval caused by collisions with molecules of the second kind. In the same way, one can obtain the number of m1 -molecules that enter the velocity set, dv1 . For this purpose, the m1 -molecules whose velocities after encounters lie in the range, dv1 , must be considered. In such inverse collisions, the molecular velocities are changed from v1c , vc2 to v1 , v 2 . Using the above analysis, one can obtain the total gain of m1 -molecules to the first velocity set, dv1 , for the time interval, dt , in the volume element, dr , owing to inverse encounters. This is given by: N12 dv1drdt

ª ³ ³ f1cf 2cg cD12 T , g c d : dvc2 º dv1c drdt , ¬ ¼

(3-12)

where f1c and f 2c stand for f1 v1c , r , t and , f 2 vc2 , r , t respectively. Now, one transforms the integrand in Eq. (3-12). First, since g g c , one can obtain:

g cD12 T , g c

gD12 T , g .

(3-13)

Next, one considers two subsystems of molecules the velocities of which, before an encounter, lie in the ranges, dv1c and dvc2 . If the collision parameters of these molecules lie in the ranges, db and d H , then their velocities after encounters will lie in the ranges, dv1 and dv 2 . In

34

Analytical Methods for Problems of Molecular Transport

accordance with Liouville’s theorem [1], the phase elements of these subsystems are connected by the relationship:

dv1c dr1cdvc2 dr2c

(3-14)

dv1dr1dv 2 dr2 .

Since the collisions between the two molecular sets occur in the same volume element, dr , Eq. (3-14) may be written as:

dv1c dvc2

(3-15)

dv1dv 2 .

Taking into account Eqs. (3-13) and (3-15), one can express Eq. (3-12) in the form: N12 dv1drdt

ª ³ ³ f1cf 2cgD12 T , g d : dv 2 º dv1drdt , ¬ ¼

(3-16)

where N12 is equal to the rate of gain of the first set molecules, per unit volume and velocity interval, due to encounters. Now, it is straightforward to obtain the collision operator for encounters between molecules of the first set and of the second kind:

§ G f1 · ¨ Gt ¸ © ¹2

N12 N12

³ ³ f1cf 2c f1 f 2 gD12 T , g d : dv 2

.

(3-17)

This form of the collision operator is usually called the Boltzmann collision integral. For a simple gas m1 m2 m, v1 o v, v 2 o v1 , Eq. (3-17) may be written as:

Gf Gt

³ ³ f cf1c ff1 gD T , g d : dv1

,

(3-18)

where G f G t denotes the rate of variation of the distribution function due to encounters at a fixed point, r . The rate of variation of any molecular property, I v, r , t , per unit volume, due to encounters, is defined by:

n'I

³ ³ ³ I f cf1c ff1 gD T , g d : dv1dv

.

(3-19)

This very important quantity may be called the moment of the collision integral associated with the property, I .

35

Chapter 3. The Collision Operator

3.

THE TRANSFORMATION OF SOME INTEGRALS.

Let the variables of integration in Eq. (3-19) be changed from v, v1 to vc, v1c . Then, this integral may be written as:

n'I c

³ ³ ³ I c ff1 f cf1c g cD T , g c d : dvcdv1c

.

(3-20)

Since the integration over all values of vc, v1c is equivalent to integration over all encounters, one can conclude that n'I c n'I . Taking into account Eqs. (3-15), (3-19) and (3-20), one can obtain:

n'I

1 2

³ ³ ³ I I c f fc1c ff1 gD T , g d : dvdv1

.

(3-21)

Since the variables, v1 and v , both refer to the same molecules, an interchange of them does not alter the value of this integral. After making this interchange of variables, Eq. (3-21) may be expressed in the form:

n'I

1 2

³ ³ ³ I1 I1c f cf1c ff1 gD T , g d : dvdv1

.

(3-22)

Then, combining Eqs. (3-21) and (3-22), one can obtain the following integral relation:

n'I

1 4

³ ³ ³ I I1 I c I1c f cf1c ff1 gD T , g d : dvdv1

.

(3-23)

PROBLEMS 3.1. Evaluate the time duration of an encounter between two molecules. Solution: Let r0 be the radius of action of the molecular forces and let V be the mean velocity of a molecule. The time duration of an encounter, W coll , is then given by W coll ~ r0 V . For rigid-sphere molecules, this may be written as W coll ~ V V where V is a molecular diameter. 3.2. Derive the condition of applicability of the Boltzmann equation if n is the gas number density. Solution: The mean distance between molecules is proportional to n 1 3 . Let r0 be the radius of action of the molecular forces. The binary interaction

36

Analytical Methods for Problems of Molecular Transport

hypothesis holds true if the molecular force action sphere contains not more than one molecule. This condition may be expressed: r03 n | 1 . 3.3. Prove by direct calculation that dv1c dvc2 dv1dv 2 during molecular encounters. Solution: First, derive the expressions; dv1dv 2 dGdg 21 and dv1c dvc2 dGdgc21 in which dg 21 g 2 dgd : . The desired proof follows directly if one takes into account the relationship g g c . 3.4. For a rigid-sphere gas in which all of the molecules have velocities of the same magnitude, determine the molecular mean free path assuming that all velocity directions are equally probable (number density, n , molecular diameter, V ). Solution: The total number of collisions occurring between molecules per unit volume and time is given by:

³ ³ ³ ff1 gD T , g d : dv1dv

N11

,

(P-1)

12

where g v 2 v12 2 vv1 cos F and F is the angle between v and v1 . The distribution function (Problem 2.4) may be expressed in the form:

f

n G c 1 , 4S v03c 2

where c obtain:

N11

v v0 and c

S 21 2 V 2 n 2 v0

4S

2

(P-2)

c . Substituting Eq. (P-2) into Eq. (P-1), one can

12

³ ³ 1 cos F

d Z1d Z ,

(P-3)

where d Z sin F d F d E and d Z1 sin F1 d F1d E1 . After integration over all directions of c and c1 , Eq. (P-3) becomes N11 43 SV 2 n 2 v0 . The mean free path is then given by:

O v0

n N11

4 3

SV 2 n

1

.

37

Chapter 3. The Collision Operator

3.5. For a two-dimensional rigid-sphere gas (number density, n , molecular diameter, V ), determine the molecular mean free path using the conditions in Problem 3.4. Solution: All molecular velocity vectors for this gas lie in the same plane. Therefore, it is convenient to use polar coordinates in the velocity space. The distribution function, in dimensionless variables, is given by (Problem 2.5):

f

n G c 1 . 2S v02 c

The total number of molecular collisions per unit volume and time is then:

S 21 2 V 2 n 2 v0

N11

2S

2

2S

12

³ 1 cos I 0

2S

dI

³ dI1

4V 2 n 2 v0 ,

0

where I is the angle between v and v1 , and I1 is the polar coordinate of v1 . The mean free path may then be expressed as:

O v0

n N11

4V n 2

1

.

REFERENCES 1. Landau, L.D. and Lifshitz, E.M., Mechanics (Pergamon, Oxford, 1960). 2. Bogoliubov, N.N., Problems of a Dynamical Theory in Statistical Physics (State Technical Press, Moscow, 1946). English translation by Gora, E.K. in Studies in Statistical Mechanics, Vol. 1, Eds. de Boer, J. and Uhlenbeck, G.E. (Wiley, New York, 1964). 3. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 4. Lifshitz, E.M. and Pitaevskii, L.P., Physical Kinetics (Pergamon, Oxford, 1960). 5. Anselm, A.I., Principles of Statistical Physics and Thermodynamics (in Russian) (Nauka, Moscow, 1973).

Chapter 4 THE UNIFORM STEADY-STATE OF A GAS

1.

THE BOLTZMANN H-THEOREM.

Consider a simple gas whose molecules possess only energy of translation, and are subject to no external forces. Let the state of the gas be uniform so that the distribution function, f , is independent of r . For this case, the Boltzmann equation may be expressed in the following simple form: wf wt

³ ³ f cf1c ff1 gD T , g d : dv1

.

(4-1)

Let the following H-function be introduced:

H

³ f ln f dv

.

(4-2)

The time derivative of the H-function may be written as:

wH wt

wf

³ 1 ln f wt dv ³ ³ ³ 1 ln f f cf1c ff1 gD T , g d : dv1dv

Now, taking into account Eq. (3.23), one can obtain:

(4-3)

.

40

Analytical Methods for Problems of Molecular Transport

wH wt

1 4

³³³

ln ff1 f cf1c f cf1c ff1 gD T , g d : dv1dv .

(4-4)

In this relation, ln ff1 f cf1c is always opposite in sign to f cf1c ff1 and, therefore, the integral is either negative or zero. As a result of this: wH d0 . wt

(4-5)

The latter differential relation indicates that the H-function can never increase. This fundamental inference is known as Boltzmann’s H-theorem and gives the time direction of processes for the uniform state of a gas [1-7].

2.

THE MAXWELLIAN VELOCITY DISTRIBUTION.

A very important question arises regarding whether there exists a limiting state for a gas after it has been disturbed from an initial uniform state. To show that there is, in fact, a limiting state, it is necessary to prove that the Hfunction cannot decrease indefinitely (i.e., that the H-function is bounded below). It is clear that H o f only if the integral given by Eq. (4-2) is divergent. If v o f , then the distribution function tends to zero and ln f o f . Since the integral, ³ 12 mv 2 fdv , expressing the energy of translation of the molecules, must be convergent, the H-function would be divergent only if ln f tended to infinity more rapidly than c 2 , where 12 c m 2kT v is the dimensionless velocity. In this case, however, f must tend to zero more rapidly than exp c 2 and, if this occurs in the integral given by Eq. (4-2), then the integral is convergent and the Hfunction is bounded below. Since it is bounded below, the H-function cannot decrease indefinitely but must tend to some limit, corresponding to a state of the gas, in which wH wt 0 . From Eq. (4-4), one may then conclude that the distribution function for the gas in this state may be found from the relation, f cf1c ff1 . This equation is equivalent to:

ln f c ln f1c ln f ln f1 .

(4-6)

If the distribution function satisfies Eq. (4-6), then, from Eq.(4-1), one can obtain that wf wt 0 also, so that such a state of the gas is steady as well as uniform.

41

Chapter 4. The Uniform Steady-State of a Gas

Now, consider the form of this distribution function. Eq. (4-6) shows that ln f is a summational invariant of encounters and, therefore, must be a linear combination of the three summational invariants. It may be expressed in the form [1,2]: 1 2 3 ln f D a mv D 12 mv 2 ,

(4-7)

or:

f

D 0 exp D 3 12 m V c

2

a V c v 3 . D 2

;

(4-8)

The constants, D , a , and D , can be expressed in terms of the number density, n , the mean velocity, u , and the temperature, T [1,2]. Using definitions of these quantities given by Eqs. (1-18)-(1-20), one can obtain: 0

2

0 § 2S · n D ¨ 3 ¸ © mD ¹

3

32

2 a

, u

D

3

3 , D

1 . kT

Then, the uniform steady-state of a gas is described by:

f

§ m · n¨ ¸ © 2S kT ¹

32

§ m v u 2 exp ¨ ¨ 2kT ©

· ¸ . ¸ ¹

(4-9)

This function is usually called the Maxwellian distribution function. Some mean values for the Maxwellian state of a gas may be calculated quite easily as they involve only simple integrations. For example, the mean value of the magnitude of the peculiar molecular velocity is given by: 12f

V

4 § 2kT · ¨ ¸ S1 2 © m ¹

3 2 ³ C exp C dC 0

12

§ 8kT · ¨ Sm ¸ © ¹

.

(4-10)

The number of molecules per unit time crossing a unit area surface element that is normal to the axis Ox and which moves with the same mean velocity as the gas is given by:

Nx

³ Vx fdv

,

(4-11)

42

Analytical Methods for Problems of Molecular Transport

where the notation, , signifies that the range of integration is over all values of v for which Vx is positive. After integration, one can obtain:

Nx

1 4

(4-12)

nV .

The last expression is very frequently used in the elementary kinetic analysis of transport phenomena [8-10].

3.

THE MEAN FREE PATH OF A MOLECULE.

Consider collisions between pairs of molecules with masses, m1 and m2 , in a binary gas mixture having a uniform steady-state. The number of collisions per unit volume and time between the two molecular sets having collision parameters in the ranges, db and d H , is given by:

dN12

f1 f 2 gD12 T , g d : dv1dv 2 .

f1 f 2 gbdbd H dv1dv 2

(4-13)

The total number of collisions per unit volume and time, N12 , is obtained by integrating Eq. (4-13) over all values of the collision parameters and the molecular velocities. For a gas having a uniform steady-state, the Maxwellian distribution function is used in the integrand. Then, N12 for rigid-sphere molecules may be written as [1]: 32

N12

S n1n2 m1m2 V 122

2S kT 3

§ m1v12 m2 v22 · exp ³ ³ ¨ 2kT ¸ gdv1dv 2 . © ¹

(4-14)

Let the variables of integration be changed from v1 , v 2 to G , g g g 21 as introduced in Section 2-5. Due to the following relationships between Jacobians:

w G, g

w v1 M 2 g , g

w v1 , g

w v1 , v 2

w v1 , v 2

w v1 , v 2

w v1 , v 2 v1

w v1 , v 2

w v1 , v 2

w v1 , v 2

1,

the velocity space element, dv1dv 2 , may be replaced by dGdg . Then, Eq. (4-14) becomes:

43

Chapter 4. The Uniform Steady-State of a Gas 32

S n1n2 m1m2 V 122

N12

2S kT 3

§ m0 G 2 M 1M 2 g 2 u³ ³ exp ¨ ¨ 2kT ©

·¸ gdGdg .

(4-15)

¸ ¹

Having performed the integration in spherical coordinates for G and g , one obtains: 12

N12

§ 2S kTm0 · 2n1n2V 122 ¨ ¸ © m1m2 ¹

.

(4-16)

The number of encounters solely between molecules of the first-kind is given by: 12

N11

§ S kT · 4n12V 12 ¨ ¸ © m1 ¹

.

(4-17)

The average number of collisions of a molecule of the first-kind per unit volume and time is usually called the collision frequency and may be written as:

Q1

N11 N12 . n1

(4-18)

The mean time between collisions (the collision interval) is then defined by:

W1

1

Q1

n1 . N11 N12

(4-19)

The mean distance between collisions is called the mean free path. This quantity is:

O1 V1W 1

12 ª m1 · º 12 2 2 § « 2 S n1V 1 S n2V 12 ¨ 1 ¸ » « © m2 ¹ »¼ ¬

1

.

(4-20)

44

Analytical Methods for Problems of Molecular Transport

For a simple gas m1 m2 m, V 1 V 2 path can be expressed in the form:

ª 21 2 S nV 2 º ¬ ¼

O

1

V , n1 n, n2

.

0 , the mean free

(4-21)

Using Eq. (4-21) the mean free path in a simple gas can now be estimated for various conditions. Assuming that one has nitrogen under conditions of Standard Temperature and Pressure (STP; 0 °C and 1 atm), the appropriate density and molecular diameter are n ~ 2.7 u 1019 cmí3 and V ~ 3 u 108 cm, respectively. Using these values in Eq. (4-21), the calculated mean free path for nitrogen at STP is then approximately 105 cm.

PROBLEMS 4.1. For the Maxwellian gas state, determine the number of molecules per unit volume with velocities in the ranges, dvz and dvU , where vz is the molecular velocity component along the axis, OZ , and vU is that along any perpendicular axis. Solution: In cylindrical coordinates, the distribution function is given by:

§ m · n¨ ¸ © 2S kT ¹

f

32

§ mv 2 exp ¨ © 2kT

· ¸ vU . ¹

The number of molecules per unit volume in dvU and dvz may be written as:

dn vU , vz

ª 2S º « ³ fd E » dvU dvz «¬ 0 »¼

§ m · 2S n ¨ ¸ © 2S kT ¹

32

§ mv 2 · exp ¨ ¸ vU dvU dvz . © 2kT ¹

4.2. For the Maxwellian gas state, determine the number of molecules per unit volume, n vx t v0 x , having an x -velocity component in the interval, vx t v0 x . Solution: The number of molecules in the above range may be determined from:

45

Chapter 4. The Uniform Steady-State of a Gas

n vx t v0 x

§ m · n¨ ¸ © 2S kT ¹ f

n

S

32 f

³

f

dv y

f

³

f

³ exp cx dcx 2

§ mv 2 · exp ¨ ¸ dvx ³ © 2kT ¹ v0 x f

dvz 1 2

c0 x

n ª1 erf c0 x º ¬ ¼

1 2

n cerf c0 x .

4.3. For the Maxwellian gas state, determine the distribution function of the kinetic energy of translation. Solution: Starting with the Maxwellian distribution function for the magnitude of the molecular velocity use the relationship 12 mv 2 E to transform the differential elements of the relationship f v dv f E dE . Substitute this transformation and the original energy relationship into the molecular velocity distribution to obtain the following distribution for the kinetic energy of translation:

f E

2

S

n kT

3 2

§ E · exp ¨ ¸ E . © kT ¹

4.4. For the Maxwellian gas state, determine the number of molecules impinging upon a unit surface element per unit time and lying in the angular range, dT , where T is the angle with respect to the surface element normal. Solution: From the Maxwellian distribution function in spherical coordinates, after integration over all velocities and the azimuthal angle, one can obtain: 12

dNT

§ 2kT · n¨ ¸ cos T sin T dT . © Sm ¹

4.5. For the Maxwellian gas state, determine the number of molecules impinging on a unit surface element per unit time and having molecular velocity magnitudes in the range, dv . Solution: This flux of molecules is given by:

dN v

§ m · 2S n ¨ ¸ © 2S kT ¹ § m · ¸ © 2S kT ¹

S n¨

32

32

§ mv 2 exp ¨ © 2kT

§ mv 2 exp ¨ © 2kT

· 3 S2 ¸ v dv ³ cos T sin T dT ¹ 0

· 3 ¸ v dv . ¹

46

Analytical Methods for Problems of Molecular Transport

4.6. A narrow molecular beam exits a system of slots into a vacuum. Determine the mean velocity and the root mean square (rms) velocity of the beam molecules. Solution: The distribution function of the beam molecules may be expressed in the form:

f v dv

n

dN v , N

where dN v has been found in Problem 4.5 and N is defined by Eq. (4-12). Then, any mean value for the beam molecules may be determined by means of this distribution function with the following results: 12

V

§ 9S kT · ¨ 8m ¸ © ¹

12

and

V

§ 4kT · ¨ m ¸ © ¹

2

.

4.7. For an equilibrium gas, prove that the ratio of the mean velocity of molecules crossing a surface element to the mean velocity of other molecules is 83 S . Solution: Using Eq. (4-10) and the result from Problem 4.6 yields:

VS V

12

§ 9S kT · ¨ 8m ¸ © ¹

12

§ 8kT · ¨ Sm ¸ © ¹

3 8

S .

4.8. Consider molecules reflecting from a differential surface element, dS c , on one surface and subsequently passing through a differential surface element, dS , on another surface. Assume that the reflected molecules at the first surface are being diffusely reflected and that the gas has a number density of nc and a temperature of T c . Moreover, assume that no molecular collisions occur while a molecule is transiting the space between the two differential surface elements. Determine the rate at which molecules reflected from dS c cross dS . Use the geometry in Fig. 4-1. Solution: The number of molecules crossing dS per second in a given direction may be expressed in the form (see Problem 4.4): 12

dNT

§ 2kT · nc ¨ ¸ © Sm ¹

dS cos T sin T dT

ncV cdS cos T 4S

d:

where d : is the solid angle subtended by dS c at dS . Taking into account that d : dS c cos T c r 2 , one can obtain:

47

Chapter 4. The Uniform Steady-State of a Gas

Figure 4-1. The geometry to be used in Problem 4.8 in determining the rate at which molecules reflected from one differential surface element cross another.

dNT

ncV cdSdS c cos T cos T c 4S r 2

12

§ 8kT c ·

Vc ¨ ¸ © Sm ¹

.

4.9. Two vessels containing the same gas at the pressures and temperatures, p1 , T1 and p2 , T2 , respectively, are connected by a small slot possessing a typical dimension of, L , which is much less than the mean free path of the gas molecules in either chamber. Derive the condition of kinetic equilibrium. Solution: If O !! L , molecules are leaving the vessels independently of each other. The condition of kinetic equilibrium requires that the net number of molecules crossing the area of the slot per second be equal to zero. This condition may be expressed as 14 n1V1 14 n2 V2 0 . From this relationship, one can obtain:

p1

p2

T1

T2

.

4.10. A binary gas mixture exits a chamber through a small slot by means of effusion. The escaping gas is collected in another chamber without loss. Calculate the ratio of the gas densities in the collection chamber.

48

Analytical Methods for Problems of Molecular Transport

Solution: Let n1 , m1 and n2 , m2 be the number densities and molecular masses, respectively, of the gases in the first chamber. The total number of molecules of each constituent in the collection chamber may be expressed as:

N ic

1 4

ni Vi S 't ,

where Vi is the mean velocity of each constituent, S is the area of the slot, and 't is the collection time interval. From this, the ratio of the gas densities can be written as: n1c n2c

N1c N 2c

n1 V1 . n2 V2

Equation (4-10) then yields:

n1c n2c

12

n1 § m2 · ¨ ¸ n2 © m1 ¹

.

4.11. Electrons evaporating from a hot wire form an electron gas with a number density, n . These electrons exit the chamber in which they are generated by means of a system of slots which serve to direct the electrons in one specific direction forming a beam. While exiting the chamber, this electron beam must pass through a hindering potential, M , which stops some of the electrons. Determine the flux density of the electrons that are able to overcome this hindering potential. Solution: The flux density of electrons having velocities in the interval, dv , is given by (Problem 4.5):

§ m · ¸ © 2S kT ¹

dN v

S n¨

32

§ mv 2 · 3 exp ¨ ¸ v dv . © 2kT ¹

The flux density of electrons overcoming the hindering potential may be written as: f

N

³ dNv

v0

12

n § kT · § eM · § eM · 1¸ exp ¨ ¨ ¸ ¨ ¸ . m kT 2S © ¹ © ¹ © kT ¹

49

Chapter 4. The Uniform Steady-State of a Gas

where

1 2

mv02

eM .

4.12. The work function of a metal is W . Determine the current density of the emitted electrons if the temperature is T and the electron number density inside the metal is n . Solution: Let OX be the axis normal to the surface of the metal. The current density along this axis is given by: 12 f

jx

§ m · ne ¨ ¸ © 2S kT ¹

§ mvx2 v exp ¨ ³ x ¨ 2kT © v0 x

· ¸¸ dvx ¹

1 4

§ W · neV exp ¨ ¸ , © kT ¹

where 12 mv02x W . This is equivalent to the classical Richardson thermionic emission formula. 4.13. Determine the number density variation for the steady-state of a gas that is situated in a force field described by the potential, M r . Solution: Let the distribution function be represented in the form:

f

§ · m n r ¨ ¨ 2S kT r ¸¸ © ¹

32

§ mv 2 · exp ¨ . ¨ 2kT r ¸¸ © ¹

Substituting this function into the Boltzmann equation, one obtains:

w w · § ¨ v wr F wv ¸ ln f 0 . © ¹ This expression may be reduced to:

° w § n v ® ln ¨ 3 2 ¯° wr © T

2 · mv wT m °½ F¾ 0 . ¸ 2 ¹ 2kT wr kT °¿

which is an identity in v . As such, the following system of equations may be obtained: wT wr

0 and

w ª § n ln ¨ wr ¬« © T 3 2

º · m M r » ¸ ¹ kT ¼

0 .

From the above system of equations, it is easy to obtain:

50

Analytical Methods for Problems of Molecular Transport

§ mM r · n r n0 exp ¨¨ ¸ . kT ¸¹ © where n0 is the number density at the point for which M r 0 and T is the constant temperature. 4.14. A gas rotates around a fixed axis with an angular speed, Z . Determine the number density distribution in the gas. Solution: The rotation of a gas is equivalent to an external force field having the potential: M r 12 Z 2 r 2 , where r is the distance from the rotational axis. From Problem 4.13, the variation in the number density is known to be:

§ mM r · n r n0 exp ¨¨ ¸ . kT ¸¹ © Substituting the rotational potential into this expression yields:

§ mZ 2 r 2 · n r n0 exp ¨ ¸ . © 2kT ¹ 4.15. Determine the number density distribution, n r , of a gas contained in a cylindrical vessel of radius, R , and length, L , that is rotating about its axis with an angular velocity, Z . The total number of molecules in the vessel is N . Solution: Using the number distribution obtained in Problem 4.14, one can obtain:

n r

§ mZ 2 r 2 · ª § mZ 2 R 2 NmZ 2 exp ¨ exp ¸« ¨ 2S L © 2kT ¹ ¬« © 2kT

· º ¸ 1» ¹ ¼»

1

.

4.16. Evaluate the mass of the Earth’s atmosphere assuming that almost all of the mass is contained in the thin layer near the Earth’s surface. Solution: The potential energy of a molecule near the Earth’s surface is given by: U h mgh . The total mass of the atmosphere is then approximated by: f

§ mgh · M | 4S R02 mn ³ exp ¨ ¸ dh © kT ¹ 0

4S R02 nkTg 1 ,

Chapter 4. The Uniform Steady-State of a Gas

51

where R0 is the radius of the Earth. Using typical values for the quantities involved yields M 5.3 u 1018 kg.

REFERENCES 1. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 2. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 3. Huang, K., Statistical Mechanics (John Wiley and Sons, New York, 1963). 4. Cercignani, C., Theory and Application of the Boltzmann Equation (Scottish Academic Press, Edinburgh, U.K., 1975). 5. Brush, S.G., Selected Readings in Physics. Vol. 2. Irreversible Processes (Pergamon Press, New York, 1966). 6. Klein, M.J., Paul Ehrenfest (North-Holland, Amsterdam and Elsevier, New York, 1970). 7. Uhlenbeck, G.E. and Ford, G.W., Lectures in Statistical Mechanics (American Mathematical Society, Providence, RI, 1963). 8. Kennard, E.H. , Kinetic Theory of Gases (McGraw-Hill, New York and London, 1938). 9. Knudsen, M. , The Kinetic Theory of Gases, third edition (Methuen & Co., London, and Wiley, New York, reprinted 1952). 10. Loeb, L.B. , The Kinetic Theory of Gases (Dover, New York, 1961).

Chapter 5 THE NON-UNIFORM STATE FOR A SIMPLE GAS

1.

EXPANSION IN POWERS OF A SMALL PARAMETER.

Consider the unsteady, non-uniform state of an infinite gas. For this general case, the Boltzmann equation has the form [1]:

Df

J ff1 ,

(5-1)

where:

Df

wf wf wf v F , wt wr wv

and:

J ff1

³ ³ ³ ff1 f cf1c gbdbd H dv1

.

The Boltzmann equation may be readily converted into a dimensionless form by introducing the following quantities: 12

§ 2kT · § m v ¨ ¸ c , f n¨ © m ¹ © 2kT

t W t , r Lr , F wF ,

· ¸ ¹

32

f ,

b V 21 4 b , H

SH ,

54

Analytical Methods for Problems of Molecular Transport

where W and L are typical values for the time and length variations of the macroscopic quantities of a gas. In terms of the above dimensionless quantities, Eq. (5-1) becomes: 12

L § m · wf wf wmL wf c F ¨ ¸

W © 2kT ¹ wt 2kT wr wc 1 J f f1 ,

D f

(5-2)

where O L . The parameter, , is usually called the typical parameter of non-uniformity of a gas but, since L is not necessarily the typical dimension of a boundary problem, is not the Knudsen number [2]. Let the macroscopic values vary slightly in the mean free path length, so that the following relations are satisfied:

O ln n 1 , O ln T 1 , O ln u 1 . Moreover, the typical values W and w are assumed to be confined by the 12 and w d 2kT mL . Then, the basic conditions W t L m 2kT dimensionless parameter, 1 . This makes it quite natural to seek a solution to Eq. (5-2) in the form of a power series in . Taking into account only the linear terms of this expansion, the distribution function may be approximated by:

f

0

1 f f ,

(5-3a)

or:

f

0 1 f f ,

(5-3b)

in ordinary variables. Substituting Eq. (5-3a) into Eq. (5-2), and retaining only the main terms in , one obtains:

0 D f

1

0

0

0

1

1 0 J f f1 J f f1 J f f1

^

` .

(5-4)

After equating coefficients of like powers of , the following relationships may be written:

55

Chapter 5. The Non-Uniform State for a Simple Gas

0

0 J f f1

(5-5)

0 ,

^

0 D f

0

1

1 0 J f f1 J f f1

` .

(5-6)

These integral equations are the basic ones used to obtain analytical expressions for the first and second-order correction terms to the distribution function.

2.

THE FIRST APPROXIMATION. In the usual variables, Eq. (5-5) has the form:

³ ³ f

0

0 0 0 c f1 c f f1 gbdbd H dv1

0 .

The general solution of this integral equation is defined by [1]: 0 0 f c f1 c

0 0 f f1 ,

and therefore, ln f is a linear combination of the summational invariants of molecular encounters. By a simple transformation one can obtain: 0

f

0

§ m · n¨ ¸ © 2S kT ¹

32

§ m v u 2 exp ¨ ¨ 2kT ©

· ¸ , ¸ ¹

(5-7)

where n , u , and T denote arbitrary values related to each of the summational invariants. These values can be identified with number density, mean velocity, and temperature at a point, r ,t . This means that the Maxwellian distribution function in which n n r , t , T T r , t , and u u r ,t is chosen as the first approximation to f . This function is usually called the local Maxwellian distribution function, which, for the general, non-uniform, non-steady-state of a gas, may be written as: 0 f r , t

§ · m n r , t ¨ ¨ 2S kT r , t ¸¸ © ¹

32

§ m v u r , t 2 exp ¨ ¨ 2kT r , t ©

· ¸ . ¸ ¹

(5-8)

56

Analytical Methods for Problems of Molecular Transport

As a consequence of the identification of n , u , and T with the density, 1 mean velocity, and temperature at the point, r ,t , the function f must satisfy the following relations:

³f

1 i

\ dv 0 ,

(5-9)

where \ 1, mV , 12 mV 2 . These conditions must be satisfied in the formulation of any expressions for the second-order correction term to the distribution function. i

3.

A GENERAL FORMAL SOLUTION FOR THE SECOND CORRECTION. Eq. (5-6) in ordinary variables may be written in the form:

0 1 1 0 J f f1 J f f1

0 D f .

(5-10)

This is the linear non-homogeneous integral equation for the second-order 1 1 1 correction to the distribution function, f . Let F and F be the particular and general solutions of the non-homogeneous and homogeneous equations, respectively. Then, the general solution can be represented by:

f

F F .

1

1

1

(5-11)

Substituting this expression into Eq. (5-10), one can obtain the following form of the homogeneous integral equation:

I M 1

(5-12)

0 .

Here, the correction M is given by the relation, F f M and I M is the standard linearized collision operator which may be expressed as: 1

1

0 0 I F n 2 ³ ³ f f1 F F1 F c F1c gbdbd H dv1 .

0

1

1

(5-13)

Eq. (5-12) means that M must be the summational collision invariant which may be written as: 1

57

Chapter 5. The Non-Uniform State for a Simple Gas

M 1 D 1,1 D 2,1 mv D 3,1 12 mv 2 ,

(5-14)

where D are arbitrary functions of r and t . These functions may be found from the conditions given by Eq. (5-9). Now, to find the necessary conditions of solubility for Eq. (5-10), one uses the integral theorem [1] given by: i ,1

³\

i ª

0 1 1 0 J f f1 J f f1 º» dv 0 . ¬« ¼

Using this theorem in Eq. (5-10), one can obtain the following conditions:

³\

i

0 D f dv 0 .

(5-15)

These relations, when i 1, 2,3 , mean that the values n , u , and T 0 contained in f must satisfy the continuum equations obtained by means of the use of the first-order distribution function [1]. 1 0 1 1 Let F be represented in the form, F f ) . Then, Eq. (5-10) gives the following basic equation for this correction:

n2 I ) 1

0 D f .

(5-16)

From general conditions given by Eq. (5-9), one can conclude that the 1 function, ) , must satisfy the following additional conditions:

³\

i 0

4.

f

) M dv 1

1

0 .

(5-17)

THE TRANSFORMATION OF THE NONHOMOGENEOUS TERM.

If the variables V, r , t are used, the right-hand-side of Eq. (5-16) may be expressed in a form analogous to that given in Eq. (2.6): 0 Df

0 0 f ln f

D Du · w w § w wu ½ V: ¾ , u® V ¨ F ¸ Dt ¹ wV wV wr © wr ¿ ¯ Dt

(5-18)

58

Analytical Methods for Problems of Molecular Transport

where: D Dt

w w . u wt wr

The space and time derivatives of n , u , and T are connected by the equations of the continuous medium in which the pressure tensor and the heat flux must be calculated by use of the local Maxwellian distribution function given in Eq. (5-8). This results in the following expressions: 0 Pik

0 pG ik , q

nkT G ik

(5-19)

0 .

Substituting these quantities into Eqs. (2.7)-(2.9), one obtains: Dn Dt

n

Du Dt

F

DT Dt

23 T

w u , wr

(5-20)

1 wp , U wr

(5-21)

w u . wr

(5-22)

Eqs. (5-20) and (5-22) result in: D ln nT 3 2 Dt

0 , ln nT 3 2

const .

(5-23)

Using the second relationship in Eq. (5-23), one can write:

0 ln f

const

mV 2 . 2kT

Taking into account Eq. (5-22), the first term in Eq. (5-18) may be expressed in the form:

59

Chapter 5. The Non-Uniform State for a Simple Gas

D 0 ln f Dt

1 2

mV 2 DT kT 2 Dt

13

mV 2 w u kT wr

wu mV 2 G ik i . kT wxk

13

The last term in Eq. (5-18) may be written as:

V : wu

0 w ln f

wV

wu m ViVk i . kT wxk

wr

The sum of the first and last terms of Eq. (5-18) is then:

wui m ViVk 13 V 2G ik kT wxk

m $ wu VV : . kT wr

(5-24)

The sum of the two middle terms of Eq. (5-18) is given by: 0 w w § f § w · 0 V ¨ ln f ln nkT ¸ V ln ¨ wr wr ©¨ nkT © wr ¹

· ¸ ¸ ¹

§ w w ln T mV 2 wT · § 1 mV 2 5 · . V ¨ ln T 5 2 12 2¸V ¸ ¨2 2 wr kT wr ¹ © kT © wr ¹

(5-25)

Using Eqs. (5-24) and (5-25), Eq. (5-18) can be represented as: $ w ln T wu °½ 0 ° f ® C 2 52 V 2 CC : ¾ . wr wr °¿ °¯

0 Df

(5-26)

The basic relation expressed by Eq. (5-16) can now be written in the form:

n2 I ) 1

5.

$ w ln T wu °½ 0 ° f ® C 2 52 V 2 CC : ¾ . wr wr ¿° ¯°

(5-27)

THE SECOND APPROXIMATION.

Since Eq. (5-27) is a linear integral equation and the right-hand-side of this equation is linear in w ln T w r and wu w r , one may seek a solution of this equation in the form of a sum of the two terms each of which is

60

Analytical Methods for Problems of Molecular Transport

proportional to the appropriate space non-uniformity. Thus, one can write [1,3-5]:

) 1

A

w ln T wu , 2B : wr wr

(5-28)

where A is a vector and B is a tensor. The quantities, A and B , introduced here are identical to those obtained from the Chapman-Enskog theory for multiple gas mixtures in the limiting case of a simple gas [1]. Substituting Eq. (5-28) into Eq. (5-27), one can separate out equations for the unknown functions, A and B . These functions are then particular solutions for the following equations:

n2 I A

0 f C 2 52 V ,

(5-29)

n2 I B

$ 0 f CC ,

(5-30)

which, in the coordinate form, can be written as:

n 2 I Ai

n 2 I Bik

0 f C 2 52 Vi ,

(5-31a)

0 f Ci Ck 13 C 2G ik .

(5-31b)

One now assumes that the solutions of Eqs. (5-31a) and (5-31b) are of the following form: A

AC C ,

B

B C CC ,

$

(5-32)

(5-33)

where A C and B C are two new unknown functions of n , T , and C C . The general formal solution of Eq. (5-27) then has the form:

61

Chapter 5. The Non-Uniform State for a Simple Gas $ w ln T wu ° 0 2 B C CC : f ® A C C wr wr °¯

f 1

D

1,1

D

2,1

mV D

3,1

1 2

(5-34)

2½

mV ¾ dv . ¿

where the constants D in Eq. (5-34) are chosen such that the function 1 f satisfies the following equations: i ,1

³f

1 i

\ dv 0 ; i 1, 2,3 .

(5-35)

Neglecting vanishing integrals, these equations have the form:

³f ³f ³f

0

D

1,1

0 §

3,1 D 12 mV 2 dv 0 ,

¨¨ A C ©

0

D

1,1

w ln T 12 2,1 · 2mkT D ¸¸ C 2 dv 0 , wr ¹

3,1 D 12 mV 2

1 2

mV 2 dv 0 .

The first and third of the above relations show that D D 0 . The 2,1 second shows that D ~ w ln T wr and, hence, that this term may be 2,1 absorbed into the first term. In order that D 0 also, the function A C must satisfy the additional relation: 1,1

³ AC C

2

0 f dv 0 .

3,1

(5-36)

Now, the second approximation to the distribution function is given by:

f 1

0 1 1 f ) M

$ w ln T wu ½° ° 0 f ® A C C 2 B C CC : ¾ . wr wr ¿° ¯°

(5-37)

62

Analytical Methods for Problems of Molecular Transport

6.

THE FIRST-ORDER CHAPMAN-ENSKOG SOLUTION FOR THERMAL CONDUCTION.

Suppose that the function A can be expressed as a convergent series of the form: f

¦ ar a r

A

,

r a

r 0

where S3 2 C 2 defined by: r

S m x n

S3 2 C 2 C , r

(5-38)

are the Sonine polynomials. The polynomial S m x is n

n

¦ x p 0

p

m n n p p n p

,

where:

m n n p m n m n 1 ... ª¬ m n n p 1 º¼

,

and m n 0 1 . Two special cases exist where:

S m x 1 , 0

S m x m 1 x . 1

These polynomials satisfy the following integral relations:

f

p q m ³ exp x Sm x Sm x x dx 0

0 ; pzq , ° ° ® ° * m p 1 ; p q , °¯ p

where * k is the gamma function. The additional condition expressed by Eq. (5-36) gives:

63

Chapter 5. The Non-Uniform State for a Simple Gas 2 2 ³ exp C C A C dC f ° f ½° p 4S ³ C 4 exp C 2 ® ¦ a p S3 2 C 2 ¾ dC °¯ p 0 °¿ 0

f

f

p 0

0

2S ¦ a p ³ exp C 2 S3 2 C 2 S3 2 C 2 C 3 d C 2 2S a0 *

52

0

p

0 ,

and therefore a0 0 . To get acceptable accuracy [1], only one term of the series in Eq. (5-38) need to be taken into account, i.e.:

A

a1 CS3 2 C 2 . 1

1

Substituting this expression into Eq. (5-29), one obtains:

n 2 a1 I CS3 2 C 2 1

1

0 f C 2 52 V

1 0 f VS3 2 C 2 .

(5-39)

After multiplying this equation by CS3 2 C 2 and integrating over all values of v , Eq. (5-39) can be expressed in the form: 1

n 2 a1 ³ CS3 2 C 2 I CS3 2 C 2 dv D1 , 1

1

1

1 or a1 a11 D1 where:

a11

ªCS 1 C 2 , CS 1 C 2 º , 32 ¬ 32 ¼

12

§ 2kT · ¸ © m ¹

D1 n 2 ³ f 0 VS312 C 2 CS312 C 2 dv 154 n 1 ¨

.

The expression for the bracket integral a11 is calculated in Appendix A. Using Eq. (A-9), the following solution of Eq. (5-39) can be obtained for rigid-sphere molecules:

64

Analytical Methods for Problems of Molecular Transport 12

§ 2kT · 154 n 1 ¨ ¸ © m ¹

a1 1

1 a11

15 SO . 16

(5-40)

The vector, A , for the first-order Chapman-Enskog solution is given by:

A

15 S O CS3 2 C 2 . 16 1

7.

(5-41)

THE FIRST-ORDER CHAPMAN-ENSKOG SOLUTION FOR VISCOSITY.

To simplify the following calculations, consider the viscosity equation for only one component of the tensor, B . The equation for Bxy may be written as:

n2 I CxC y B C

0 f C x C y .

(5-42)

Since the tensor, B , is a non-divergent tensor of molecular velocity components, it can be expressed in the general form [1]: f

$ $ r 1 B CC B C CC ¦ br S5 2 C 2 , r 1

(5-43)

where the coefficients, br , are constants to be determined. When only one tensor component, Bxy , is being considered, it can be expressed as:

Bxy

f

r 1 C x C y ¦ br S5 2 C 2 . r 1

(5-44)

To obtain the first-order Chapman-Enskog solution (retaining only one term of the expansion), this tensor is given by:

Bxy

b1 C x C y . 1

Substituting Eq. (5-45) into Eq. (5-42) one can obtain:

(5-45)

65

Chapter 5. The Non-Uniform State for a Simple Gas

n 2b1 I C x C y

1

0 f Cx C y .

(5-46)

After multiplying both sides of Eq. (5-46) by C x C y dv and integrating over all velocities one obtains:

c b1 b11 1

(5-47)

E1c ,

where:

c b11

¬ªC x C y , C x C y ¼º ,

E1c n 2 ³ C x2C y2 f 0 dv

1 4

n 1 .

c and E1c introduced here differ It is important to note that the values of b11 from the corresponding values in [1] by a factor equal to 101 . Using Eq. (Ac , one obtains: 12) for b11 b1 1

12

1 cn 4b11

5 8

§ m · ¸ © 2kT ¹

(5-48)

.

SO¨

Both Eqs. (5-40) and (5-48) have been obtained for rigid-sphere molecules. From the preceding analysis, in general, the components of the tensor for the first-order Chapman-Enskog solution, Bik , have the form: 12

Bik

5 8

§ m · ¸ © 2kT ¹

SO¨

C C i

k

13 G ik C 2 .

(5-49)

Using Eqs. (5-41) and (5-49) and introducing the mean free path, the firstorder Chapman-Enskog distribution function is then given by:

f

w ln T 0 ° O S S312 C 2 Ci f ®1 15 16 wxi °¯

12

§ m · O S¨ ¸ © 2kT ¹ 5 4

C C i

k

1 2 C ik 3

G

wui ½° ¾ . wxi °¿

Here, molecules have been assumed to act as rigid spheres.

(5-50)

66

Analytical Methods for Problems of Molecular Transport

8.

THE THERMAL CONDUCTIVITY AND VISCOSITY COEFFICIENTS.

Since the distribution function is known, the coefficients of viscosity and thermal conductivity may be readily calculated. The thermal conductivity coefficient can be obtained from the relation:

qi

m ³ V 2Vi fdV

1 2

N

wT . wxi

(5-51)

Having performed the very simple integration in Eq. (5-51), one obtains the following expression for the thermal conductivity coefficient of a monatomic gas: 12

N

75 64

k § kT · ¨ ¸ V 2 ©Sm ¹

.

(5-52)

The molecules here have been assumed to behave as rigid spheres. To derive a similar expression for the viscosity coefficient, the simplest case of u ^0, u ,0` and u u x is considered. In this case:

Pxy

m ³ VxVy fdV

P

wu . wx

Using Eq. (5-50) in this expression and integrating yields: 12

P

5 16

V

2

§ mkT · ¨ S ¸ © ¹

.

(5-53)

The general expression for the pressure tensor [1] is given by:

Pik

ª§ wu wu · º pG ik P «¨ i k ¸ 23 G ik u » . «¬© wxk wxi ¹ »¼

(5-54)

Chapter 5. The Non-Uniform State for a Simple Gas

9.

67

THE FIRST-ORDER APPROXIMATION FOR ARBITRARY INTERMOLECULAR POTENTIAL.

In the first approximation, the thermal conductivity and viscosity coefficients depend only on the bracket integrals, a11 and b11 , respectively. These bracket integrals for arbitrary intermolecular potentials may be expressed in the form [1]:

a11

ªCS 1 C 2 , CS 1 C 2 º 32 ¬ 32 ¼

b11

$ ª $ º CC , CC « » ¬ ¼

2,2 4: ,

(5-55)

and: 2,2 4: .

(5-56)

where the b11 given here is in the same form as that used by Chapman and Cowling [1]. i, j The : -integrals can be evaluated only if the form of the intermolecular interaction is specified. All potential functions may be characterized by two parameters, H and V , and it is convenient to represent the mutual potential energy of two molecules in the form:

M r H f r V ,

(5-57)

where f r V is the same function for all gases described by the specific interaction model. The Lennard-Jones (6-12) potential model has found the widest use in applied transport problems. This model is described by:

M r

ª§ V ·12 § V ·6 º 4H «¨ ¸ ¨ ¸ » , © r ¹ »¼ «¬© r ¹

(5-58)

where H is the depth of the potential well and V r at the point where M r 0 . This parameter may be called an effective molecular diameter. The two parameters of this model, H k (K) and V , have been tabulated for many gases by a comparison of the theory with experimental data. All of the : -integrals are expressible in the form [3,5]:

68

Analytical Methods for Problems of Molecular Transport

: i, j

ª: i , j º : i , j å , ¬ ¼ r .s.

(5-59)

i, j where ª : º is the integral for rigid-sphere molecules that is given by: ¬ ¼ r .s.

ª: i , j º ¬ ¼ r .s.

12

§ kT · SV ¨ ¸ ©Sm ¹ 2

i 1 1 º » . 1 2 i 1 » «¬ ¼

j 1 «ª 2

(5-60)

i, j å The reduced integral, : , depends only on the reduced temperature which is defined as:

T

kT

H

.

(5-61)

The numerical values of the reduced : -integrals for various models of the intermolecular potential may be found in [3,5,6]. Taking into account Eq. (5-59), one can express the first approximation to the transport coefficients for a monatomic gas as follows:

> P @1

5 16

>N @1

25 32

S mkT 1 2

SV 2 : 2,2 å T

S mkT 1 2

SV 2 : 2,2 å T

,

cV ,

(5-62)

(5-63)

where cV is the specific heat per unit mass at constant volume which is (for a monatomic gas, for example) cV 3k 2m .

10.

THE SECOND-ORDER APPROXIMATION FOR ARBITRARY INTERMOLECULAR POTENTIAL.

The parameters of the second-order Chapman-Enskog solutions may be found from the following algebraic systems [1]:

Chapter 5. The Non-Uniform State for a Simple Gas

69

12

2 2 a1 a11 a2 a21

§ 2kT · 154 n 1 ¨ ¸ © m ¹

2 2 a1 a12 a2 a22

0 ,

,

and: 2 2 b1 b11 b2 b21

5 2

2 2 b1 b12 b2 b22

0 ,

n 1 ,

All of the parameters are expressible in terms of the bracket integrals defined in [1]. In the second approximation, the thermal conductivity and viscosity coefficients can be shown to be:

>N @2 >N @1 fN 2 T

,

(5-64)

> P @2 > P @1 f P 2 T

,

(5-65)

and:

2 2 where the functions, fN T and f P T , are given by:

77 112v 80t , 28 64v 2 80t

(5-66)

301 336v 240t . 154 48t

(5-67)

2 fN T

and: 2 f P T

Here, the following notations have been introduced:

2,3 å v : T

: 2,2 å T ,

(5-68)

70

Analytical Methods for Problems of Molecular Transport

Table 5-1. Functions for calculating the second-order transport coefficients. 2 2 2

fN

T

0.30 0.50 0.75 1.00 1.25 1.50 2.00 2.50 3.00 4.00 5.00 10.0 50.0 100.0 400.0

t

T

fP

1.00189 1.00009 1.00004 1.00000 1.00012 1.00055 1.00214 1.00380 1.00516 1.00737 1.00867 1.01096 1.01160 1.01167 1.01170

: 2,4 å T

T

1.01030 1.01020 1.02181 1.02775 1.03074 1.03149 1.02832 1.02580 1.02448 1.02105 1.01948 1.01582 1.01419 1.01410 1.01383

a2

T

0.02680 0.00588 –0.00369 –0.00030 0.00680 0.01419 0.02801 0.03719 0.04325 0.05158 0.05590 0.06279 0.06465 0.06483 0.06492

: 2,2 å T .

2 b2 T

0.01724 0.00376 –0.00237 –0.00019 0.00441 0.00922 0.01825 0.02426 0.02824 0.03370 0.03653 0.04103 0.04222 0.04233 0.04239

(5-69)

The second coefficients of the Chapman-Enskog solutions are found to be: 2 a2

32v 28 , 77 112v 80t

(5-70)

2 b2

96v 84 . 301 336v 240t

(5-71)

and:

2 2 2 2 2 2 where a2 a2 a1 and b2 b2 b1 . For the Lennard-Jones (6-12) 2 2 2 2 potential, the functions, fN T , f P T , a2 T , and b2 T , are given in Table 5.1.

PROBLEMS 5.1. Show that if:

f

0

§ · m n r, t ¨ ¨ 2S kT r, t ¸¸ © ¹

32

§ m v u r, t 2 exp ¨ ¨ 2kT r, t ©

· ¸ , ¸ ¹

71

Chapter 5. The Non-Uniform State for a Simple Gas

then the collision operator associated with this function is equal to zero. Solution: Very simple transformations result in the following expression which occurs in the collision operator: 0 0 0 0 f f1 f c f1 c

§ m v 2 v12 · m ° ¸ A r, t ®exp ¨ v v u 1 ¨ ¸ kT 2kT °¯ © ¹

§ m vc2 v1c2 ·½ m ° vc v1c u ¸ ¾ 0 . exp ¨ ¨ ¸° 2kT kT © ¹¿ That this is equal to zero (and hence the collision operator is equal to zero) follows from the conservation of energy and momentum during molecular encounters.

5.2. Obtain the second-order Chapman-Enskog solution for the thermal conduction problem by using the following polynomial expansion: A

^

` .

1 2 2 2 a1 C S3 2 C 2 a2 S3 2 C 2

Assume that the molecules in this problem are rigid spheres. 1 Solution: Multiplying Eq. (5-29) alternately by the terms, CS3 2 C 2 2 2 and CS3 2 C , one can obtain the following algebraic system:

2 2 a1 a11 a2 a21

12

§ 2kT · 154 n 1 ¨ ¸ © m ¹

where a12 a21 14 a11 and a22 solution of this system yields: 2 a1

45 1 a 44 1

2 , a2

1 1 a 11 1

2 2 , a1 a12 a2 a22

45 a 16 11

2 , a2

0 ,

for rigid-sphere molecules [1]. The

4 45

.

5.3. Derive the expression for the thermal conductivity coefficient for the second-order Chapman-Enskog solution for rigid-sphere molecules. Solution: From Eqs. (5-37) and (5-51), one can obtain:

72

Analytical Methods for Problems of Molecular Transport

qi

§ 2kT · 12 mnS 3 2 ¨ ¸ © m ¹ 3 2

mnS

1 3

§ 2kT · mnS 1 2 ¨ ¸ © m ¹ 12

2kT · ¸ © m ¹

5§ 4¨

kna1

1 wT C 2Ci2 A C exp C 2 dC T wxi ³

32

1 wT T wxi

32

1 wT T wxi

§ 2kT · ¨ m ¸ © ¹

1 6

32

wT wxi

N

2 2 2 ³ 52 C C A C exp C dC f

³x

32

0

§ f · p 1 S3 2 x ¨ ¦ a p S3 2 x ¸ exp x dx ¨ ¸ ©p 1 ¹

wT . wxi

This implies that the thermal conductivity coefficient is proportional to a1 only and may be written as: 12

§ 2kT · 54 ¨ ¸ © m ¹

N

kna1 .

For the second-order Chapman-Enskog solution (Problem 5.2), one can obtain:

N

2

12

2kT · ¸ © m ¹ 5§ 4¨

2 kna1

45 44

N 1 ,

where the molecules have been assumed to be rigid spheres.

5.4. Obtain the second-order Chapman-Enskog solution for the viscosity problem by using the following polynomial expansion:

Bxy

^

` .

2 2 1 b1 C x C y 1 b2 S5 2 C 2

Assume that the molecules in this problem are rigid spheres. Solution: Multiplying Eq. (5-30) alternately by the terms, C x C y and 1 Cx C y S5 2 C 2 , one can obtain the following algebraic system:

2 c b2 2 b21 c b1 b11

4n 1

2 c b2 2 b22 c b1 b12

0 ,

,

Chapter 5. The Non-Uniform State for a Simple Gas

c b21 c 14 b11 c and b22 c where b12 solution of this system yields: 2 b1

205 1 b 202 1

2 , b2

1 6 b 101 1

205 c b 48 11

73

for rigid-sphere molecules [1]. The

2 , b2

12 205

.

5.5. Derive the expression for the viscosity coefficient for the second-order Chapman-Enskog solution for rigid-sphere molecules. Solution: Taking into account the definition of the pressure tensor and Eq. (5-37), one can obtain:

Pxy

4S 3 2 nT

wu C x2C y2 B C exp C 2 dC wx ³

S 1 2 nkT 16 15

wu f 6 C B C exp C 2 dC wx ³0

158 S 1 2 nkT

§ f · wu f 5 2 0 p 1 x S x ¨¨ ¦ bp S5 2 x ¸¸ exp x dx 52 ³ wx 0 ©p 1 ¹

nkTb1

wu . wx

This implies that the viscosity coefficient is proportional to b1 only and does not contain any other coefficients from the polynomial expansion. Then the viscosity coefficient can be written as P nkTb1 . For the second-order Chapman-Enskog solution (Problem 5.4), one can obtain:

P 2

2 kTb1

205 202

P 1 .

REFERENCES 1. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 2. Uhlenbeck, G.E. and Ford, G.W., Lectures in Statistical Mechanics (American Mathematical Society, Providence, RI, 1963). 3. Hirschfelder, J.O., Curtiss, C.F. and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954). 4. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, New York, 1969). 5. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 6. Maitland, G.C., Rigby, M.A., Smith, E.B., and Wakeham, W.A., Intermolecular Forces (Oxford University Press, Oxford, 1981).

Chapter 6 REGIMES OF RAREFIED GAS FLOWS

1.

THE KNUDSEN NUMBER.

Consider the steady-state flow of an infinite stream of a rarefied gas over a body having a characteristic dimension, R , in the absence of external forces. In this case, the Boltzmann equation may be expressed in the form:

v

wf wr

³ ³ ³ f cf1c ff1 gbdbd H dv1

.

(6-1)

Let this equation be transformed into dimensionless variables by means of the following: 12

§ 2kT · v ¨ ¸ c, f © m ¹

§ m · n¨ ¸ © 2kT ¹

32

f , r

21 4 V b , H

Rr , b

SH .

In these variables, Eq. (6-1) becomes:

c

wf wr

R

O

J f f1

Kn 1 J f f1 ,

(6-2)

where Kn O R is the Knudsen number and J f f1 is the standard dimensionless collision operator. The Knudsen number is the natural parameter characterizing the degree of rarefaction of a gas for a given boundary problem. The Knudsen number is defined as the ratio of the mean free path of a molecule to a characteristic external geometrical dimension of

76

Analytical Methods for Problems of Molecular Transport

the boundary problem. The Knudsen number is not a property of the gas alone as it explicitly involves the system dimension, R . It is, however, the basic parameter for any given boundary value transport problem.

2.

A GENERAL ANALYSIS OF THE DIFFERENT GAS FLOW REGIMES.

The problem of the steady-state flow of a rarefied gas past a body has a characteristic dimensionless parameter. This parameter is the Knudsen number described in Section 6.1. Large or small values of the Knudsen number permit simplification or modification of the Boltzmann equation for each particular case. Corresponding to certain ranges of the Knudsen number, there are four regimes in the Kinetic Theory of Gases. This classification has a schematic character. For conditions when Kn o 0 ( Kn 0.01 , the continuum regime), one may use the Navier-Stokes equations with the usual hydrodynamic boundary conditions (Eqs. (2-7)-(2-9) together with Eqs. (5-51) and (5-54)). If Kn 1 ( 0.01 Kn 0.1 , the slip-flow regime), the gas may be described by the Navier-Stokes equations but the boundary conditions should be modified by the introduction of slip conditions on the surface of the body. The slipflow regime will be described in detail in Chapter 10. The other regions are those in which the Knudsen number is greater than unity. Two sub-regions can be distinguished here. The limiting regime, where Kn o f , is usually called the free-molecular regime. If the dimensions of the body are insignificant compared with the mean free path, one says that the flow past the body is of the free-molecular type. For such flows the presence of the body does not disturb the distribution function of the incident molecules of the gas in the neighborhood of the body. In this case, the molecules that rebound from the surface of the body collide with other molecules at distances on the order of the mean free path from the body. Therefore, in a free-molecular flow, the collisions between the molecules that rebound and those coming from the gas occur at a great enough distance from the body that the influence of these collisions can be neglected. The free-molecular distribution function is constant along the molecular trajectories. On the surface of the body the distribution function of incident molecules is the same as that in the surrounding volume of gas far from the body. The distribution function of molecules rebounded from the body can be found from the boundary conditions which will be considered in Section 6.3. The region where Kn 1 is usually called the near free-molecular regime. For flow in this regime, Knudsen iteration [1] is a well known

Chapter 6. Regimes of Rarefied Gas Flows

77

approximate solution of the Boltzmann equation. The distribution function is assumed to have the form:

f

0 1 f Kn 1 f .

(6-3)

The function, f , may be found from the non-homogeneous linear partial differential equation that contains the free-molecular distribution function inside the collision integral. The solutions of several specific problems of interest in this regime may be found in [2-9]. The region where Kn ~ 1 is the most difficult to describe theoretically. In this region the flow has features typical of both the free-molecular and continuum regimes. In this transition region, it is often convenient to use the various moment methods that will be discussed in Chapter 11 in order to find the solutions to boundary problems of interest. 1

3.

THE BOUNDARY CONDITIONS.

Both the mathematical formulation and the solution of all boundary value transport problems, by both the direct and the moment methods, requires knowledge of the applicable microscopic boundary conditions. These conditions on the surface of a body can be formulated for the distribution function of the reflected molecules by means of a dispersion kernel [10-15]. Within this framework only a general description may be provided at present, which cannot be used to solve practical boundary problems because no theory currently exists from which the necessary kernels can be derived. The absence of a strict theory for the molecular interaction of a gas with a solid surface results in the need to use definite models to postulate the distribution function for the reflected molecules. This function contains a certain number of free parameters which are usually called accommodation coefficients. These coefficients may be determined by a comparison of the theory constructed from a given model with an appropriate experiment. Maxwell was the first to propose the boundary model that has been widely used in various modified forms. At present, this model still appears to be the most convenient and correct formulation for the various boundary value transport problems; particularly those in which one assumes a noncondensable gas for which the boundary surface is impenetrable. The Maxwellian boundary model is constructed on the assumption that some fraction 1 DW of the incident gas molecules are reflected from the surface specularly, while the remaining fraction, DW , are reflected diffusely with a Maxwellian distribution. This supposition can be expressed as:

78

Analytical Methods for Problems of Molecular Transport

f v, rS , t DW f r 1 DW f vc, rS , t ,

(6-4)

where f r are the distribution functions of the molecules reflected and incident on the surface at a point, rS , vc v 2 v n n , n is the unit normal vector to the surface at the given point, and f r is the Maxwellian distribution function given by:

fr

§ · m nr rS , t ¨ ¨ 2S kT r , t ¸¸ r S © ¹

32

§ · mv 2 exp ¨ . ¨ 2kT r , t ¸¸ r S © ¹

(6-5)

There are two parameters in this distribution function, nr and Tr , which correspond to the number density and temperature of some fictitious gas emitted by the surface. Also, in this form of the reflected distribution function, the surface element is assumed to be stationary. Now, consider the physical meaning of the coefficient, DW . The tangential momentum transferred to a unit surface element per second by incident molecules is given by:

PW

m

f vW v n dv ,

³

(6-6)

vn 0

and that carried away by the reflected molecules may be expressed as:

PW

m

³

f vW v n dv

1 DW PW

.

(6-7)

vn ! 0

From Eqs. (6-6) and (6-7), the following relation can be written:

DW PW

PW PW ;

0 d DW d 1 .

Therefore, DW gives the fraction of the tangential momentum of incident molecules transmitted to the surface by all molecules. This parameter is usually called the tangential momentum accommodation coefficient. The unknown parameters, nr and Tr , in Eq. (6-5) may be found if the impenetrability condition for the surface and the energy accommodation coefficient, DT , are introduced by means of the relations:

N N

0,

(6-8)

Chapter 6. Regimes of Rarefied Gas Flows

79

and:

DT

Q Q Q Qeq

;

0 d DT d 1 ,

(6-9)

where N # and Q # are the fluxes of the number of molecules and energy for the incident and reflected molecules across a unit element of the surface at the point rS per unit time. The value Qeq is the energy flux which would be carried away by the reflected molecules if the gas were in equilibrium with the surface, i.e., when Tr Tw , where Tw is the temperature of the surface at the point rS . The energy accommodation coefficient can be interpreted as the fraction of the equilibrium heat flux transmitted to the surface by the gas molecules. The details of the calculational technique associated with this formulation of the boundary problem are given in detail in Problem 6.3. The accommodation coefficients, DW and DT , may be found via experiment. The classical Millikan experiments [16], for instance, have shown that for most cases DW # 1 . This would seem to indicate that the model of diffuse reflection is generally the most appropriate model to employ in practice. The boundary conditions needed for a condensable gas are significantly different from those considered above. On a liquid surface, for instance, the evaporation (condensation) coefficient, D m , should be introduced for the gas (the vapor phase of the liquid comprising the surface) via the relationship:

Dm

N N N NS

,

(6-10)

where N S is the reflected flux of molecules which have the parameters of a saturated vapor at the applicable surface temperature. This coefficient should be interpreted to mean that a fraction of the incident flux of molecules at the surface, D m , is reflected from the surface having the parameters of a saturated vapor and that the remaining fraction of the incident molecules, 1 D m , leaves the surface without having undergone a phase transition. This latter fraction forms what is, in essence, a non-condensable gas for which the parameters may be specified by an impenetrability condition for the surface and an energy accommodation coefficient, DT , as was done above for the case of a purely non-condensable gas. One must keep in mind, however, that the use of an energy accommodation coefficient in such a scenario makes sense only for the non-

80

Analytical Methods for Problems of Molecular Transport

condensable part of the gas. The evaporating molecules must have the equilibrium parameters of a saturated vapor. To be consistent with the definition of the evaporation coefficient, the reflected distribution function can be expressed in the form:

f v, rS , t § · m D m nS Tw rS , t ¨¨ ¸¸ © 2S kTw rS , t ¹

32

§ · m 1 D m nr rS , t ¨ ¨ 2S kT r , t ¸¸ r S © ¹

§ · mv 2 exp ¨ ¨ 2kT r , t ¸¸ w S © ¹

32

(6-11)

§ · mv 2 exp ¨ , ¨ 2kT r , t ¸¸ r S © ¹

where nS Tw rS , t is the saturated vapor density at the given surface temperature Tw rS , t which, in the general case, is a function of the specific location on the surface, rS , and the time, t . The unknown parameters associated with the non-condensable fraction of the reflected molecules may be found using the standard impenetrability and energy accommodation conditions. A detailed description of this technique is given in Problem 6.4.

4.

THE BOUNDARY DISPERSION KERNEL.

The Maxwellian model of the boundary conditions is frequently used to analyze boundary value transport problems. This model gives a simple expression for the distribution function of molecules reflected from the surface of the wall. For a non-condensable gas, the diffusely reflected part of this distribution function contains two unknown parameters which may be defined by introducing the energy accommodation coefficient and the impenetrability condition for the surface. Note that the expression for the energy accommodation coefficient contains an additional unknown parameter connected with the number density of the molecules that are reflected and which are in thermal equilibrium with the wall which is assumed to be located at x 0 and to have a constant temperature, T0 . Hence, three additional integral conditions must be used to obtain the unknown parameters. For the linearized case, if evaporation (condensation) is absent at the wall, a general form of the boundary condition may be derived in which these additional relations are taken into account by the dispersion kernel, W c, c1 :

81

Chapter 6. Regimes of Rarefied Gas Flows

) f c,0

³ W c, c1 ) f c1 ,0 dc1

,

(6-12)

where f r c, x f 1 ) rf c, x and ) rf c, x are the full corrections to the distribution function which, in general, contain the Chapman-Enskog distributions. For the correction to the reflected distribution function, the Maxwellian boundary condition may be written as: 0

1 DW ) f cc,0 DW ª¬Q r c 2 32 W r º¼

) f c,0

,

(6-13)

where:

cc c 2 c n n Using Eq. (6-13) and Eqs. (6-8) and (6-9), one can obtain:

Qr

Wr

^ `^2

S 1 ³ cx exp c 2

1 2

1 DT 2 c 2 `) f dc

1 DT S 1 ³ cx 2 c 2 exp ^c 2 `) f dc

,

.

Then, Eq. (6-13) may be written as:

1 DW ) f cc,0 DW S 1 ³ c1 x exp ^c12 `

) f c,0

(6-14)

^

u 2 1 DT 2 c

2

2 c12

`

) f c1 ,0 dc1 ,

Allowing for Eq. (6-12), one can express the dispersion kernel in the form:

W c, c1

1 DW G ª¬c1 c 2 c n n º¼

^ `^2 1 D 2 c 2 c ` ,

S 1DW c1 x exp c12

2

T

where G x is the Dirac delta function [17].

2 1

(6-15)

82

Analytical Methods for Problems of Molecular Transport

Figure 6-1. The variation of the mean velocity of a gas near a moving wall. Here, uw is the velocity of the wall, 'u is the slip velocity, and u 0 is the real mean velocity at the wall.

5.

FEATURES OF THE BOUNDARY CONDITIONS FOR SMALL KNUDSEN NUMBER.

Consider the gas flow along a plane wall under conditions in which the mean velocity has only a tangential component that is a linear function of the normal coordinate beyond the wall. This linear dependence may be obtained from the solution of both the Boltzmann equation and Navier-Stokes equations. Since the Navier-Stokes distribution function is a solution of the Boltzmann equation only at large distances from the wall, the gas state in the Knudsen layer must be investigated in order to establish the boundary conditions for Navier-Stokes equations [18-26]. Let the solid curve (Fig. 6-1) be the variation of the mean velocity of a gas that has been obtained by the solution of the Boltzmann equation. Then, the linear extrapolation of the asymptotic solution (outside of the Knudsen layer) of this equation gives a fictitious value of mean velocity at the wall, uw 'u . If one uses this fictitious velocity, uw 'u , on the wall as a boundary condition for the Navier-Stokes equations, then the same velocity profile may be obtained as from the solution of the Boltzmann equation outside of the Knudsen layer. The difference between the fictitious velocity and the actual wall velocity is called the slip velocity. The most reliable expression

Chapter 6. Regimes of Rarefied Gas Flows

83

for the slip velocity has been obtained by Ivchenko, Loyalka and Tompson [27]:

'u

§ wu · cm OP ¨ ¸ , © wx ¹f

(6-16a)

where:

cm

2 DW

DW

5 8

ª 0.6690 0.1775 DW º », «¬ 0.7549 0.09714 DW »¼

S«

(6-16b)

is the isothermal-creep coefficient, the molecules have been assumed to be 205 rigid spheres, and OP 202 O for the second-order Chapman-Enskog solution. In the very important case when DW 1 , one obtains:

'u

wu · ¸ . © wx ¹f

1.1006 OP §¨

(6-17)

The temperature-jump may be introduced in the same way (Fig. 6-2). The most reliable analytical expression for the temperature-jump has been obtained by Loyalka [18]. Loyalka’s formula, if the tangential momentum accommodation coefficient is taken into account, can be expressed as [28]:

'T

§ wT · cT ON ¨ ¸ , © wx ¹f

(6-18a)

where:

cT

75 128

S

2 DW DT ª DW DT 2 DW § 52 H T 1 · º », «1 DW DT ¬« 2 DW DT ¨© 25 S 2 ¸¹ »¼

(6-18b)

45 is the temperature-jump coefficient, ON 44 O , and H T 0.9378 for rigidsphere molecules if the second-order Chapman-Enskog solution is employed. Consider a gas having a tangential temperature gradient located over a non-uniform heated plane surface (Fig. 6-3). Under these conditions, the gas has a mean velocity in the direction of the temperature gradient, i.e. the gas

84

Analytical Methods for Problems of Molecular Transport

Figure 6-2. The variation of the gas temperature near a wall. Here, Tw is the wall temperature, 'T is the temperature-jump, and T 0 is the real temperature of the gas at the wall

slips along the wall. The gas velocity beyond the wall is the thermal-creep velocity or the thermal-creep velocity. The most accurate analytical expression for the thermal-creep velocity has been obtained by Ivchenko, Loyalka and Tompson [28,29]:

usl

cTslQ

w ln T , wy

(6-19)

where:

cTsl

3 2

0.4354 0.2179 DW 0.8518 0.1096 DW

,

is the thermal-creep coefficient, Q ON OP Q , Q P U is the kinematic viscosity, and molecules have been assumed to be rigid spheres. For two specific cases of interest, this coefficient may be written as:

cTsl

1.0193 ; DW ® ¯0.7667 ; DW

1, 0 .

(6-20)

85

Chapter 6. Regimes of Rarefied Gas Flows

Figure 6-3. The thermal-creep geometry.

The slip boundary conditions may be obtained only from the solution of the Boltzmann equation with microscopic boundary conditions for the distribution function. These problems will be considered in Chapter 8.

PROBLEMS 6.1. Using the Maxwellian boundary model, determine the form of the reflected distribution function if the boundary surface is moving and uc is the wall velocity. Solution: The mean velocity of the diffusely reflected molecules is equal to uc . Therefore, the reflected distribution function may be written as:

f

v, rS

§ · m DW nr rS ¨¨ ¸¸ © 2S kTr rS ¹

32

§ m v uc 2 exp ¨ ¨ 2kTr rS ©

1 DW f vx u cx , v y , vz , rS

· ¸ ¸ ¹

,

where x is a normal coordinate.

6.2. Determine the form of the dispersion kernel which is specified by:

f v, rS

³

W v1 , v f v1 , rS dv1 ,

v1 n 0

for the Maxwellian boundary model. Solution: The distribution function of reflected molecules is given by Eqs. (6-4) and (6-5). The condition of wall impenetrability gives:

86

Analytical Methods for Problems of Molecular Transport 12

§ · m nr rS 2 S ¨ v1 n f v1 , rS dv1 . ³ ¨ 2kT r ¸¸ r S ¹ © v1 n 0 Using this expression, one can obtain:

W v1 , v

D W 2S

2

§ m · § mv 2 · ¨¨ ¸¸ exp ¨¨ ¸¸ v1 n © kTr rS ¹ © 2kTr rS ¹ 1 DW G ª¬ v1 v 2 v n n º¼ .

6.3. Write down the full system of boundary conditions at the solid impenetrable wall if the wall temperature is T0 . Solution: The number density, nr , and the temperature, Tr , in Eqs. (6-4) and (6-5) are defined by N N 0 and:

DT

Q Q Q Qeq

,

where N r and Q r are normal components of the relative fluxes. The flux, , is determined by means of the reflected distribution function in which Qeq nr neq and Tr T0 . The unknown equilibrium number density, neq , may be found from N eq N 0 .

6.4. At a liquid surface, write down the boundary conditions for reflected molecules of the same liquid vapor if the temperature of the surface is slightly different from the equilibrium temperature T0 , Tw T0 1 W w . Solution: The reflected distribution function may be presented in the form:

f

v,0

§ m · D m nS Tw ¨ ¸ © 2S kTw ¹

32

§ mv 2 · exp ¨ ¸ © 2kTw ¹

§ m · 1 D m nr ¨ ¸ © 2S kTr ¹

32

§ mv 2 · exp ¨ ¸ , © 2kTr ¹

where the first term describes molecules evaporated from the liquid surface while the second term is related to molecules reflected from the surface without a phase transition. It should be noted that, for this latter group of

87

Chapter 6. Regimes of Rarefied Gas Flows

molecules, it is appropriate to treat the liquid surface as if it were impenetrable and to employ an appropriate surface condition. The unknown parameters nr and Tr , may be determined from the relations 1 D m N N r 0 and:

DT

1 D m Q Qr 1 D m Q Qreq

,

where Qreq contains an additional unknown parameter related to the density of a fictitious gas reaching equilibrium with the surface if Tr Tw . This parameter may be found from 1 D m N N req 0 . The fluxes, N r , N req , Qr , and Qreq , are calculated by using the second term of the reflected distribution function. The number density of the saturated vapor at the temperature, Tw , may be found by using the Clausius-Clapeyron equation [30] which yields nS Tw nS T0 ª¬1 ] 1 W w º¼ where ] h fg kT0 and h fg is the latent heat of condensation for a vapor molecule.

6.5. Generalize Eq. (6-14) to obtain an expression for the distribution function of the rebounding molecules for a case when the equilibrium temperature of the wall is given by Tw T0 'T where 'T T0 . Solution: The boundary condition for a correction to the distribution function may be written as:

) f c,0

1 DW ) f cx , c y , cz ,0 DW ª¬Q r c 2 32 W r º¼

,

where the two boundary quantities, Q r and W r , may be found by introducing the energy accommodation coefficient, DT , together with the condition of impenetrability of the wall for molecules of the gas. This yields:

) f c,0

1 DW ) f cx , c y , cz ,0 ª 2G · º 2G ½° §G ° DW ® 2 c 2 « DT W w 1 DT ¨ 2 1 ¸ » 1 ¾ , S ¹ ¼ S ¿° ©S ¬ ¯°

where:

Ww G1

'T , T0 2 ³) f c,0 cx exp c dc

,

G2

³) f c,0 cx c

2

exp c 2 dc .

88

Analytical Methods for Problems of Molecular Transport

6.6. Generalize Eq. (6-14) to obtain an expression for the distribution function of the rebounding molecules for a case when the evaporation/condensation processes at the wall are taken into account and the equilibrium temperature of the surface is given by Tw T0 'T where 'T T0 and T0 is the equilibrium vapor temperature. Solution:

) f c,0 D m ªQ w c 2 32 W w º 1 D m ªQ r c 2 32 W r º , ¬

¼

¬

¼

where Q w is the correction to the number density of the saturated vapor at the temperature, Tw T0 1 W w , and Q r and W r , may be found by introducing the energy accommodation coefficient, DT , and using the impenetrability condition for molecules reflected without a phase transition. This yields:

) f c,0 D m ªQ w c 2 32 W w º ¬

¼

ª 2G · º 2G °½ §G ° 1 D m ® 2 c 2 « DT W w 1 DT ¨ 2 1 ¸ » 1 ¾ , S ¹ ¼ S °¿ ©S °¯ ¬

where:

Ww G1

'T , T0 2 ³) f c,0 cx exp c dc

,

G2

³) f c,0 cx c

2

exp c 2 dc .

The correction, Q w , may be found by using the Clausius-Clapeyron equation [30] which yields nS Tw nS T0 ª¬1 ] 1 W w º¼ where ] h fg kT0 and h fg is the latent heat of condensation (evaporation) for a vapor molecule.

REFERENCES 1. Keller, J.B., “On the Solution of the Boltzmann Equation for Rarefied Gases,” Comm. Pure App. Math. 1(3), 275-284 (1948). 2. Rose, M.H., “Drag on an Object in Nearly-Free-molecular Flow,” Phys. Fluids 7(8), 12621269 (1964).

Chapter 6. Regimes of Rarefied Gas Flows

89

3. Liu, V.C., Pang, S.C., and Jew, H., “Sphere Drag in Flows of Almost Free Molecules,” Phys. Fluids 8(5), 788-796 (1964). 4. Brock, J.R., “The Thermal Force in the Transition Region,” J. Colloid Interface Sci. 23(3), 448-452 (1967). 5. Brock, J.R., “Highly Non-equilibrium Evaporation of Moving Particles in the Transition Region of Knudsen Number,” J. Colloid Interface Sci. 24, 344-351 (1967). 6. Brock, J.R., “The Diffusion Force in the Transition Region of Knudsen Number,” J. Colloid Interface Sci. 27(1), 95-100 (1968). 7. Ivchenko, I.N. and Yalamov, Yu.I., “On the Thermophoresis of Aerosol Particles in the Almost-Free-molecular Regime,” Izv. AN SSSR, M. Zh. G. (3), 3 (1970). 8. Kelly, G.E. and Sengers, J.V., “Droplet Growth in a Dilute Vapor,” J. Chem. Phys. 61(7), 2800-2807 (1974). 9. Barrett, J.C. and Shizgal, B., “Condensation and Evaporation of a Spherical Droplet in the Near Free Molecule Regime,” in Rarefied Gas Dynamics: Physical Phenomena, Progress in Astronautics and Aeronautics Series, V. 117, edited by Muntz, E.P., Weaver, D.P., and Campbell, P.H. (AIAA, Washington, D.C., 1989) pp. 447-459. 10. Gross, E.P., Jackson, E.A., and Ziering, S., “Boundary Value Problems in Kinetic Theory of Gases,” Annals of Physics 1(2), 141-167 (1957). 11. Cercignani, C., Theory and Application of the Boltzmann Equation (Scottish Academic Press, Edinburgh, U.K., 1975). 12. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, NY, 1969). 13. Barantsev, R.G., Rarefied Gas Interactions with Streamlined Surfaces (Nauka, Moscow, 1975). 14. Goodman, F.O. and Wachman, H.Y., Dynamics of Gas-Surface Scattering (Academic Press, New York, 1976). 15. Kušþer, I., “Phenomenology of Gas-Surface Accommodation,” in Proceedings of the Ninth International Symposium on Rarefied Gas Dynamics, edited by Becker, M. and Fiebig, M. (DFVLR Press, Pozz-Wahn, 1974). Vol. 2, p. E.1-1. 16. Millikan, R.A., “Coefficients of Slip in Gases and the Law of Reflection from the Surfaces of Solids and Liquids,” Phys. Rev. 21(1), 217-238 (1923). 17. Dirac, P.A.M., The Principles of Quantum Mechanics, 4-th edition revised (Clarendon Press, Oxford, reprinted 1993). 18. Ivchenko, I.N. and Yalamov, Yu.I., “The Hydrodynamic Method of Calculation of the Thermophoresis of Aerosol Particles,” J. Phys. Chem. (Russia) 45(3), 577-582 (1971). 19. Loyalka, S.K., “Momentum and Temperature-Slip Coefficients with Arbitrary Accommodation at the Surface,” J. Chem. Phys. 48(12), 5432-5436 (1968). 20. Hidy, G.M. and Brock, J.R., Topics in Current Aerosol Research, part 2 (Pergamon Press, 1972). 21. Derjaguin, B.V., Ivchenko, I.N., and Yalamov, Yu.I., “About Construction of Solutions of the Boltzmann Equation in the Knudsen Layer,” Izv. AN SSSR, M. Zh. G. (4), 167-172 (1968). 22. Derjaguin, B.V. and Yalamov, Yu.I., “The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and Their Experimental Testing,” in Topics in Current Aerosol Research, part 2, edited by Hidy, G.M. and Brock, J.R. (Pergamon Press, 1972). 23. Loyalka, S.K. and Lang, H., “On Variational Principles in the Kinetic Theory,” in Proceedings of the Seventh International Symposium on Rarefied Gas Dynamics (Editrice Tecnico Scientifica, Pisa, Italy, 1970).

90

Analytical Methods for Problems of Molecular Transport

24. Loyalka, S.K., “Slip and Jump Coefficients for Rarefied Gas Flows: Variational Results for Lennard-Jones and n(r)-6 Potentials,” Physica A163, 813-821 (1990). 25. Rolduguin, V.I., Application of the Non-Equilibrium Thermodynamics Method in Boundary Problems of the Kinetic Theory of Gases (M.S. Thesis, Moscow, 1979). 26. Savkov, S.V., The Slip-Flow Boundary Conditions for the Non-Uniform Binary Gas Mixture and an Application of Them in Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 27. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “The Precision of Boundary Models in the Gas Slip Problem,” High Temp. 31(1), 127-129 (1993). 28. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “On the Use of Conservation Laws in Plane Slip Problems,” Teplofizika Vysokikh Temperatur (in Russian), 33(1), 66-72 (1995). 29. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “On One Boundary Model for the Thermal Creep Problem,” Fluid Dynamics 28(6), 876-888 (1993). 30. Landau, L.D. and Lifshitz, E.M., Statistical Physics (Pergamon, London, Addison Wesley, Reading, Mass., 1958).

Chapter 7 THE FREE-MOLECULAR REGIME

1.

THE FREE-MOLECULAR DISTRIBUTION FUNCTION.

The theoretical treatment of gas flow problems for extremely high Knudsen numbers (the free-molecular regime) is very different from that typically encountered in conjunction with the hydrodynamics of media having the higher number densities most people are used to working with. The theoretical frameworks encountered at these extremes are connected via various fundamental relationships between the Kinetic Theory of Gases and the mechanics of media that exhibit continuum type behaviors (the continuum regime). The continuum equations may be obtained from the Boltzmann equation if one uses the summational invariants of encounters as molecular properties with which to construct a system of moment equations. This moment system, however, is indeterminate unless one knows the relationships between the basic hydrodynamic values and the stress tensor or the thermal flux vector. These relationships may be derived by using the Chapman-Enskog method of solution of the Boltzmann equation. It is very important in the continuum regime that such gas characteristics as the viscosity and thermal conductivity coefficients be determined independently of the boundary conditions. The boundary problems for high Knudsen numbers are quite different. The kinetic analysis may only be used for the description of gas flows. The general solution of the Boltzmann equation requires usage of appropriate boundary conditions for the distribution function itself which ultimately results in a dependence of the hydrodynamic functions on the boundary parameters of the specific problem under consideration. Hence, in this

92

Analytical Methods for Problems of Molecular Transport

regime there are no general relationships between the various macroscopic quantities. The free-molecular regime will be examined in this chapter. A general characteristic of gas flows in this regime is that the presence of a body does not effectively disturb the distribution function of the impinging molecules since the collisions between molecules being reflected and those coming from infinity occur at distances on the order of O from the surface of the body. Since O R , where R is the typical dimension of the body, the influence of these collisions can be neglected. Owing to this simplification, the theory describing free-molecular flows has been sufficiently developed that several useful solutions to applied transport problems have been obtained [1-6]. A rigorous theory of transport problems must start with the Boltzmann equation which, for stationary conditions, when Kn o f , may be written as:

v

wf wr

vx

wf wf wf vy vz wx wy wz

0 ,

or:

0 .

(7-1)

Eq. (7-1) is a linear partial differential equation of the first-order. The auxiliary system of ordinary differential equations, usually called the characteristic or Lagrange system, is:

dx vx

dy vy

dz vz

df 0

dt ,

(7-2)

where t is an auxiliary parameter used to ease the solution of this system of equations. The characteristics of these equations are the straight lines specified by:

r r0 v t t0 ,

(7-3)

where t denotes the value of the parameter on the characteristic (i.e. t is not time). These characteristics may be interpreted to be the molecular trajectories associated with this problem [7]. Eq. (7-2) means that:

Chapter 7. The Free-Molecular Regime

93

Figure 7-1. The cone of influence for a sphere.

df dt

0 .

(7-4)

which implies that f is constant along the molecular trajectories. For the molecular trajectories that start at infinity and at the surface of the body the distribution functions are, therefore, equal to the values of the distribution

94

Analytical Methods for Problems of Molecular Transport

functions at infinity in the undisturbed flow and at the point, P rS , on the surface. These functions are not the same and, hence, the free-molecular distribution function is discontinuous in velocity space. Consider the form of the distribution function at the point, M r , in various velocity domains when a spherical model of the particle is assumed. At each spatial point, M r , there exists a cone of influence, the surface of which is formed by the extensions of the tangents to the surface of the particle passing through the point M r (see Fig. 7-1). This conical surface in the velocity space is described by F F 0 , F 0 arcsin R r arcsin 1 r where R is the radius of the particle and F is the angle between vectors c and r . In the velocity space, the surface of the cone of influence is the surface of discontinuity for the distribution function. This surface divides the molecules into two groups with distribution functions f1 and f 2 ( f 2 describes the molecules for which vr ! vr where vr v cos F 0 , and f1 describes the remaining molecules). The free-molecular distribution function in the full velocity space may be expressed as [8]:

f

1 2

f1 f 2 12 f 2 f1 sign vr vr

.

(7-5)

This relationship is a generalization of expression previously proposed for the planar geometry [9,10]. On the particle surface, vr 0 , and the distribution functions of the reflected and incident molecules may be written as:

° f v, Rn f v, Rn ® °¯ f v, Rn

f2

;

vr ! 0 ,

f1

;

vr 0 ,

where n is a unit vector normal to the surface element and pointing into the gas. These distribution functions are connected by the boundary condition given in Eq. (6-4). Having derived the distribution function one may calculate the flux of any molecular property across the particle surface. For example, the force, F , on the particle is given by integrating the net momentum transferred to the particle by the gas molecules over the particle surface, S , per unit time:

F

³ dS ¦ ³ mv v n f r v, rS dv , r r

(7-6)

95

Chapter 7. The Free-Molecular Regime

Figure 7-2. The geometry of the sphere drag problem.

where dS is the area of the surface element, f and f indicate the distribution functions at the particle surface for both incident v n 0 and reflected v n ! 0 molecules, respectively, and rS is the position vector of an arbitrary point on the surface. For evaporation (condensation) problems one should know the flux of the number of molecules across the particle surface, S , per unit time. This flux may be given by:

N

r ³ dS ¦r ³ v n f v, rS dv

,

(7-7)

r

where N is positive or negative depending on whether the process is connected with evaporating or condensing particles.

96

Analytical Methods for Problems of Molecular Transport

2.

THE FORCE ON A PARTICLE IN A UNIFORM GAS FLOW.

Consider a small spherical particle R O in a uniform gas flow for which the velocity at infinity is a constant value, u (Fig. 7-2). A coordinate system is introduced in which the polar axis has the same direction as the vector, u . Then, u ur e r uT eT where ur u cos T and uT u sin T . From the symmetry of this problem, it is clear that the force has only one component in the direction of the vector u . This component may be given by:

Fu

S ° ½° 2S R 2 m ³ sin T ®¦ ³ vr vr cos T vT sin T f r dv ¾ dT , (7-8) 0 ¯° r r ¿°

where the integration over all values of the azimuthal angle has been performed. 12 Suppose that u 2kT m . In this case of low velocity of the undisturbed flow, the following approximation for the distribution function of the incident molecules may be used:

f

0 f 1 2u c

^

`

0 f 1 2u ª¬cr cos T cT sin T º¼ ,

(7-9)

where: 12

u

12

§ m · § m · ¨ 2kT ¸ u , c ¨ 2kT ¸ © ¹ © ¹

0 v , f

§ m · n0 ¨ ¸ © 2S kT ¹

32

exp c 2 .

For the distribution function of the reflected molecules, one can use the Maxwellian model of the boundary conditions given by Eq. (6-4) in which the density of molecules reflected diffusely may be written as: nr rS n0 1 Q rS , where Q rS is proportional to the small parameter u which is a linearization parameter. The linearized boundary condition may be expressed in the form:

97

Chapter 7. The Free-Molecular Regime

f v, rS

1 DW f cr , cT , cI , rS § m · DW n0 ¨ ¸ © 2S kT ¹

32

exp c 2 1 Q rS .

(7-10)

The unknown function Q rS may be found from the condition that there is no evaporation or condensation of molecules at the surface element surrounding the point, rS (i.e. the particle surface is impenetrable for the gas molecules). This condition may be written as:

n ³ cf dv n ³ cf dv 0 .

(7-11)

Having performed the integration in Eq. (7-11), one obtains:

Q rS S u n .

(7-12)

Substituting Eqs. (7-9), (7-10), and (7-12) into Eq. (7-8), the following expression may be written:

Fu

2S R

2

U § 2kT ·

S

¨ ¸ u sin T S 3 2 © m ¹ ³0

° u ®2 1 DW ³ cr ª¬cr2 cos 2 T cT2 sin 2 T º¼ exp c 2 dc °¯

DW S ³ cr ª¬ cr cos 2 T cT sin T cos T º¼ exp c 2 dc

½° 2 2 ³ cr ª¬cr cos T cT sin T º¼ exp c 2 dc ¾ dT . °¿

After some straightforward integrations, one obtains:

F

8 3

12

R 2 n0 2S mkT

1 18 DW S u .

(7-13)

The same expression for the free-molecular force has been obtained by Waldmann [11].

98

Analytical Methods for Problems of Molecular Transport

3.

CALCULATION OF MACROSCOPIC VALUES IN THE FREE-MOLECULAR REGIME.

Each macroscopic quantity may be expressed as the mean value of a molecular property, I v, r , by the relation:

³ I fdv

nI

.

(7-14)

For the free-molecular regime, this integral contains the discontinuous distribution function given by Eq. (7-5). Substituting Eq. (7-5) into Eq. (714), one obtains:

nI

³

I f1dv

1 2

³ I f 2 f1 dv

.

(7-15)

2

where, in the first integral, the integration extends over all values of the molecular velocity while, in the second integral, the integration extends only over that part of the velocity range for which vr ! vr (region 2 in the velocity space). For spherical geometries it is convenient to use spherical coordinates in the velocity space to perform the integration over region 2. These coordinates are given by:

cr

c cos F ,

cT

c sin F sin E ,

cI

c sin F cos E .

The integration over region 2 may be represented as:

³

2

dc

f

2S

w

0

0

0

2 ³ c dc ³ d E ³ sin F d F ,

(7-16)

where w arcsin 1 r and r r R . The main subtleties here are connected with the integration of the diffusely and specularly reflected parts of the distribution function. These parts always contain the scalar product of n S n rS and the typical flow vector, u . To complete the integration over region 2 one must find the connection between the unit vector n S and the molecular velocity vector represented by the characteristic PM (see Fig. 7-1). The vector n S can be expressed as:

99

Chapter 7. The Free-Molecular Regime

nS

r c PM . R c R

The value PM may be found from the triangle OPM (see Fig. 7-1) by solution of the quadratic equation:

PM 2 2r cos F PM r 2 R 2

0 ,

that may be written as: PM

r cos F r R 1

r 2 sin 2 F . R2

In this formula, only the minus sign may be used in order that, as r o f and F o 0 , then PM r R . Then, the expression for n S n S r , c may be written as:

n S r , c

º r c ª r r 2 « cos F 1 2 sin 2 F » . R c «R R »¼ ¬

(7-17)

Now, as an example, the integration required for the problem of freemolecular flow past a sphere that was discussed in Section 7.2 will be considered in order to calculate the number density of the flow. The distribution function may be written as:

f

f1 ° ° ® f2 ° ° ¯

1 D f 1 2cc u D f ª1 S n u º ; ¬ ¼ 0 f 1 2cr ur 2cT uT ; 0

cr cr ,

(7-18)

W

0

W

S

cr !

cr

,

where cc c 2n S n S c and for integration of the reflected distribution function one should use Eq. (7-17) for n S . To simplify what would otherwise be a very complicated set of integrations the conclusion of this calculation will be shown for the particular case when molecules are assumed to undergo only diffuse reflection at the surface ( DW 1 ). Then, the number density is expressed by:

100

Analytical Methods for Problems of Molecular Transport

n

f1dv

³

1 2

³ f 2 f1 dv

,

(7-19)

2

where the integrand of the second term has the form:

0 f ª S n S r, c u 2cr ur 2cT uT º . ¬ ¼

f 2 f1

After integration of the first term in Eq. (7-19) one obtains:

i1

f1dv

³

n0 .

1 2

The second integral may be written as:

i2

f

n0

2S

w

0

0

2 2 ³ c exp c dc ³ d E ³ sin F

S3 2

0

^ S

ur

ª r sin ¬«

2

F cos F

1 r 2 sin 2 F º» ¼

(7-20)

`

2ur c cos F d F . Basic integration gives:

i2

n0ur F r ,

(7-21)

where:

F r

1 2

^

`

S r 1 a 13 r 1 a 3 r 2 ª¬ 13 2S 1 º¼ ,

and a 1 r 2 . The vanishing integrals have been eliminated from Eq. (720). Now, the number density may be expressed in the form:

n r n0 ª¬1 u F r cos T º¼ .

(7-22)

For the limiting case when r o f the number density n r o n0 which is equivalent to an undisturbed flow.

101

Chapter 7. The Free-Molecular Regime

4.

THERMOPHORESIS OF PARTICLES IN THE FREE-MOLECULAR REGIME.

It has been observed that a temperature gradient in a gas causes small particles suspended in the gas to migrate in direction of decreasing temperature. The force arising from the temperature gradient causing the migration of the particles is commonly termed the thermal force and the phenomenon is usually known as thermophoresis. Previously, several theoretical descriptions of the thermophoresis of particles in the free-molecular regime, Kn o f , have been given [11-16]. In these papers, calculations are made of the thermal force from the net momentum transferred by gas molecules striking and reflecting from the particle surface. The physical model that is used is a single spherical particle of radius, R , in an infinite gas with a temperature gradient, T , which is constant at large distances from the particle. It is assumed that the gas as a whole is in mechanical equilibrium so that at large distances from the particle there is no pressure gradient. Introduce a spherical coordinate system r ,T , I with its origin at the center of the stationary spherical particle and a polar axis, x , in the direction of the vector, T f (see Fig. 7-3). Then, the temperature distribution at large distances from the particle has the form:

T

T r T0 T x r cos T T0 1 q r ,

(7-23)

where q (T )f T0 . The condition, O ln T 1 , is assumed to be satisfied at large distances from the particle. Then, the distribution function is the ChapmanEnskog distribution, that is given by:

f

^

` ,

f eq 1 a1 q c S3 2 c 2 1

1

(7-24)

where a1 15 S O for rigid-sphere molecules and f eq is the local 16 Maxwellian distribution that may be written as: 1

f eq

§ · m n r ¨ ¨ 2S kT r ¸¸ © ¹

32

§ mv 2 · exp ¨ . ¨ 2kT r ¸¸ © ¹

(7-25)

102

Analytical Methods for Problems of Molecular Transport

Figure 7-3. The geometry of the thermal force problem.

From the condition, n0 kT0 represented as:

n r

n0T0 T r

n r kT r , n r in Eq. (7-25) may be

n0 . 1 q r

For r ~ O the term q r 1 and Eq. (7-25) may be represented as a power series in q r . Considering only the linear term in this expansion, Eq. (725) may be written as:

f eq where:

0 f ª1 c 2 52 q r º , ¬ ¼

(7-26)

103

Chapter 7. The Free-Molecular Regime 0 f

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

§ mv 2 · exp ¨ ¸ . © 2kT0 ¹

From the symmetry of the problem under consideration, it is clear that the thermal force has only one component in the direction of the vector, q . This component may be given by: S

Fq

2S R 2 m ³ sin T dT 0

(7-27)

° °½ u ®¦ ³ vr vr cos T vT sin T f r dv ¾ , °¯ r r °¿

where f r are the distribution functions of the reflected and incident molecules, at the surface of the particle, respectively. Using Eqs. (7-24) and (7-26), and the boundary condition given by Eq. (6-4), one can obtain the following expression for the distribution function at the surface of the particle:

f

f ° ° °° f ® ° ° ° °¯

1 0 1 f ª1 c 2 52 q n R a1 q c S3 2 c 2 º ¬ ¼ 0 1 DW f ª¬1 c 2 52 q n R

1 1 a1 g cr cos T cT sin T S3 2 c 2 º ¼

(7-28)

0 DW f ª1 Q r rS c 2 32 W r rS º . ¬ ¼

The unknown functions, Q r rS and W r rS , are determined from the two boundary conditions given by:

° ½° n ® ³ vf dv ³ vf dv ¾ 0 , ¯° ¿°

(7-29)

and:

§ wTp · ° °½ n ® ³ 12 mv 2 vf dv ³ 12 mv 2 vf dv ¾ N p ¨ ¸ °¯ °¿ © wr ¹ r

, R

(7-30)

104

Analytical Methods for Problems of Molecular Transport

where Tp is the temperature of the particle and N p is the thermal conductivity coefficient of the particle. The distribution of temperature inside the particle, in spherical coordinates, may be expressed in the form:

Tp

f

T0 ¦ Cn r n Pn cos T .

(7-31)

n 1

Here, Cn are the coefficients of the Legendre polynomials, Pn cos T . The left-hand-side of Eq. (7-30) contains the terms that are proportional only to cos T , and therefore there is only one term in this sum and the temperature distribution inside the particle may be written as:

Tp

T0 1 W p r

§ C r cos T · T0 ¨¨1 1 ¸¸ , T0 © ¹

(7-32)

where C1 is an unknown constant. Before using the boundary conditions, one can transform the distribution function given by Eq. (7-28). Small terms of O > R O @ may be neglected in this regime. Eliminating these small terms, the following expression for the distribution function at the surface of the particle is obtained:

f

f ° ° °f ® ° ° ° ¯

1 0 1 f ª1 a1 q c S3 2 c 2 º , ¬ ¼

1 DW f 0

(7-33)

1 1 u ª1 a1 g cr cos T cT sin T S3 2 c 2 º ¬ ¼ 0 ª 2 3 1 Q r rS c 2 W r rS º . DW f ¬ ¼

The parameters of diffusely reflected molecules are assumed to deviate only slightly from the equilibrium state. The boundary condition, Eq. (7-29), gives:

Q r 12 W r

(7-34)

0 .

Eq. (7-29) is not altered in the case when the gas molecules are reflected from the surface of the particle under equilibrium conditions W r W p . In this case one has:

105

Chapter 7. The Free-Molecular Regime

Q p 12 W p

(7-35)

0 .

After performing the integrations in Eq. (7-30), one obtains:

Q r 32 W r 85 S a11 q cos T

N p C1 cos T , DW SD

(7-36)

where: 1 2

D

§ 2kT0 · ¸ © m ¹

S 3 2 n0 m ¨

32

.

Using the definition of DT in Eq. (7-30) yields:

DT Q p 32 W p 85 S a11 q cos T

N p C1 cos T , DW SD

(7-37)

where:

Wp

C1 R cos T . T0

From Eqs. (7-34)-(7-37) it follows that:

ª ¬

W r rS 85 S a11 q n «1 DT

1 º , 1 H »¼

(7-38)

where:

H

DW DT N g R , 75SN p O

and N g is the thermal conductivity coefficient for the gas. Molecules are assumed to be rigid spheres. In the free-molecular regime the quantity H may be neglected. Then the following expressions specify the boundary parameters, Q r rS and W r rS :

106

Analytical Methods for Problems of Molecular Transport

Q r rS

5 16

S 1 DT a11 q n ,

(7-39)

and:

W r rS 85 S 1 DT a11 q n .

(7-40)

Using these relations, one may transform Eq. (7-27) into the form:

4n0 kT0 R 2 a1 q S 1

Fq

S f

³ sin T dT 0

° 1 u ®1 DW ³ cr cr2 cos 2 T cT2 sin 2 T S3 2 c 2 exp c 2 dc °¯ 0

165 S DW 1 DT cos 2 T

(7-41)

f ªf º u « ³ cr2 exp c 2 dc 2 ³ cr2 c 2 32 exp c 2 dc » «¬ 0 »¼ 0 f °½ 1 ³ cr cr2 cos 2 T cT2 sin 2 T S3 2 c 2 exp c 2 dc ¾ . °¿ 0

Having performed the necessary integrations, the expression for the thermal force may be written as:

F

54 S R 2 n0 k O T f ª¬1 DW 1 D T 325 S º¼ .

(7-42)

This expression contains a new term, namely DW 1 DT 325 S , as compared with the well known results [11-13]. This term for the particular case when DW 1 has been obtained by Ivchenko and Yalamov [13] and later by Talbot et al. [14].

5.

CONDENSATION ON A SPHERICAL DROPLET.

Consider the growth of a single droplet in an infinite expanse of gas that consists only of the gas phase of the liquid which forms the droplet. At large distances from the droplet the gas is assumed to have a constant temperature and to be supersaturated. The distribution function for these conditions has the form:

107

Chapter 7. The Free-Molecular Regime 0 f 1 Q 0 ,

f

(7-43)

where:

f

0

§ m · n0 T0 ¨ ¸ © 2S kT0 ¹

32

§ mv 2 · exp ¨ ¸ , © 2kT0 ¹

n0 T0 is the number density of the saturated vapor at temperature, T0 , and

Q 0 is the supersaturation. The distribution function on the surface of the

droplet is assumed to be the Maxwellian function for the reflected molecules with the unknown number density and temperature. Let these parameters be deviated only slightly from the equilibrium state, and therefore one may use the linearized form of the distribution function. The distribution functions of incident and reflected molecules at the droplet may be expressed in the form:

f

f ° °f ° ® ° ° ° ¯

0 f 1 Q 0 , 0 f ª¬1 ) c, rS º¼ D m f 0 ª1 Q S c 2 32 W S º ¬ ¼

(7-44)

0 1 D m f ª1 Q r c 2 32 W r º . ¬ ¼

where Q S is the correction to the saturated vapor density at the temperature, TS T0 1 W S . The unknown parameters, Q r and W r , in the reflected distribution function may be found from the energy accommodation and impenetrability conditions which have sense only for those molecules reflected from the particle surface without a phase transition. The perfect representation of the reflected distribution function was obtained previously from Problem 6.6 and is expressible in the form:

) c, rS D m ªQ S c 2 32 W S º ¬

¼

ª 2G · º 2G ½° ° §G 1 D m ® 2 c 2 « DT W S 1 DT ¨ 2 1 ¸ » 1 ¾ , S ¹ ¼ S ¿° ©S ¬ ¯°

(7-45)

where:

G1

12 SQ 0 and G 2

SQ 0

(7-46)

108

Analytical Methods for Problems of Molecular Transport

The correction to the equilibrium number density may be represented by:

QS

2J m aW S , U1kT0 R

(7-47)

where J is the surface tension coefficient, U1 is the density of the drop, R is the drop radius, a 9 1 , 9 h fg kT0 , and h fg is the latent heat of condensation per molecule. This equation is the sum of the linearized Kelvin and Clausius-Clapeyron terms [17,18]. The temperature, W S , may be found from the equation of energy conservation, which may be written as: Qr

(7-48)

h fg N r ,

or:

DT Qr , S

(7-49)

h fg D m N r , S .

Eq. (7-48) expresses the fact that, for stationary conditions, the heat of condensation must be transferred to the ambient gas by thermal conductivity. It also shows that the growth rate of the particle is controlled by the rate at which the liberated heat of condensation can be dissipated into the environment. Eq. (7-49) gives:

WS

D mQ 0c

9 2 , 2D m a D m9 a 12 2DT 1 D m 3 2

(7-50)

where:

Q 0c

2J m Q . U1kT0 R 0

The total flux of the number of molecules across the droplet surface may be written as:

N

4S R 2 N r

4S R 2D m N r , S 12

4S R 2D m

n0 § 2kT0 · ¨ ¸ 2 S © m ¹

ªQ 0c a 12 W S º . ¬ ¼

(7-51)

109

Chapter 7. The Free-Molecular Regime

Substituting W S in Eq. (7-51), one obtains: 12

N

§ 8S kT0 · ¨ m ¸ © ¹

R 2 n0Q 0c[ D m ,DT , 9 ,

(7-52)

where:

[ D m ,DT , 9 4D m

D m DT 1 D m

2

D m 29 39 2 4D T 1 D m

.

The growth rate of the drop is determined by the equation:

dR dt

mN . 4S R 2 U1

(7-53)

Integration of this equation gives the dependence, R form:

R t , that has the

12

t

U1 § 2S · ° b § Q 0 R b · °½ ¨ ¸ ® R R0 ln ¨ ¸¾ , Q 0 © Q 0 R0 b ¹ ¿° n0Q 0[ © mkT0 ¹ ¯°

(7-54)

where:

b

6.

2J m . U1kT0

NON-STATIONARY GAS FLOWS.

The properties of non-stationary flows of a rarefied gas may be described by solving the Cauchy problem for Boltzmann’s equation that may be written in the form: wf wf wf wf vx vy vz wt wx wy wz

0 .

The initial condition for this problem is given by:

(7-55)

110

Analytical Methods for Problems of Molecular Transport

f v, r ,0

fi v, r .

(7-56)

The solution of this mathematical formulation begins with the construction of a general solution of Lagrange’s auxiliary system of ordinary differential equations which may be expressed in the form:

dt 1

dx vx

dy vy

dz vz

df . 0

(7-57)

The general solution of Eq. (7-57) is:

f

D

r vt

const ,

r0

(7-58)

const .

(7-59)

By virtue of Eqs. (7-58) and (7-59), the distribution function may be shown to be expressible in the following general form:

f v, r , t < r0 , where < is an arbitrary function. The initial condition then yields:

f v, r ,0 < r0

fi v, r0 .

(7-60)

Allowing for this relation, one can obtain the following solution of the Cauchy problem for the distribution function:

f v, r , t

fi v, r vt .

(7-61)

The knowledge of the distribution function allows one to calculate any macroscopic quantity characteristic of the gas flow. It is, however, very important to note that the calculation technique is very different from that described in Section 7.3. For a non-stationary flow, there is no cone of influence in the velocity space since the molecular trajectories passing through a fixed point, M r , at various moments of time, t , must start from different initial points, r0 . The moments of the distribution function may be found by means of direct integration of Eq. (7-61) over all velocity space. The mean value of any molecular property, I v , is given by:

111

Chapter 7. The Free-Molecular Regime

I

n 1 ³ I v f i v, r vt dv .

(7-62)

To calculate this integral, it is convenient to change the variable of integration from v to a new variable, r c , which is related to v by:

r c r vt .

(7-63)

The volume elements of this transformation are connected through the Jacobian in the following manner:

dv

w vx , v y , vz

dr c

w xc, y c, z c

t 3 dr c .

(7-64)

Finally, the mean value of I v may be expressed by [6]:

nI

§ r r c · § r r c · t 3 ³ I ¨ , r c ¸ dr c , ¸ fi ¨ © t ¹ © t ¹

(7-65)

where integration is extended over the domain occupied by the gas at t

0.

PROBLEMS 7.1. Determine the temperature of a sphere located in an infinite stream of a monatomic gas the velocity of which, at large distances from the body, is given by a vector, u . Consider the case of arbitrary values of the ratio, 12 u 2kT m , if the particle is assumed to be a perfect heat conductor and molecules of a gas are considered to be reflected by the sphere surface with equilibrium conditions. Solution: The distribution function at the surface of the particle can be expressed in the form: f

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

§ m v u 2 exp ¨ ¨ 2kT0 ©

· ¸ , ¸ ¹

112

Analytical Methods for Problems of Molecular Transport

f

m · ¸ © 2S kTS ¹

§

32

§ mv 2 exp ¨ © 2kTS

DW n rS ¨

· ¸ 1 DW f cr , cT , cI , rS ¹

,

where n rS denotes the density of a fictitious gas at the point, rS , and TS is the constant temperature of the particle. The unknown function, n rS , and the quantity, TS , can be specified from two conditions expressing the impenetrability of the surface for gas molecules and the heat balance:

n S ³ vfdv

0 ,

³ dS ³ 12 mv

2

vfdv

0 .

S

From the first condition, one can obtain: 12

n rS

§T · n0 ¨ 0 ¸ © TS ¹

^exp u

2 r

` ,

S ur 1 erf ur

where ur u n S u cos T . The direction of the vector, u , is assumed to be along the polar axis. Taking into account both conditions, one can obtain:

TS

T0

1

1 2 u 2

1 2

S

5 u 2

u 3 Q u

1 S u Q u

,

where: 1

Q u

³ t erf u t dt 0

1

³ exp u 0

2 2

t

dt

, and u

§ 2kT0 · u¨ ¸ © m ¹

1 2

.

A more detailed analysis is contained in [3,19].

7.2. Using the same conditions as in Problem 7.1, determine the drag force on the sphere.

113

Chapter 7. The Free-Molecular Regime

Solution: Substituting the distribution function described in Problem 7.1 in Eq. (7-8), one obtains:

FD

8 3

12

2

R n0 2S mkT0

° u ®Q 1 u DW °¯

ª § T ·1 2 º½ « 81 S ¨ S ¸ Q 2 u » °¾ , « © T0 ¹ »° ¬ ¼¿

where u is the dimensionless velocity of the gas, DW is the tangential momentum accommodation coefficient and the following notations have been introduced: 1

³ ª«¬t

Q1 u

3 2

2

0

Q 2 u

exp u 2t 2

1 2

º § t · 2u t 3 ¸ erf u t » dt ,

©u ¹ ¼

1 ª 32 ³ « t 2 12 exp u 2 t 2 0¬

S¨

º § t · 12 S ¨ 12 u t 2u t 3 ¸ erf u t » dt . © u ¹ ¼

The temperature of the sphere, TS , was determined in Problem 7.1.

7.3. Determine the mean velocity distribution (both components) for a linearized uniform flow of a gas past a sphere if the mean gas velocity far from the sphere is u . Consider the specific case when DW 1 and 12 u 2kT m . Solution: For this problem the distribution function given in Eq. (7-18) is employed. The component of the mean velocity vector may be calculated from Eq. (7-15). For example, the radial component may be determined in the following manner:

ur r ,T f

1 cr f1dv ur S 3 2 ³ n0 1 2

^

³ dE 0

w

³ cos F sin F

u³ c3 exp c 2 0

2S

0

12

u S r sin 2 F cos F ª¬1 r 2 sin 2 F º¼

2c cos F `dcd F ,

114

Analytical Methods for Problems of Molecular Transport

where w arcsin 1 r and ur u cos T . An analogous relationship exists for the other component, uT r ,T . Having performed the indicated integrations, one obtains:

ur r ,T u 1 g r cos T , and uT r ,T u 1 h r sin T . Here, the functions, g r and h r , may be written as:

g r

1 4

ª § 1 r 1 · º 1 3 r 3 r « 14 r 3 18 r 1a 2 18 a 4 ln ¨ ¸ » 2 1 a , and: a © ¹ ¼» ¬«

hr

1 2

° ½° ª 1 1 1 2 § 1 r 1 · º 3 ¸ » 2 1 a ¾ , ® g r r « 2 r 2 a ln ¨ «¬ © a ¹ »¼ ¯° ¿°

where a

12

1 r 2

. More detailed analyses may be found in [7,20].

7.4. Prove that the velocity field, as defined in Problem 7.3, is not the potential field. Solution: For the potential flow field, the following equation must be satisfied: u u r 0 . This quantity, for the free-molecular velocity field described in Problem 7.3, may be expressed in the form:

er r sin T

eT r sin T

eI

w wr ur r ,T

w wT ruT r ,T

w wI 0

2

u u r ,T

r

ª h r g r dh º eI u sin T « »z0 r dr ¼ ¬ From this one can see that the velocity field has no potential.

7.5. In Problem 7.3, determine the pressure and temperature distribution. Check the validity of the relationship: p nkT .

115

Chapter 7. The Free-Molecular Regime

Solution: The same distribution function and calculational techniques as in Problem 7.3 yield:

^

`

p0 1 12 S ur \ 1 r \ 2 r ,

p

¼

\ 1 r r ª1 a 13 1 a 3 º , ¬

\ 2 r

1 3

r 2 1 8S 1 , and:

T r ,T T0 1 13 S 1 2ur r 2 ,

12

where a 1 r 2 and ur u cos T . It is easy to see that the equation of state is not altered for this free-molecular flow.

7.6. A sphere suspended in a gas has radius, R , and temperature, Determine the heat flux and the temperature T0 'T 'T T0 . distribution of a gas if its number density and temperature far from the sphere are n0 and T0 , respectively. Solve this problem for the freemolecular regime. Solution: A distribution function should be sought in the form:

f1

0 f ,

f2

1 DW f 0 DW f 0 ^1 Q r c 2 32 W r `

,

where: f

0

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

§ mv 2 · exp ¨ ¸ , © 2kT0 ¹

is the distribution function at large distances from the heated sphere. The quantities, Q r and W r , may be determined from the boundary relationships

116

Analytical Methods for Problems of Molecular Transport

Figure 7-4. The geometry to be used for Problem 7.7 in determining the force on a round disk perpendicular to a uniform gas flow.in determining the force on a round disk perpendicular to a uniform gas flow

N r 0 , N r eq 0 , and Qr DT Qr eq . Performing some simple integrations and algebraic transformations, one can obtain: 12

Q

'T § 2kT0 · DW DT 4S R 2 n0 kT0S 1 2 ¨ , ¸ T0 © m ¹

where Q is the heat flux through the sphere surface. The temperature distribution can be expressed in the form:

T r T0 ®1 34 DW DT ª1 1 r 2 ¬« ¯

12

º 'T ½ . ¾ ¼» T0 ¿

7.7. Calculate the force on a round disk the surface of which is perpendicular to the velocity vector of a uniform gas flow at infinity. The radius of the disk, R , is assumed to be much less than the mean free path 12 where u and T are the mean gas velocity and and u 2kT m temperature, respectively. Use the geometry in Fig. 7-4. Solution: The force on the left and right sides of the disk may be expressed as:

F1,2

# S R 2 ¦ ³ mv vn1,2 f1,2 v dv ,

(P-1)

# #

where n i is the external (pointing into the gas) normal vector to the disk surface. The reflected distribution functions are given by:

§ m · f1 v DW n1 ¨ ¸ © 2S kT ¹

32

exp c 2 1 DW f1 cx , c y , cz ,

117

Chapter 7. The Free-Molecular Regime

Figure 7-5. The geometry to be used for Problem 7.8 in determining the drag on a plate parallel to a steady, free-molecular gas flow.

§ m · f 2 v DW n2 ¨ ¸ © 2S kT ¹

32

exp c 2 1 DW f 2 cx , c y , cz ,

where the notation (±) refers to those regions in the velocity space in which cx ! 0 and cx 0 , respectively. The incident distribution function may be written as:

f1 v

§ m · f 2 v n ¨ ¸ © 2S kT ¹

32

exp c 2 1 2cx u .

The number density, ni n 1 Q i , may be found from the impenetrability condition of the disk surface which yields: Q 1,2 r S u . Then, after some simple integration, the expression for the force becomes:

F

12

4 R 2 n 2S kTm

ª1 DW 18 S 12 º u . ¬ ¼

7.8. A plate of area, S , is situated in a steady gas flow and is oriented parallel to the velocity vector, u , as shown in Fig. 7-5. In the freemolecular regime, determine the drag on the plate for arbitrary values of the 12 speed ratio, u 2kT m , and of the tangential momentum accommodation coefficient. Solution: In this problem, the drag is given by:

F 2S ¦ ³ mvvx f r v dv . r r

The incident and reflected distribution functions associated with the upper side of the plate are specified by:

118

Analytical Methods for Problems of Molecular Transport

Figure 7-6. The geometry to be used for Problem 7.9 in determining the drag on a plate oriented at an angle to a steady, free-molecular gas flow.

f

f

32

v

§ m · n¨ ¸ © 2S kT ¹

v

§ m · DW n ¨ ¸ © 2S kT ¹

§ m v u 2 exp ¨ ¨ 2kT ©

32

m · 1DW n ¨§ ¸ © 2S kT ¹

· ¸ , and: ¸ ¹

§ mv 2 · exp ¨ ¸ © 2kT ¹ 32

ª m vx 2 v y u exp « ¬ 2kT

2

º vz2 » . ¼

Using these, the drag is found to be: 12

§ 2m · F DW pS ¨ ¸ u. © S kT ¹ 7.9. A plate of area, S , is situated in a steady gas flow and is oriented such that the gas stream is incident on the plate at an angle T as shown in Fig. 76. In the free-molecular regime, determine the drag on the plate for arbitrary 12 values of the speed ratio, u 2kT m , and of the tangential momentum accommodation coefficient. Solution: The net drag component in direction of the flow vector, u , is given by: Fu

S ¦ ³ m vx cos T v y sin T vx f1r f 2r d v , r r

119

Chapter 7. The Free-Molecular Regime

where the indices 1 and 2 refer to the upstream and downstream sides of the plate, respectively. The distribution functions of incident and reflected molecules may be expressed in the form:

§ m · f 2 v n ¨ ¸ © 2S kT ¹

f1 v

32

§ m exp ¨ v u 2 ·¸ , kT 2 © ¹

32

f1

v

32

§ m · § m 2· § m · v ¸ 1 DW n ¨ DW n1 ¨ ¸ exp ¨ ¸ © 2S kT ¹ © 2kT ¹ © 2S kT ¹ 2 2 m ª ½ u exp ® vx u cos T v y u sin T vz2 º ¾ , « » ¼¿ ¯ 2kT ¬

and:

§ m f 2 v DW n2 ¨ © 2S kT m u exp ® ¯ 2kT

· ¸ ¹

32

§ m 2· § m · exp ¨ v ¸ 1 DW n ¨ ¸ 2 kT © ¹ © 2S kT ¹

ª v u cos T 2 v u sin T y ¬« x

2

32

½ vz2 º ¾ . ¼» ¿

The values, n1 and n2 , are specified by the impenetrability condition. The drag is found to be: 12

F

§ 2m · pS ¨ u 2 DW cos T ª cos T exp u 2 cos 2 T ¸ ¬ © S kT ¹ º S 1 2 2 u cos T erf u cos T » DW ª¬ 12 S cos 2 T 2

u »¼

^

`

sin 2 T exp u 2 cos 2 T S u cos T erf u cos T º» , ¼ where u

m

12

2kT

u.

7.10. Determine the force on a round disk located in a rarefied gas if the surfaces of the disk have different temperatures, T1 and T2 . The disk has a radius of R which is assumed to be much less than the mean free path and T1 T2 T0 where T0 is the gas temperature. Use the geometry in Fig. 7-7.

120

Analytical Methods for Problems of Molecular Transport

Figure 7-7. The geometry to be used for Problem 7.10 in determining the force on a round disk in a rarefied gas when the two sides of the disk have different temperatures.

Figure 7-8. The geometry to be used for Problem 7.11 in determining the thermophoretic force on a thin disk located in a rarefied gas having a constant temperature gradient.

Solution: Using the same technique as in Problem 6.5, one can obtain:

; W 1w

T1 T0 , T0

; W 2w

T2 T0 , T0

)1

DW DT 2 c 2 W 1w

) 2

DW DT 2 c 2 W 2 w

where )1 and ) 2 are the corrections to the reflected distribution functions associated with the left- and right-hand surfaces of the disk, respectively, and )1 ) 2 0 . The force on the disk may be calculated from the equation given in the solution to Problem 7.7 [Eq. (P-1)]. This gives:

Fx

1 4

S R 2 p0DW DT

T1 T2 . T0

7.11. Determine the thermophoretic force on a thin absolutely thermal conductive disk located in a rarefied gas with a constant temperature gradient, T . The radius of the disk is R and is assumed to be much less than the mean free path. Use the geometry in Fig. 7-8. Solution: The corrections to the reflected distribution functions may be written as (Problem 6.5):

121

Chapter 7. The Free-Molecular Regime

)1 c,0

1 DW )1 cx , c y , cz ,0 ° DW ® 2 c 2 ¯°

ª

««1 D ¬

§ G 21

T ¨ ¨

© S

1 1 2G1 · º 2G1 ½° ¸» ¾ , S ¹¸ »¼ S ° ¿

and:

) 2 c,0

1 DW ) 2 cx , c y , cz ,0 ° DW ® 2 c 2 ¯°

ª § G 2 2G 2 «1 DT ¨ 2 1 ¨ S S «¬ ©

· º 2G1 2 ½° ¸» ¾ . ¸» S °¿ ¹¼

Here, )1 c,0 , ) 2 c,0 , G1 , and G 2 , are defined by: i

i

0 , G 2

2 0 , G 2

)1 c,0 a11 qcx S312 c 2 , G11

) 2 c,0 a11 qcx S312 c 2 , G1 2

1

5 8

S 3 2 a11 q , and:

5 8

S 3 2 a11 q ,

with:

q T01

wT , wx

where the indices 1 and 2 are associated with the left- and right-hand surfaces of the disk, respectively. Substituting these into the force equation from Problem 7.7 [Eq. (P-1)], one can obtain:

F

158 S R 2 n0 k O T f ª¬1 12 DW DW 1 DT 325 S º¼ .

7.12. Determine the number density of a gas at a point, A , located behind a round disk of radius, R , which is much less than the mean free path, i.e. R O . The mean gas velocity, u , is assumed to be much less than the thermal molecular velocity. Use the geometry in Fig. 7-9. Solution: Since n S const for all molecular trajectories beginning at the disk surface, vc vx , v y , vz . Therefore, the distribution function has a very simple form for arbitrary values of the tangential accommodation

122

Analytical Methods for Problems of Molecular Transport

Figure 7-9. The geometry to be used for Problem 7.12 in determining the number density behind a round disk perpendicular to a free-molecular gas flow in which the mean gas velocity is much less than the thermal molecular velocity.

coefficient. The distribution function outside and inside of the cone of influence is expressed as:

f1

0 f 1 2cx u ,

f2

DW f 0 1 S u 1 DW f 0 1 2cx u .

The number density is then given by:

° 2 n x n0 ®1 S °¯

f

F0

2 2 ³ c exp c dc ³ sin F d F 0

0

½ ª DW S u 2 2 DW u c cos F º °¾ ¬ ¼° ¿ 12 ° § m · ª S§ x n0 ®1 u ¨ ¨¨1 ¸ «DW 2 © R2 x2 © 2kT0 ¹ ¬« ¯°

2 DW

· ¸¸ ¹

R 2 º °½ »¾ . S R 2 x 2 ¼ ¿°

1

7.13. A gas with a number density, n0 , is flowing into vacuum across a round hole of radius, R O . Determine the number density at a point, A . Use the geometry in Fig. 7-10.

123

Chapter 7. The Free-Molecular Regime

Figure 7-10. The geometry to be used for Problem 7.13 in determining the number density of a free-molecular flow exiting through a round hole.

Solution: The distribution function outside and inside of the cone of influence may be expressed in the form:

f1

0 ,

f2

§ m · n0 ¨ ¸ © 2S kT ¹

32

exp c 2 .

From this, the number density at A is determined to be:

n x

1 2

ª x n0 «1 2 «¬ R x2

º » . »¼

7.14. A gas with number density, n0 , passes outward through a small slot by means of effusion (pressure driven motion of a gas from a vessel to a vacuum through a small hole). Determine the gas number density at a large distance, r , from the slot ( r S0 , where S0 is the slot area ). Use the geometry in Fig. 7-11. Solution: The distribution functions outside (1) and inside (2) of the cone of influence are given by:

§ m · f1 0 and f 2 n0 ¨ ¸ © 2S kT ¹

32

exp c 2 , respectively.

124

Analytical Methods for Problems of Molecular Transport

Figure 7-11. The geometry to be used for Problem 7.14 in determining the number density of an effusing gas at large distances from a small slot.

Taking into account the condition, r R , after integration one can obtain:

n r

n0 R 2 4 r 2

n0 S 4S r 2

n0 S0 cos T 4S r 2

.

7.15. Determine the number density of a gas at a point, A , behind a round disk of radius, R O . The mean gas velocity, u , is assumed to be arbitrary. The flow geometry is given in Problem 7.12 (see Fig. 7-9). Consider the particular case for which: 12

u 2kT0 m

12

and cos F 0 2kT0 m u 1 .

Solution: The distribution functions outside and inside of the cone of influence are given by:

f1

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

exp §¨ c u ©

2

· , ¸ ¹

125

Chapter 7. The Free-Molecular Regime

f2

§ m · ¨ ¸ © 2S kT0 ¹

^

32

u DW nr exp c 2 1 DW n0 exp §¨ cx u ©

2

`

c 2y cz2 ·¸ . ¹

The number density, nr , of reflected molecules may be specified from the impenetrability condition which yields:

nr

n0 ª« exp u 2 S u 1 erf u º» . ¬ ¼

Having employed standard integration techniques (see Section 7.3), one can obtain:

n x n0

^

1 2

2 DW 12 DW 1 cos F 0

u ª« exp u 2 S u 1 erf u º» ¬ ¼

12 DW cos F 0 exp u 2 sin 2 F 0 1 2

2 DW ª¬cos F 0 erf u cos F 0

`

u exp u 2 sin 2 F 0 erf u º , ¼ where:

cos F 0

x 2

R x2

.

In the particular case when u cos F 0 1 , this expression reduces to:

n x n0 cos F 0 exp u 2 sin 2 F 0 . 7.16. Determine the frictional force on the lower plate in a system where the upper of two parallel plates is moving horizontally. Use the geometry in 12 Fig. 7-12 and assume that u 2kT m .

126

Analytical Methods for Problems of Molecular Transport

Figure 7-12. The geometry to be used for Problem 7.16 in determining the frictional force on a plate due to the horizontal motion of a second, parallel plate.

Solution: The distribution function for vx ! 0 and vx 0 may be 0 expressed as f r f 1 a r c y where a r const . Using the Maxwellian boundary condition for each plate, one can obtain:

a

2 1 DW 2 DW

u , a

12

§ m · ¨ 2kT ¸ u . © ¹

2 u , u 2 DW

The force of friction per unit surface area is given by: 12

Fy

¦ ³ mvx v y f r dv r r

DW § m · 2 DW ¨© 2S kT ¸¹

pu ,

where p is the hydrostatic pressure.

7.17. Assume that a gas is flowing in the y -direction between two infinite planes separated in the x -direction by a distance, d , where d O and O is the mean free path of the gas molecules. Assume that the number density of the gas is maintained at different steady values at two different locations along the y -axis and that the temperature of this system is uniform. Determine the number of molecules per unit area per second flowing in y direction. Solution: The distribution function may be chosen in the form:

f

0 n ) c, x , where g n f 1 yg

1 n0 dn

dy

The correction to the distribution function, ) c, x , which describes the gassurface interactions may be specified from the Boltzmann equation such that:

127

Chapter 7. The Free-Molecular Regime

cx

w) cy gn 0 . wx

The correction, ) c, x , may be expressed in the form:

) c, x

1 2

a

x a x c y 12 a x a x c y sign cx

The half-range moment equations are obtained by multiplying the Boltzmann equation by c y 1 r sign cx exp c 2 dc and integrating over all velocities. These equations are found to be:

da r x r S gn . dx The boundary conditions are given by a r # d 2 simple calculations one finds: 12

Jy

n0u y

1 2

§ 2kT · n0 ¨ ¸ © m ¹

a

a

2 DW

DW

1 DW a # # d 2 . 12

1 8

§ 8kT · ¸ © Sm ¹

Sd¨

After

wn . wy

7.18. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at different temperatures, T c and T cc with T c ! T cc . At the beginning t 0 the vessels are filled with different gases to the same pressure. When the capillary is opened a pressure difference, G p pcc pc , will arise. Determine the time dependence of this pressure difference. Solution: To describe this transport problem one should start with the basic equations of balance for the number of molecules of each of the two gases which are given by: V c

dnic dt

J zi ,

(P-2)

V cc

dnicc dt

J zi ,

(P-3)

128

Analytical Methods for Problems of Molecular Transport

where J zi is the i -th constituent molecular number flux across the tube cross section per second. For the free-molecular regime, J zi , may be expressed in the form [21-23]:

J zi

2 D iW

D iW

2 3

S R3

8k d n T , S mi dz i

where D iW is the tangential momentum accommodation coefficient of the i th constituent. This flux may be approximately represented by:

J zi

ai

ai nicc T cc nic T c , where:

2 D iW 2S R 3 D iW 3L

8k

S mi

.

Then, the basic equations describing kinetics of this system take the form:

dnic a Tc a T cc nic i nicc , i dt Vc Vc

(P-4)

dnicc ai T c a T cc nic i nicc . dt V cc V cc

(P-5)

The necessary initial conditions are given by:

n1c 0 0 , n1cc

p 0 p 0 , n2c 0 , and n2cc 0 . kT cc kT c

This formulation then yields:

Chapter 7. The Free-Molecular Regime

129

G p t k ª¬T cc n1cc n2cc T c n1c n2c º¼ p 0 °§ V c T cc · § T c T c · ¸ ¨ ® 1 G 0 °¯©¨ V cc T c ¹¸ ©¨ T cc T cc ¹¸ · °½ § V c T c · § T cc exp Z2t exp Z1t ¸¸¾ , ¨ ¸ ¨¨ © V cc T cc ¹ © T c ¹ ¿° where:

G0

ª Tc Tc Vc T cc º and Zi ai « ». V cc »¼ T cc V cc ¬« V c

7.19. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at different temperatures, T c and T cc with T c ! T cc . At the beginning t 0 both vessels are filled with a simple gas to the same pressure. When the capillary is opened a pressure difference, G p t pcc pc , will arise. Determine the time dependence of this pressure difference and the steady-state pressure difference that develops. Solution: The pressure difference may be obtained from solution of Problem 7.18. in which one should substitute Z1 Z2 Z . This yields:

G p t

p 0 § T cc · § V c T c · ª1 exp Zt ¼º , where: ¨¨ 1 ¸ G0 © T c ¹¸ ¨© V cc T cc ¸¹ ¬ 1/ 2

2 DW 2S R 3 § 8kT cc · Z 3L ¨© S m ¸¹ DW

§ 1 1 Tc · ¨¨ ¸¸ . © V cc V c T cc ¹

The steady-state (or stationary) pressure difference is then given by:

'pstat

G0

G p f

Tc Vc . T cc V cc

p 0 § T cc · § V c T c · , where: ¨¨ 1 ¸ G0 © T c ¸¹ ¨© V cc T cc ¸¹

130

Analytical Methods for Problems of Molecular Transport

7.20. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at the same constant temperature, T c T cc T and are filled at the beginning t 0 with a simple gas to different initial pressures, pc 0 p and pcc 0 p 'p . When the capillary is opened the pressure difference, G p t pcc pc , will change as a function of time. Determine the time dependence of this pressure difference. Derive a formula for the tangential momentum accommodation coefficient if the pressure difference, G p t , is known at a particular time, t . Solution: Using the following initial conditions:

nc 0

p p 'p and ncc 0 kT kT

in Eqs. (P-4) and (P-5) (Problem 7.18), one can obtain:

G p t 'p exp Zt where:

Z

2 DW

DW

12

Z d , Zd

2S R 3 § 8kT · 3LV ¨© S m ¸¹

and V

V cV cc . V c V cc

The expression for 'p yields:

DW

§ 'p 1 ° ¨ 2 ®1 ln

¨ G p t t Z d °¯ ©

1

·½ ¸ °¾ . ¸° ¹¿

7.21. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at the same constant temperature, T c T cc T and are filled at the beginning t 0 with different gases to the same initial pressure. When the capillary is opened a pressure difference, G p pcc pc , will develop as a function of time. Determine the time dependence of this pressure difference and the maximum value of the pressure difference that develops.

131

Chapter 7. The Free-Molecular Regime

Solution: The time dependence of the number densities in the two vessels is described by Eqs. (P-4) and (P-5) (Problem 7.18). The initial conditions are given by:

p 0

n1c 0 0 , n1cc 0

kT

, n2c 0

p 0 kT

, and n2cc 0 0

This formulation yields a pressure difference of:

G p t p 0 ª¬exp Z1t exp Z2t º¼ where: 12

Zi

2 D iW 2S R 3 § 8kT · ¨ ¸ D iW 3LV © S mi ¹

V cV cc . V c V cc

and V

The subtraction of the two decreasing exponential terms results in the existence of a pressure difference maximum, 'p | G p |max . This phenomenon is usually called the diffusion-pressure effect. The value, 'p , is found to be:

'p p 0

f Z

1 · § ¨1 ¸ exp u . © Z ¹

The time at which the maximum pressure difference, 'p , occurs is given by t u Z2 . Here, the following notations have been introduced:

Z

D 2W 2 D1W m2 and u D1W 2 D 2W m1

.

ln Z

Z 1

7.22. For the free-molecular regime, determine the gas temperature between two parallel plates having temperatures T1 and T2 . Molecules are assumed to be reflected diffusely with arbitrary accommodation of energy at each surface. Use the geometry in Fig. 7-13: Solution: The distribution functions for the sets of molecules having positive ( vx ! 0 ) and negative ( vx 0 ) motions are expressible as:

f

§ m · n1 ¨ ¸ © 2S kT1r ¹

32

§ mv 2 · exp ¨ ¸ and: © 2kT1r ¹

132

Analytical Methods for Problems of Molecular Transport

Figure 7-13. The geometry to be used in Problem 7.22 in determining the gas temperature between two parallel plates in the free-molecular regime.

f

§ m · n2 ¨ ¸ © 2S kT2 r ¹

32

§ mv 2 · exp ¨ ¸. © 2kT2 r ¹

The number densities, n1 , and n2 , may be found from the relations:

Nx

³ vx f

dv ³ vx f dv 0 and n n1 n2 ,

where n is the gas number density. These yield: n1

a n and n2 1 a

12

§T · 1 n where a ¨ 2 r ¸ . 1 a © T1r ¹

The quantities, T1r , and T2r , are specified by the boundary conditions:

DT

Qx >Q1x @ eq

Qx where: >Q2 x @ eq

>Q1x @eq

1 2

ª º m « ³ v 2 vx f eq dv ³ v 2 vx f dv » and: «¬ »¼

>Q2 x @eq

1 2

ª º m « ³ v 2 vx f dv ³ v 2 vx f eq dv » . ¬« ¼»

133

Chapter 7. The Free-Molecular Regime

The additional unknown parameters, [n1 ]eq and [n2 ]eq , which are included in f eq and f eq can be expressed in terms of T1r , and T2r using the impenetrability conditions [ N1x ]eq 0 and [ N 2 x ]eq 0 . The gas temperature is found to be:

T

m v 2 fd v 3nk ³

T rT r 1 2 1

2

12

1

2 DT ª¬T1 1 DT T2 T2 1 DT T1 º¼

.

7.23. Determine the free-molecular heat flux in a monatomic gas between two coaxial cylinders the radii and temperatures of which are R1 , T0 'T , and R2 , T0 respectively, and where R1 R2 . Assume that 'T T0 and that the number density of the gas is n0 at r R1 . The tangential momentum and energy accommodation coefficients are assumed to be different for the inner and outer cylinders. Solution: The zone of influence around a cylindrical body is analogous to the cone of influence found around a spherical body or associated with a circular orifice. It is formed by the intersection of two planes that lie tangent to the cylinder in question. The distribution functions outside and inside the zone of influence of the inner cylinder may be chosen in the form:

f1

0 f 1 Q 1 c 2 32 W 1 and f 2

0 f 1 Q 2 c 2 32 W 2 .

The boundary conditions are given by:

D1T

Qr 1 Qr 1 ª¬Qr 1 º¼

N r 1 0 , n 1

N r 1 ª¬ N r 1 º¼ eq

, D 2T eq

Qr z Qr z ª¬Qr z º¼

eq

n0 , N r z 0 ,

0 , and N r z ª¬ N r z º¼

eq

0,

where z R2 R1 . In calculating the fluxes using this formulation, one has to allow for the discontinuity of the distribution functions. For example, the heat flux, if the reflected molecules have the same temperature as the surface of the outer cylinder, may be calculated in the following way:

134

Analytical Methods for Problems of Molecular Transport 12

Qr z ª¬Qr z º¼

p0 § 2kT0 · D 2W S 3 2 ¨© m ¸¹

eq

f

f

° u ®SQ eq ³ cU2 exp cU2 dcU ³ cU2 cz2 exp cz2 dcz °¯ 0 f

S D ªD u « ³ )1 c, z sin T dT ³ ) 2 c, z sin T dT «¬ 0 D S º ½° ³ )1 c, z sin T dT »¾ , S D ¼» °¿

where ) i c, z Q i ª cU2 cz2 32 º W i and D arccos z 1 . The heat flux ¬ ¼ per unit length of the cylinders is found to be: 1

QFM

12

ª 1 · º § 2k · 1 1 § 1 1 ¸ » ¨ « ¨ ¸ «¬ 2S R1 D1W D1T 2S R2 © D 2W D 2T ¹ »¼ © S mT0 ¹

p0 'T .

7.24. Determine the free-molecular heat flux in a monatomic gas between two concentric spheres the radii and temperatures of which are R1 , T0 'T , and R2 , T0 , respectively, and where R1 R2 . Assume 'T T0 and that the number density of the gas is n0 at r R1 . The tangential momentum and energy accommodation coefficients are assumed to be different for the inner and outer spheres. Solution: The distribution functions outside and inside the cone of influence of the inner sphere may be chosen in the form:

f1

0 f 1 Q 1 c 2 32 W 1 and f 2

0 f 1 Q 2 c 2 32 W 2 .

The same boundary conditions are used as were used in Problem 7.23, but the integration technique is necessarily different from that used with the cylindrical geometry. An example of this integration is given by:

135

Chapter 7. The Free-Molecular Regime

Figure 7-14. The geometry to be used for Problem 7.25 in determining the torque on a flat, rotating disk in the free-molecular regime.

12

Qr

p 2kT z ª¬Qr z º¼ eq 302 §¨ 0 ·¸ D 2W S © m ¹ w f ° ª 5 2 u ® « SQ eq 2S ³ c exp c dc ³ ) 2 c, z sin F cos F d F °¯ «¬ 0 0

S 2

º ½°

³ )1 c, z sin F cos F d F »» ¾ ,

¼ °¿

w

where ) i c, z Q i c 2 32 W i and w arcsin z 2 . The heat flux is then found to be: 1

QFM

12

ª 1 · º § 2k · 1 1 § 1 1 ¸ » ¨ « ¸ 2 2 ¨ ¬« 4S R1 D1W D1T 4S R2 © D 2W D 2T ¹ »¼ © S mT0 ¹

p0 'T .

7.25. A flat disk of radius, R O , revolves in a rarefied gas with a constant 12 angular speed, Z , where Z R 2kT m . Determine the torque on the disk, K . Use the geometry in Fig. 7-14. Solution: The torque, K , in this geometry has only one component, K x , which may be expressed as:

136

Analytical Methods for Problems of Molecular Transport R

Kx

R

4S ³ r 2 dr ¦ ³ mvI vx f r v dv ,

2³ rdFI 0

r r

0

where the (±) notation refers to vx ! 0 and vx 0 , respectively, and dFI is the force component acting on one side of the disk only. The distribution function of reflected and incident molecules may be written as:

12 0 0 f c DW f 1 2m kT cI Z r 1 DW f cr , cI , cx ,

f c

0 f .

Substituting these functions into the above expression, one can obtain:

Kx

12

12 DW Z R 4 n 2S mkT

.

7.26. An infinite cylinder of radius, R O , revolves in a rarefied gas with a 12 constant angular speed, Z , where Z R 2kT m . Determine the torque per unit length of the cylinder, K . Solution: The x -component of the torque is specified by:

Kx

2S R 2 ¦ ³ mvr vI f r v dv , r r

where:

f

f

12

DW f 0 1 2m kT cI Z R 1 DW f 0 cr , cI , cx ,

0 f .

and the direction of the x -axis is the same as the direction of Ȧ . Simple integrations then yield: Kx

12

DW Z R 3 n 2S mkT

.

137

Chapter 7. The Free-Molecular Regime

7.27. A sphere of radius, R O , revolves in a rarefied gas with a constant 12 angular speed, Z , where Z R 2kT m . Determine the torque, K , on the sphere if the polar axis, x , is in the same direction as Ȧ . Solution: The vector, K , has one component, K x , which may be found from: S

Kx

2S R 3 ³ sin 2 T dT 0

f

f

¦ ³ mvr vI f r v dv

, where:

r r

12 DW f 0 1 2m kT cI Z R sin T 1 DW f 0 cr , cT , cI ,

0 f .

After integration, one obtains:

Kx

12

43 DW Z R 4 n 2S mkT

.

7.28. An equilibrium gas occupies a semi-infinite expanse, x 0 and is contained by an infinite planar surface at x 0 . Determine the number density of the gas as a function of time in the neighborhood of the surface ( x O ) if the gas is expanding into a vacuum (the region of x ! 0 ) after the sudden removal of the surface (which disappears at t 0 ). Solution: If x O , molecular encounters may be neglected. Therefore, the Boltzmann equation is given by: wf wf vx wt wx

0 .

The Lagrange auxiliary system is then:

dt 1

dx vx

df . 0

The two independent solutions of this system may be written as x vx t x0 , and f D , where x0 and D are arbitrary constants. The general solution of the Boltzmann equation is then expressed as ) x0 ,D 0 where ) x0 ,D is an arbitrary function. This yields:

138

Analytical Methods for Problems of Molecular Transport

< x0 < x vx t .

f

Taking into account the initial condition, one can represent the distribution function in the form:

° f 0 ® °¯ 0

f

x0 0 x0 ! 0

; vx ! x t ; vx x t

, ,

where:

f

0

v

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

§ mv 2 · exp ¨ ¸ . © 2kT0 ¹

Now, the number density may be calculated according to the following:

n x, t

§ m · n0 ¨ ¸ © 2S kT0 ¹ n0

S

f

³

32 f f f

2 ³ ³ ³ exp c dv y dvz dvx

f f x t

exp cx2 dcx

x m t 2 kT0

§x n0 § m ¨1 erf ¨ ¨ t 2kT 2 ¨© 0 ©

·· ¸¸ ¸ . ¸ ¹¹

7.29. A sphere of radius, R O contains a gas with a constant density, n0 , and constant temperature, T0 . At some initial time, t 0 , the sphere vanishes releasing the gas into an infinite vacuum. Determine the number density, n r , t . Solution: Given Eq. (7-65) in the text, the number density may be expressed as: § r r c · n r , t t 3 ³ f i ¨ , r c ¸ dr c , © t ¹ where:

fi

§ m · n0 r c ¨ ¸ © 2S kT0 ¹

32

§ m § r r c · 2 · exp ¨ ¸ , ¨ 2kT0 ¨© t ¸¹ ¸ © ¹

139

Chapter 7. The Free-Molecular Regime

n n0 r c ® 0 ¯0

; rc d R , ; rc ! R .

This integration may be performed easily in spherical coordinates when the polar angle is taken as the angle between r and r c . Having integrated over all of the angular variables, one obtains:

n r, t

n0 § m ¨ r ¨© 2S kT0t 2

12R

· ¸¸ ¹

³ rc 0

ª § § m m 2· 2 ·º u «exp ¨¨ r r c ¸¸ exp ¨¨ r r c ¸¸ » dr c . 2 2 © 2kT0t ¹ © 2kT0t ¹ ¼» ¬« Further integration with respect to r c yields:

n r, t

1 2

where D

n0 ®erf D r R erf D r R ¯ 1

ª exp D 2 r R 2 exp D 2 r R 2 º ½ , ¾ ¼» ¿ D S r ¬«

12

m 2kT t 2

0

.

REFERENCES 1. Kennard, E.H., Kinetic Theory of Gases (McGraw-Hill, New York, 1938). 2. Heineman, M., “Theory of Drag in Highly Rarefied Gases,” Comm. Pure Appl. Math. 1(3), 259-273 (1948). 3. Patterson, G.N., Molecular Flow of Gases (Wiley and Sons, New York, 1956). 4. Schaaf, S.A. and Chambré, P.L., Flow of Rarefied Gases (Princeton University Press, Princeton, New Jersey, 1961). 5. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, NY, 1969). 6. Bird, G.A., Molecular Gas Dynamics (Clarendon Press, Oxford, 1976). 7. Szymanski, Zd., “Some Flow Problems of Rarefied Gases,” Arch. Mech. Stos. (Warsaw) 8, 449-470 (1956). 8. Ivchenko, I.N., “Generalization of the Lees Method in Boundary Problems of Transfer,” J. Coll. Interface Sci. 135(1),16-19 (1990). 9. Wang-Chang, C.S. and Uhlenbeck, G.E., “Transport Phenomena in Very Dilute Gases,” Report CM-579, UMH-3-F, University of Michigan, 1949. 10. Ziering, S., On Transport Theory of Rarefied Gases (Ph.D. Dissertation, Syracuse University, 1958).

140

Analytical Methods for Problems of Molecular Transport

11. Waldmann, L., “Uber die Kraft Eines Inhomogenen Gases auf Kleine Suspendierte Kugeln,” Z. fur Naturforsch. 14a, 589-599 (1959). 12. Brock, J.R., “The Thermal Force in the Transition Region,” J. Coll. Interface Sci. 23(3), 448-452 (1967). 13. Ivchenko, I.N. and Yalamov, Yu.I., “On the Thermophoresis of Aerosol Particles in the Almost-Free-molecular Regime,” Izv. AN SSSR, M. Zh. G. (3), 3-7 (1970). 14. Talbot, L., Cheng, R.K., Schefer, R.W., and Willis, D.R., “Thermophoresis of Particles in a Heated Boundary Layer,” J. Fluid Mech. 101(4), 737-758 (1980). 15. Williams, M.M.R., “On the Motion of Small Spheres in Gases, II: Thermo-phoresis, Diffusio-phoresis and Related Phenomena,” Z. fur Naturforsch. 27a(12), 1804-1811 (1972). 16. Mason, E.A. and Malinauskas, A.P., Gas Transport in Porous Media: The Dusty-Gas Model (Elsevier, Amsterdam, Oxford, N.Y., 1983). 17. Fuchs, N.A., Evaporation and Droplet Growth in Gaseous Media (Pergamon Press, Oxford, 1959). 18. Mason, B.J., Clouds, Rain and Rainmaking (Cambridge University Press, 1962). 19. Williams, M.M.R., “On the Motion of Small Spheres in Gases III: Drag and Heat Transfer,” J. Phys. D: Appl. Phys. 6, 744-758 (1973). 20. Wang-Chang, C.S., “Transport Phenomena in Very Dilute Gases, II,” Report CM-654, NORD 7924, UMH-3-F, University of Michigan, 1950. 21. Smoluchowski, M., “Zur Kinetischen Theorie der Transpiration und Diffusion Verdünnter Gase,” Annalen der Physik 33(16), 1559-1570 (1910). 22. Waldmann, L. and Schmitt, K.H., “Über das bei der Gasdiffusion auftretende Druckgefälle,” Z. fur Naturforsch. 16a, 1343-1354 (1961).

Chapter 8 METHODS OF SOLUTION OF PLANAR PROBLEMS

1.

MAXWELL’S METHOD.

For the last thirty years, the most progress in investigating transport problems has been achieved for one-dimensional, stationary transport problems. In this context, several analytical methods have been developed which have been applied to the classical slip problems. This has resulted in several analytical expressions for the slip coefficients. The recent development of direct numerical solutions for these same problems allows one to estimate the accuracies of the various analytical methods. For simplicity, only planar transport problems are described in this chapter and all molecules are assumed to act as rigid spheres. Although the methods that are discussed in this chapter may be generalized for arbitrary intermolecular potential models, this chapter does not include the specific details necessary for such generalizations. However, some generalization details for certain selected methods are included in Chapter 9. The reason that the details of the generalization process have been limited in this book to selected methods is that the details necessary for the generalization of each method will typically be specific to that method and some of these generalizations can be very complex and difficult to achieve. The half-range moment method is a case in point and no attempt has been made in this book to generalize it although the basic method is described in Section 8.3. In the case of planar transport problems there exist some simple approximate methods based on the use of conservation laws as exact moment solutions of the Boltzmann equation. First, consider the approximate method proposed by Maxwell to solve some boundary transport

142

Analytical Methods for Problems of Molecular Transport

Figure 8-1. The geometry of slip problems.

problems [1]. The main features of this method will be analyzed by the solving of the slip boundary problems. First of all, the velocity-slip and thermal-creep problems will be examined. One considers a semi-infinite expanse of a gas bounded by a flat plate located at x 0 , and lying in the y z plane. Far from the plate the gas is maintained at a constant velocity gradient wu wx f , normal to the plate and at a constant temperature gradient wT wy f , tangential to the wall. The geometry of these slip-flow problems is illustrated in Fig. 8-1. These gradients are assumed to be small such that the relations Ou u 1 and O ln T 1 are satisfied. Then, at large distances from the wall, the gas is described by the Chapman-Enskog distribution function. The distribution function near the wall is a discontinuous function in the velocity space. Maxwell was the first to notice this feature of the distribution function and to show that by a simple consideration of the constancy of the stress and heat flux, some meaningful results for the slip terms could be obtained. The influence of the wall near its surface may be taken into account by the introduction of the discontinuous term, ) r v, r , in the distribution function. This distribution function, for all planar and linearized transport problems, may be written as:

fr

^

`

f eq 1 < v, r ) r v, r ,

(8-1)

where f eq is the local Maxwellian distribution function given by Eq. (5-8) and < v, r is the Chapman-Enskog correction to the distribution function. For the slip-flow and thermal-creep problems the distribution functions f eq and < v, r may be written as:

f eq where:

0 y 2c y u x º , f ª1 c 2 52 yq ¬ ¼

(8-2)

Chapter 8. Methods of Solution of Planar Problems 0 f

§ m · n0 ¨ ¸ © 2S kT0 ¹

143

32

exp c 2

and:

< c, r < 1 c, r < 2 c, r c yIT c q y cx c yIu c h , where q y T01 wT wy f , T0 is the temperature in the plane y h wu wx . The remaining quantities are given by:

(8-3)

0 , and

f

12

§ m · u x ¨ ¸ u x , © 2kT0 ¹

IT c A c , and: 12

§ 2kT ·

Iu c 2 ¨ ¸ B c . © m ¹ The first approximations to the functions, A c and B c , have been found in Chapter 5 to be:

15 A c 16 S O S3 2 c 2 , 1

and: 12

B c

5 8

§ m · ¸ © 2kT ¹

SO¨

.

where molecules are assumed to be rigid spheres. The Navier-Stokes equations are used in the Maxwellian analysis to obtain the dependence of the gas velocity on the normal coordinate. These equations are the moment equations which for the current problem may be written as:

144

Analytical Methods for Problems of Molecular Transport

w vx fdv wx ³

0 ,

(8-4a)

w vx2 fdv 0 , ³ wx

(8-4b)

w vx v y fdv 0 . wx ³

(8-4c)

If the influence of the wall could be neglected ( ) r 0 in Eq. (8-1)), then from Eq. (8-4c) one would have a linear dependence of the mean velocity of the gas on the x -coordinate. Then, allowing for the velocity of the wall equal to zero, one would obtain: Pxy

const ,

wu

wx

u x

const ,

f

x wu wx

f

.

(8-5)

Now, the equation for the correction to the distribution function, Substituting Eq. (8-1) into the Boltzmann equation, one obtains:

) r c, x , may be derived.

c y c 2 52 q y 2cx c y h cx

w) r wx

J < 1 < 2 ) r .

(8-6)

Moreover, one has the following equations for the functions, IT c and Iu c :

c y c 2 52

2cx c y

J c I

c

y T

,

J cx c yIu c ,

(8-8)

where: 12

J M

n2 § m · 00 ¨ ¸ f © 2kT ¹

(8-7)

I M ,

Chapter 8. Methods of Solution of Planar Problems

145

and I M is the standard collision operator given by Eq. (5-13). Allowing for these equations, one may derive the basic equation for the correction, ) c, x :

cx

w) wx

J ) .

(8-9)

Now, consider the boundary condition for this equation. From the general form of the boundary conditions given by Eq. (6-4) one obtains for these slip problems:

) c,0 DW< 1 2 DW < 2 1 DW ) cc,0 ,

^

(8-10)

`

where cc cx , c y , cz . It is useful here to introduce a Hilbert space defined by the scalar product:

U1 x, c , U2 x, c ³ U1 x, c exp^c2 ` U2 x, c dc .

(8-11)

The Maxwellian analysis takes into account the influence of the wall by the choice of a definite form for the distribution function. Maxwell used the following approach for the distribution function of incident molecules:

) c,0 a0 c y ,

) c, f ) c, f a0 c y , where a0 const . This says that the incident distribution function at the wall is identical to the distribution function far from the wall which implies that, in the Knudsen layer, collisions between the incident and reflected molecules have been neglected. Now, to find an approximate solution for a0 , one proceeds in the following way. Taking the scalar product of Eq. (8-9) with c y (in the manner of Eq. (8-11)) gives:

w cx c y , ) c, x wx

i.e.:

0 ,

146

Analytical Methods for Problems of Molecular Transport

c c

, ) c, x

x y

const .

(8-12)

This constant can be evaluated by the use of the asymptotic values, a0 c y , for ) r c, f and thus, the relationship expressing conservation of tangential momentum may be written as:

c c

, ) r c,0

x y

c c

x y

, ) r c, f

0 .

(8-13)

The constant a0 may be found by the use of the boundary condition, Eq. (8-10), in the scalar product, Eq. (8-13). It is the main supposition of the Maxwellian analysis that the discontinuity of the distribution function is allowed for only by using the boundary distribution function for reflected molecules. The Maxwellian distribution function does not, itself, satisfy the boundary condition exactly, but rather, the boundary condition is satisfied only in an integral sense. Having performed the integration in Eq. (8-13), one obtains:

a0

2 DW

DW

5 8

§ wu · 1 § wT · 15 ¸ 32 S OT0 ¨ ¸ . © wy ¹f © wx ¹f

SO ¨

(8-14)

This expression results in relations for the slip-flow and thermal-creep coefficients. In accordance with the definitions of these coefficients, the mean velocity of the gas at large distances from the wall may be presented in the form:

c

u f

§ wT · § wu · cTslQ T01 ¨ ¸ , ¸ © wx ¹f © wy ¹f

O x ¨

m P

(8-15)

where Q QON OP and Q is the kinematic viscosity. Here, two other representations of the mean free path have been introduced; specifically OP and ON . Later, it will be useful to have the following expressions for these: 12

OP and:

8 5

P § 2kT · ¨ ¸ S p© m ¹

OP P 1 ,

(8-16a)

Chapter 8. Methods of Solution of Planar Problems 12

ON

N § mT · ¨ ¸ S p © 2k ¹

64 75

ON N 1 ,

147 (8-16b)

where P and N are the viscosity and thermal conductivity coefficients obtained from the first-order approximation in the Chapman-Enskog theory. This choice for the Maxwellian analysis provides independence of the slipflow coefficients on the order of the Chapman-Enskog solution. Since only the first-order Chapman-Enskog solution is employed in this section ( OP ON O ), the slip factors in Eq. (8-15) are given by: 1

1

cm

2 DW

DW

5 16

S ,

(8-17a)

and: 3 4

cTsl

,

(8-17b)

Next, consider the temperature-jump problem. One may use the Maxwell method to solve this problem. Consider a semi-infinite expanse of gas bounded by a plate oriented in the y z plane and located at x 0 . Assume that, far from the plate, the temperature of the gas depends only on the x -coordinate and that the temperature gradient is small such that the relationship O ln T 1 is satisfied. If this condition holds one may use the Chapman-Enskog distribution function at large distances from the wall and w ln T wx may be expressed as qx T01 wT wx , where T0 is the equilibrium temperature. The temperature is assumed to deviate only slightly from the equilibrium value. To write the correct form of the distribution function far from the wall one should take into account the conservation law obtained from the Boltzmann equation which, for this problem, is given by:

w vx v 2 fdv 0 , wx ³

(8-18)

This implies that:

³ vx v

2

fdv const .

(8-19)

148

Analytical Methods for Problems of Molecular Transport

From this equation, if x o f such that f o f eq 1 cxIT c qx , then (wT / wx )f const . The distribution function for this problem may be written as:

fr

^

`

0 x cxIT c qx c 2 52 a1r x , f 1 c 2 52 xq

(8-20)

where a1r x is a correction to the distribution function which takes into account the influence of the wall. The following approach has been used in the Maxwellian analysis:

a1r f a1 , a1 0 a1 .

(8-21)

The distribution function of the molecules incident on the wall is assumed to be the same as that far from the wall. The distribution function of the reflected molecules on the wall may be found by the use of a boundary condition, which may be written as:

f v,0

1 DW f vx , v y , vz ,0

(8-22)

0 DW f 1 Q r c 2 32 W r ,

where Q r and W r are unknown functions which may be determined by using the impenetrability condition and introducing the thermal accommodation coefficient. For the equilibrium conditions at the wall, W w 0 , but Q w z 0 . This correction to the equilibrium distribution function may be found from the condition:

³ vx f v,0 dv

°

°½

DW ® ³ vx f v,0 dv ³ vx f 0 1 Q w dv ¾ 0 . ¯°

(8-23)

¿°

From Eq. (8-23) one may obtain:

Qw

12 a1 .

(8-24)

The conservation law given by Eq. (8-19) may be written as: 2 2 ³ vx v f v,0 dv ³ vx v f v, f dv

,

149

Chapter 8. Methods of Solution of Planar Problems

and, using the boundary conditions, one can transform this relation to the form:

°

½°

DW D T ® ³ vx v 2 f v,0 dv ³ vx v 2 f 0 1 Q w dv ¾ ¯°

¿°

2

³ vx v f v, f dv

(8-25)

.

Performing the integrations in Eq. (8-25), one can obtain:

'T

a1T0

2 DW DT

DW DT

75 128

§ wT · ¸ . © wx ¹f

(8-26)

SO ¨

Thus, the Maxwellian method may be used successfully to obtain expressions for the slip coefficients. Within the framework of the moment theory, this method may be classified as a one-moment approach because only one moment has been used. The distribution function for the incident molecules at the wall is the same as that in the ambient gas at large distances from the wall and, therefore, the distribution function is independent of the normal coordinate. This does not permit one to obtain spatial distributions of macroscopic values for the gas in the Knudsen layer near the wall; a serious situation which comprises the main deficiency of the Maxwellian method.

2.

LOYALKA'S METHOD.

The Maxwellian method for determining the distribution function of incident molecules at the wall may be generalized by a simple modification suggested by Loyalka [2]. Consider the details of this generalization in the solutions of the slip-flow and temperature-jump problems. To obtain simple solutions for the slip coefficients, one considers a gas flow bounded by a flat wall for the conditions described in Section 8.1. The distribution function for these slip problems has the form:

f

^

0 y 2c y xh c yIT c q y f 1 c 2 52 yq

`

cx c yIu c h ) r c, x ,

(8-27)

150

Analytical Methods for Problems of Molecular Transport

where q y T01 wT wy f , h wu wx , and c yIT c q y and cx c yIu c h f are the Chapman-Enskog corrections to the distribution function for thermal conductivity and viscosity, respectively. The correction that allows for the influence of the wall may be found from the equation:

cx

w) r wx

J )r .

(8-28)

For x o f , the functions ) r c, x have the form: lim ) r c, x a0 c y ,

x of

where a0 is a constant. If the two-moment approach is used, then the correction, ) r c, x , may be expressed as:

) r c, x a0r x c y .

(8-29)

The Maxwellian analysis is constructed on the assumption that a0r f a0 0 a0 . Loyalka has proposed a simple generalization of Maxwell’s method which assumes that the distribution function of incident molecules has the following form at the wall:

) c,0 a0 0 c y

(8-30)

aw c y ,

where aw is an unknown constant. From a physical viewpoint, the relation aw z a0 implies that the incident stream has a mass velocity different from that of the hydrodynamic solution at the wall. The unknown constants, a0 and aw , may be found by the consideration of some scalar products of Eq. (8-28). First, the scalar product given by Eq. (8-13) is used. Next, the scalar product of Eq. (8-28) with cx c yIu c gives:

w 2 cx c yIu c , ) r c, x wx

c c I

x y u

J ª¬c c I c º¼ , ) x y u

r

c, x

c , J ª¬) r c, x º¼

2 cx c y , ) r c, x

0 ,

and, therefore:

c c I

2 x y u

c , ) r c, x

const .

(8-31)

151

Chapter 8. Methods of Solution of Planar Problems

To obtain this expression, Eq. (8-8) and the commutative property of the standard bracket integrals (which is due to the self-adjointness of the collision operator) have been used. Now there are two equations that may be represented as:

c c

x y

, ) r c,0

0 ,

(8-32)

and:

c c I

2 x y u

c , ) r c,0

c c I

2 x y u

c , ) r c, f

.

(8-33)

Using the boundary condition given by Eq. (8-10) with Eqs. (8-32) and (8-33), one can obtain:

aw

15 32

S Oqy

2 DW

DW

5 8

(8-34)

SO h ,

and:

a0

1 12 DW 1532

S Oqy

2 DW § 4 S · 5 1 DW ¸ 8 SO h . DW ¨© 2S ¹

(8-35)

Taking into account Eq. (8-15), the following expressions for the slip coefficients may be written:

cm

cTsl

2 DW § 4 S · 5 1 DW ¸ 16 S , DW ¨© 2S ¹

3 4

1 12 DW .

In particular, for the very important case when DW given by:

'u

wu · ¸ , © wx ¹f

1.116 O §¨

(8-36)

(8-37)

1 , the slip velocities are

(8-38)

152

Analytical Methods for Problems of Molecular Transport

usl

9 8

§ wT · ¸ . © wy ¹f

Q T01 ¨

(8-39)

The analytical expressions for the slip coefficients are not altered for arbitrary models of the intermolecular potential. This assumption is correct within the framework of this method if one uses the first-order ChapmanEnskog solutions for thermal conductivity and viscosity to derive these formulas. Next, consider the solution of the temperature-jump problem by Loyalka’s method. The same conditions are used for the gas at large distances from the wall as were used in the Maxwellian analysis. Therefore, the distribution function may be written as:

fr

^

`

0 x cxIT c qx ) r c, x , f 1 c 2 52 xq

(8-40)

where qx T01 wT wx f and ) r c, x is the correction that allows for the influence of the wall. This correction is assumed to have the form:

) r c, x a1r x c 2 52 ,

(8-41)

where:

lim ) r c, x constu c 2 52 .

x of

The following approach has been used in the Loyalka analysis:

a1r f a1 ,

a1 0 a1w ,

a1 z a1w ,

such that there are now two independent constants, a1 and a1w , instead of the single constant, a1 , in the Maxwellian analysis. This generalization gives a more correct form for the incident molecular distribution function at the wall than that used by Maxwell. The unknown constants, a1 and a1w , may be obtained by the use of two scalar products of the Boltzmann equation, Eq. (8-28), and the boundary conditions for the correction to the distribution function. First, the scalar product given by Eq. (8-18) results in one equation for the two unknown constants:

153

Chapter 8. Methods of Solution of Planar Problems

c c

2

x

, ) r c,0

c c x

2

, ) r c, f

0 .

(8-42)

Next, the scalar product of Eq. (8-28) with cxIT c yields:

w 2 cx IT c , ) r c, x wx

c I c , J ª¬) c, x º¼ ) c, x , J ª¬c I c º¼ ) c, x , c c c c , ) c, x 0 . r

x T r

x

T

2

r

5 2

x

2

r

x

This implies that:

c I

2 x T

c , ) r c, x

const .

(8-43)

Thus, the second equation for the unknown functions, a1 and a1w , may be written as:

c I

2 x T

c , ) r c,0

c I

2 x T

c , ) r c, f

.

(8-44)

The corrections to the distribution function of the reflected molecules at the wall may be found by using the Maxwellian model of the boundary conditions which, for this problem, may be expressed in the form:

) c,0 2 DW cxIT c qx 1 DW ) cc,0

DW ªQ r c 2 32 W r º , ¬ ¼

^

(8-45)

`

where cc cx , c y , cz , and Q r and W r are corrections to the number density and temperature for the diffusely reflected molecules. The corrections, Q r and W r , may be obtained from the three additional conditions: ³ vx f v,0 dv ³ vx f v,0 dv

0 ,

(8-46)

154

Analytical Methods for Problems of Molecular Transport ³ vx f v,0 dv ³ vx f w v,0 dv

DT

0 ,

(8-47)

Qx . Qxw

(8-48)

In these expressions, one must use the general form of the boundary condition (Eq. (8-22)). Then, using Eq. (8-41), after integration, one obtains:

Qr

75 256 S 1 DT O qx 12 2 DT a1w ,

Wr

75 128

(8-49)

S 1 DT O qx 1 DT a1w .

(8-50)

Now, Eqs. (8-42) and (8-44) may be employed to derive the relations for a1 and a1w . Substituting Eqs. (8-49) and (8-50) into Eq. (8-45), and performing the integrations indicated by Eqs. (8-42) and (8-44), one can obtain:

a1w

a1

75 128

S

2 DW DT

DW DT

O qx ,

ª D D 2 DW a1w «1 W T 2 DW DT «¬

(8-51)

52 25

º

»

S 1 12 » .

(8-52)

¼

Taking into account these expressions, one can obtain the following formula for the temperature-jump:

'T

§ wT · cT O ¨ ¸ . © wx ¹f

(8-53)

Here, cT is the temperature-jump coefficient given by:

cT

75 128

S

2 DW DT ª DW DT 2 DW «1 DW DT ¬« 2 DW DT

52 25

º

»

S 1 12 » . ¼

(8-54)

Chapter 8. Methods of Solution of Planar Problems

155

This expression for the temperature-jump coefficient given in Eq. (8-54) is more general than that reported in [2]. For the specific case when DT 1 , however, both expressions are identical. Eq. (8-54) is applicable for arbitrary models of the intermolecular potential if one uses the first-order Chapman-Enskog solution. A more complete description of the temperature-jump phenomenon may be obtained by solving Problem 8.6. The preceding analysis of the slip-flow transport problems shows that Loyalka’s method may be classified as a two-moment approximation. Although this method eliminates the main deficiency of the Maxwellian analysis, it does not permit one to obtain the dependence of macroscopic values on the normal coordinate in the Knudsen layer. Nevertheless, it is very important to note that the conservation laws used in Loyalka’s analysis are the exact moment solutions of the Boltzmann equation. Each exact solution for the correction to the distribution function must satisfy these moment relationships which are the most general properties of the Boltzmann equation. The use of exact moment solutions allows one to anticipate highly accurate results based on what is a reasonably simple analytical analysis. Moreover, for this method there would appear to be no difficulties in generalizing the analysis for arbitrary models of the intermolecular potential. To this end, it is necessary only that ChapmanEnskog solutions of sufficiently high-order be employed.

3.

THE HALF-RANGE MOMENT METHOD.

Another approximate method was first proposed by Wang-Chang and Uhlenbeck [3] and was then further developed by Gross and Ziering, among others, to solve planar boundary transport problems [4-13]. Known initially as the half-range moment method it is also now commonly referred to as the Gross-Ziering method after those principally responsible for its further development. To demonstrate the various features of this method as described in [13], the velocity-slip problem is examined. The conditions for the gas are assumed to be the same as described in Section 8.1. For the linearized problem the distribution function may be given by:

fr

½° § wu · § wu · 0 ° r f ®1 2c y x ¨ c c c I ¸ ¨ ¸ ) c, x ¾ , x y u © wx ¹f © wx ¹f ¯° ¿°

(8-55)

where ) r c, x is the correction to the distribution function which accounts for the influence of the wall and Iu c is specified by Eq. (8-3). The correction, ) r c, x , may be found from the linearized Boltzmann equation:

156

Analytical Methods for Problems of Molecular Transport

cx

w) wx

J ) ,

(8-56)

where:

J )

³ bdbd H ³ c1 c

0 f1 )1c ) c )1 ) dv1 .

In the half-range moment method, the distribution function is approximated by half-range polynomials in velocity space and the spacedependent coefficients are determined by solving half-range moment equations. It might appear that the most natural systematic scheme for construction of half-range moment equations is to introduce a complete orthonormal set of polynomials, defined over the half-range in velocity space, such that the distribution function may then be expanded in terms of these polynomials and characterized by the space-time dependent expansion coefficients. Unfortunately, this scheme requires the use of a large number of expansion terms which results in great mathematical difficulty. The complexity of the calculations increases rapidly with the number of terms and, therefore, this calculation scheme is not always practical for use in transport problems. However, the distribution function can sometimes satisfy the boundary conditions exactly such that very accurate results can be obtained by low-order approximations employing only a few terms in the expansion. Such approximations can give the exponential variation of the macroscopic values in the Knudsen layer near the wall. When the boundary conditions are not satisfied exactly, it is necessary to use the integral forms of the boundary conditions. For the current slip-flow problem, assume that the correction, ) r c, x , may be represented in the half-range velocity space as:

) r c, x a0r x c y a1r x cx c y .

(8-57)

If the following function is introduced:

1 ; cx ! 0 , sign cx ® ¯1 ; cx 0 , the correction to the distribution function, ) c, x , is conveniently expressed for the full velocity space by:

157

Chapter 8. Methods of Solution of Planar Problems

) r c, x ) a0

1 sign cx 2 2

a0

cy

a0

) 2

a0

1 sign cx 2

c y sign cx

a1 a1 cx c y 2

(8-58)

a1 a1 cx c y sign cx . 2

Substituting this correction to the distribution function into Eq. (8-56), multiplying both sides alternately by:

c y 1 r sign cx exp c 2 dc , and:

cx c y 1 r sign cx exp c 2 dc , and, finally, integrating over all velocities, one obtains the following system of half-range moment equations:

da0 dx

a0 A1 a0 A1 a1 A2 a1 A3 ,

(8-59a)

da0 dx

a0 A1 a0 A1 a1 A3 a1 A2 ,

(8-59b)

da1 dx

a0 B1 a0 B1 a1 B2 a1 B3 ,

(8-59c)

da1 dx

a0 B1 a0 B1 a1 B3 a1 B2 ,

(8-59d)

where the coefficients, Ai and Bi , may be expressed as linear combinations of some of the moments of the collision integral:

158

Analytical Methods for Problems of Molecular Transport

A1

b ª¬ I1 12 S I 2 º¼ ,

A2

b ª¬ I 2 12 S I 3 I 4 º¼ ,

A3

b ª¬ I 2 12 S I 3 I 4 º¼ ,

B1

b ª¬ I 2 12 S I1 º¼ ,

B2

b ª¬ I 3 I 4 12 S I 2 º¼ ,

B3

b ª¬ I 3 I 4 12 S I 2 º¼ ,

where:

b

4 . S 4 S

Introducing the following notation: ª¬) v ,< v º¼

2 ³) v J < v exp c dc .

the bracket integrals, I j , may be written as:

I1

ª¬ c y sign cx , c y sign cx º¼ ,

I2

ª¬c y sign cx , cx c y º¼ ,

159

Chapter 8. Methods of Solution of Planar Problems

I3

ª¬cx c y , cx c y º¼ ,

I4

ª¬cx c y sign cx , cx c y sign cx º¼ .

The following additional bracket integrals that one might expect to encounter are identically zero and do not appear in this moment system:

ª¬ cx c y , cx c y sign cx º¼

ª¬c y sign cx , cx c y sign cx º¼

0 .

The bracket integrals, I j , correspond to the original notation used by Gross and Ziering [5] and are different from the Chapman-Enskog bracket integrals usually encountered in kinetic theory. The relationship between these two kinds of bracket integrals is: 12

>) ,< @

§ m · nS 3 2 ¨ ¸ © 2kT ¹

>) ,< @

,

(8-60)

where >) ,< @ is the standard bracket integral. The values of the integrals, I j , are dependent on the intermolecular interaction potentials. A method of calculation of these integrals, for rigidsphere molecules, is given in Appendix A and is substantially simpler than previous methods [14, 15]. For rigid-sphere molecules, this method results in the following bracket integrals:

I1

1 8 2 S , I2 12 O

I3

52

S , I4 O

3S 8 2 S , S O 32

46 17 2 S . O 120

The value for the integral, I 4 , originally reported by Gross and Ziering [5] is in error. The absolute value of the Gross-Ziering integral is about 9.3 times greater than that reported here. The same analytical expressions being used in this analysis were previously reported by Porodnov and Suetin [9]. Now, the solution of the moment system, Eqs. (8-59a)-(8-59d), takes the form:

160

Analytical Methods for Problems of Molecular Transport

a0 x

A D 0 B exp D x ,

(8-61a)

a0 x

A B exp D x ,

(8-61b)

a1 x D1 B exp D x ,

(8-61c)

a1 x D1 B exp D x ,

(8-61d)

where:

4.19294 , 0.52413 ,

D 0 1

D

D1 D

4.12697 , 2.20153 O 1 .

The constants of integration, A and B , may be found from the boundary conditions (Eq. (8-10)) which may be written as:

a0 0

1 DW a0 0

,

§ wu · a1 0 2 DW Iu c ¨ ¸ 1 DW a1 0 . w x © ¹f

(8-62)

From these boundary conditions one then obtains:

§ wu · 2 DW 1 DW D 0 , ¸ © wx ¹f DW D1 1 DW D1

A

5 4

SO¨

B

5 4

SO¨

§ wu · 2 DW , ¸ x w © ¹f D1 1 DW D1

The solution of the moment system, Eqs. (8-61a)-(8-61d), can now be employed in the following expression for the mean velocity of the gas:

161

Chapter 8. Methods of Solution of Planar Problems

u x n 1 ³ v y f v, x dv 12

§ 2kT · ¨ ¸ © m ¹

º ½° 1 ° § wu · 1 ª a1 a1 » ¾ . ® x ¨ ¸ 4 « a0 a0 S ¬ ¼ °¿ °¯ © wx ¹f

(8-63)

which, at large distances from the wall, has the linear profile: 12

u x

x of

§ wu · § 2kT · x ¨ ¸ 14 ¨ ¸ w x © ¹f © m ¹

a

0

f a0 f

.

(8-64)

The solution of the Navier-Stokes equations, if one uses the slip boundary conditions, has the following form:

u x

wu · ¸ . © wx ¹f

cm O x §¨

(8-65)

A comparison of Eqs. (8-64) and (8-65) indicates that: cm

2 DW

DW

5 8

S

3.193 DW 3.603 0.524 DW

.

(8-66)

For the particular case when DW 1 , one has cm 1.125 . Additionally, for this case, the mean velocity of the gas may be expressed as:

u x

ª § 2.202 x · º § wu · « x 1.125 O 0.345 O exp ¨¨ ¸¸ » ¨ ¸ . O © ¹ ¼» © wx ¹f ¬«

(8-67)

In summary, the four-moment approximation to the half-range moment method described in this section satisfies the boundary conditions exactly and results in an exponential profile for the mean velocity of the gas in the Knudsen layer (near the wall). This equates to substantially more information about the state of the gas near the wall than is obtained from the methods of Maxwell and Loyalka described in Sections 8.1 and 8.2.

162

Analytical Methods for Problems of Molecular Transport

4.

FEATURES OF THE BOUNDARY CONDITIONS FOR THE MOMENT EQUATIONS.

In Sec. 8.3, the boundary slip-flow problem was solved by using the Maxwellian boundary conditions for the distribution function. Unfortunately, these boundary conditions have an essential deficiency connected with the execution of the conservation laws for the Boltzmann equation. It was stated in Section 8.2 that a correction to the distribution function is needed to satisfy the two conservation laws expressed by Eqs. (832) and (8-33). The satisfaction of these conditions is a general requirement of the Boltzmann equation in planar boundary transport problems. If the exact form of the boundary conditions is used, the conservation law expressed in Eq. (8-33) is not satisfied. Therefore, one must use another form of the boundary conditions. To explore this problem in more detail, consider a microscopic boundary condition with the following form: f

Af .

(8-68)

The Boltzmann equation near the flat wall may be represented by [11]:

vx

wf wx

Gf G x ª¬K vx vx f K vx vx Af º¼ , Gt

(8-69)

where G x is the Dirac delta function and: 1 ; v x ! 0 , ¯0 ; v x 0 .

K vx ®

Introducing the boundary condition into the right-hand-side of Eq. (8-69) and performing the integration of this equation over x from 0 to H ! 0 gives: H

lim ³ vx

H o0

0

wf dx vx f wx

° vx Af ; vx ! 0 , ® ; vx 0 . °¯vx f

(8-70)

From this relation one may conclude that to obtain the boundary conditions to the moment equations one should use the following form:

163

Chapter 8. Methods of Solution of Planar Problems

³ vxIi v f

³ vxIi v Af

dv

dv ,

(8-71)

where Ii are molecular properties which are used to construct the moment system. It would appear that there is now a complete set of boundary conditions for the moment equations. Unfortunately, these boundary conditions do not simultaneously satisfy both conservation laws which are expressed as Eqs. (8-32), (8-33) and (8-42), (8-44), for the slip-flow and thermal transport problems, respectively. These conservation equations are general properties of the Boltzmann equation and, therefore, to obtain the correct results, the two conservation laws must be taken into account. One possible solution to this problem is to use these conservation laws as integral expressions of the boundary conditions. To examine this idea, consider the slip-flow problem for which there exists a solution by the half-range moment method. Instead of the boundary conditions expressed by Eq. (8-62) there are two integral relations that may be written as:

c c

x y

, ) r c,0

c c I

2 x y u

0 ,

c , ) r c,0

(8-72)

c c I

2 x y u

c , ) r c, f

.

(8-73)

The correction to the reflected distribution function at the wall surface has the following form:

§ wu · ¸ © wx ¹f

) c,0 2 DW cx c yIu c ¨ 1 DW

ª a0 ¬

0 c y 0 cx c y º¼ a1

(8-74)

,

12

where Iu c 2 2kT m B c and the function, B c , corresponds to the second-order Chapman-Enskog solution. The second-order terms in this quantity improve the accuracy of the analysis. This function, for rigidsphere molecules, may be written as [17]:

¼

Iu c 54 S OP ª1 b2 2 S512 c 2 º , ¬

(8-75)

164

Analytical Methods for Problems of Molecular Transport

205 where b2 0.05854 , and OP 202 O. The correction to the distribution function has been obtained in Section 8.4 and is given by Eqs. (8-61a)-(8-61d). The use of the new boundary conditions, however, results in new values for the constants, A and B . By using the new boundary model of Eqs. (8-72)-(8-73), one obtains: 2

A

2 DW '1 DW '

5 4

S OP h ,

'2 '

B

5 4

S OP h ,

(8-76)

where '1 , ' 2 , and ' may be represented for rigid-sphere molecules by the following expressions: '1

ª¬ 0.6690 DW 0.1775 º¼ ,

(8-77a)

'2

0.3413 DW 0.1706

(8-77b)

'

,

ª¬ 0.7549 DW 0.09714 º¼ .

(8-77c)

The mean velocity of the gas is expressed by Eq. (8-63). Using Eqs. (877), the relation for the mean velocity may be given by:

§ wu · u x ¬ª cm OP x OP ud x ¼º ¨ ¸ © wx ¹

,

(8-78)

f

where cm is the isothermal-creep coefficient and ud x is the dimensionless velocity defect that describes the deflection of the mean velocity from the linear profile in the Knudsen layer. The expressions for these quantities, with an arbitrary tangential momentum accommodation coefficient, may be written as: cm

2 DW

DW

5 8

S

'1 , '

ud x 1.4229

§ 2.2015 x · '2 exp ¨¨ ¸¸ . ' O © ¹

(8-79)

(8-80)

165

Chapter 8. Methods of Solution of Planar Problems

Table 8-1. Values of the slip-flow coefficient,

cm

DW 0.2 0.4 0.6 0.8 1.0 †

G%

aa

Eq. (8-79) 9.0710 4.1310 2.4650 1.6185 1.1006

cm

Eq. (8-81) 9.0277 4.0976 2.4400 1.6006 1.0884

a u 100 ; a

cm .

G

†

0.48 0.81 1.01 1.11 1.11

cm

G

Eq. (8-66) 9.1235 4.1759 2.5027 1.6496 1.1254

†

0.80 1.51 2.14 2.71 3.22

cm (from Eq. (8-79).

Table 8-1 contains the comparison of results which are obtained by the use of the various methods for several values of the coefficient, DW . Results obtained by Loyalka’s analysis take into account the correction of the second-order to the Chapman-Enskog distribution function, and therefore the expression for the slip coefficient differs from that represented by Eq. (835). For this case, Loyalka’s formula has the form:

cm

2 DW § 4Z1 S · 5 1 DW ¸ 16 S ; Z1 1 b2 2 174 b2 2 ¨ DW © 2S ¹

2

.

(8-81)

The Gross-Ziering analysis does not permit the use of the second-order Chapman-Enskog solution owing to the form of the boundary conditions. The comparison of these results shows a good agreement between the values of cm calculated by the use of Eq. (8-79) and by Loyalka’s formula. Equation (8-66), which was obtained using the common forms of the boundary conditions, yields results that do not agree so well with Loyalka’s formula. However, this comparison does not allow one to draw any rigorous conclusions regarding the accuracy of any these models because the exact analytical solutions are not known. It should be possible, however, to draw conclusions regarding the accuracy of the different methods by comparing their results to reasonably exact direct numerical solutions of this slip-flow problem. This will be considered in Section 9.3, where a complete analysis of the accuracy of the various methods and boundary models will be made.

5.

SOLUTION OF THE THERMAL-CREEP PROBLEM BY THE HALF-RANGE MOMENT METHOD.

The flow of a gas in the y -direction, over a plane surface lying at x 0 , is considered for the conditions described in Section 8.1. Far from the plate

166

Analytical Methods for Problems of Molecular Transport

the gas is maintained at a constant temperature gradient, wT wy f . For these conditions, the thermal-creep problem is described by the following distribution function:

f

0 y c y A c q y ) r c, x º , f ª1 c 2 52 yq ¬ ¼

(8-82)

where A c may be written as [17]:

1 2 2 2 A c a1 ª S3 2 c 2 a2 S3 2 c 2 º , ¬ ¼ 45 where a2 0.08889 , a1 15 S ON , and ON 44 O . In this formula 16 molecules are assumed to be rigid spheres. The integral form of the boundary conditions allows one to employ the same correction to the distribution function for the thermal-creep problem as in the slip-flow problem. This form of the distribution function takes into account the viscosity effects in a sufficiently correct form. Let the correction to the distribution function be assigned by the relation: 2

2

) r c, x a0r x c y a1r x cx c y ,

(8-83)

where the functions, air may be written as:

a0 x C1 D 0 C2 exp D x ,

(8-84a)

a0 x C1 C2 exp D x ,

(8-84b)

a1 x D1 C2 exp D x ,

(8-84c)

a1 x D1 C2 exp D x .

(8-84d)

The values, D ir , D , for rigid spheres are given by:

D 0 D1

4.19294 , 0.52413 ,

D1 D

4.12697 ,

2.20153 O 1

.

167

Chapter 8. Methods of Solution of Planar Problems

The constants C1 and C2 may be obtained from boundary conditions which have the following form:

c c

x y

, ) r c,0

c c I

2 x y u

0 ,

(8-85a)

c , ) r c,0

c c I

2 x y u

c , ) r c, f

.

(8-85b)

where ) c,0 is represented by:

¼

) c,0 DW a1 2 q y c y ª S312 c 2 a2 2 S3 22 c 2 º ¬

(8-86)

1 DW ª¬ a0 0 c y a1 0 cx c y º¼ , and the standard notation for the scalar product from Eq. (8-11) has been used. In the case of Iu c , Eq. (8-75) is used. This integral form of the boundary conditions makes allowance for the features of the Boltzmann equation near a flat wall. Having performed integration in Eqs. (8-85) one obtains: C1

'1 , '

C2

'2 , '

where '1 , ' 2 , and ' may be written as:

'1

2 a1 q y ª¬ 0.4354 0.2179 DW º¼ ,

'2

0.3022 DW a1 2 q y

'

,

0.8518 0.1096 DW

.

For this problem it is convenient to represent the mean velocity of the gas in the form:

168

Analytical Methods for Problems of Molecular Transport

u x cTslQ q y ½ ª º' 1 D1 D1 » 2 exp D x ¾ , u ®1 12 « D 0 1 S ¬ ¼ '1 ¯ ¿

(8-87)

where cTsl is given by: 3 2

cTsl

0.4354 0.2179 DW 0.8518 0.1096 DW

(8-88)

.

The dimensionless velocity defect may be defined as:

ud x

u f u x

Q q y

(8-89)

.

This quantity may then be expressed by:

ud x DW

§ 2.2015 x · 0.5822 exp ¨¨ ¸¸ O 0.8518 0.1096 DW © ¹

.

(8-90)

For the particular case when DW 1 , one obtains cTsl 1.0193 . The full analysis of results obtained here will be described later. It is of considerable interest to investigate the influence of the various forms of the boundary conditions on the thermal-creep coefficient and the velocity defect. This analysis will be described in the next section.

6.

INFLUENCE OF THE BOUNDARY MODELS ON THE THERMAL-CREEP COEFFICIENT.

Here, expressions are derived for the thermal-creep coefficient and the velocity defect for the case when one uses the ordinary moment form of the boundary conditions that may be represented by: 2 2 ³ cx c y) c,0 exp c dc ³ cx c y A) c,0 exp c dc

,

(8-91a)

169

Chapter 8. Methods of Solution of Planar Problems 2 2 2 2 ³ cx c y) c,0 exp c dc ³ cx c y A) c,0 exp c dc

,

(8-91b)

where:

1 2 2 2 A) c,0 DW a1 q y ª S3 2 c 2 a2 S3 2 c 2 º ¬ ¼

(8-92)

1 DW ª¬ a0 0 c y a1 0 cx c y º¼ . Substituting the functions given by Eqs. (8-84a)-(8-84d) into Eqs. (8-91a) and (8-91b), one obtains the constants C1 and C2 which determine the correction to the distribution function. This analysis yields: cTsl

3 2

1.5464 1.1741 DW 3.0927 0.4499 DW

ud x DW

,

§ 2.2015 x · 3.4149 exp ¨¨ ¸¸ O 3.0927 0.4499 DW © ¹

(8-93)

.

(8-94)

For example, if DW 1 , then cTsl 1.1519 . This value is about 15% greater than that given by Eq. (8-96). But a reliable analysis of the accuracy of the model considered here, as was indicated previously, may only be performed by making a comparison with the numerical solution for this boundary transport problem. The accuracy of the various methods and boundary models for the thermal-creep problem will be investigated in Section 9.5.

PROBLEMS 8.1. Determine the mean velocity profile, u x , and the pressure tensor component, Pxy , for Couette flow by means of the Maxwell method. Derive an expression for the apparent viscosity coefficient, P c , that is determined by:

Pc

uw d

P

wu . wx

170

Analytical Methods for Problems of Molecular Transport

Figure 8-2. The geometry to be used in Problem 8.1 in determining the mean velocity profile, the pressure tensor component, and the apparent viscosity for Couette flow.

Use the flow geometry in Fig. 8-2. Solution: For this analysis the distribution function should be chosen in the form: 0 y , f 1 cx c yIu c A 2 Axc

^

f

`

which provides a linear profile for the mean velocity of the gas. Conservation of tangential momentum at the wall may be written as:

c c

x y

, ) r c, 12 d

0 .

Using the Maxwellian boundary condition for the rebounded distribution function, one can obtain: uw x] Kn , d

u x

] Kn

§ 2 DW 5 O · S ¸ ¨1 DW 8 d ¹ ©

Pxy

uw , d

P c

and P c P] Kn .

1

,

171

Chapter 8. Methods of Solution of Planar Problems

8.2. Determine the mean velocity profile, u x , and the pressure tensor component, Pxy , for Couette flow by means of Loyalka’s method. Derive an expression for the apparent viscosity coefficient, P c , that is determined by:

Pc

uw d

P

wu . wx

Use the same flow geometry as in Problem 8.1 (see Fig. 8-2). Solution: One should use a distribution function of the form:

f

^

`

0 f 1 ) r c, x ,

where ) r c, x 12 a0 x a0 x c y 12 a0 x a0 x c y sign cx . As a result of this choice, the boundary conditions are satisfied exactly. The conservation laws may then be expressed in the form cx c y , ) r c, x A1 and cx2 c yIu c , ) r c, x 2 A1 x A2 where A1 and A2 are constants which may be found from the two boundary conditions. From the conservation laws, one obtains:

a0 x uw

5 8

1

S

O

2

d 2 DW

5 8

DW

x d

S

,

O d

and:

a0 x uw

5 8

1

S

O

2

x d

5 8

S

d 2 DW

DW

O

.

d

Using these expressions, one can obtain the same relations for u x and Pxy as in Problem 8.1. However, the distribution function obtained here describes the Knudsen limit regime better than that obtained in Problem 8.1. 8.3. Determine the mean velocity profile, u x , and the pressure tensor component, Pxy , for Couette flow by means of the Gross-Ziering method for the two-moment approach. Take the correction to the distribution function in the form ) r c, x a0r x c y . Derive an expression for the apparent viscosity coefficient, P c , that is determined by:

172

Analytical Methods for Problems of Molecular Transport

Pc

uw d

Pxy .

Use the same flow geometry as in Problem 8.1 (see Fig. 8-2). Solution: For this two-moment approach, the half-range moment system is obtained by multiplying the Boltzmann equation, Eq. (8-56), by c y ª¬1 r sign cx º¼ exp c 2 dc and integrating over all velocities. The solution of this moment system yields:

uw x] Kn , d

u x

] Kn

§ 2 DW S · ¨1 ¸ DW I1d ¹ ©

12

p S § m · I1 ¨© 2kT ¸¹

Pxy

I1

1

,

uw ] Kn , d

1 8 2 S , 12 O

and P c P] Kn where P

0.2909 U v O .

8.4. Using the full-range moment equations given by: d vx Q v fdv dx ³

n' Q v ,

where Q v vx and vx v y , determine the mean velocity, u x , and the pressure tensor component, Pxy , for the Couette flow. Solve this problem for the two-moment approach using the following form of the correction to the distribution function ) r c, x a0r x c y . Derive an expression for the apparent viscosity coefficient, P c , that is determined by:

Pc

uw d

Pxy .

Chapter 8. Methods of Solution of Planar Problems

173

Use the same flow geometry as in Problem 8.1 (see Fig. 8-2). Solution: The moment equations are obtained by multiplying the Boltzmann equation, Eq. (8-56), by the terms, c y and cx c y , and integrating over all velocities. The solution of the moment system then yields: uw x] Kn , d

u x

§

] Kn ¨1 ©

2 DW S 3 2 · ¸ DW 2 I 2 d ¹ 12

Pxy

I2

1 2

pS § m · I 2 ¨© 2kT ¸¹

1

,

uw ] Kn , d

3S 8 2 S , S O 32

and P c P] Kn where P 0.5100 U v O . In this solution, molecules have been assumed to be rigid spheres. 8.5. Determine the temperature distribution between parallel plates by the Maxwell method. Derive an expression for the apparent thermal conductivity coefficient N c , that is given by:

Nc

2'T d

N

wT . wx

Use the flow geometry in Fig. 8-3. The correction, 'T , to the temperature, T0 , is assumed to be small such that the relationship 'T T0 is satisfied. Solution: A distribution function should be sought in the form:

f

^

1 0 f 1 cx S3 2 c 2

15 16

`

S O A c 2 52 Ax .

Using the boundary conditions obtained in Problem 6.5 and conservation of the heat flux as given by:

174

Analytical Methods for Problems of Molecular Transport

Figure 8-3. The geometry to be used in Problem 8.5 in determining the temperature distribution and apparent thermal conductivity coefficient between parallel plates having different temperatures.

c c x

2

, ) c, 12 d

0 ,

one can obtain W x qx] T Kn ,

q

]T

2'T , T0 d

§ 75 2 DW DT O · ¨1 64 S ¸ DW DT d ¹ ©

1

,

and N c N] T Kn where W x is a correction to the temperature. 8.6. Derive analytical formulas for the boundary temperature and pressure jumps for a planar surface in a monatomic non-condensable gas using Loyalka’s method. Use the second-order Chapman-Enskog solution for rigid-sphere molecules. Assume that molecules reflect diffusely at the surface. Solution: The following distribution function should be used to describe this boundary problem:

f

0 x cx A c qx ) r c, x º . f ª1 c 2 52 xq ¬ ¼

The correction, ) r c, x , should be taken in the form:

175

Chapter 8. Methods of Solution of Planar Problems

) r c, x a r x c 2 52 b r x . The limiting values of the functions introduced here are given by: lim a r x a f , lim b r x b f ,

x of

x of

a 0 aw , and b 0 bw . These limiting values are specified by the boundary-jump effects. The unknown constants, a f , b f , aw , and bw , may be determined from the conservation laws of Eq. (8-9) which are given by:

c

x

, ) r c,0

c c x

2

0 , cx2 , ª¬) r c,0 ) r c, f º¼

, ) r c,0

0 ,

0 , and cx2 A c , ª¬) r c,0 ) r c, f º¼

0 .

The distribution function of reflected molecules (see Problem 6.5) is given by:

ª 2G · º 2G §G ) c,0 cx A c qx 1 DT « 2 c 2 ¨ 2 1 ¸ » 1 , S ¹¼ S ©S ¬

where:

G1 12 S aw 12 bw , and G 2 S

5 8

S a1 2 qx aw 12 bw .

This formulation yields:

Tasy 0 T0 T0

cT ON qx , and

pasy 0 p0 p0

The jump coefficients are then given by:

c p ON qx .

176

Analytical Methods for Problems of Molecular Transport

cT

75 128

S

2 DT ª DT § 52 H T 1 · º » , «1 DT ¬ 2 DT ¨© 25 S 2 ¸¹ ¼

cp

75 256

S

2 DT ª D T § 8 H1 1 · º » , «1 DT ¬ 2 DT ¨© 5 S 2 ¸¹ ¼

and:

2

,

2 23 2 433 where H T 1 26 a2 208 a2 for rigid-sphere molecules.

H1 1 14 a2 2 , and a2 2

0.08889

8.7. Derive analytical expressions for the boundary temperature and pressure jumps in a monatomic vapor over a planar, liquid surface using Loyalka’s method. Use the second-order Chapman-Enskog solution for rigid-sphere molecules. Consider the case of an arbitrary evaporation coefficient. Solution: The distribution function used should be:

0 x 2cx u cx A c qx ) r c, x º , f ª1 c 2 52 xq ¬ ¼

f

12

where u m 2kT0 u x . The correction, ) r c, x , should be taken in the The limiting values of the form ) r c, x a r x c 2 52 b r x . functions introduced here are given by:

lim a r x a f , lim b r x b f ,

x of

x of

a 0 aw , and b 0 bw . These limiting values are specified by the boundary-jump effects. The unknown constants, a f , b f , aw , and bw , may be determined from the conservation laws of Eq. (8-9) which are given by:

c

x

, ) r c,0

c c x

2

, ) r c,0

0 , cx2 , ª¬) r c,0 ) r c, f º¼

0 ,

0 , and cx2 A c , ª¬) r c,0 ) r c, f º¼

0 .

Chapter 8. Methods of Solution of Planar Problems

177

The boundary conditions (see Problem 6.6) for the reflected distribution function may be expressed as:

) c,0 cx A c qx 1 D m ª 2G · º 2G °½ §G ° u ® 2 c 2 «1 DT ¨ 2 1 ¸ » 1 ¾ , S ¹ ¼ S ¿° ©S ¬ ¯°

where:

G1

12 S aw 12 bw S u

and:

G2 S

5 8

S a1 2 qx aw 12 bw 54 S .

This formulation yields:

Tasy 0 T0

T u cT ^D m ` ON qx cT ^D m ` u ,

T0 and:

pasy 0 p0 p0

T u cp ^D m ` ON qx cp ^D m ` u ,

where the jump coefficients are given by: T cT ^D m `

75 128

S

D § 52 H T 1 · º 2 D ª , «1 D ¬ 2 D ¨© 25 S 2 ¸¹ »¼

ª 4 3D 16 H1 º u cT ^D m ` 18 S « 5 » , S¼ ¬ D

178

Analytical Methods for Problems of Molecular Transport T cp ^D m `

75 256

S

D § 8 H1 1 · º 2 D ª , «1 D ¬ 2 D ¨© 5 S 2 ¸¹ »¼

and:

ª 4 3D m 1 § 4 3D u cp ^D m ` 12 S « 8¨ © D ¬ Dm Here, D 2 and a2

·º ¸» . ¹¼ 2

,

2 23 2 433 1 26 a2 208 a2 0.08889 for rigid-sphere molecules.

D m DT 1 D m , H T

H1 1 14 a2 2 ,

REFERENCES 1. Maxwell, J.C., “On Stresses in Rarefied Gases Arising from Inequalities of Temperature,” Phil. Trans. Roy. Soc. (London) 170, 231 (1879). 2. Loyalka, S.K., “Approximate Method in the Kinetic Theory,” Phys. Fluids 14(11), 22912294 (1971). 3. Wang-Chang, C.S. and Uhlenbeck, G.E., Transport Phenomena in Very Dilute Gases (VMH-3-F, University of Michigan, 1949). 4. Gross, E.P., Jackson, E.A., and Ziering, S., “Boundary Value Problems in Kinetic Theory of Gases,” Ann. Phys. 1(2), 141-167 (1957). 5. Gross, E.P. and Ziering, S., “Kinetic Theory of Linear Shear Flow,” Phys. Fluids 1(3), 215-224 (1958). 6. Gross, E.P. and Jackson, E.A., “Kinetic Theory of the Impulsive Motion of an Infinite Plane,” Phys. Fluids 1(4), 318-328 (1958). 7. Gross, E.P. and Ziering, S., “Heat Flow Between Parallel Plates,” Phys. Fluids 2(6), 701712 (1959). 8. Bakanov, S.P. and Derjaguin, B.V., “On the State of a Gas Moving Near a Solid Surface,” Dokl. AN SSSR 139(1), 71-74 (1961). 9. Porodnov, B.T. and Suetin, P.E., “Rarefied Gas Flow Between Two Parallel Plates,” Izv. AN SSSR, M. Zh. G. (6), 93-98 (1967). 10. Ivchenko, I.N. and Yalamov, Yu.I., “The Kinetic Theory of a Gas Flow Over a Solid Wall in a Velocity Gradient Field,” Izv. AN SSSR, M. Zh. G. (6), 139-143 (1968). 11. Rolduguin, V.I., Application of the Non-Equilibrium Thermodynamics Method in Boundary Problems of the Kinetic Theory of Gases (M.S. Thesis, Moscow, 1979). 12. Savkov, S.A., Slip Boundary Conditions for Non-Uniform Binary Gas Mixtures and their Application to Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 13. Yalamov, Yu.I. and Ivchenko, I.N., “Isothermal Heat Flux in the Knudsen Layer,” J. Phys. Chem. (Russian) 43(10), 2622-2624 (1969). 14. Derjaguin, B.V., Ivchenko, I.N., and Yalamov, Yu.I., “About Construction of Solutions of the Boltzmann Equation in the Knudsen Layer,” Izv. AN SSSR, M. Zh. G. (4), 167-171 (1968). 15. Derjaguin, B.V. and Yalamov, Yu.I., “The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and Their Experimental Testing,” In Topics in

Chapter 8. Methods of Solution of Planar Problems

179

Current Aerosol Research, vol. 3, part 2, edited by Hidy, G.M. (Pergamon Press, Oxford, 1972). 16. Loyalka, S.K. and Lang, H., “On Variational Principles in the Kinetic Theory,” in Proceedings of the Seventh International Symposium on Rarefied Gas Dynamics (Editrice Tecnico Scientifica, Pisa, Italy, 1970). 17. Loyalka, S.K., “Slip and Jump Coefficients for Rarefied Gas Flows: Variational Results for Lennard-Jones and n(r)-6 Potentials,” Physica A 163, 813-821 (1990).

Chapter 9 THE VARIATIONAL METHOD FOR THE PLANAR GEOMETRY

1.

ANOTHER FORM OF THE BOLTZMANN EQUATION.

In the previous chapter the basic analytical methods for the planar boundary transport problems were considered. All of these schemes are considered to be moment methods. It is very important to note that the simple analysis of some general properties of the Boltzmann equation related to the conservation of moments results in sufficiently accurate expressions for the velocity-slip and temperature-jump coefficients. The Maxwell method and its generalization proposed by Loyalka allow one to analyze some integral properties of a gas, for which one can expect reliable results. The physical basis of these methods is not questionable. But within the framework of these methods, as has already been noted, it is impossible to describe the gas behavior in the Knudsen layer. Nevertheless, Loyalka’s method of taking into account two conserving moments may be used to calculate the gas parameters both at large distances from the wall and directly on its surface. In this chapter a simple and accurate method for dealing with planar boundary transport problems will be described by applying a variational technique to the linearized Boltzmann equation. The analysis will be confined to a consideration of the slip and thermal-creep problems which are the classical problems used to analyze the main features and accuracy of this approach. Since the main interest is in linearized transport problems, the distribution function can be written as:

182

Analytical Methods for Problems of Molecular Transport 0 f ª¬1 < c, r ) r c, x º¼ ,

f

(9-1)

where < c, r is the Chapman-Enskog correction for the appropriate problems and x is the normal coordinate. The correction to the distribution function, ) r c, x , taking into account the influence of the wall, may be specified from the linearized Boltzmann equation that is given by Eq. (8-9). The specific features of the variational method are related to the integral form of Boltzmann’s equation. To derive the integral equation for the distribution function it is convenient to introduce another form of Boltzmann’s equation that, for stationary problems, is defined by:

wf wr

v

1 2 J J ,

(9-2)

where the collision operator is presented as a sum of two separate terms which are given by:

J 1

2 J

³ bdbd H ³ gf cf1cdv1

,

(9-3)

f ³ bdbd H ³ gf1dv1 .

(9-4)

For molecules of infinite range of interaction these integrals diverge, since they include among the colliding molecules some which interact at arbitrarily large distance with infinitely small changes of state [1]. These integrals converge for the rigid-sphere molecules, and, therefore, can be calculated separately. If the interaction potential falls off rapidly enough, distant collisions may be neglected by means of the use of the definite model for a ‘truncated’ potential having a finite interaction range. For planar, stationary, linearized transport problems, Eq. (9-2) may be expressed in the form:

cx

w ) c, x V c ) c, x wx

H) c, x .

(9-5)

Here, V c is a function depending only on the magnitude of a molecular velocity vector and H H c, cc is an operator defined by the relation:

H) c, x H ³ dcc exp cc2 K c, cc ) cc, x ,

(9-6)

183

Chapter 9. The Variational Method for the Planar Geometry

in which K c, cc is a symmetric kernel. The parameter H is defined as:

H

1

1

2S O

(9-7)

.

It is very important to emphasize that both the function, V c , and the kernel, K c, cc , can be defined by analytical expressions if molecules are assumed to be rigid spheres. A very simple integration in Eq. (9-4) gives:

V c HQ c

1

ª º 1·1 § 2 «exp c ¨ 2c c ¸ 2 S erf c » . O 2S ¬ © ¹ ¼

(9-8)

One can find an analytical expression for K c, cc in [2, 3]. This expression for the kernel, K c, cc , is not utilized in the variational analysis and, thus, it is sufficient to know only the general definition for the operator, H .

2.

THE VARIATIONAL TECHNIQUE FOR THE SLIP-FLOW PROBLEM.

In this section the classical slip-flow problem will be analyzed by means of a variational technique that was described by a consideration of the planar boundary transport problems in Loyalka’s works [4-7]. The statement of this problem is given in Sec. 8.1 and hence, here, the main focus will be on investigating the features and the accuracy of the variational technique. Let the distribution function be described by:

fr

^

`

0 cx c yIu c h ) r c, x . f 1 2c y xh

(9-9)

12

where Iu c 2 2kT m B c . The correction, ) r c, x , to the distribution function, which takes into account the perturbation of a gas flow by the wall, has the following asymptotic behavior: lim ) r c, x 2c y u0 ,

x of

(9-10)

where u0 is a dimensionless constant. Then, one uses the Maxwellian boundary model for a description of the gas-surface interaction. This boundary model is given by:

184

Analytical Methods for Problems of Molecular Transport

) c,0 2 DW cx c yIu c h 1 DW ) cx , c y , cz ,0 .

(9-11)

Now, one can use the form of the Boltzmann equation given in Eq. (9-5) to derive the integral equation for ) r c, x . Taking into account Eqs. (910) and (9-11) and performing a formal integration of this linear nonhomogeneous differential equation of the first-order, one obtains the following integral equation:

) c, x L) c, x pu c, x ,

(9-12a)

where L is an integral operator and:

§ V c · pu c, x K cx 2 DW cx c yIu c h exp ¨¨ x ¸¸ , © cx ¹

(9-12b)

1 ; x ! 0 , ¯0 ; x 0 .

K x ®

The operator, L , is defined by the relation:

° 1 x § V c · L ) c, x K cx ® ³ H) c, xc exp ¨¨ xc x ¸¸ dxc © cx ¹ ¯° cx 0 1 DW

f § V c · ½ 1 H ) c, xc exp ¨¨ xc x ¸¸ dxc¾° ³ cx 0 © cx ¹ °¿

(9-13)

f § V c · ½ ° 1 K cx ® ³ H) c, xc exp ¨¨ xc x ¸¸ dxc°¾ , © cx ¹ ¿° ¯° cx x

where H c, cc RH H cx , c y , cz , cc and R is the reflection operator. If the operator, H , is applied to the function, ) c, x , one obtains another function, H ) c, x , that is given by:

H ) c, x H ³ dcc exp cc2 K c, cc ) cc, x , where the following notation is introduced in this integrand:

(9-14)

185

Chapter 9. The Variational Method for the Planar Geometry

K c, cc

RK c, cc .

Since K c, cc is a symmetric function of c, cc one can conclude that:

K c, cc K cc, c .

(9-15)

From both the symmetry of this kernel and Eq. (9-14), it is shown [8] that the operator, H , is self-adjoint, i.e. H H , in the spaces defined by the scalar products:

] 1 c, x , ] 2 c, x ³ dc] 1 exp c2 ] 2

,

(9-16)

f

ª] 1 c, x , ] 2 c, x º ¬ ¼

³ dx ] 1 , ] 2 .

(9-17)

0

Now, in order to construct a variational principle, one must first express u0 as an integral of ) c, x . For this purpose one employs the conservation of moments of the Boltzmann equation as was done in the previous analysis. These properties of the Boltzmann equation can be expressed as:

c c

x y

, ) c, x

c c I

2 x y u

(9-18)

0 ,

c h , ) c,0

c c I

2 x y u

c h , ) c, f

.

(9-19)

Using the asymptotic solution and Eq. (9-19) one can obtain:

u0

^

1 2

cx2 c y u

I c h , cy

(9-20)

`

u 2 DW cx3c 2yIu2 c h 2 , K cx ª¬ H ) c, x , pu c, x º¼ . To obtain an integral relation for the dimensionless velocity, the first-order Chapman-Enskog solution for Iu c is used. This simplifies the algebraic calculations in Eq. (9-20) and results in the following integral relation:

186

Analytical Methods for Problems of Molecular Transport

u0

2 DW 85 O h

8

5S 2 O h ³

exp c 2 dc (9-21)

f

u ³)

c, xc pu c, xc dxc

,

0

where ) c, x H ) c, x . Now, one may construct the functional:

ª¬)

I )

, ) L) 2 pu º ¼

F ) ,) .

As is shown in Appendix D, this functional is stationary for ) ) , where ) is a solution of Eq. (9-12a). If the stationary value of I ) is introduced in Eq. (9-21) one can obtain:

u0

2 DW 85 O h

8 I st ) 5S 2 O h

(9-22)

cm O h .

Consider this functional in detail. It does not involve the real function,

) c, x , and, therefore, the variational principle can be immediately

employed to determine the best value of a given trial function. Since the main purpose is to obtain a solution for the mean velocity, u0 , which is directly related to the stationary value taken by I ) , one may expect that a sufficiently accurate approximation for u0 can be determined by considering even some simple trial functions. One simple trial function that takes into account the asymptotic behavior of ) c, x is:

) c, x D c y ,

(9-23)

) c, x DV c c y ,

(9-24)

where D is a ‘so-called’ variational parameter. Eq. (9-24) can be obtained directly when the trial function, D c y , is substituted into Eq. (9-5). Having inserted this trial function into the functional and performed a very simple integration, one can obtain the following expression:

I D

1 4

S ª¬D 2DW D 2 DW 54 SO h º¼ .

(9-25)

187

Chapter 9. The Variational Method for the Planar Geometry

The stationary value for I D occurs if dI dD

D

2 DW

DW

5 8

0 , so that:

SO h .

(9-26)

From Eq. (9-22) one can obtain the expression for the slip coefficient that may be written as:

cm

2 DW § 4 S · 5 1 DW ¸ 16 S . DW ¨© 2S ¹

(9-27)

It is very important to note that this formula is exactly the same as the expression derived by Loyalka’s method. Now, consider how one might describe the gas flow in the Knudsen layer. It should be emphasized that the variational method, even employed in its simplest form, permits one to perform this analysis. The main interest here is to investigate the velocity defect expression that is derived by taking a moment of the distribution function. The correction to the distribution function, ) c, x , obtained from Eq. (9-12) has the following form for a given trial function:

§ V c · x ¸¸ © cx ¹

°

) c, x K cx ® 2 DW cx c yIu c h exp ¨¨ °¯

ª § V c · º °½ x ¸¸ » ¾ K cx D c y . D c y «1 DW exp ¨¨ © cx ¹ ¼» °¿ ¬«

^

(9-28)

`

To obtain the right asymptotic form for this correction, insert D 2u0 into Eq. (9-28) instead of the value of D given by Eq. (9-26). This gives:

§ V c · x ¸¸ © cx ¹

) c, x 2u0 c y K cx exp ¨¨ u ª¬ 2 DW cx c y

5 4

S Oh

(9-29)

2u0 c y W

D º¼ .

Allowing for Eq. (9-29), one can obtain the following expressions for both the mean velocity and the velocity defect of a gas:

188

Analytical Methods for Problems of Molecular Transport

2c x 5 º § wu · ª u x ¨ ¸ O «cm DW 3m2 J1 x 2 DW J 2 x » , O 4S S © wx ¹f ¬ ¼

(9-30)

º § wu · ª 2c ud x ¨ ¸ O «DW 3m2 J1 x 2 DW 54 S 1J 2 x » . © wx ¹f ¬ S ¼

(9-31)

Here, the functions, J1 x and J 2 x , are defined by: f ª § V c · § V c ·º J1 x S ³ c 4 exp c 2 « E2 ¨¨ x ¸¸ E4 ¨¨ x ¸¸ » dc , ¹ © c ¹ ¼» 0 ¬« © c

(9-32)

f ª § V c · § V c ·º J 2 x S ³ c5 exp c 2 « E3 ¨¨ x ¸¸ E5 ¨¨ x ¸¸ » dc , c c « ¹ © ¹ »¼ 0 ¬ ©

(9-33)

where the functions, En z , are the exponential integrals that can be expressed in the form: 1

En z

³t 0

n2

§ z· exp ¨ ¸ dt . © t¹

(9-34)

As can be easily seen, the variational method allows one to investigate the Knudsen layer. The accuracies of this method and the other methods described previously will be considered in the next section.

3.

DISCUSSION OF THE SLIP-FLOW RESULTS.

There are several expressions for the mean velocity and slip coefficient that have been obtained here by employing various methods and using various boundary models. To establish the accuracy of these methods it is necessary to compare the approximate analytical results with direct numerical results. For this purpose, we have found it convenient to use the numerical results of Loyalka and Hickey [9]. The analytical results that we are comparing are those that take into account the second-order corrections to the Chapman-Enskog distribution functions and, moreover, the molecules are assumed in the analysis to be rigid spheres.

Chapter 9. The Variational Method for the Planar Geometry

189

It is easy to see that both Loyalka’s method and the variational method give the same results for the slip coefficient if the trial function given by Eq. (9-23) is employed in the variational analysis. These methods result in the following relation:

cm

2 DW

DW

5 16

S ª¬1 0.1086 DW º¼ .

(9-35)

The velocity defect cannot be obtained by Loyalka’s method with the exception of one point for which x 0 . Since the expression for aw given by Eq. (8-34) is not altered, if the second-order Chapman-Enskog solution is used, one can obtain the velocity defect at the wall given by:

ud 0

5 32

ª 4Z1

S 2 DW «

¬ S

2

1 S

1 2 b 2 2

º»¼

0.2759 2 DW

. (9-36)

2

where Z1 1 b2 174 b2 . The half-range moment method (four-moment approach) gives different results for various boundary models. The boundary model proposed by Ivchenko, Loyalka, and Tompson [10] (the ILT boundary model) results in the following expressions for cm and ud x : 2

cm

2 DW

DW

ud x

5 8

2

ª 0.6690 0.1775 DW º » , ¬« 0.7549 0.09714 DW ¼»

(9-37)

0.3413 0.1706 DW 0.7549 0.09714 DW

(9-38)

S«

1.4229

x · § exp ¨ 2.2015 ¸ . O¹ ©

The same method for the Maxwellian boundary model gives:

cm

2 DW

ud x

DW

5 8

º 3.1929 DW » «¬ 3.6028 0.5241 DW »¼

ª

S«

1.4229

,

2 DW x · § exp ¨ 2.2015 ¸ . O¹ 3.6028 0.5241 DW ©

(9-39)

(9-40)

190

Analytical Methods for Problems of Molecular Transport

Table 9-1. The slip-flow coefficient,

Four-Moment Approach ILT Maxwellian Boundary Model [10] Boundary Model [10] cm G cm G †

Numerical Values [9]

†

DW

cm

0.2 0.4 0.6 0.8 1.0

9.0458 4.1147 2.4531 1.6096 1.0940

G%

a anum

cm .

9.0710 4.1310 2.4650 1.6185 1.1006

DW

cm

0.2 0.4 0.6 0.8 1.0

9.0458 4.1147 2.4531 1.6096 1.0940

a anum

0.86 1.49 2.02 2.48 2.87

cm .

Numerical Values [9]

G%

9.1235 4.1759 2.5027 1.6496 1.1254

anum u 100 ; a cm

Table 9-2. The slip-flow coefficient,

†

0.28 0.40 0.49 0.55 0.60

Loyalka Method & Variational Results, Eq. (9-35) G † cm 9.0277 4.0976 2.4400 1.6006 1.0884

0.20 0.42 0.53 0.56 0.51

anum u 100 ; a cm

To compare the accuracy of the various expressions for the slip coefficient the analytical and numerical results are presented together in Tables 9-1 and 9-2. It can be seen that a very good agreement exists between analytical and numerical values of cm for the ILT boundary model. The discrepancy of these results is less than 0.6% for all DW . The maximum deviation of the other analytical data amounts to 0.56% and 2.87% for the variational method and the Maxwellian boundary model, respectively, and, therefore, Eqs. (9-35) and (9-37) are the most accurate analytical formulas for the slip coefficient. Another macroscopic parameter which has a significant influence on the description of the gas flow in the Knudsen layer is the velocity defect at the wall. The values of this parameter are presented in Tables 9-3 and 9.4. For this parameter the best agreement with numerical values in the full region of DW is obtained from the ILT boundary model. The maximum difference for this model, for all DW , is about 5.5%, while the discrepancies associated with the Loyalka and variational methods are 8.4% and 18.5% respectively for DW 1.0 , and range up to 16.9% and 42.2% respectively for DW 0.2 . The numerical and analytical data for the dimensionless velocity defect are presented in Table 9-5.

191

Chapter 9. The Variational Method for the Planar Geometry

Table 9-3. The velocity defect at the wall, Numerical Values [9]

†

Four-Moment Approach ILT Maxwellian Boundary Model [10] Boundary Model [10]

DW

ud 0

G †

ud 0

G

ud 0

0.2 0.4 0.6 0.8 1.0

0.5976 0.5180 0.4422 0.3701 0.3013

0.5645 0.4895 0.4181 0.3500 0.2851

5.5 5.5 5.5 5.4 5.4

0.6908 0.5972 0.5085 0.4245 0.3448

15.6 15.3 15.0 14.7 14.5

G%

a anum

anum u 100 ; a ud 0

Table 9-4. The velocity defect at the wall, Numerical Values [9]

†

ud 0 .

ud 0 .

Loyalka Method Eq. (9-36)

Variational Method Eq. (9-31)

DW

ud 0

G †

ud 0

G

ud 0

0.2 0.4 0.6 0.8 1.0

0.5976 0.5180 0.4422 0.3701 0.3013

0.4967 0.4414 0.3863 0.3311 0.2759

16.9 14.8 12.6 10.5 8.4

0.3452 0.3283 0.3061 0.2784 0.2455

42.2 36.6 30.8 24.8 18.5

G%

a anum

anum u 100 ; a ud 0

An analysis of Table 9-5 shows that for the ILT boundary model a satisfactory agreement with the numerical results occurs if 0 xc 0.5 . The maximum deviation from the numerical values for this domain of xc is about 11.5%. When xc ! 0.5 the velocity defect decreases more rapidly than the corresponding numerical value. In this region the discrepancy of the results is very large. This occurs because the asymptotic velocity defect behavior for this approach is described by one exponential term which does not characterize a real dependence on the normal coordinate [11]. This dependence is more complicated than a simple exponential drop, as one can see from Eq. (9-12). Moreover, when xc ! 0.5 , the velocity defect values become extremely small so that the moment approach employed does not have a sufficient accuracy to describe this very subtle quantity. The variational method does not accurately describe the velocity defect because a very simple trial function has been used. The results for the mean velocity profile are shown in Table 9-6. For the ILT model the maximum deviation is 2.92%. For the Maxwellian boundary model this difference is 3.77%.

192

Analytical Methods for Problems of Molecular Transport

Table 9-5. The dimensionless velocity defect,

ud xc , ( DW

Dimensionless Normal Coordinate

Numerical Values [9]

Four-Moment Approach ILT Maxwellian Boundary Model [10] Boundary Model [10]

xc

ud xc

ud xc

ud xc

0.3013 0.1997 0.1560 0.0888 0.0430 0.0233 0.0133

0.2851 0.2226 0.1738 0.0827 0.0240 0.0070 0.0020

0.3448 0.2692 0.2102 0.1000 0.0290 0.0084 0.0024

†

0.0 0.1 0.2 0.5 1.0 1.5 2.0 †

5 8

x

1.12423 O xc

S OP xc

Table 9-6. The mean velocity,

†

0.0 0.1 0.2 0.5 1.0 1.5 2.0 †

x

#

G%

5 8

4.

S OP xc

u xc , ( DW

1 ).

Four-Moment Approach ILT Maxwellian Boundary Model [10] Boundary Model [10]

Numerical Values [9]

xc

1 ).

u xc

u xc

G #

u xc

G

0.7924 1.0049 1.1594 1.5589 2.1586 2.7322 3.2961

0.8155 0.9888 1.1484 1.5718 2.1844 2.7553 3.3142

2.92 1.60 0.95 0.83 2.58 2.31 0.55

0.7806 0.9670 1.1368 1.5793 2.2042 2.7787 3.3386

1.49 3.77 1.95 1.31 2.11 1.70 1.29

a anum

1.12423 O xc

anum u 100 ; a u xc

THE VARIATIONAL SOLUTION FOR THE THERMAL-CREEP PROBLEM.

The thermal-creep problem has been stated in detail in the previous chapter, and therefore, only a brief outline will be given here. Let the distribution function be described by:

f

^

`

0 y c yIT c q y ) r c, x . f 1 c 2 52 yq

Here, IT c A c and may be written as:

(9-41a)

Chapter 9. The Variational Method for the Planar Geometry

IT c A c

15 16

¼

S ON ª S312 c 2 a2 2 S3 22 c 2 º , ¬

193 (9-41b)

45 where ON 44 O and a2 0.08889 . The correction, ) r c,0 , to the distribution function can be specified from Eq. (9-5) and its boundary conditions which can be expressed as: 2

) c,0 DW c yIT c q y 1 DW ) cx , c y , cz ,0 ,

(9-42)

lim ) r c, x 2c y u0 .

(9-43)

x of

As it is shown in Section 9.2, one can obtain the following integral equation for ) r c,0 :

) c, x L) c, x pT c, x .

(9-44)

Here the integral operator, L) c, x , is given by Eq. (9-13) and pT c, x is defined by the relation:

ª § V c ·º pT c, x K cx «DW c yIT c q y exp ¨¨ x ¸¸ » . «¬ © cx ¹ »¼

(9-45)

This function represents the external non-uniformity of a gas in Eq. (9-44). Now, taking into account the conservation of moments of the Boltzmann equation given by Eqs. (9-18) and (9-19) one can derive the following integral equation for the mean velocity:

u0

DW q y h cx2 c 2yIu c IT c , K cx I

2h c y , cx2 c yIu c

,

(9-46)

where I ª¬ H ) c, x , pu c, x º¼ and pu c, x is the known function given by Eq. (9-12a) that is proportional to the non-uniformity of a gas for the slip-flow problem. The standard notations given by Eqs. (9-16), (9-17), and (9-14) are used in this expression. The functional contained in Eq. (9-46) is a linear form with respect to the trial function ) c, x and, therefore, it cannot be immediately used for solution of a variational problem. A new functional must be constructed on

194

Analytical Methods for Problems of Molecular Transport

the basis of the square form of the trial functions and which has a stationary value equal to I . Let this new functional be the sum of two functionals, the first of which is equal to [() , pu (c, x ))] . Since pu is proportional to O h and ) ~ V (c)O q y ~ q y , the integrand of the second functional must be proportional to O q y h , and moreover this term is equal to zero if the trial function coincides with the real correction, ) c, x . This indicates that the integrand of the second functional has to contain the factor ) L) pT , which is proportional to O q y . Only the unique square form for this integrand, which is the product of ) u , where ) u is a trial function for the slip-flow problem, and ) L) pT , can be constructed so that it should be proportional to V c O 2 hq y . This very simple dimensional analysis shows that the following form can be used for this variational problem:

F ) u ,)

ª ) , pu º ª ) u , ) L) pT º . ¬ ¼ ¬ ¼

(9-47)

The stationary value of this functional is equal to I . Now, the asymptotic solutions of Eqs. (9-12) and (9-44) can be chosen as the trial functions for F ) u ,) which can be expressed in the form:

) D1c y

, )

V c D1c y ,

(9-48a)

) u D 2 c y , ) u V c D 2 c y .

(9-48b)

It is very important to note that different constants, D1 and D 2 , are used for these trial functions as they are asymptotic solutions of different equations. The first-order Chapman-Enskog solutions for Iu c and IT c are assumed to be employed in a later analysis and one can use unity instead of Iu c as it is easy to see from Eq. (9-46). In accordance with this suggestion one can obtain the following relationship:

F D1 , D 2

(9-49)

A1D1 A2D1D 2 A3D 2 .

Here, the following notations are introduced: A1

2 DW 18 S 3 2

, A2

DW 14 S , A3

DW

15 128

S 3 2O q y .

(9-50)

The desired values of D1 and D 2 for which this functional is stationary can be defined from the following equations:

Chapter 9. The Variational Method for the Planar Geometry

wF wD1

wF wD 2

0 ,

195

0 .

The stationary value of the functional can be expressed in the form:

Fst ) u ,)

15 S 2O q y 2 DW 254

.

(9-51)

Using Eq. (9-51) in Eq. (9-46), one can obtain:

u0

(9-52)

cTslQ q y .

The thermal-creep coefficient in this formula is given by:

cTsl

3 4

1 12 DW .

(9-53)

This expression coincides exactly with the one obtained by Loyalka’s method. Substituting the asymptotic value of the mean velocity 2u0 c y instead of D1c y into an expression of the distribution function, one can obtain the relation for the velocity defect defined by:

ud x

ud x

2cTsl

DW ®

Q qy

¯S

32

J1 x

3

S

32

½ J 3 x ¾ . ¿

(9-54)

Here, J1 x is given by Eq. (9-32) and J 3 x is defined by: f

J 3 x S ³ c 4 0

5 2

c 2 exp c 2

ª § V c · § V c ·º x ¸¸ E4 ¨¨ x ¸¸ » dc , u « E2 ¨¨ ¹ © c ¹ ¼» ¬« © c

(9-55)

where the notations in Eq. (9-55) are the same as in Eqs. (9-32) and (9-33).

196

Analytical Methods for Problems of Molecular Transport

5.

DISCUSSION OF THE THERMAL-CREEP RESULTS.

The accuracy of the thermal-creep problem results obtained by various approximate methods will be discussed here. This analysis can be performed by means of a comparison of analytical values with the numerical results obtained by Loyalka [12]. First, the accuracy of the various expressions of the thermal-creep coefficient will be examined. All of the analytical expressions for this coefficient and the velocity defect will be given here in the most convenient form for this analysis. Loyalka’s method and the variational solution give the same results for the thermal-creep coefficient. If the second-order Chapman-Enskog solution is used, these methods both yield:

cTsl

3 4

ªH1 H 2 12 H1 DW º , ¬ ¼

(9-56)

The variational where H1 1 14 a2 and H 2 1 72 b2 72 a2 b2 . velocity defect obtained by means of the simplest trial functions is given by Eq. (9-54). The velocity defect at the wall obtained by Loyalka’s method can be written in the form: 2

ud 0

3 4

2

2

2

(9-57)

DW H 2 .

In the previous chapter the various solutions of the thermal-creep problem were obtained within the framework of the half-range moment method for the different boundary models. The boundary model proposed by Ivchenko, Loyalka, and Tompson [13, 14] results in the following expressions for cTsl and ud x : cTsl

ud x

3 2

0.4354 0.2179 DW 0.8518 0.1096 DW

(9-58)

,

0.5822 DW x · § exp ¨ 2.2015 ¸ O¹ 0.8518 0.1096 DW ©

.

(9-59)

For the integral boundary model that is usually employed in transport problems, these relations can be written as:

197

Chapter 9. The Variational Method for the Planar Geometry

Table 9-7. The thermal-creep coefficient,

Numerical Values [12] #

† #

DW

cTsl

0.2 0.4 0.6 0.8 1.0

0.8234 0.8733 0.9207 0.9657 1.0089

a anum

G%

cTsl .

Four-Moment Approach ILT Integral Boundary Model [13] Boundary Model [13] Eq. (9-58) Eq. (9-60) G %† G% cTsl cTsl 0.8223 0.8752 0.9255 0.9735 1.0193

0.1 0.2 0.5 0.8 1.0

anum u 100 ; a cTsl

cTsl . Loyalka Method & Variational Results, Eq. (9-56) G %† cTsl

Numerical Values [12] #

#

1.7 5.8 9.1 11.8 14.2

These additional numerical values have been obtained by Loyalka.

Table 9-8. The thermal-creep coefficient,

†

0.8395 0.9240 1.0041 1.0799 1.1519

DW

cTsl

0.2 0.4 0.6 0.8 1.0

0.8234 0.8733 0.9207 0.9657 1.0089

G%

a anum

0.8120 0.8573 0.9027 0.9480 0.9933

1.4 1.8 2.0 1.8 1.5

anum u 100 ; a cTsl

These additional numerical values have been obtained by Loyalka.

cTsl

ud x

3 2

1.5464 1.1741 DW 3.0927 0.4499 DW

,

3.4149 DW x · § exp ¨ 2.2015 ¸ O¹ 3.0927 0.4499 DW ©

(9-60)

.

(9-61)

Now, consider the accuracy of the various analytical results for this problem. The numerical and analytical data for the thermal-creep coefficient versus the accommodation coefficient, DW , are presented in Tables 9-7 and 9-8. It is easy to see that the most accurate values of cTsl are predicted by the moment method for the ILT boundary model. In this case, the maximum discrepancy between the numerical and analytical results is about 1.0%, while Loyalka’s method and the variational method give results that differ from the numerical results by 1.4% for DW 0.2 up to 2.0% when DW 0.6 .

198

Analytical Methods for Problems of Molecular Transport

Table 9-9. The velocity defect at the wall,

Numerical Values [12] #

† #

ud 0 .

Four-Moment Approach ILT Integral Boundary Model [13] Boundary Model [13] Eq. (9-59) Eq. (9-61)

DW

ud 0

ud 0

G %†

ud 0

G%

0.2 0.4 0.6 0.8 1.0

0.1596 0.3078 0.4458 0.5748 0.6960

0.1333 0.2600 0.3807 0.4958 0.6056

16.5 15.5 14.6 13.7 13.0

0.2146 0.4174 0.6093 0.7913 0.9640

34.5 35.6 36.7 37.7 38.5

G%

a anum

anum u 100 ; a ud 0

These additional numerical values have been obtained by Loyalka.

Table 9-10. The velocity defect at the wall, Numerical Values [12] #

† #

ud 0 .

Loyalka Method Eq. (9-57)

Variational Method Eq. (9-54)

DW

ud 0

ud 0

G %†

ud 0

G%

0.2 0.4 0.6 0.8 1.0

0.1596 0.3078 0.4458 0.5748 0.6960

0.1220 0.2440 0.3660 0.4880 0.6100

23.6 20.7 17.9 15.1 12.8

0.0825 0.1800 0.2925 0.4200 0.5625

48.3 41.5 34.4 26.9 19.2

G%

a anum

anum u 100 ; a ud 0

These additional numerical values have been obtained by Loyalka.

The integral boundary model values deviate greatly from the numerical results (from 1.7% when DW 0.2 up to 14.2% when DW 1.0 ). In Tables 9-9 and 9-10 values of the velocity defect at the wall are presented. A comparison of these velocity defect values shows that the discrepancy between the analytical and numerical results is noteworthy for all of the analytical methods discussed. Loyalka’s method and the ILT boundary model have the best agreement with the numerical values. The maximum deviations of these values are 23.6% and 16.5%, respectively, when DW 0.2 and are seen to decrease to only 12.8% and 13.0%, respectively, when DW 1.0 . Values of the velocity defect versus the normal coordinate for DW 1.0 are presented in Table 9-11. It can be seen that the relative deviations of all of the analytical values from the corresponding numerical values are very large for xc t 0.5 . The velocity defect behavior inside the Knudsen layer seems to be more complicated [15] than the behavior predicted by these analytical formulae and, moreover, the accuracy of these methods appears insufficient overall to describe this extremely subtle quantity.

199

Chapter 9. The Variational Method for the Planar Geometry

Table 9-11. The dimensionless velocity defect, Numerical Values [12]

xc

ud xc

ud xc

ud xc

0.6960 0.3840 0.2539 0.1747 0.1229 0.0632 0.0349

0.6056 0.3052 0.0933 0.0366 0.0143 0.0022 0.0003

0.9640 0.3783 0.1485 0.0583 0.0229 0.0035 0.0005

†

x

15 16

S ON xc

1.69944 O xc

Table 9-12. Numerical values of the velocity defect,

ud xc .

Velocity Defect,

xc

†

1 ).

Four-Moment Approach ILT Boundary Integral Boundary. Model [13] Model [13] Eq. (9-59) Eq. (9-61)

Dimensionless Normal Coordinate 0.00 0.25 0.50 0.75 1.00 1.50 2.00 †

ud xc , ( DW

x

†

DW =0.2

DW =0.4

DW =0.6

0.00 0.25 0.50 0.75 1.00 1.50 2.00

0.1596 0.0840 0.05499 0.03768 0.02645 0.01355 0.007017

0.3078 0.1639 0.1076 0.0738 0.05184 0.02659 0.01404

0.4458 0.2400 0.1579 0.1085 0.07629 0.03915 0.02069

15 16

S ON xc

ud xc

DW =0.8 0.5748 0.3126 0.2062 0.1419 0.09984 0.05127 0.02712

DW =1.0 0.6960 0.3840 0.2539 0.1747 0.1229 0.0632 0.0349

1.69944 O xc

The results for the mean velocity profile are shown in Fig. 9-1. For the ILT model the maximum deviation of the mean velocity is approximately 13% which occurs when xc 0 , while for the integral model this discrepancy is 40%. It is clearly necessary to use even more exact approaches if one is to succeed in obtaining a good overall agreement between the analytical and numerical results for the mean velocity profile. In Table 9-12 the numerical results for the velocity defect are presented. These data have been obtained by Loyalka by the use of the numerical technique reported in [12]. The accuracy of any approximate method can be examined by using these data.

200

Analytical Methods for Problems of Molecular Transport

1.2

Dimensionless Mean Velocity

1.0

0.8

0.6

0.4

0.2

0.0 0.0

0.5

1.0

1.5

2.0

Dimensionless Normal Coordinate

Figure 9-1. The dimensionless, mean velocity profile, u ( xc) u ( xc)/Q q as a function of the dimensionless normal coordinate, xc , where x (1.69944)O xc . Numerical values [12] are shown as black circles, ILT Boundary Model results [13] are shown as gray diamonds, and Integral Boundary Model results [13] are shown as open squares. Curves are shown for continuity only in order to help the reader visualize the shape of the profile.

6.

SLIP-FLOW AND TEMPERATURE-JUMP COEFFICIENTS FOR THE LENNARD-JONES (612) POTENTIAL MODEL.

In the framework of the variational and generalized Maxwellian methods all the slip and temperature-jump coefficients depend upon a potential model if one uses the second- and higher-order Chapman-Enskog approximations to describe the gas state beyond the wall. For the second-order ChapmanEnskog approximation both these methods yield:

201

Chapter 9. The Variational Method for the Planar Geometry

Slip Coefficient Values (dimensionless)

1.14

1.12

1.1

cm

1.08

1.06

1.04

c Tsl 1.02 0.1

1

10

100

1000

Reduced Temperature (dimensionless)

Figure 9-2. The dependence of the slip coefficients, cm and cTsl , on the reduced

temperature, T , for the Lennard-Jones (6-12) potential model )(assuming that DW 1 ).

cm

cTsl

cT

2 DW

DW

3 4

5 16

§

S ¨1 ©

4Z1 S DW 2S

· ¸, ¹

ªH1 H 2 12 H1 DW º , ¬ ¼

2 DW DT

DW DT

75 128

ª

§

¬

©

S «1 ¨ 52 25

H T 1 · DW DT 2 DW º » S 2 ¸¹ 2 DW DT ¼

(9-62)

(9-63)

(9-64)

where the following notations have been introduced:

Z1 1 b2 2 174 b2 2

2

,

(9-65)

202

Analytical Methods for Problems of Molecular Transport

Temperature Jump Coefficient (dimensionless)

2.15

2.10

cT

2.05 0.1

1

10

100

1000

Reduced Temperature (dimensionless)

Figure 9-3. The dependence of the temperature-jump coefficient, cT , on the reduced

temperature, T , for the Lennard-Jones (6-12) potential model (assuming that DW DT 1 ).

H1 1 14 a2 2 ,

(9-66)

H 2 1 72 b2 2 72 a2 2 b2 2 ,

(9-67)

2 23 2 433 H T 1 26 a2 208 a2

2

.

(9-68)

The slip and jump coefficients depend upon the reduced temperature (see 2 2 Section 5.9), T * , through their dependence on a2 and b2 . Using the data given in Tab. 5-1, one can calculate the appropriate values of cm , cTsl , and cT for the Lennard-Jones (6-12) potential model.

203

Chapter 9. The Variational Method for the Planar Geometry

The variation of cm , cTsl , and cT with the reduced temperature, T * , is illustrated in Figs. 9-2 and 9-3. All the curves have a maximum when T * is about unity. The relative deviations of the maximum values from those, when T * is large, are about 10%, 3% and 2% for cTsl , cT , and cm , respectively. This analysis shows that the thermal-creep coefficient which is dependent on the cross effects (viscosity and thermal conductivity) is quite sensitive to the parameters of the Lennard-Jones (6-12) potential. Now, examine dependence of the slip and jump coefficients on the intermolecular potential model. The analysis is restricted to consideration of the rigid-sphere and Lennard-Jones (6-12) models. For rigid-sphere molecules the potential well depth, H kT T , is equal to zero and, therefore, all results depending on intermolecular interactions are very close to those for the Lennard-Jones (6-12) model in which T o f . This shows that only the thermal-creep coefficient is fairly sensitive to the potential model. The maximum relative discrepancy between cTsl for these two models is about 12% if DW 1 and T is about unity. It is very important to note that cTsl is independent of the temperature, T , for the rigid-sphere model.

PROBLEMS 9.1. Determine the slip and thermal-creep coefficients using the Maxwellian method for the second-order Chapman-Enskog solution. Solution: The distribution function may be written as:

f

^

`

f eq 1 cx c yIu c h c yIT c q y ) r c, x ,

where ) c,0 ) r c, f a0 c y and the functions Iu c and IT c are given by Eqs. (8-75) and (9-41b), respectively. The constant, a0 , may be found from the relation cx c y , K cx A) c,0 K cx ) c,0 0 . Having performed some simple integrations, one can obtain:

cm

2 DW

DW

5 16

S ,

and cTsl 34 H1 , where H1 1 14 a2 1.02222 . Thus, for a Maxwellian analysis, the usage of the second-order Chapman-Enskog solution results in an increase of the thermal-creep coefficient by 2.2% while the slip-flow coefficient is not altered. 2

204

Analytical Methods for Problems of Molecular Transport

9.2. Determine the temperature-jump at the wall using the Maxwellian method for the second-order Chapman-Enskog solution. Consider the case when DW 1 . Solution: The distribution function is given by:

f

^

`

0 x c 2 52 cxIT c qx ) r c, x , f 1 xq

where IT c

and:

S ON S312 c 2 a2 2 S3 22 c 2

15 16

) r c, f ) c,0 a1 c 2 52 . The temperature-jump may be calculated by employing the general moment solution of the Boltzmann equation given by cx c 2 , ) r c,0 0 . Since:

2 2 k 2 2 ³ cx c S3 2 c exp c dc

0 ;

kt2 ,

the temperature-jump is not altered for arbitrary orders of the ChapmanEnskog solution and may be expressed in the form:

'T

2 DT

DT

75 128

§ wT · ¸ © wx ¹f

SON ¨

;

ON

45 44

O .

9.3. Determine the temperature-jump coefficient using the Maxwellian method and the general form of the boundary conditions given by Eqs. (612) and (6-15). Use the second-order Chapman-Enskog solution. Solution: Following the solution of Problem 9.2, one obtains:

cT

2 DW DT

DW DT

75 128

S .

9.4. Derive expressions for the velocity defect at the wall for both the slipflow and thermal-creep problems by using the Maxwellian method. Solution: The velocity defect at the wall for these problems is given by:

ud 0 OP wu wx f

cm

u 0

OP wu wx f

,

205

Chapter 9. The Variational Method for the Planar Geometry

and:

ud 0

Q qy

cTsl

u 0

Q qy

; Q Q

ON . OP

Using Maxwellian boundary conditions to calculate the mean velocity at the wall, u 0 , one can obtain for the slip and thermal-creep problems:

ud 0 OP wu wx f

5 32

ª ¬

S 2 DW «1

2

S

1

1 2 b 2 2

º»¼ ,

and:

ud 0

Q qy

3 8

DW H1 ,

respectively, where:

H1

1

1 4

2 a2

9.5. Determine an expression for the temperature defect at the wall by using the Maxwellian method for the first- and second-order Chapman-Enskog solutions. Obtain the result for DW 1 and compare with the values from [12]. Solution: The temperature defect at the wall is given by:

§ wT · Td 0 Tasy 0 T 0 cT ON ¨ ¸ T0W 0 , © wx ¹f

where W 0 23 S 3 2 ³ c 2 32 ) r c,0 exp c 2 dc . Having performed the necessary integrations and algebraic transformations, one obtains:

Td 0

Td 0 ON wT wx f

75 256

¼

2 28 1 a º» . S 2 DW ª«1 15 S 1 13 28 2

¬

If DW 1 in this expression, then Td 0 0.3509 , while the value reported in [12] is Td [12] 0 0.6374 . The temperature defect for the first-order

206

Analytical Methods for Problems of Molecular Transport

Chapman-Enskog approximation may be formally obtained from the general 2 expression if a2 0 . This then results in Td 0 0.3735 which is an improvement in agreement with the value reported in [12]. 9.6. Derive a formula for the temperature-jump coefficient by Loyalka’s method for the second-order Chapman-Enskog solution and using the general form of the boundary conditions given by Eqs. (6-12) and (6-15). Solution: The correction to the distribution function is given by:

) r c, f a1 c 2 52 , and ) c,0 aw c 2 52 , where a1 z aw . The two constants, a1 and aw , may be found from the two conservation laws given by Eqs. (8-42) and (8-43). Boundary conditions are described by Eqs. (6-12) and (6-15). Integration yields the following expression:

cT

75 128

where H T

S

2 DW DT ª DW DT 2 DW «1 DW DT ¬ 2 DW DT

23 2 1 26 a2

433 208

a 2 2

2

52 25

º

S 1H T 12 » ,

0.9378

¼

for rigid-sphere molecules.

9.7. Generalize Loyalka’s expression for the temperature defect at the wall using two accommodation coefficients and the second-order ChapmanEnskog solution. Compare your analytical results with the numerical results [12] for the first- and second-order Chapman-Enskog approximations when DW 1 . Solution: Employing the same techniques as in Problem 9.5, one can obtain:

Td 0

75 256

¼

2 28 1 S 2 DW ª« 104 S 1H T 15 S 1 13 a º» , 25 28 2

¬

where H T 0.9378 . If DW 1 , then the values obtained for the first- and second-order Chapman-Enskog approximations are Td 0 0.6719 and Td 0 0.5734 , respectively. These values represent deviations from the numerical value reported in [12] of 5.4% and 10%, respectively.

207

Chapter 9. The Variational Method for the Planar Geometry Table 9-13. Velocity defect data obtained during the solution of Problem 9.8.

DW

1

0.8

0.6

0.4

0.2

ud >[email protected] 0

0.3013

0.3701

0.4422

0.5180

0.5976

0.2850

0.3420

0.3990

0.4560

0.5130

5.4

7.6

9.8

12.0

14.2

ud

0

deviation, %

9.8. Using the integral representation of the distribution function given by Eq. (9-12a), derive an expression for the velocity defect at the wall for the slip-flow problem if a trial function containing two independent constants is chosen in the form:

) K cx ª¬D c y º¼ K cx ª¬D c y E c y exp V x cx º¼ . For different values of DW , compare these results with those obtained in [9]. Use the second-order Chapman-Enskog solution. Solution: To obtain an approximate solution for the distribution function one can employ the approximate method described in [16,17]. Having substituted the trial function into Eq. (9-12a), one can obtain:

) c, x K cx ª¬ 2 DW cx c yIu c h exp V x cx D c y 1 DW exp V x cx 1 DW E c y exp V x cx º¼ K cx ª¬D c y E c y exp V x cx º¼ . The velocity defect at the wall may then be written as:

ud 0 OP wu wx f

5 8

2 DW Z1 12

,

where:

Z1 1 b2 2 174 b2 2

2

0.95602

.

For various values of DW , the velocity defect data are shown in Table 9-13.

208

Analytical Methods for Problems of Molecular Transport

9.9. For Problem 9.8, derive an expression for the velocity defect. Solution: The mean velocity is given by: 12

§ 2kT · 3 2 u x ¨ c y) c, x exp c 2 dc . ¸ S ³ m © ¹

After integration of the distribution function presented in Problem 9.8, one can obtain:

ud x OP wu wx f

5 4

ª J1 x

2 DW «

¬S

32

2Z1

J 2 x º » , S ¼

where Z1 0.95602 and the functions, J1 x and J 2 x , are given by Eqs. (9-32) and (9-33), respectively. 9.10. Using the same trial function as in Problem 9.8, derive an expression for the velocity defect for the thermal-creep problem. In particular, determine the velocity defect at the wall. Solution: Integration of the distribution function given in Problem 9.8 results in:

ud x

ud x

Q q

ª J x J x º 3DW « 3 3 2 H 2 1 3 2 » , S ¼ ¬S

where the functions, J1 x and J 3 x , are given by Eqs. (9-32) and (9-55), respectively. The velocity defect at the wall may then be written as 2 2 2 This ud 0 34 DW H 2 where H 2 1 72 b2 72 a2 b2 0.81332 . expression is exactly the same as that obtained via Loyalka’s method.

REFERENCES 1. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, NY, 1969). 2. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “On the Collision Kernels for Gas Mixtures,” Ann. Nucl. Energy 23(18), 1489-1495 (1996). 3. Loyalka, S.K. and Hickey, K.A., “Plane Poiseuille Flow: Near Continuum Results for a Rigid Sphere Gas,” Physica A160, 395-408 (1989). 4. Loyalka, S.K., “Momentum and Temperature-Slip Coefficients with Arbitrary Accommodation at the Surface,” J. Chem. Phys. 48(12), 5432-5436 (1968). 5. Loyalka, S.K. and Ferziger, J.H., “Model Dependence of the Temperature Slip Coefficient,” Phys. Fluids 11(8), 1668-1671 (1968).

Chapter 9. The Variational Method for the Planar Geometry

209

6. Loyalka, S.K., “Slip in the Thermal Creep Flow,” Phys. Fluids 14(1), 21-24 (1971). 7. Loyalka, S.K., “The Slip Problems for a Simple Gas,” Z. Naturforsch. 26a, 964-972 (1971). 8. Morse, P.M. and Feshbach, H., Methods of Theoretical Physics (McGraw-Hill, NY, 1953). 9. Loyalka, S.K. and Hickey, K.A., “The Kramers Problem: Velocity Slip and Defect for a Hard Sphere Gas with Arbitrary Accommodation,” ZAMP 41, 246-253 (1990). 10. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “The Precision of Boundary Models in the Gas Slip Problem,” High Temperature 31(1), 127-129 (1993). 11. Loyalka, S.K., “Velocity Profile in the Knudsen Layer for the Kramer’s Problem,” Phys. Fluids 18(12), 1666-1669 (1975). 12. Loyalka, S.K., “Temperature Jump and Thermal Creep Slip: Rigid Sphere Gas,” Phys. Fluids A 1(2), 403-408 (1989). 13. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “A Boundary Model for the Thermal Creep Problem,” Fluid Dynamics 28(6), 876-878 (1993). 14. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “On the Use of Conservation Laws in Plane Slip Problems,” Teplofizika Vysokikh Temperatur (Russian) 33(1), 66-72 (1995). 15. Loyalka, S.K., “Velocity Profile in the Thermal Creep Slip Problem,” Phys. Fluids 19(10), 1641-1642 (1976). 16. Loyalka, S.K., “An Approximate Method in Transport Theory,” Phys. Fluids 24(10), 1912-1914 (1981). 17. Loyalka, S.K. and Cipolla, J.W.Jr., “On Choice of Trial Functions in Integro-Differential Variational Principles of Transport Theory,” Nucl. Sci. Engineering 99, 118-122 (1988).

Chapter 10 THE SLIP-FLOW REGIME

1.

BASIC EQUATIONS.

In this chapter, the description of a particle in a rarefied gas is considered under conditions where the Knudsen number is small. The analysis is restricted to the usual conditions assumed for aerosol particle motion in nonuniform gases. These conditions will be discussed later in detail. The classical sphere drag and thermal force problems are solved as important practical applications of the theory and techniques described here. In this regime, an aerosol particle disturbs the surrounding gas out to a distance on the order of the particle radius. The behavior of this particle can be described using the Navier-Stokes continuum equations except for the thin Knudsen layer in the neighborhood of the particle where the disturbance is greatest. Since, in the slip-flow regime, the typical dimension of the system is much greater than the mean free path, the continuum equations are assumed to be suitable for all distances. The local features of the gas in the Knudsen layer are then taken into account in the boundary conditions which have a special form discussed in Section 6.4. The closed system of continuum equations derived in Chapters 2 and 5 may be written as:

wu Dn n i Dt wxi

0 ,

(10-1)

212

Analytical Methods for Problems of Molecular Transport

wPij

Dui · § 0 , U ¨ Fi Dt ¸¹ wx j ©

DT Dt

2 § wui wqi · ¨ Pij ¸ , Nk n ¨© wx j wxi ¸¹

Pij

§ wu wu j 2 wuk · pG ij P ¨ i G ¸ , ¨ wx j wxi 3 ij wxk ¸ © ¹

Qi

N

wT , wxi

(10-2)

(10-3)

(10-4)

(10-5)

where P and N are the viscosity and thermal conductivity coefficients that are given by Eqs. (5-53) and (5-52), respectively, D Dt is a time-derivative following the motion, as in hydrodynamics, and N is the number of degrees of freedom of a molecule. For monatomic gases, N 3 , while for other gases, N ! 3 . Each of these equations is a non-linear partial differential equation, and therefore, there are many difficulties in solving this system in the most general form. Thus, one considers possible methods that might be used to linearize the equations. In this analysis, the non-linear and linear terms are estimated using the technique described in [1]. The force on a stationary particle is proportional to the external parameters of the surrounding nonuniform gas flow. Such parameters are the mean gas velocity for the sphere drag problem and the temperature gradient for the thermal force problem The following conditions are assumed to be satisfied in the statement of the above mentioned problems: 1. the gas mean velocity is much less than the heat velocity of a molecule; 2. the relative alteration of the temperature is much less than unity for a distance of the same order as the mean free path; 3. the relative temperature variation is small for a typical dimension distance, R . These conditions governing the external parameters can be expressed mathematically as:

213

Chapter 10. The Slip-Flow Regime

U 1, V

O T f T

1 ,

R T f T

1 .

(10-6)

As mentioned above, to evaluate the order of the partial derivatives contained in Eqs. (10-1)-(10-3) the technique described in [1] is used. It is necessary to note that the typical distance of a variation of all the disturbed parameters is the characteristic dimension of the suspended particle, R . The disturbance of the pressure in the sphere drag problem, and that of the local thermal flow velocity in the thermal force problem, can be evaluated by dimensional analysis which gives:

'p ~ UU 2 ,

U th ~

P T f . UT

(10-7)

The ratio of the terms in the continuity equation for each problem can be expressed as:

ui wn wxi U 'p R U 2 1 , ~ ~ n wui wxi kTR nU V 2 and:

R T f ui wn wxi U th p0 T f kTR 1 , ~ ~ n wui wxi p0U th T kT 2 where the pressure in the thermal force problem is assumed to be constant. By analogy, one obtains the following for the appropriate ratio in Eq. (10-2):

ui wui wxi

Q 2ui

~

U 2 R 2 UR ~ Q R QU

Re 1 ,

where Q is the kinematic viscosity and Re is the Reynolds number. For the thermal force problem, one needs to use the third moment equation which contains two non-linear terms. The first ratio of the nonlinear to linear terms can be evaluated by the relation:

n kui wT wxi

N 2T

~

nkTU th R 2 U th R R T f 1 , ~ ~ R NT VO T

214

Analytical Methods for Problems of Molecular Transport

where N ~ UV O cv ~ nkV O and U th ~ V O T f T . For the other non-linear term the appropriate evaluation is given by:

Pij wui wx j

N 2T

~

R T f nkTU th R 2 1 . ~ R nkV OT T

As one can see, if the basic assumptions defined by Eq. (10-6) are satisfied, the ratio of the non-linear to the linear terms is much less than unity for all the moment equations and, therefore, this system can be linearized by the simple expedient of neglecting the non-linear terms. The stationary linearized form of this system can be expressed as:

u 0 ,

(10-8)

P 2u p ,

(10-9)

2T

0 ,

(10-10)

where the external force, Fi , is assumed to be equal to zero. Having obtained the basic equations describing the gas flow over the particle, one is now ready to consider, in detail, the various transfer problems of the slipflow regime.

2.

THE SPHERICAL DRAG PROBLEM.

An understanding of translational motion of small single particles in a gas is required in disciplines as diverse as nano-phase materials synthesis, environmental physics, clean-room technology, cloud physics, and nuclear reactor safety [2,3]. While these particles, excepting liquid drops, are generally non-spherical, studies of spherical particles are necessary for the intrinsic fundamentals involved as well as for applied reasons. The drag problem is of great significance and, since the work of Millikan [4], there has been a substantial body of related experimental [5] and theoretical work [6-9] generated. Consider the short statement of the spherical drag problem. The origin of a spherical coordinate system is assumed to be at the center of a stationary

Chapter 10. The Slip-Flow Regime

215

spherical particle and the direction of the polar axis is the same as that of the mean velocity of the uniform flow at large distances from the particle, U (Fig. 7-2). The gas state can be described by Eqs. (10-8) and (10-9) which, for the spherical geometry and azimuthally symmetric problems, are given by [1]: 2 1 w rvr 1 w 2 vr cot T wvr 2vr 2 2 r wr 2 r wT 2 r 2 wT r 2 § wv · R wp 2 ¨ T vT cot T ¸ , T w r © ¹ P wr

(10-11)

2 1 w r vT 1 w 2 vT cot T wvT 2 r wr 2 r wT 2 r 2 wT v 2 wv R 1 wp 2 r 2 T2 , r wT r sin T P r wT

(10-12)

wvr 2vr 1 wvT vT cot T 0 . r r wT r wr

(10-13)

The special features of the problem associated with the slip-flow regime are accounted for in the boundary conditions specified at the particle surface. For large distances from the particle, the radial and tangential components of the mean velocity of the gas are given by:

ur f U cos T and uT f U sin T .

(10-14)

At the sphere surface, however, the tangential component of the mean velocity is not equal to zero as it is for the continuum regime. This implies that the gas is slipping along the surface. Since this slipping is the consequence of momentum conservation at the surface, the slip boundary conditions take the form:

vT

cm

vr

0 ,

O P , P rT

(10-15)

(10-16)

216

Analytical Methods for Problems of Molecular Transport

where cm is the isothermal-slip coefficient. This coefficient, from a physical point of view, is the relative rate of communication of the tangential momentum, per unit area of the particle surface, to the surrounding gas. For this regime the force on the particle is given by:

Fu

³ Prr cos T PrT sin T nr dS ;

r

R ,

(10-17)

S

where nr is a unit normal vector at the surface of the sphere which points into the gas. The pressure tensor components, Prr and PrT , are defined by [1]:

wvr , wr

Prr

p0 2P

PrT

§ 1 wvr wvT vT · P ¨ ¸ . © r wT w r r ¹

(10-18a)

(10-18b)

Again, the standard method involving separation of variables [1] is used to obtain the following solution for the system of Eqs. (10-11)-(10-13):

vr

§ A B · ¨ 3 r U ¸ cos T , ©r ¹

(10-19)

vT

B § A · ¨ 3 2r U ¸ sin T , 2 r © ¹

(10-20)

p

p0 P

B cos T , r 2

(10-21)

where the two constants, A and B , may be found from boundary conditions given by Eqs. (10-15) and (10-16). Simple calculations result in the following expression for the force on the particle:

Fu

6SP R U

1 2cm Kn , 1 3cm Kn

where Kn OP R and OP is given by Eq. (8-16a).

(10-22)

217

Chapter 10. The Slip-Flow Regime

Table 10-1. The isothermal-slip coefficient,

DW

1.00 1.1006

cm

0.95 1.2106

cm , for different values of DW . 0.90 1.3321

0.85 1.4673

0.80 1.6185

The most reliable analytical expression for the slip coefficient is given by Eq. (8-79) that can be presented in the form: cm

2 DW

DW

5 8

S

0.6690 0.1775 DW 0.7549 0.09714 DW

.

(10-23)

Table 10.1 gives the slip coefficient, cm , for a number of values of the tangential momentum accommodation coefficient, DW . Compare the analytical expression given by Eq. (10-22) with the Millikan experimental data for this regime. It is convenient to use the ‘socalled’ drag ratio value, D , which is given by: D{

Fu Kn F fm

,

where F fm is the drag force in the free-molecular regime that is defined by Eq. (7-13). From Eq. (10-22) one can obtain:

Dth

45 8

S Kn 1 2cm Kn . 8 DW S 1 3cm Kn

(10-24)

The Millikan data for the full range of Kn can be described by means of the use of the Cunningham slip correction factor the most exact form of which has been proposed by Buckley and Loyalka [8]. The use of this factor gives: Dexp

1.617 Kn CC Kn

,

(10-25)

where CC Kn is defined by [8]:

ª § 0.425 · º CC Kn 1 Kn «1.099 0.518 exp ¨¨ ¸» . Kn ¸¹ »¼ «¬ ©

(10-26)

218

Analytical Methods for Problems of Molecular Transport

Table 10-2. Comparison of experimental and theoretical values of drag on a sphere.

Kn

1

1 2 4 6 8 10 12 14 16 18 20

Dth

Dexp

DW =1.00

DW =0.95

DW =0.90

DW =0.85

DW =0.80

0.6633 0.4870 0.3113 0.2265 0.1774 0.1456 0.1234 0.1071 0.0946 0.0847 0.0766

1.1803 0.6284 0.3368 0.2331 0.1790 0.1455 0.1227 0.1061 0.0935 0.0836 0.0756

1.1883 0.6315 0.3384 0.2344 0.1802 0.1466 0.1237 0.1070 0.0943 0.0844 0.0763

1.1970 0.6348 0.3401 0.2358 0.1814 0.1477 0.1247 0.1079 0.0952 0.0852 0.0771

1.2063 0.6383 0.3417 0.2371 0.1825 0.1487 0.1257 0.1089 0.0961 0.0860 0.0778

1.2163 0.6420 0.3434 0.2384 0.1837 0.1498 0.1266 0.1098 0.0969 0.0868 0.0786

The Millikan experimental data and analytical results are presented in Table 10-2. A comparison of experimental results with the theory for the slip-flow regime is illustrated by Fig. 10-1. It can be easily seen that the best agreement between the theoretical and experimental data for this regime is reached when DW 0.95 . For this case the discrepancy of results is less than 0.5% for Kn 0.1 . This excellent agreement of data is an additional confirmation of the sufficiently high accuracy of Eq. (10-23).

3.

THE THERMAL FORCE PROBLEM.

In Section 7.4 the problem of thermal forces (thermophoresis) on aerosol particles in the free-molecular regime was considered. Here, this problem will be examined in the slip-flow regime. A quantitative knowledge of thermophoresis is of great importance in many areas such as optical fiber fabrication, nuclear reactor safety, micro-contamination control, etc. Thermophoresis has been studied extensively, both experimentally [1018] and theoretically [19-29]. In spite of the numerous articles associated with this problem, both the experimental data and the theory remain controversial. The former due to the difficulties in measuring the relatively small effect because of competing phenomena, and the latter because of the approximations employed in the theoretical efforts. None of the previous work, with the exception of Loyalka [29], has attempted to solve this problem in its general form. Loyalka’s work contains the most general numerical analysis that may be used to obtain results for all Knudsen numbers.

219

Chapter 10. The Slip-Flow Regime

1.2

Sphere Drag Ratio (dimensionless)

1.0

0.8

0.6

0.4

0.2

0.0 0.0

5.0

10.0

15.0

20.0

Inverse Knudsen Number (dimensionless)

Figure 10-1. Comparison of the experimental and theoretical values of the drag on a sphere. Theoretical values are given by the solid line. Experimental points are obtained from the interpolation formula of Buckley and Loyalka [8] for DW 0.95 .

Consider the various aspects of this problem within the slip-flow framework. Let a single spherical particle of radius, R , be located in an infinite gas with a temperature gradient T , which is constant at large distances from the particle. All the other conditions have been specified in Section 7.4. For the slip-flow regime, it is shown here that the NavierStokes equations may be applied, with boundary conditions appropriate for Kn 1 , to the quantitative description of the thermal force. The momentum and continuity equations are given by Eqs. (10-11)-(10-13). The temperature distribution, both inside and outside of the particle, is described by the thermal conductivity equation which may be written as:

1 w § 2 wT · 1 w r r 2 wr ¨© wr ¸¹ r 2 sin T wT

wT · § ¨ sin T ¸ 0 . wT ¹ ©

(10-27)

The proper slip-flow boundary conditions for the motion of the gas at the particle surface are defined by the relations:

220

Analytical Methods for Problems of Molecular Transport

vr

0 ;

r

R ,

vT

ª w §v cm OP « r ¨ T ¬ wr © r

(10-28)

· 1 wvr º

1 § wT · ¸ r wT » cTslQ RT ¨ wT ¸ ; r ¹ ¹ ¼ 0 ©

R ,

(10-29)

where cTsl is the thermal-creep coefficient. The most exact analytical expression for cTsl is given by Eq. (8-88). At large distances from the particle we have:

vr

vT

0 .

(10-30)

The appropriate boundary conditions for the thermal conductivity equation have the form:

Ng

wTg wr

Tg Tp

Np

wTp wr

cT ON

; r

wTg wr

R ,

; r

R ,

Tg o T f r cos T ; r o f ,

(10-31)

(10-32)

(10-33)

where cT is the temperature-jump coefficient. The most exact expression for cT is given by Loyalka’s formula, Eq. (8-50). The solutions of Eqs. (1011)-(10-13) and (10-27) can be presented in the form:

vr

§ A B· ¨ 3 r ¸ cos T , ©r ¹

vT

B· § A ¨ 3 ¸ sin T , 2r ¹ © 2r

221

Chapter 10. The Slip-Flow Regime

B cos T , r 2

p

p0 P

Tg

T0 T f r cos T

Tp

T0 C2 r cos T ,

C1 cos T , r 2

where the constants A , B , C1 , and C2 can be found from boundary conditions. By means of Eq. (10-17), the following expression is obtained for the thermal force:

F

12S

P 2 R T f cTsl ] m] T , U T0

(10-34)

where the quantities ] m and ] T are defined by:

]m

1 , 1 3cm Kn

]T

N N c Kn 1 2 N N 2c Kn g

p

g

T

p

.

T

The trustworthiness of the results obtained here may be analyzed by comparing the thermal force values given by Eq. (10-34) with Loyalka’s numerical values [29]. For this comparison, the thermal force expression of Eq. (10-34) must be transformed into the form used by Loyalka. It is convenient to introduce a dimensionless thermal force, FT , by means of the relation:

F

p0 R 2

O T f T0

FT ,

(10-35)

where, FT , may be written as:

FT

75 32

S 2 KncTsl ] m] T ,

(10-36)

222

Analytical Methods for Problems of Molecular Transport

Table 10-3. The reduced thermal force,

Kn

FT .

FT

N p N g 10.0

1

Eq. (10-36) 2.0175 1.7303

1.2463 1.6617

N p N g 100.0

Loyalka [29] 2.2056 1.7965

Eq. (10-36) 1.9936 1.6972

Loyalka [29] 2.1378 1.7099

Table 10-4. The reduced thermal force on NaCl aerosol particles.

FT

FT experimental [15] 2.4300 2.1552 1.9115 1.6935 1.5036 1.3336

Kn 1 1.00 1.25 1.50 1.75 2.00 2.25

theoretical Eq. (10-36) 2.2059 1.9879 1.8006 1.6387 1.4977 1.3742

Kn 1 2.50 2.75 3.00 3.25 3.50 3.75

experimental [15] 1.1828 1.0490 0.9304 0.8252 0.7319 0.6491

theoretical Eq. (10-36) 1.2654 1.1690 1.0832 1.0066 0.9378 0.8759

and molecules are assumed to be rigid spheres. It is important to note that FT depends on the nature of the aerosol particle through the accommodation coefficients, the Knudsen number, and the ratio of the thermal conductivity coefficients of the gas and the particle. Let the accommodation coefficients, DW DT 1 . For this case, the slip and jump coefficients are specified by: cTsl

1.0193

,

cm

1.1006

,

cT

2.0633

.

A comparison of the results obtained from Eq. (10-36) with Loyalka’s numerical results [29] is given in Table 10-3. It should be noted that the results reported here as Loyalka’s were transformed in a simple fashion to conform to the current notation. One can see that the discrepancy between these results is very small. One may further compare the theoretical results with the experimental values for FT that have been reported by Jacobsen and Brock [15]. The experimental results for sodium chloride aerosols in argon can be fitted, for 0 Kn 1 4 , by the expression:

F

T

exp

5 4

§ W · ¸ , © Kn ¹

S exp ¨

(10-37)

where the constant W 0.48 . The theoretical and experimental results are presented in Table 10-4. A graphical comparison of these values is shown in

223

Chapter 10. The Slip-Flow Regime

Reduced Thermal Force (dimensionless)

2.5

2.0

1.5

1.0

0.5 1.0

1.5

2.0

2.5

3.0

3.5

4.0

Inverse Knudsen Number (dimensionless)

Figure 10-2. A comparison of experimental and theoretical reduced thermal forces for NaCl. Theoretical values from Eq. (10-36) are shown as a solid line and experimental values reported in [15] are shown as solid circles.

Fig. 10-2. One can easily see that the agreement of the theory with the experiment is within 10% for the very narrow region of inverse Knudsen number, 0 Kn 1 3 . A substantial discrepancy occurs outside this region, however, and further experimental and theoretical work is clearly necessary before an acceptable understanding of the whole slip-flow region is possible. One contributing factor to the discrepancy between theory and experiment is that the theory does not take into account all of the first-order boundary effects. A strict theory must account for corrections that are proportional to Kn for the thermal-creep. These additional terms occur due to surface curvature and to Barnett’s thermal stresses [30]. The influence of all the corrections that are proportional to Kn on the thermophoresis of aerosol particles has been analyzed in [30]. Unfortunately, the accuracy of the additional terms was not sufficiently discussed and, moreover, a kinetic theory analysis of these phenomena leads to very complicated calculations that are of the same order of difficulty as an appropriate solution for all Knudsen numbers. Consequently, it stands to reason that the most reliable

224

Analytical Methods for Problems of Molecular Transport

theory must ultimately be based on the direct solution of the Boltzmann equation by use of different moment methods.

PROBLEMS 10.1. A sphere suspended in a gas has radius, R , and temperature, Determine the heat flux and the temperature T0 'T 'T T0 . distribution for the surrounding gas if its number density and temperature far from the sphere are n0 and T0 , respectively. Solve this problem for the slipflow regime. Solution: The temperature distribution can be determined from the Laplace equation:

1 w § 2 wT · ¨ r ¸ 0 . r 2 wr © wr ¹ The boundary conditions are:

T

T0 'T cT O

wT wr

and T

T0 ;

r o f .

r R

The solution of this boundary problem may then be expressed as:

T r T0 'T ]

R ; ] r

O· § ¨1 cT R ¸ © ¹

1

.

The heat flux then takes the form:

Q

4S R 2N

wT wr

4S RN'T ] . r R

10.2. To maintain the constant surface temperature, T0 'T 'T T0 , of a sphere (radius R ), there is a heat source of power, Q , located at the center of the sphere. Determine the thermal conductivity coefficient of the surrounding gas if its number density and temperature far from the sphere are n0 and T0 , respectively. Solve this problem for the slip-flow regime. Solution: Using the expression for the heat flux obtained in Problem 10.1, one has that:

225

Chapter 10. The Slip-Flow Regime

N

Q ; ] 4S R'T ]

O· § ¨1 cT R ¸ © ¹

1

.

10.3. One method of measuring the thermal conductivity of a gas is to determine the loss of heat through the gas from a hot sphere of radius, R . For the same conditions used in Problem 10.2, determine a calculation error for the thermal conductivity coefficient of the gas if the temperature-jump at the surface of the sphere is not taken into account. Solution: In the continuum regime, the thermal conductivity coefficient is given by: Q . 4S R'T

N0

Using the result of Problem 10.2, one can obtain:

G

N N0 N

1

1

]

cT

O R

.

10.4. Determine the heat flux and temperature distribution of a gas between two concentric spheres the radii and temperatures of which are R1 , T0 'T and R2 , T0 'T , respectively. Assume 'T T0 . The equilibrium number density of the gas is n0 . Solve this problem for the slip-flow regime. Solution: The temperature distribution is given by:

T r

A B . r

The two constants, A and B , are determined from the boundary conditions:

T R1

T0 'T cT O

wT , and: wr

T R2

T0 'T cT O

wT . wr

The temperature distribution and the thermal flux may be expressed as:

O · R1 § O · º °½ ° 'T ª 2 R1 § ] « T r T0 ®1 ¨1 cT ¸ ¨ 1 cT ¸» ¾ , T0 «¬ r R1 ¹ R2 © R2 ¹ »¼ ¿° © ¯°

226

Analytical Methods for Problems of Molecular Transport

Q

8S R1N'T ] ,

]

ª§ O · R1 § O ·º «¨1 cT ¸ ¨1 cT ¸» R1 ¹ R2 © R2 ¹ »¼ «¬©

1

.

10.5. Determine the heat flux and temperature distribution of a gas between two coaxial cylinders the radii and temperatures of which are R1 , T0 'T and R2 , T0 for the inside and outside cylinders, respectively. Assume 'T T0 . Solve this problem for the slip-flow regime. Solution: The temperature distribution may be found from the following boundary problem:

1 w § wT · r 0 , r wr ¨© wr ¸¹

T R1

T0 'T cT O

T R2 T0 cT O

wT , wr

wT . wr

After some calculations, one can obtain:

ln r R2 cT O R2 ° 'T °½ T r T0 ®1 ¾ , and: T0 ln R2 R1 cT O 1 R1 1 R2 ¿° ¯°

Q

ª §R 2SN'T « ln ¨ 2 ¬« © R1

· § 1 1 ·º ¸ cT O ¨ ¸» ¹ © R1 R2 ¹ ¼»

1

.

10.6. One method of measuring the thermal conductivity of a gas is to determine the loss of heat through the gas from a long hot wire enclosed by a cylinder having a radius much greater than that of the wire. If the thermal power generated per unit length of the wire, Q , and the temperature difference, 'T , are measured during the experiment, determine the thermal

Chapter 10. The Slip-Flow Regime

227

conductivity coefficient of the ambient gas. Assume that the temperature of the external cylinder is T0 and that slip-flow conditions are satisfied during this experiment. Solution: Using the expression for the heat flux from Problem 10.5 and allowing for R r , one can obtain:

N

Q 2S'T

ª §R· Oº «ln ¨ r ¸ cT r » . ¬ © ¹ ¼

10.7. For the conditions given in Problem 10.6, determine the relative error of the predicted values of the thermal conductivity coefficient of the ambient gas if the temperature-jump at the surface of the wire is neglected. Solution: In the continuum regime, the thermal conductivity coefficient is given by:

N0

Q §R· ln ¨ ¸ . 2S'T © r ¹

From this one may generate the following relative error:

G

N N0 N

cT O r . ln R r cT O r

10.8. Two flat parallel round disks of equal radius, R , are separated by a small distance, h R . While the lower disk is stationary, the upper disk moves toward it with a velocity, u . Determine the resistance force on the upper disk for the slip-flow regime. Use the geometry in Fig. 10-3.

228

Analytical Methods for Problems of Molecular Transport

Figure 10-3. The geometry to be used in Problem 10.8 in determining the resistance force on a flat, round disk in the slip-flow regime due to its approach to an identical, parallel, stationary disk.

Solution: The motion of the gas in this problem is axially symmetric such that uI 0 and, therefore, the Navier-Stokes system of equations in cylindrical coordinates may be written as [1]:

wu 1 w r z ru r wr wz

§

P ¨ 'ur ©

P'u z

'f

ur · ¸ r 2 ¹

0 ,

wp , wr

wp , where: wz 1 w § wf · w 2 f r . r wr ¨© wr ¸¹ wz 2

It may readily be shown that u z ~ u and ur ~ R h u and, therefore, that ur u z . For the partial derivatives, one can obtain the following estimates:

wur Ru ~ wr Rh

u , h

wur Ru ~ hh wz

R u , h2

229

Chapter 10. The Slip-Flow Regime

P'u z ~ UV O

u , h2

wp U kT UV 2 ~ ~ . mh h wz Consequently:

P'u z wp wz

~

wu wr h u O 1 , and r ~ 1 . V h wur wz R

Taking into consideration these estimates, one can reduce the Navier-Stokes system to the form:

wu 1 w r z ru r wr wz

P

w 2 ur wz 2

wp wz

0 ,

(P-1)

wp , wr

(P-2)

0 .

(P-3)

The boundary conditions may be expressed as:

z

0 , uz

0 , ur

z

h , uz

u , u r

r

R , p

p0 .

wur , wz

cm

O P P rz

cm O

cm

O P P rz

cm O

wur , wz

From Eqs. (P-2) and (P-3) and the boundary conditions, one can obtain:

230

Analytical Methods for Problems of Molecular Transport

ur

1 dp ª z z h cm O h º¼ 2 P dr ¬

.

Integration of Eq. (P-1) yields: h

u

1 w r dz ru r wr ³0

h3 d § dp · § O· r ¸ ¨ 1 6cm ¸ . ¨ 12P r dr © dr ¹ © h¹

From this relation, the pressure may be expressed in the form:

p

p0

3P u R 2 r 2 . h3 1 6cm O h

The resistance force on the upper disk is then given by:

Fz

3 2

S

P R 4] u h3

where ]

O· § ¨ 1 6cm h ¸ © ¹

1

.

10.9. Determine the velocity distribution and the net mass transport of a gas flowing along a cylindrical tube owing to a pressure gradient, dp dx . The tube radius is R . Solution: The gas velocity has only an x -component which depends only on the radial coordinate, r . The Navier-Stokes system may be written as:

1 d § du · r r dr ¨© dr ¸¹ wp wr

1 wp , P wx

0 .

The slip-flow boundary condition is given by: u R cm O

wu . wr

The solution of this boundary problem can be written as:

231

Chapter 10. The Slip-Flow Regime

u r

º dp 1 dp ª 2 § O· R ¨ 1 2cm ¸ r 2 » , with « dx 4 P dx ¬ © R¹ ¼

const .

The net mass transfer is given by: R

M

2SU ³ rudr 0

SU R 4 dp § O· ¨1 4cm ¸ . R¹ 8P dx ©

(P-4)

10.10. Determine the pressure difference at the ends of a cylindrical tube of radius, R , and length, L , if the net mass transfer through the tube is M . The gas flow is assumed to be isothermal. Solution: The pressure gradient may be expressed in the form (see Problem 10.9):

dp dx

8P kTM , S mR p 4cm p1 O1 R 4

where p1 and O1 are the pressure and mean free path at the entrance of the tube. Integrating this expression, one can obtain:

p1 p2

12 ª º ½° 16P kTML ° » ¾ . p1 1 4cm O1 R ®1 «1 2 4 2 ° «¬ S mR p1 1 4cm O1 R »¼ ° ¯ ¿

10.11. Two vessels containing the same gas at the different temperatures, T1 and T2 , are connected by a long cylindrical tube of radius, R . There exists a pressure difference between the vessels owing to the phenomenon of thermal-creep. Determine this pressure difference and the velocity distribution of the gas in the tube. Solution: The net mass transfer per unit time, ignoring transfer in the Knudsen layer, may be written as (see Problem 10.9):

M

18 S

U R4 § O · dp 1 dT cTslQUS R 2 1 4cm ¸ . ¨ P © R ¹ dx T dx

The mechanical equilibrium condition, M

O · dp § p ¨ 1 4cm ¸ R ¹ dx ©

8k P 2 cTsl dT . mR 2 dx

0 , gives the following relation: (P-5)

232

Analytical Methods for Problems of Molecular Transport

If one takes into consideration that O ~ 1 p and P ~ T , this equation may be transformed to:

O1 · dp § ¨ p 4cm p1 R ¸ dx © ¹

8k P12 cTsl dT T . dx mR 2T1

Integrating this expression, one obtains: p2 p1

]

p1] 1

1 A T

O1 · § ¨ 1 4cm R ¸ © ¹

2 2

1

and A

T12 1 where:

8k P12] 2 cTsl . mT1 p12 R 2

The velocity distribution is then given by:

u

º dp 1 ª 2§ O· 1 dT cTslQ R ¨1 2cm ¸ r 2 » . R¹ dx T dx 4 P «¬ © ¼

Substituting dp dx into this relation from Eq. (P-4), one obtains:

§ 2r 2 · 1 dT u r cTslQ] ¨ 2 1¸ . © R ¹ T dx For r

0 and r

R , the gas velocity has opposite signs.

10.12. For the same conditions used in Problem 10.11, determine the pressure difference if the variation in the mean velocity due to thermal-creep in the Knudsen layer is taken into account. Solution: The mean velocity due to the temperature gradient is given by Eq. (8-87) and may be expressed in the form:

uth r cTslQ ª¬1 \ exp D R r º¼

\

0.3882 DW 0.4354 0.2179 DW

,

1 dT where: T dx

233

Chapter 10. The Slip-Flow Regime

Figure 10-4. The geometry to be used for Problem 10.13 in determining the net mass transfer in a closed capillary loop with different radii and end temperatures.

and D 2.2015 O 1 . Taking into account the mass transport in the Knudsen layer, Eq. (P-5) from Problem 10.11 may be cast in the form: 1

O ·§ 2\ · dp § p ¨ 1 4cm ¸¨1 R ¹© D R ¸¹ dx ©

8k P 2 cTsl dT . mR 2 dx

The pressure difference is given by:

p1] 11

p2 p1

]1

1 A T 1

§ § 2\ · O1 · ¨1 ¨ 4cm ¸ DO ¸¹ R ¹ © ©

2 2

T12 1 where:

1

and A1

8k P12] 12 cTsl . mT1 p12 R 2

10.13. Two capillary tubes having an equal length, L , but different radii, R1 and R2 where R1 R2 , are connected at each end to form a closed loop (see the figure). The ends of the capillary tubes are maintained at the temperatures, T0 and T0 'T where 'T T0 , as shown in Fig. 10-4. Determine the net mass transfer through any cross section of either tube that is due to a circular motion of the gas. The gas parameters, P and U , are assumed to be constant and the isothermal end connections are assumed to make no contribution to the flow. Solution: The solution for the net mass transfer from Problem 10.9 may be presented in the form:

M]i Ri4

SU dp cTsl SP] i 1 dT where: 8P dx Ri2 T0 dx

234

Analytical Methods for Problems of Molecular Transport

Figure 10-5. The geometry to be used in Problem 10.14 in determining the torque on a rotating sphere in the slip-flow regime.

]i

§ O· ¨1 4cm ¸ Ri ¹ ©

1

.

Integrating over all values of x along a closed path, one can obtain:

M

cTsl P

'T S R12 R22 ] 1 R22 ] 2 R12 . T0 L ] 1 R24 ] 2 R14

10.14. A sphere of radius, R , is rotating with an angular velocity, Z , such 12 that Z R 2kT m . Determine the torque, K , on the sphere in the slipflow regime. Use the geometry in Fig. 10-5. Solution: Owing to the axial symmetry of this problem, the mean velocity of the gas has only one component, uI r,T u r,T . Additionally, one also has that p const . The Navier-Stokes system of equations thus reduces to a single equation given by [1]:

235

Chapter 10. The Slip-Flow Regime

1 w § 2 wu · 1 w r r 2 wr ¨© wr ¸¹ r 2 sin T wT

wu · u § ¨ sin T ¸ 2 2 wT ¹ r sin T ©

0 .

The boundary conditions may be expressed as:

u R Z R sin T cm

O P R and u f 0 where: P rI

§ wu u · PrI R P ¨ ¸ © wr r ¹ r

. R

Let u r ,T be expressed in the following form u r,T f r sin T . Then, the function, f r , may be determined from the equation:

d § 2 df · 2f r dr ¨© dr ¸¹

0 .

The solution of the above boundary problem is then:

u r,T

Z R 3] r 2

sin T ,

PrI R 3PZ] sin T ,

dFI where ]

3PZ R 2] sin 2 T dT dI ,

1 3cm O R 1 . 2S

Kx

The net torque, K x , is then:

S

R ³ dI ³ sin T dFI dT 0

8SPZ R 3] .

0

10.15. An infinite cylinder of radius, R , is rotating with an angular velocity, 12 Z , such that Z R 2kT m . Determine the torque per unit length in the slip-flow regime.

236

Analytical Methods for Problems of Molecular Transport

Figure 10-6. The geometry to be used for Problem 10.16 in determining the average speed of a steady gas flow between two, parallel, planar surfaces in the slip-flow regime.

Solution: The mean velocity of the gas possesses only one component, uI u r . Additionally, p const . The velocity distribution may be found from the following boundary problem:

1 w § wu · u r r wr ¨© wr ¸¹ r 2 u R Z R cm

0 ,

O P R and u f 0 , where: P rI

§ wu u · ¸ . PrI R P ¨ © wr r ¹ The solution of this problem may be expressed as:

u

Z R 2] r

and PrI R 2 PZ] where ]

1 2cm

O R

.

The torque is then given by K x 4SPZ R 2] where the direction of the angular velocity vector, Ȧ , is in the direction of the x -axis. 10.16. In the steady flow of a gas between two parallel planar surfaces, if d O 1 (the slip-flow regime) determine the average gas speed, u , owing to a constant pressure gradient, dp dy const , where the velocity is assumed 1-dimensional in the y -direction. Use the geometry in Fig. 10-6. Solution: The gas flow may be described by Eqs. (10-8) and (10-9) which, for this particular case, take the form:

wu y wy

§ w 2u · wp 0, P¨ 2 ¸ , wp wx 0 , and u y u x . © wx ¹ wy

Chapter 10. The Slip-Flow Regime

237

The boundary conditions are given by u d 2 cm O wu wx u d 2 cm O wu wx . This statement of the problem then yields:

and

d 2

u d 1

³

d 2

u x dx 121

d2 § O · dp 1 6cm ¸ . d ¹ dy P ¨©

REFERENCES 1. Landau, L.D. and Lifshitz, E.M., Fluid Mechanics (Pergamon Press, Oxford, 1987). 2. Hidy, G.M. and Brock, J.R., Topics in Aerosol Research, vols. 1-3 (Pergamon Press, Oxford, 1971-1973). 3. Williams, M.M.R. and Loyalka, S.K., Aerosol Science: Theory and Practice, with Special Applications to Nuclear Industry (Pergamon Press, Oxford, 1991). 4. Allen, M.D. and Raabe, O.G., “Re-evaluation of Millikan’s Oil Drop Data for the Motion of Small Particles in Air,” J. Aerosol Sci. 13(6), 537-547 (1982). 5. Cercignani, C., Pagani, C.D., and Bassanini, P., “Flow of Rarefied Gas Past an Axisymmetric Body. II. Case of a Sphere,” Phys. Fluids 11(7), 1399-1403 (1968). 6. Phillips, W.F., “Drag on a Small Sphere Moving Through a Gas,” Phys. Fluids 18(9), 1083-1089 (1975). 7. Lea, K.C. and Loyalka, S.K., “Motion of a Sphere in Rarefied Gas,” Phys. Fluids 25(9), 1550-1557 (1982). 8. Buckley, R.L. and Loyalka, S.K., “Cunningham Correction Factor and Accommodation Coefficient: Interpretation of Millikan’s Data,” J. Aerosol Sci. 20(3), 347-349 (1989). 9. Loyalka, S.K., “Motion of a Sphere in a Gas: Numerical Solution of the Linearized Boltzmann Equation,” Phys. Fluids A 4(5), 1049-1056 (1992). 10. Rosenblatt, P. and La Mer, V.K., “Motion of a Particle in a Temperature Gradient. Thermal Repulsion as a Radiometer Phenomenon,” Phys. Rev. 70(5-6), 385-395 (1946). 11. Saxton, R.L. and Ranz, W.E., “Thermal Force on an Aerosol Particle in a Temperature Gradient,” J. Appl. Phys. 23(8), 917-923 (1952). 12. Schadt, C.F. and Cadle, R.D., “Thermal Forces on Aerosol Particles in a Thermal Precipitator,” J. Colloid Sci. 12, 356-362 (1957). 13. Schadt, C.F. and Cadle, R.D., “Thermal Forces on Aerosol Particles,” J. Phys. Chem. 65(10), 1689-1694 (1961). 14. Schmitt, K.H., “Untersuchungen an Schwebstoffteilchen im Temperaturfeld,” Z. Naturforschg. 14a, 870-881 (1959). 15. Jacobsen, S. and Brock, J.R., The Thermal Force on Spherical Sodium Chloride Aerosols,” J. Colloid Sci. 20(6), 544-556 (1965). 16. Waldmann, L. and Schmitt, K.H., “Thermophoresis and Diffusiophoresis of Aerosols,” in Aerosol Science, edited by Davies, C.N. (Academic Press, London, 1966). 17. Derjaguin, B.V., Storozhilova, A.I., and Rabinovich, Ya.I., “Experimental Verification of the Theory of Thermophoresis of Aerosol Particles,” J. Colloid and Interface Sci. 21(1), 35-58 (1966). 18. Keng, E.Y.H. and Orr, C.Jr., “Thermal Precipitation and Particle Conductivity,” J. Colloid and Interface Sci. 22, 107-116 (1966). 19. Epstein, P.S., “Zur Theorie des Radiometers,” Z. Physik 54(4), 537-563 (1929).

238

Analytical Methods for Problems of Molecular Transport

20. Brock, J.R., “On the Theory of Thermal Forces Acting on Aerosol Particles,” J. Colloid Sci. 17, 768-780 (1962). 21. Dwyer, H.A., “Thirteen-Moment Theory of the Thermal Force on a Spherical Particle,” Phys. Fluids 10(5), 976984 (1967). 22. Ivchenko, I.N. and Yalamov, Yu.I., “The Hydrodynamic Calculation Method of the Thermophoresis of Sufficiently Large Nonvolatile Aerosol Particles,” J. Phys. Chem. (Russia) 45(3), 577-582 (1971) (in Russian). 23. Derjaguin, B.V. and Yalamov, Yu.I., “The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and their Experimental Testing,” in Topics in Current Aerosol Research, vol. 3, part 2, edited by Hidy, G.M. (Pergamon Press, Oxford, 1972). 24. Sone, Y. and Aoki, K., “Negative Thermophoresis: Thermal Stress Slip Flow Around a Spherical Particle in a Rarefied Gas,” in Rarefied Gas Dynamics, vol. 74, part I, edited by Fisher, S.S. (American Institute of Aeronautics and Astronautics, 1981). 25. Yamamoto, K. and Ishihara, Y., “Thermophoresis of a Spherical Particle in a Rarefied Gas of a Transition Regime,” Phys. Fluids 31(12), 3618-3624 (1988). 26. Bakanov, S.P., “Thermophoresis in Gases at Small Knudsen Numbers,” Aerosol Sci. Tech. 15(1), 77-92 (1991). 27. Loyalka, S.K., “Mechanics of Aerosols in Nuclear Reactor Safety: A Review,” Prog. Nucl. Energy 12(1), 1-56 (1983). 28. Loyalka, S.K., “Rarefied Gas Dynamics Problems in Environmental Sciences,” in Rarefied Gas Dynamics, XVI Symposium, edited by Boffi, V. and Cercignani, C. (Teubner, Stuttgart, 1986). 29. Loyalka, S.K., “Thermophoretic Force on a Single Particle-I. Numerical Solution of the Linearized Boltzmann Equation,” J. Aerosol Sci. 23(3), 291-300 (1992). 30. Poddoskin, A.B., Yushkanov, A.A., and Yalamov, Yu.I., “The Thermophoresis Theory for Sufficiently Large Aerosol Particles,” J. Tech. Phys. (Russian) 52(11), 2253-2261 (1982) (in Russian).

Chapter 11 BOUNDARY VALUE PROBLEMS FOR ALL KNUDSEN NUMBERS

1.

THE MOMENT EQUATIONS IN ARBITRARY CURVILINEAR COORDINATES.

The most advanced analysis possible for boundary value transport problems involving arbitrary Knudsen number and for various molecular interaction laws must be founded on the solution of the Boltzmann equation. In the transition regime, the effects of intermolecular collisions become dominant and their influence on the gas behavior is accurately described only by the full Boltzmann form of the collision operator found in the Boltzmann equation. For this reason, different model forms of the collision integral and model equations will not be considered here. There are two kinds of difficulties associated with the solution of the Boltzmann integro-differential equation [1]. The first difficulty is connected with the direct solution of this equation in which the distribution function depends, in the general case, on seven independent variables. It is well known that the difficulty in solving differential equations increases very rapidly with the number of independent variables. The second difficulty is associated with the very complicated non-linear structure of the collision term in the Boltzmann equation. The first difficulty may be overcome by replacing the Boltzmann equation with a system of equations for macroscopic values (moments). Such a moment system involves, in general, functions of only four independent variables and, thus, the difficulty of solution does not increase to the same level as it would with the Boltzmann equation due to this lesser

240

Analytical Methods for Problems of Molecular Transport

number of independent variables. Additionally, for specific transport problems involving stationary or symmetry conditions, the number of independent variables may even be less than four. The second difficulty may be overcome, for some applied transport problems, through the process of linearization. This may be done, however, only if the appropriate linearization conditions are physically satisfied. In order to investigate boundary value transport problems for all Knudsen numbers a moment method is considered. Instead of seeking exact solutions of the Boltzmann equation, this equation may be satisfied in a certain average sense. For boundary value transport problems, one is often not particularly interested in the distribution function itself since only certain lower moments of this function have the practical significance of being connected with the macroscopic values that describe the gas. Since one is mainly interested in mean flow quantities rather than the distribution function itself, the Maxwell integral transport equations, or moment equations, will be taken as the basis of the theory of transport phenomena for arbitrary Knudsen number. The general moment equation can be obtained by multiplying the Boltzmann equation by the function, Q v , of the components of the molecular velocity and integrating over the velocity space. The general form of this equation was derived in [2]. The transport moment equation of the molecular property, Q v , in arbitrary, orthogonal, curvilinear coordinates, has the following form:

w wt

3

³ fQ v dv h h1 h ¦ wwD h h ³ fv Q v dv j k

1 2 3 i 1

3 § 3 F w · ° 1 ³ f ¨ ¦ i Q v ¸ dv ³ f ¦ ® hh i 1° © i 1 m wvi ¹ ¯ i j

wh 1 § 2 whk vi vk i ¨ vk wD k hi hk © wD i

i

i

§ wh j wh vi v j i ¨ v 2j ¨ wD i wD j ©

· ½° w Q v dv ¸¾ ¹ °¿ wvi

· ¸ ¸ ¹

(11-1)

n' Q v ,

where D1 , D 2 , D 3 are the curvilinear coordinates, hi D1 , D 2 ,D 3 denotes the metric coefficient such that dli hi dD i , Fi is the component of the external force acting on a molecule, and the moment of the collision operator has been expressed in standard notation as n'Q( v ) . For the spherical and cylindrical geometries these metric coefficients can be written as:

D1

r

h1 1

D2 T

D3 I ,

h2

h3

r

r sin T ,

Chapter 11. Boundary Value Problems for All Knudsen Numbers

241

and:

D1

r

h1 1

z ,

D2 T

D3

h2

h3 1 .

r

For the stationary problem with radial symmetry, in the absence of external forces, the general moment equation, Eq. (11-1), takes the following form [2,3]:

w 1 d j 1 r ³ fvr Q v dv ³ f ® ¬ªvT2 j 1 vI2 ¼º Q v j wvr r r dr ¯ w ª¬ j 1 cot T vI2 vT vr º¼ Q v wvT

j 1 ª¬vI vr cot T vI vT º¼

½° w Q v ¾ dv wvI °¿

(11-2)

Rn'Q v ,

where, for the cylindrical problem, j 1 , for the spherical problem, j 2 , r r R is a dimensionless radial vector and R is the radius of the sphere or cylinder. If Q v is successively selected to be the collisional invariants, m , mvr , and 12 mv 2 , for which n'Q ( v ) 0 , one can obtain the ordinary continuity equation, and the radial momentum and energy transport equations. These moment equations are usually employed in all boundary value transport problems. Other equations can be obtained by using the arbitrary functions, Q v . It is necessary to note, however, that such a system of moment equations will never be closed as the number of the unknown moments is always greater than the number of moment equations. A common feature of all moment methods is their inclusion of some technique for closing the moment system. Typically, the technique for closing the moment system consists of making a special choice of the distribution functions so that they contain a certain number of unknown quantities that depend upon the various coordinates and time. The number of these unknown quantities must be equal to the number of moment equations and thus is a characteristic of the order of a given moment approximation. Two ways may be distinguished by which to close the moment system. The first is based on the use of polynomial expansions in velocity space for the distribution function. The coefficients in this expansion are unknown functions of time and space but may be determined from the moment system

242

Analytical Methods for Problems of Molecular Transport

if one employs a final number of expansion terms and uses an equal number of moment equations [4-7]. The second method of closing the system of moment equations involves making a special choice for the distribution function that is pertinent to the specific transport problem under consideration [2,3,8-12]. Simple approximate functions may be chosen if one takes into account conditions specific to the problem. However, one must insure that the distribution function chosen has the features normally characteristic of both the freemolecular and continuum regimes. Mott-Smith [13], to describe planar shock waves, was the first to suggest the use of the discontinuous Maxwellian distribution function. This idea was then generalized by Lees and Liu [2,3] for arbitrary curvilinear geometries. At present, the distribution functions proposed by Mott-Smith and Lees are considered to be the most suitable for the solution of boundary value transport problems. The Lees method will be discussed in the next section.

2.

THE TWO-SIDED MAXWELLIAN DISTRIBUTION FUNCTIONS.

As mentioned above, for arbitrary Knudsen number it is most convenient to use a moment method. Any moment method is an integral method. Just as in any other integral method, the distribution function that is employed is not necessarily an exact solution of the original Boltzmann equation. Nevertheless, this function must have features common to both the freemolecular and continuum distributions. It is also desirable that the distribution function be adaptable in such a fashion as to satisfy the boundary conditions. The best distribution function satisfying these conditions for the Maxwellian boundary model is that proposed by Lees and Liu [2,3]. In the Lees method the distribution function is assumed to be twosided Maxwellians which are discontinuous on the surface of the ‘so-called’ ‘cone of influence’ in the velocity space. Consider the arbitrary point, M r , where r is a dimensionless position vector. For each specified point there exists a cone of influence the surface of which is formed by tangents to the particle surface passing through the point, M r . At this point it is necessary to distinguish two regions in the molecular velocity space. In the body coordinates all the outwardly directed molecular velocity vectors lying within the cone of influence are considered to belong to the second region in the velocity space. All other molecules belong to the first velocity set, denoted as region 1. For these velocity regions, the simplest distribution function having a discontinuous structure and being capable of exhibiting a

243

Chapter 11. Boundary Value Problems for All Knudsen Numbers

smooth transition between the free-molecular and continuum regimes is the two-sided Maxwellian distribution given by: 32

f1

ª º m n1 r, t « » ¬« 2S kT1 r, t ¼»

f2

ª º m n2 r, t « » ¬« 2S kT2 r, t ¼»

§ m v u r, t 2 1 exp ¨ ¨ 2kT1 r, t ©

· ¸ , ¸ ¹

(11-3a)

and: 32

§ m v u r, t 2 2 exp ¨ ¨ 2kT2 r, t ©

· ¸ , ¸ ¹

(11-3b)

where ni r, t , Ti r, t , and ui r, t are ten unknown functions of position and time. These ‘so-called’ characteristic densities, temperatures, and mean velocities form the basic set of parameters over which all of the moments of the distribution function and, therefore, all of the macroscopic quantities for a gas, may be calculated. As one can see, the two-sided Maxwellian distribution function is a straightforward generalization of the distribution function for free-molecular flow. Such a choice for the distribution function gives a chance to close the moment system but, unfortunately, it limits the number of moment equations and consequently the accuracy of this analysis. An improvement in the accuracy can be achieved by the use of the ‘so-called’ modified two-stream Maxwellians proposed by Krook [14] and described in detail by Kogan [15]. For various regions of the velocity space these modified two-stream Maxwellians can be represented by:

fi

ª m º ni « » ¬ 2S kTi ¼

32

§ m v u 2 i exp ¨ ¨ 2kTi ©

· ¸ 1 Ak i vk Akl i vk vl , (11-4) ¸ ¹

where, i 1 for the first region and i 2 for the second. The functions ni , i i ui , Ti , Ak , and Akl depend on r and t in the general case. To facilitate the employment of the discontinuous distribution functions given by Eqs. (11-3a), (11-3b), and (11-4), one should attempt to represent these functions by analytical formulas that might be suitable for the entire velocity space [16]. For two specific geometries of interest, this type of representation will be discussed in detail. For the spherical geometry, at each space point, M r , there exists a cone of influence, the surface of which is described by the equation F F 0 , F 0 arcsin R r , where R is the radius of the particle, r Rr , and F is

244

Analytical Methods for Problems of Molecular Transport

Figure 11-1. The cone of influence in cylindrical coordinates.

the angle between vectors v and r (Fig. 7.1). In velocity space, the cone of influence divides the molecules into two groups with distribution functions f1 and f 2 ( f 2 describes the molecules for which vr ! vr , vr v cos F 0 ; f1 describes all other molecules). In the full velocity space the distribution function may be expressed as [16]:

f

1 2

f1 f 2 12 f 2 f1 sign vr vr

.

(11-5)

This is a representation of the distribution function that is very convenient to use in a moment system.

Chapter 11. Boundary Value Problems for All Knudsen Numbers

245

Now, consider the corresponding representation for the cylindrical geometry shown in Fig 11-1. While still traditionally termed a ‘cone of influence,’ the region is actually tent-shaped and is bounded by the two planes passing through the point M r and lying tangent to the surface of the cylinder. The cylinder, which nominally has a dimensional radius of R , is shown in Fig. 11-1 in dimensionless form where r r R and thus the cylinder has a dimensionless radius of unity. Let vT , vr , vz and vU ,T , vz be Cartesian and cylindrical coordinates in velocity space, respectively. Now consider an arbitrary velocity, v vT v r v z . From Fig. 11-1 one can see that v U vT v r with vr vU sin T and vT vU cos T . The velocity component, v z , which is the same in both coordinate systems, is not shown in Fig. 11-1 which is only a two-dimensional representation, but it may be envisioned as projecting normally upward from the plane of the page perpendicular to the components, vT and v r . If one defines vr vU sin D for the special case when v U and v both lie on the surface of the cone of influence, i.e. when T D arccos 1 r , then from Fig. 11-1 one can conclude that all molecules for which vr ! vr belong to the second region of the velocity space in which the angle, T , varies over the domain D d T d S D . For all of the remaining molecules, vr vr and the molecules belong to the first region for which S D T 2S D . Eliminating the angular dependence, vr is given in terms of vU and r by:

vr

vU 1 r 2

12

.

(11-6)

This analysis allows one to write an expression for the distribution function in the full velocity space in the same form as that given by Eq. (11-5) which is a form of the distribution function that allows one to find all of the moments necessary to construct the moment system. The next section contains an analytical method for the calculation of these moments for both the spherical and cylindrical geometries.

3.

MOMENTS OF DISCONTINUOUS DISTRIBUTION FUNCTIONS.

Let value Q v be any function of the molecular velocity, v . Then, the mean value of Q v , being a moment of the distribution function, can be calculated by the formula:

246

Analytical Methods for Problems of Molecular Transport

nQ v

³ Q v f1dv ³ Q v f 2 dv

1

³

2

Q v f1dv

1 2

³

Q v f 2 dv

2

³

Q v f1dv ,

(11-7)

2

where ³ i and ³ 1 2 are interpreted to mean integration over the specified region of the velocity space and the entire velocity space, respectively. For the spherical geometry, integration over velocities lying within the cone of influence proceeds according to the formula:

³

dv

2

f

2S

F0

0

0

0

2 ³ v dv ³ d E ³ sin F d F ,

(11-8)

where F 0 arcsin 1 r and r is the dimensionless radial coordinate of the point, M r . The relationships between spherical and Cartesian coordinates in velocity space are:

vr

v cos F , vT

v sin F sin E , vI

v sin F cos E .

Thus, the same integration for the cylindrical geometry proceeds according to the formula: f

³

2

dv

³ vU dvU 0

S D

f

³

f

dvz

³

dT ,

(11-9)

D

where D arccos 1 r and vr vU sin T , vT vU cos T , and vz vz . For example, one may calculate the gas number density in the fourmoment approach for which the Lees distribution function is given by:

fi

ª º m ni r , t « » «¬ 2S kTi r, t »¼

32

§ mv 2 · exp ¨ ¸¸ ; i 1; 2 , ¨ © 2kTi r, t ¹

(11-10)

Using Eq. (11-7) with Q v 1 , one obtains for the spherical geometry:

Chapter 11. Boundary Value Problems for All Knudsen Numbers

n r, t n1 r, t

n2 r, t f

1 2

0

n1 r, t

S3 2

f

0

0

2S

F0

0

0

³ c1 exp c1 dc1 ³ d E ³ sin F d F 2

2

0

n1 n2 12 n2 n1 x 12

F0

³ d E ³ sin F d F

2 2 ³ c2 exp c2 dc2

S3 2

2S

247

(11-11)

, 12

and ci m 2kTi v . The appropriate expression for where x 1 r 2 the cylindrical geometry is then given by: 1 2

n

where D

4.

n1 n2 12 n2 n1

2D

S

(11-12)

,

arccos 1 r . Other moments can be calculated in the same way.

ANALYTICAL EXPRESSIONS FOR THE BRACKET INTEGRALS.

Let this analysis be confined to a description of linearized transport problems in which the four-moment approach is used. The characteristic densities and temperatures in Eqs. (11-3a) and (11-3b), for these linearized problems, may be written as: ni r, t n0 ª¬1 Q i r, t º¼ ,

(11-13a)

Ti r, t T0 ª¬1 W i r, t º¼ .

(11-13b)

The corrections, Q i and W i , are assumed to satisfy the following conditions Q i 1 and W i 1 . The two-sided Maxwellian distribution function takes the form:

^

where:

f

`

0 f 1 Q i r, t c 2 32 W i r, t ,

fi

0

§ m · n0 ¨ ¸ © 2S kT0 ¹

32

exp c 2 .

(11-14)

248

Analytical Methods for Problems of Molecular Transport

In the full velocity space, for the spherical and cylindrical geometries, this distribution function may be expressed, according to Eq. (11-5), as:

^

0 f 1 12 Q 12 c 2 23 W

f

ª 12 Q 12 c 2 23 W º sign cr cr ¬ ¼

(11-15)

` ,

where Q r Q 2 r Q 1 and W r W 2 r W 1 . If the specific molecular properties, 1, cr , c 2 , c 2 cr , are substituted into Eq. (11-2) instead of Q v , then four-moment equations will be produced. The right-hand-side of the fourth equation will contain the following bracket integrals:

ªsign cr cr , c 2 cr º , ¬ ¼

ªc 2 sign cr cr , c 2 cr º . ¬ ¼

Consider the first of the above bracket integrals when molecules are assumed to act as rigid spheres. For the spherical geometry, this integral has the following analytical form:

12

ªsign cr cr , c 2 cr º ¬ ¼ 2S

S

u³ sin T dT 0

f

³ 0

V 2 § kT0 · S 3 ¨© m ¸¹ 2S

S

d H ³ sin D dD

0

u ³ exp Gr2 dGr

f

f

f

f

³ dE 0

³ exp GT dGT

f

1 2

2

(11-16)

u ³ exp GI2 dGI ³ g 3 exp g 2 dg f

^

0

u 2 Gr g r2 g rc2 GT g r gT g rc gTc GI g r gI g rc gIc § u sign ¨ Gr g r ¨ ©

`

12 1 · · 2 § G g 1 ¨ 2 ¸ ¸¸ . © r ¹ ¹

Here, the relations given in Appendix A are used to specify the components of the relative velocity vector, g , before and after a collision. From Eq. (11-16) one can see that this integral depends only on the radial coordinate, r . Two limiting values of this integral may easily be found. On

249

Chapter 11. Boundary Value Problems for All Knudsen Numbers

the particle surface, when r 1 , the integrand has the usual form encountered in planar transport problems. This integral may be expressed as [16]:

ªsign cr cr , c 2 cr º ¬ ¼r

1

12

2kT0 · V2 . ¸ © m ¹ 2§ 3¨

ªsign cr , c 2 cr º ¬ ¼

In the other limit, when r o f , sign Gr g r G g 1 and, therefore, the integrand is an odd function with respect to the components of the molecular velocity, Gr , GT , GI . Consequently, in this limit one has:

lim ªsign cr cr , c 2 cr º ¼

r of ¬

0 .

For arbitrary values of r , these multiple integrations result in very cumbersome algebraic calculations. After integration with respect to the variables T , H , and g , and after introducing the new variables Gr xg , GT yg , GI zg , and cos D t , this integral becomes:

12

ªsign cr cr , c 2 cr º ¬ ¼

96V 2 § kT0 · ¨ ¸ 2S 2 © m ¹

I1> @ r , 1

(11-17)

where:

I1> @ r 1

1

2S

f

³ dt ³ d E

1

0

F t , E , y, z, x

³

f

f

f

dy ³ dz ³ F t , E , y , z , x dx , f

x0

x t 2 13 yt 1 t 2 cos E zt 1 t 2 sin E

1 x

2

2

y z

2

5

and:

x0

t ª«1 t 2 y 2 z 2 2 y 1 t 2 cos E ¬ 12

2 z 1 t 2 sin E º» ¼

r

2

12

1

.

,

250

Analytical Methods for Problems of Molecular Transport

For the cylindrical geometry the integrand is the same, but it is convenient to perform the first integration over all values of z as the lower limit of the integral with respect to x does not depend on z . In this case, the bracket integral may be written as:

12

96V 2 § kT0 · ¨ ¸ 2S 2 © m ¹

ªsign cr cr , c 2 cr º ¬ ¼

I 2> @ r , 1

(11-18)

with: 1

>[email protected]

I2 r

2S

f

f

f

³ dt ³ d E ³ dy ³ dx ³ F t , E , y, z, x dz

1

0

f

x0c

,

f

and:

x0c

12

t y 1 t 2 cos E r 2 1

.

In the same way, analogous expressions may be obtained for the second bracket integral that is contained in the right-hand-side of the heat flow transport equation. For the spherical geometry this integral may be represented as:

12

240V 2 § kT0 · ¨ ¸ 2S 2 © m ¹

ª c 2 sign cr cr , c 2 cr º ¬ ¼

2 I1> @ r ,

(11-19)

with: 2 I1> @ r

1

2S

³ dt ³ d E

1

0

f

³

f

f

f

dy ³ dz ³ F t , E , y, z , x ) t , E , y, z , x dx , f

x0

and:

) t , E , y, z, x 1 x 2 y 2 z 2 2 xt 2 y 1 t 2 cos E 2 z 1 t 2 sin E

1 x

2

y2 z2

.

251

Chapter 11. Boundary Value Problems for All Knudsen Numbers

For the cylindrical geometry, the appropriate integral is given by:

12

240V 2 § kT0 · ¨ ¸ 2S 2 © m ¹

ª c 2 sign cr cr , c 2 cr º ¬ ¼

I 2> @ r , 2

(11-20)

with: 1

I 2> @ r 2

2S

³ dt

³

1

f

dE

0

³

f

f

f

dy ³ dx ³ F t , E , y, z , x ) t , E , y, z , x dz , x0c

f

The numerical values of the functions, I i> @ r , are necessary to obtain results of practical interest for arbitrary Knudsen numbers. The application of these functions will be more convenient if one introduces some special functions depending only on a radial coordinate by some normalization procedure. Using the values of the bracket integrals for the planar geometry [16], one can introduce the following notation for these functions: j

ªsign cr cr , c 2 cr º ¬ ¼

12

2kT0 · V 2 I i >[email protected] r , ¸ © m ¹

ª c 2 sign cr cr , c 2 cr º ¬ ¼

2§ 3¨

12

2kT0 · 2 > [email protected] ¸ V Ii r , © m ¹

23 14 S §¨

(11-21)

(11-22)

j where the normalized functions, I i > @ r , may be written as:

Ii > @ r

1

72 >[email protected] I r , 2 i

S

and:

2 Ii > @ r

1440 2 I i> @ r , S 8 3S 2

and in which the index, i , assumes the value of i 1 for the spherical geometry and assumes the value i 2 for the cylindrical geometry. Now,

j the functions, I i > @ r , have the common range, [1,0], for values of r from 1 to f . These special functions allow one to represent the solutions to many boundary value transport problems in analytical forms.

252

Analytical Methods for Problems of Molecular Transport

As one can see, the calculational procedure for the bracket integrals presents the main difficulty in the application of Lees’s method. The aforementioned difficulty typically results in the use of various approximate methods which will be considered later in Sections 11.7 and 11.8. Another method for simplifying the analysis is based on a specific form of the moment system which has been generated using some general properties of the Chapman-Enskog solution (see Section 11.10). One method that may be used to alleviate the need for numerical values of the special functions described above has been discussed for cylindrical and spherical geometries in [17,18] and in Appendix B. Since it is impossible to obtain these integrals

j in analytical forms, numerical values of the special functions, I i > @ r , would be very useful for many boundary value transport problems. Some of these numerical values have been given in Tables B-1 and B-2.

5.

BOUNDARY CONDITIONS FOR MOMENT EQUATIONS.

In a general case the microscopic boundary conditions on the body surface can be formulated for reflected molecular distribution function by means of the dispersion operator, A , that may be expressed by the relation:

f v, rS Af v, rS ,

(11-23)

where rS is the position radius vector for a given point on the particle surface. For the arbitrary boundary model the approximating functions given by Eqs. (11-3a), (11-3b) and (11-4) do not coincide with the distribution function of reflected molecules, expressed by Eq. (11-23), for any values of the characteristic parameters that are contained in them. In that case the boundary condition may, on the average, be satisfied only approximately if one uses the integral boundary model which may be written as: ³ v n Q v f dv

vn !0

³ v n Q v A f dv

,

(11-24)

vn ! 0

where n is a unit normal vector to the surface directed into the gas and Q v usually represents the molecular properties used in the moment equations. For various Q v , this relation yields the necessary number of boundary conditions for the moments.

Chapter 11. Boundary Value Problems for All Knudsen Numbers

253

A question arises concerning what molecular properties must be used for a given problem on each of the sections of the boundary. There are some uncertainties connected with this choice. Nevertheless, a general recommendation might be useful to facilitate this choice. Both on inner and outer surfaces it is convenient to use the integral form of the boundary conditions constructed on the integral definition of the accommodation coefficients. For heat and tangential momentum transport problems, the following integral boundary conditions should be used: Qn rS DT ªQn rS Qn rS ¬«

PW rS DW ª PW rS PW rS «¬

º ,

» eq ¼

º .

eq » ¼

(11-25a)

(11-25b)

The second relation is the generalized form of Eq. (6-7) in which ( PW (rS ))eq is the tangential momentum that would be carried away by reflected molecules if the tangential momentum accommodation coefficient were equal to unity. Note that, for stationary surfaces, ( PW (rS ))eq 0 . Another way of constructing boundary conditions [4] may be used for transport problems for which the approximating distribution function given by Eqs. (11-3a), (11-3b) and (11-4) is adapted to the Maxwellian boundary model (for example, the single-sphere heat transport, drag and torque problems). If the two-sided Maxwellian distribution function is adapted to the boundary model, the boundary conditions may be satisfied exactly by choosing certain characteristic parameters. To simplify this analysis, one can use the general expression for the reflected distribution function (Problem 6.5) which can be written in the form:

) c, rS

1 DW ) cc, rS

2G · ½ 2G (11-26a) §G DW 2 c 2 ®DTW w 1 DT ¨ 2 1 ¸ ¾ DW 1 , S ¹¿ S ©S ¯

where ) r is a measure of the perturbation in the distribution function from an absolute Maxwellian, cc c 2n c n , W w is a correction to the temperature at the surface point, rS , and G1 and G 2 are given by:

G1

³

cn 0

) c, rS c n exp c 2 dc ,

(11-26b)

254

Analytical Methods for Problems of Molecular Transport

G2

³

) c, rS c 2 c n exp c 2 dc .

cn 0

(11-26c)

Now, equating terms in Eq. (11-26a) with the same molecular velocity factors, one can obtain the boundary conditions necessary to determine the parameters in the two-sided Maxwellian distribution function associated with a given boundary value problem.

6.

THERMAL CONDUCTION FROM A HEATED SPHERE.

In this section, the main features of the transition between gas-kinetics and gas-dynamics are analyzed by means of the four-moment approach, utilizing the two-sided Maxwellian distribution function to close the moment system. This approach is illustrated by considering, for arbitrary Knudsen number, the steady heat conduction problem from a heated sphere. Consider the following brief statement of this problem [18]. Let a sphere of radius, R , be located in an infinite expanse of a gas which, far from the sphere, has a temperature, T0 , and a number density, n0 . The temperature, T0 'T , of the sphere surface is supposed to be slightly different from that of the surrounding gas 'T T0 . For this assumption the gas is near to the equilibrium state, and, therefore, this problem may be linearized be means of using the small parameters that are proportional to 'T T0 . The problem being considered possesses radial symmetry, and, therefore, the moment equation for transport of any molecular property, Q v , is:

1 w 2 r ³ fvr Q v dv r 2 wr wQ v wQ v 1 ° ³ ® ª¬ vT2 vI2 º¼ ª¬vI2 cot T vT vr º¼ wvr wvT r °¯ wQ v °½ ¬ªvI vr vT vI cot T ¼º ¾ fdv Rn'Q v . wvI °¿

(11-27)

Alternately selecting Q v to be the collisional invariants, 1, vr , and v 2 , for which 'Q v 0 , one obtains the ordinary continuity, radial momentum, and energy equations. Since the primary interest is in the radial heat flux, one also takes Q v vr v 2 which yields a fourth moment equation involving this heat flux [2].

Chapter 11. Boundary Value Problems for All Knudsen Numbers

255

For the four-moment approach the linearized distribution function may be used in the form given by Eq. (11-15). From Eq. (11-27), the following system of moment equations may be derived:

w 2 r nur wr

(11-28a)

0 ,

w 1 Prr 2 Prr PTT PII r wr

w 2 r Qr wr

(11-28b)

0 ,

0 ,

(11-28c)

1 w 2 2 2 1 r v vr fdv ³ v 2 vT2 vI2 fdv ³ 2 r r wr

Rn' v 2 vr .

(11-28d)

All moments of this system can be calculated by means of the method described in Section 11.3. These moments are given by the relations:

n r n0 1 12 Q 12 Q x ,

Pr r

12

§ 2kT0 · ¸ 2 ¨ 2 Sr © m ¹ 1

ur r

Q

12 W ,

p0 1 12 Q 12 Q x 3 12 W 12 W x3 ,

pr

p0 1 12 Q 12 Q x 12 W 12 W x ,

12

Qr r

p0 § 2kT0 · ¨ ¸ S r2 © m ¹

Q

32 W ,

256

Analytical Methods for Problems of Molecular Transport 2 2

³ c cr fdv

5 4

n0

1

1 2

Q 12 Q x3 W W x3 ,

and: 2 2 2 ³ c cT cI f dv

n0

5 2

1

1 2

Q 34 Q x W 32 W x 14 Q x3 12 W x3 ,

12

d 1 Q 2W dr

where x 1 r 2 . The moment system given by Eqs. (11-28) for the parameters of the distribution function can be expressed as:

d Q W dr

0 ,

(11-29b)

0 ,

d 3 Q 2W dr

0 ,

d Q 2W dr

8 5

(11-29a)

(11-29c)

12

R§ m · ¨ ¸ n0 © 2kT0 ¹

n' c 2 cr ,

(11-29d)

where the moment of the collision operator, n'c 2 cr , is determined by:

n' c 2 cr

n02

^ Q 1 2

32 W ªsign cr cr , cr c 2 º ¬ ¼

`

12 W ªc 2 sign cr cr , cr c 2 º . ¬ ¼ Now, consider the boundary conditions for these moment equations. Since the distribution function that has been employed is well suited to the Maxwellian boundary model, the ordinary microscopic boundary condition on the sphere surface can be used to yield:

Q 2 1 c 2 32 W 2 1

1 DW ª¬Q 1 1 c 2 32 W1 1 º¼ DW ª¬Q r c 2 32 W r º¼

.

(11-30a)

Chapter 11. Boundary Value Problems for All Knudsen Numbers

257

This expression contains two unknown quantities, Q r and W r , that may be found from the two additional conditions:

Nr

n0ur

0 ,

DT

Qr Qr , Qrw Qr

(11-30b)

(11-30c)

where the first relationship expresses the fact that the sphere is impenetrable to the gas molecules and the second one is the thermal accommodation condition. At large distances from the sphere the number density and the temperature tend to the equilibrium values that can be written as:

Q f Q f 0 ,

(11-30d)

W f W f 0 .

(11-30e)

After the integration of Eqs. (11-29a)-(11-29d), one can obtain:

Q

12 A3 ,

(11-31a)

W

A3 ,

(11-31b)

Q

A3[ r 2 A2 A4 ,

(11-31c)

W

A3[ r A4 A2 ,

(11-31d)

where A2 , A3 , and A4 A1 0 are the arbitrary constants of integration and a function, [ r , may be expressed as:

[ r

r ª r >[email protected] º

2 c c Kn «D1 ³ I1 r dr D 2 ³ I1 > @ r c dr c» . «¬ 1 »¼ 1 1

(11-31e)

258

Analytical Methods for Problems of Molecular Transport

In this expression, the following notations were introduced:

D1

8 3 S

8 2 , D2 15S

2

,

30S

and the functions I1 > @ r and I1 > @ r are determined by Eqs. (11-21) and (11-22). Using the boundary conditions given by Eqs. (11-30a)-(11-30e) and (1131a)-(11-31e) and performing some simple transformations, one obtains the following expression for the heat flux through the sphere surface:

1

Q Q fm

2

1 , 1 ] Kn 1

(11-32a)

where: 12

n kT § 2kT0 · 'T 4S R 0 0 ¨ , DW DT ¸ T0 S © m ¹ 2

Q fm

(11-32b)

is the free-molecular heat flux and the quantity, ] , is given by:

]

1 2

ª

f

f

º

«¬

1

1

»¼

DW DT «D1 ³ I1 >[email protected] r dr D 2 ³ I1 > [email protected] r dr » ,

(11-33)

in which DW and DT are the accommodation coefficients. This analytical expression gives the exact result for the free-molecular regime when Kn o f . For another limiting regime, when Kn o 0 , Eq. (11-32a) may be written as:

Q

4S RN'T ,

where N is the thermal conductivity coefficient that, for the given approach, can be expressed in the form: 12

N

2 § 2kT0 · ¨ ¸ S© m ¹

f ª f 1 º

2 n0 k O «D1 ³ I1 > @ r dr D 2 ³ I1 > @ r dr » «¬ 1 »¼ 1

1

.

259

Chapter 11. Boundary Value Problems for All Knudsen Numbers

If Kn o 0 , this expression for the thermal conductivity coefficient must tend to that obtained by the same approach for the planar geometry [16]. Using results from [16], one obtains: f

f

1

1

D1 ³ I1 >[email protected] r dr D 2 ³ I1 > [email protected] r dr D1 D 2 . This relationship may be considered a general property of the special functions. Having employed this property, one can express Eq. (11-33) in the form:

]

1 2

DW DT D1 D 2

0.2508 DW DT

(11-34)

.

The results obtained here will be discussed later in detail together with other results that will be derived by the use of the various approximate methods.

7.

METHOD OF THE ‘SMOOTHED’ DISTRIBUTION FUNCTION.

The absence of analytical expressions for the bracket integrals containing discontinuous Lees’ distribution functions prevents one from obtaining solutions for boundary value transport problems, with the exception of cases involving the planar geometry and Maxwellian molecules. Some approximate results can be obtained by utilizing a new procedure proposed in [19] that allows one to overcome these difficulties. A continuous distribution function of the Chapman-Enskog type is assumed to be employed in the collision integral instead of the actual discontinuous function. This auxiliary function may be called a ‘smoothed’ distribution function. The unknown parameters of the ‘smoothed’ distribution function may be determined by equating its basic moments with those obtained from the actual distribution function. Employing this method to solve the problem investigated in Sec. 11.6, one assumes that the ‘smoothed’ distribution function used in the collision operator has the following form:

f

^

` , (11-35)

1 0 f 1 Q r c 2 32 W r 2cr ur r A r cr S3 2 c 2

where Q r and W r are corrections to the number density and the temperature. The quantity, A r , is determined by:

260

Analytical Methods for Problems of Molecular Transport

³ cr c

2

fdc

³ cr c

2

f dc ,

(11-36)

where, f , is given by Eq. (11-15). All unknown parameters in Eq. (11-35) may be expressed through the basic functions Q r r and W r r . This ‘smoothed’ function allows one to obtain analytical results for arbitrary molecular potential interactions. The standard bracket integrals, which are tabulated for many molecular potential models, may be utilized for these analyses. For example, the expression, n 'c 2 cr , from the moment system, Eqs. (11-29d), may be written as:

n 'c 2 cr

1 1 n02 A r ª cr S3 2 c 2 , cr S3 2 c 2 º . ¬ ¼

Using this relation in the right-hand-side of Eq. (11-29d), one can obtain (for rigid-sphere molecules) the following final formula for the heat flux:

Q Q fm

1 , 1 ] Kn 1

(11-37)

where ] DW DT 64 S 1 . 75 As one can see, this formula gives the exact expression for the heat flux in a continuum regime when Kn o 0 . This method allows one to simplify the calculation of the bracket integrals. Moreover, there are no difficulties in generalizing this analysis for arbitrary models of the intermolecular potential.

8.

THE POLYNOMIAL EXPANSION METHOD.

The above mentioned difficulties of calculating the bracket integrals in the analytical form can be overcome by means of the use of a special representation for the function, sign cr cr . In this section a method will be presented to calculate the bracket integrals in the analytical form that consists of an expansion of the discontinuous function, sign cr cr , in velocity space polynomials. First, consider the problem in a spherical geometry. In Cartesian coordinates, this function can be written as:

sign cr c cos F 0 sign §¨ cr cr2 cT2 cI2 1 r 2 ©

12

· . ¸ ¹

(11-38)

It is easy to see that this function depends upon the three velocity coordinates and on the radial coordinate of the position of a molecule.

Chapter 11. Boundary Value Problems for All Knudsen Numbers

261

Owing to this notation, one must utilize a three-dimensional expansion of the velocity components in Hermite polynomials. This expansion was first proposed by Ivchenko [16]. But it is necessary to note that the expansion given in [16] is very difficult to be realized for practical calculations as there are many difficulties connected with the use of general schemes for the three-dimensional expansions. Since the function given by Eq. (11-38) depends only on the sum of the squares of the independent variables, cT and cI , one could hope to find a more simple scheme in which the sum, cT2 cI2 9 2 , is considered as a new independent variable. As 0 d 9 f , Hermite polynomials cannot be employed for this expansion. Therefore, Sonine polynomials are used instead. A polar coordinate system is introduced for the above velocity components by means of the relations cT 9 cos E , cI 9 sin E , and dcT dcI 9 d 9 d E . Integration with respect to these variables is determined by: f f

³³

exp cT2 cI2

f f

F 9 2 dcT dcI

f

2S ³ exp 9 2 F 9 2 9 d9 0

f

S ³ F x exp x dx , 0

and therefore the integrand contains the weight function, x m 1 . This p means that m 0 and the Sonine polynomials, S0 9 2 , must used in the expansion. The orthonormal conditions for this kind of the Sonine polynomials can be expressed as:

f

p q ³ exp x S0 x S0 x dx

G pq .

(11-39)

0

This polynomial can be defined by the expression: n

S0

x p n x ¦ 2 p 0 p n p n

,

where: 0 1 2 S0 x 1 , S0 x 1 x , S0 x 1 2 x 12 x 2 , and:

(11-40)

262

Analytical Methods for Problems of Molecular Transport 3 S0 x 1 3x 32 x 2 16 x 3 .

Taking into account this analysis, one can use the following twodimensional expansion:

sign §¨ cr cr2 cT2 cI2 1 r 2 ©

12

· ¸ ¹

(11-41)

¦¦ Akn r H k cr S0 cT2 cI2 . n

k

n

Since this expansion contains only the odd Hermite polynomials, the coefficients, Akn r , are given by: f

4

Akn r

2k k S

³c 0

2

exp c 2 dc (11-42)

1

n

u ³ H k c t S0 t0

c

2

2

ª1 t º dt , ¬ ¼

12

. where t0 1 r 2 Now, consider the features of this expansion for the cylindrical geometry. As one can see from Eq. (11-6), for this case, the quantity, cr , depends only on the two molecular velocity components, cr and cT , and, therefore, the appropriate expansion is defined by:

sign §¨ cr cr2 cT2 1 r 2 ©

12

· ¸ ¹

¦¦ Aij r H i cr H j cT i

, (11-43)

j

where the coefficients, Aij r , in this expansion can be expressed as: f

Aij r

1 c9 exp c92 dc9 i j 2 ijS ³0

2S S D ° °½ u ® ³ H i cr H j cT dT 2 ³ H i cr H j cT dT ¾ , °¯ 0 °¿ D

where cT c9 cos T and cr c9 sin T . In the planar geometry, one has a simple one-dimensional expansion given by:

263

Chapter 11. Boundary Value Problems for All Knudsen Numbers

sign cx

f

¦ Ai H i cx

(11-44)

,

i 0

where the constant coefficients can be determined by:

Ai

1 i 2 i S

f

³ exp cx H i cx sign cx dcx 2

.

f

This expansion was first proposed by Savkov in [20]. Now, consider some of the features of this method associated with the spherical geometry. If the expansion given by Eq. (11-41) is used in the bracket integrals presented by Eqs. (11-21) and (11-22), one can obtain the

i following analytical expressions for the special functions, I1 > @ r : >@ I1 > @ r 32 Z 1 ¦¦ Akn r D kn ,

1

1

k

2 I1 > @ r

(11-45)

n

12 > [email protected] Z 1 ¦¦ Akn r D kn . 8 3S k n

(11-46)

>@ > @ Here, the constants D kn and D kn are the common bracket integrals: 1

2

>[email protected] D kn

ª H k cr S0 n cT2 cI2 , cr c 2 º , ¬ ¼

(11-47a)

> [email protected] D kn

ª H k cr c 2 S0 n cT2 cI2 , cr c 2 º . ¬ ¼

(11-47b)

Instead of the infinite sums in Eqs. (11-45) and (11-46) one assumes that finite sums can be employed. It is not difficult to determine that the convergence of these series is very rapid. Indeed, if three terms n >[email protected] 0 ; k 1,3,5 are used, one will find that I1 >[email protected] 1 1.0102 and I1 1 1.1198 . The relative errors of such an approach are about 1% and 12% r 1 for the first and second functions respectively. When n 0 ; 1 d k d 7 , the relative errors are less than 1% r 1 for both functions. All of the bracket integrals necessary for this analysis can be found in Table 11-1. For arbitrary r , the other terms for which n z 0 must be considered. To

i obtain some appropriate expressions for the special functions, I1 > @ r , the

264

Analytical Methods for Problems of Molecular Transport

Table 11-1. Expressions for some bracket integrals.

ª cri , cr c 2 º ¬ ¼

i a †

a 2S Z

ª cri c 2 , cr c 2 º ¬ ¼

†

a 2S Z

3

5

7

1

3

5

7

4 5

30 7

287 12

4 3

6

123 4

741 4

12

Z V 2 2kT m

following finite sums, for which n 0 ; k 1,3,5,7 and k 1; n 1, 2,3 , are employed. Moreover, to obtain the more accurate expressions for

2 I1 > @ r , an additional term k 3 ; n 1 has been included in the expansion given by Eq. (11-46). For both expansions, the exact values of the bracket integrals are given in Appendix C and the expression > [email protected] D 31 925 2S Z has been employed. Having substituted the bracket integrals into Eqs. (11-45) and (11-46), one can obtain:

I1 > @ r

53 6 53 4 2 ª¬ 105 r 84 r 163 r 2 º¼ , 280

(11-48)

2 I1 > @ r

12 2 95 4 ª 19 r 6 84 r 17 r 2 º¼ . 21 8 3S ¬ 21

(11-49)

1

These analytical expressions can be readily integrated, after which, one obtains: f

>@ ³ I1 r dr

1

f

0.9779

1

,

> [email protected] ³ I1 r dr 0.9793 .

(11-50)

1

It is necessary to note, however, that the accuracy of the approach proposed here may only be specified by a comparison of these analytical results with those obtained numerically. The deviations of the integrals given in Eq. (1150) from the corresponding numerical values are approximately 2% for both of these integrals. This very close agreement with the numerical results indicates a rapid convergence of the expansions being used. Therefore, one can see that the approach proposed here is a reasonable approximate method for solving boundary value transport problems. For the problem of thermal conduction from a heated sphere, the method of polynomial expansion yields:

Q Q fm

1 , 1 ] Kn 1

(11-51a)

265

Chapter 11. Boundary Value Problems for All Knudsen Numbers Table 11-2. Values of the reduced heat flux ratio,

Q Q fm

DW

DT

1 .

Q Q fm

Kn 1 0.00 0.25 0.50 0.75 1.00 2.00 3.00 4.00 5.00 6.00 8.00 10.0 12.0 14.0 16.0 18.0 20.0

Exact Moment Solution

Smoothed Distribution Function Method

Method of Polynomial Expansion

1.0000 0.9410 0.8886 0.8417 0.7995 0.6660 0.5706 0.4992 0.4437 0.3992 0.3326 0.2851 0.2494 0.2217 0.1995 0.1813 0.1662

1.0000 0.9364 0.8804 0.8308 0.7864 0.6480 0.5510 0.4793 0.4241 0.3803 0.3152 0.2691 0.2348 0.2082 0.1871 0.1698 0.1555

1.0000 0.9422 0.8907 0.8445 0.8029 0.6707 0.5759 0.5045 0.4489 0.4044 0.3374 0.2894 0.2534 0.2254 0.2029 0.1845 0.1692

]

0.2508

]

0.2716

]

0.2455

where, for rigid-sphere molecules:

]

0.2455 DW DT

.

(11-51b)

The reduced heat flux given by Eqs. (11-51a) and (11-51b) deviates only slightly from the solution given by Eq. (11-34). Table 11-2 gives values of the reduced heat flux ratio for DW DT 1 and Fig. 11-2 illustrates these values for the various methods employed here.

9.

SOLUTION OF ONE CLASSIC TRANSPORT PROBLEM.

In this section the accuracy of the various methods previously proposed will be analyzed. A comparison is made between the various approximate analyses and the exact moment solution for a classical problem of heat transport between two parallel plates with the temperature difference being 2'T . Let the temperature of the lower plate x L 2 be T T0 'T and that of the upper one x L 2 be T T0 'T , where T0 is the temperature

266

Analytical Methods for Problems of Molecular Transport

Reduced Heat Flux (dimensionless)

1.0

0.8

0.6

0.4

0.2

0.0 0.0

5.0

10.0

15.0

20.0

Inverse Knudsen Number

Figure 11-2. A comparison of reduced heat flux values, Q Q fm , obtained from various approximate methods assuming (DW D T 1) with the Exact Moment Solution given in Table 11-2. The Exact Moment Solution is represented as a solid line while values obtained using the Smoothed Distribution Function method are shown as crosses and values obtained using the Polynomial Expansion Method are shown as open circles.

at x 0 . The temperature difference is assumed to be small compared with T0 , and therefore, the linearized theory is used in this problem. This problem has been previously formulated and studied by WangChang and Uhlenbeck [21], Jackson [22], Gross and Ziering [23], Bassanini, Cercignani and Pagani [24], Loyalka [25], and Ivchenko [16]. Moreover, the heat conduction characteristics and density distribution were experimentally determined by Teagan and Springer [26]. The solution of this planar problem may be based on the Maxwellian integral transport equations which can be directly derived from the Boltzmann equation by means of the relation:

w vx Q v fdv wx ³

Gf

³ Q v G t

dv ,

(11-52)

Chapter 11. Boundary Value Problems for All Knudsen Numbers

267

where x is the normal coordinate to the plate. The corresponding moment equations can be obtained by setting Q v equal to 1, vx , v 2 , and vx v 2 in Eq. (11-52). These equations can be written as:

d nu 0 , dx

(11-53a)

d Pxx dx

0 ,

(11-53b)

d Qx dx

0 ,

(11-53c)

d vx2 v 2 fdv dx ³

n' v x v 2 .

(11-53d)

In order to close this system of moment equations, the distribution function is chosen to be of the form:

f

^

0 f 1 12 ªQ x c 2 23 W x º ¬ ¼

`

12 ªQ x c 2 23 W x º sign cx . ¬ ¼

(11-54)

All the moments of the distribution function in Eqs. (11-53a)-(11-53d) may be calculated from the expressions given in Sec. 11.6 by setting r 1 . The solution of the moment system then contains four arbitrary constants of integration which can be specified by boundary conditions. For this problem the Maxwellian boundary model is employed in which the tangential accommodation coefficient is assumed to be equal to unity. Utilizing the values of the bracket integrals which are given in Table 111, one obtains the exact analytical solution via a four-moment approach. The heat flux may then be written as [16]:

ª º DT < Kn 1 » Q Q fm «1 ¬ 2 DT ¼

1

,

@ r . Solution: The moment system for this cylindrical boundary value problem may be written as: d rnur 0 , dr d 1 Prr Prr PTT 0 , dr r

d rQr 0 , dr

12

1 d 1 r ³ fcr2 c 2 dv ³ fcT2 c 2 dv r dr r

§ m · R1 ¨ ¸ © 2kT0 ¹

n 'cr c 2 .

For the distribution function given in Eq. (11-5), the moments connected with this system may be calculated by means of the technique described in Section 11.3. For example, the number density, mean velocity, and heat flux are given by:

2D · § n r n0 ¨ 1 12 Q 12 Q , S ¸¹ ©

12

1 § 2kT0 · ¨ ¸ r S © m ¹

ur

1 2

Qr

p0 § 2kT0 · ¨ ¸ r S © m ¹

12

Q

Q

12 W ,

32 W ,

280

Analytical Methods for Problems of Molecular Transport

where D arccos r 1 . After some transformations, the moment system takes the same form as that given by Eqs. (11-29a)-(11-29d). Therefore, one can use the general solution described by Eqs. (11-31a)-(11-31d). The function, [ r , is given by: r r >1@ ½°

2 c c Kn ®D1 ³ I 2 r dr D 2 ³ I 2> @ r c dr c¾ . °¯ 1 °¿ 1 1 °

[ r

The boundary conditions may be written as n 1 n0 and:

DT Qrreq Qr# for r 1 and r

Qr

z

R2 . R1

The equilibrium number densities associated with Qrreq may be found from the impenetrability condition. This statement gives: 12

§ 2kT0 · 2 S R1 p0 ¨ ¸ © m ¹

Q

] z

1 2

DW DT 'T , 1 T0 1 z 1 DW DT Kn 1] z

°

z

z

°½

°¯

1

1

°¿

DW DT ®D1 ³ I 2 >1@ r dr D 2 ³ I 2 > 2@ r dr ¾ ,

where Q is the heat flux per unit length of the inner cylinder. 11.4. Using the Chapman-Enskog solution, Ar , as a molecular property in the moment system, determine the heat flux between two coaxial cylinders the radii and temperatures of which are R1 , T0 and R2 , T0 'T for the inner and outer cylinders, respectively. Assume 'T T0 . The number density of the gas is n0 at r R1 . Solution: The moment system for this problem is: d 1 Q 2W dr

d Q W dr

0 ,

0 ,

Chapter 11. Boundary Value Problems for All Knudsen Numbers

d 3 Q 2W dr

281

0 ,

12

d W dr

8 5

R1n0 § m · ¨ ¸ a1 © 2kT0 ¹

> Ar , ) @

,

where Ar is the Chapman-Enskog solution for thermal conductivity and ) is the correction to the distribution function. The basic function, [ r , is discussed in Section 11.6 and, in this case, may be written as:

[ r

8 5

R1 1 2 S ln r . a1

It is very important to note that this relation is applicable for arbitrary models of the intermolecular potential. For rigid-sphere molecules this expression becomes [ r 128 S 1 Kn 1 ln r . The heat flux per unit 75 length of the inner cylinder is then given by: 12

Q

§ 2kT0 · 2 S R1 ¨ ¸ © m ¹

p0

DW DT 'T , T0 1 z 1 1 DW DT Kn 1] z

where:

] z

64 75

S 1DW DT ln z and z

R2 . R1

11.5. Determine the general expression for the torque on one of two concentric spheres if the inner sphere of radius, R1 , is slowly rotating with an angular speed, Z , such that Z R1 (2kT m)1 2 and the outer sphere of radius, R2 , is stationary. Assume that molecules are rigid spheres. From this general expression, derive the torque for the specific limiting case when h R2 R1 R1 . Solution: To solve this problem, one can use the same distribution function as was obtained in Section 11.11. The corrections, G r r , are given by:

G r r ª A2 A1 43 ] 1 r 3 º and G r 2 S A1r . ¬ ¼

282

Analytical Methods for Problems of Molecular Transport

The constants of integration, A1 and A2 , may be specified from two boundary conditions at the sphere surfaces. It is convenient to use the integral form of the boundary conditions that are expressed by:

PrI

DW ª Pr#I PrrI

¬«

º

» eq ¼

;

r 1, z .

These relations yield:

G 1

1 2

ª§ 2m ·1 2

º

DW «¨ ¸ Z R1 G 1 G 1 » , kT ¹ ¬«© ¼»

z 4G z DW ª 12 G z G z z 4G z º . ¬ ¼ The torque on the inner sphere is found to be:

ª «1 DW « ¬

Kx

K x

K x

23 DW SUVR14Z

8 15

1

S Kn

1

z

3

1 z º » z 4 1 DW » ¼

1

where:

z4 , z 4 1 DW

12

V 8kT S m is the mean speed of the gas molecules, Kn 1 R1 O , z R2 R1 , and the x -axis is in the direction of the vector, Ȧ . In the limiting case when h R2 R1 R1 , the torque is given by: Kx

ª DW 8 1 h º S K1 x «1 » 5 O¼ ¬ 2 DW

K1 x

DW 2 SUVR14Z . 3 2 DW

1

with:

283

Chapter 11. Boundary Value Problems for All Knudsen Numbers

11.6. Determine the general expression for the torque on one of two coaxial cylinders if the inner cylinder of radius, R1 , is slowly rotating with an angular speed, Z , such that Z R1 ( 2kT m )1 2 and the outer cylinder of radius, R2 , is stationary. Assume that molecules are rigid spheres. From this general expression, derive the torque for the specific limiting case when h R2 R1 R1 . Solution: Using Q v cT and Q v cr cT B c 2 as the molecular properties in Eq. (11-2), one can obtain:

1 d 1 rncr cT ncr cT r dr r

1 d rncr2 cT B c 2 r dr

0 ,

1r nc B c 1r nc c B c

3

2

2 r T

T

12

§ m · R1 ¨ ¸ © 2kT ¹

2

n' cr cT B c 2 .

where r r R1 . For closing this moment system the two-sided Maxwellian 0 distribution function may be chosen in the form f i f 1 2cT Gi r . This choice yields:

n

ncr cT

2 Sr

ncr2 cT B c 2

ncT3 B c 2

3

G r ,

x ª º nb1 « 14 G r 12 S 1D G r 12 S 1G r 2r 2 1 » , r ¬ ¼

D x ª º nb1 « 34 G r 32 G r 12 S 1G r 2r 2 3 » , r S ¬ ¼

where D arccos r 1 system is found to be:

and x

12

1 r 2

. The solution of the moment

G r 2 S A1r , and: G r

A2 r A1r ª 2] 1 r 2 º , where: ¬ ¼

284

Analytical Methods for Problems of Molecular Transport 12

]

R1 § m · ¨ ¸ b1 © 2kT ¹

.

The constants, A1 and A2 , may be specified from the boundary conditions which are given by: PrT 1 DW ª PrT 1 PrT ¬«

º and P z D ª P z P rT W « rT rT ¬

» eq ¼

º.

» eq ¼

The torque per unit length of the inner cylinder may be expressed as:

Kx

ª z2 1 z º

« 1 4 1 » K x 1 DW 5 S Kn 3 « z 1 DW » ¬ ¼

K x

12 DW SUVR13Z

1

, where:

z3 , z 3 1 DW

12

V 8kT S m is the mean speed of the gas molecules, Kn 1 R1 O , z R2 R1 , and the x -axis is in the direction of the vector, Ȧ . In the limiting case when h R2 R1 R1 , the torque is given by: Kx

K1 x

K1 x

ª DW 8 1 h º S «1 » 5 O¼ ¬ 2 DW

1

,

DW 1 SUVR13Z . 2 DW 2

11.7. Generalize the sphere torque problem described in Section 11.11 for arbitrary models of the intermolecular potential. Derive the reduced torque for the Lennard-Jones (6-12) potential. Consider two different gases (N2 and CO2) when T 293 K. The quantity, ] , depending upon the model of the Solution: intermolecular potential, may be expressed in the form:

]

>] @RS V

2

: 2,2 å T ,

Chapter 11. Boundary Value Problems for All Knudsen Numbers

285

where the first-order Chapman-Enskog approximation for b1 is utilized, V V V RS , and T kT H is the reduced temperature ( V and H k are parameters of the potential function). In terms of the : -integral, the torque on the sphere is given by:

ª «1 DW ¬

Kn 1

Kx

K x

K x

23 DW SU vR 4Z ,

8 15

S

V

2

:

2,2 å

º T » ¼

1

,

where the Knudsen number is calculated for rigid-sphere molecules. For the specific condition of T 293 K, the reduced temperatures and : -integrals of the gases under consideration are: TN 2

TCO 2

293 91.5

293 190

3.2022 ;

: 2,2 å 3.2022 1.0218 ,

1.5491 ;

: 2,2 å 1.5491 1.2988 .

The reduced torque for the Lennard-Jones (6-12) potential can be expressed in the form:

Kx K x

ª1 DW< Kn 1 º ¬ ¼

1

,

where < N 2 0.1672 and < CO 2 0.1642 . For comparison, it should be noted that < RS 0.1698 for rigid-sphere molecules. 11.8. Determine the domains of applicability of the free-molecular and continuum formulas for the heat flux between two coaxial cylinders. Assume that the limiting formulas have to be correct to within about 5%. Consider the case when DW DT 1 . Solution: The heat flux expression derived in Problem 11.4 and the limiting expressions derived from it should be combined in the following manner for the free-molecular and continuum regimes:

Q 1 Q 1 . and 1 Q fm 1 Kn ] z Qc 1 Kn ª] z º 1 ¬ ¼

286

Analytical Methods for Problems of Molecular Transport

For ] z 64 S 1DW DT ln z , these free-molecular and continuum formulas 75 are valid to within about 5% if the following inequalities are satisfied:

Kn t 5.43249 ln z and Kn d 0.01358 ln z where Kn O R1 and z R2 R1 . 11.9. Determine the domains of applicability of the free-molecular and continuum formulas for the heat flux between two concentric spheres. Assume that the limiting formulas have to be correct to within about 5%. Consider the case when DW DT 1 . Solution: The heat flux expression derived in Problem 11.2 and the limiting expressions derived from it should be combined in the following manner for the free-molecular and continuum regimes:

Q 1 Q 1 and . Q fm 1 Kn 1] z Qc 1 Kn ª] z º 1 ¬ ¼

S 1DW DT z 1 z , these free-molecular and continuum For ] z 64 75 formulas are valid to within about 5% if the following inequalities are satisfied:

Kn t 5.43249 1 z 1 and Kn d 0.01358 1 z 1

where Kn O R1 and z R2 R1 .

REFERENCES 1. Kogan, M.N., “Recent Developments in the Kinetic Theory of Gases,” in Rarefied Gas Dynamics 6th Symposium 1, 1-39, Academic Press Inc., N.Y. (1969). 2. Lees, L., “Kinetic Theory Description of Rarefied Gas Flow,” J. Soc. Indust. Appl. Math. 13(1), 278-311 (1965). 3. Lees, L. and Liu, C.Y., “Kinetic Theory Description of Conductive Heat Transfer from a Fine Wire,” Phys. Fluids 5(10), 1137-1148 (1962). 4. Grad, H., “On the Kinetic Theory of Rarefied Gases,” Comm. Pure and Appl. Math. 2(4), 311-407 (1949). 5. Grad, H., “Principles of the Kinetic Theory of Gases,” in Handbuch der Physik, vol. 12, ed. Flügge, S. (Springer Verlag, Berlin, 1958). 6. Grad, H., “Asymptotic Theory of the Boltzmann Equation,” Phys. Fluids 6, 147 (1963). 7. Kumar, K., “Polynomial Expansions in Kinetic Theory of Gases,” Ann. Phys. 37(2), 113141 (1966).

Chapter 11. Boundary Value Problems for All Knudsen Numbers

287

8. Muckenfuss, C., “Some Aspects of Shock Structure According the Bimodal Model,” Phys. Fluids 5(11), 1325-1336 (1962). 9. Salwen, H., Grosch, C.E. and Ziering, S., “Extension of the Mott-Smith Method for a Onedimensional Shock Wave,” Phys. Fluids 7(2), 180-189 (1964). 10. Hurlbut, F.C., “Note on Conductive Heat Transfer from a Fine Wire,” Phys. Fluids 7(6),904-906 (1964). 11. Liu, C.Y. and Sigimura, T., “Rarefied Gas Flow Over a Sphere at Low Mach Numbers,” in Rarefied Gas Dynamics 6th Symposium 1, 789-794, Academic Press Inc., N.Y. (1969). 12. Phillips, W.F., “Drag on a Small Sphere Moving Through a Gas,” Phys. Fluids 18(9), 1089-1093 (1975). 13. Mott-Smith, H.M., “The Solution of the Boltzmann Equation for a Shock Wave,” Phys. Rev. 82(6), 885-892 (1951). 14. Krook, M., “Continuum Equations in Dynamics of Rarefied Gases,” J. Fluid. Mech. 6(4), 523-541 (1959). 15. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, N.Y., 1969). 16. Ivchenko, I.N., “Generalization of the Lees Method in Boundary Problems of Transfer,” J. Colloid and Interface Sci. 135(1), 16-19 (1990). 17. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “A Solution of the Problem of Heat Transfer Between Two Cylinders for Arbitrary Knudsen Number,” High Temperature 31(4), 776-783 (1993). 18. Ivchenko, I.N., Loyalka, S.K., and Tompson, R.V., “A Method for Solving Linearized Problems of the Transfer Theory for Spherical Geometry at Arbitrary Knudsen Numbers,” Fluid Dynamics 29(6), 888-893 (1994). 19. Ivchenko, I.N., “Evaporation (Condensation) Theory of Spherical Particles with All Knudsen Numbers,” J. Colloid and Interface Sci. 120(1), 1-7 (1987). 20. Savkov, S.A., The Slip Boundary Conditions of the Non-Uniform Binary Gas Mixture and an Application of Them in Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 21 Wang-Chang, C.S. and Uhlenbeck, G.E., “Heat Transport Between Parallel Plates,” University of Michigan Project M999 (1953). 22. Jackson, E.A., Boundary Value Problems in Kinetic Theory of Gases (Ph.D. Dissertation, Syracuse University, 1958). 23. Gross, E.P. and Ziering, S., “Heat Flow Between Parallel Plates,” Phys. Fluids 2(6), 701712 (1959). 24. Bassanini, P., Cercignani, C., and Pagani, C.D., “Comparison of Kinetic Theory Analyses of Heat Transfer Between Parallel Plates,” Inter. J. Heat Mass Transfer 10, 447 (1967). 25. Loyalka, S.K., “Linearized Couette Flow and Heat Transfer Between Two Parallel Plates,” in Rarefied Gas Dynamics 6th Symposium 1, 195, Academic Press Inc., N.Y. (1969). 26. Teagan, W.P. and Springer, G.S., “Heat-Transfer and Density-Distribution Measurements Between Parallel Plates in the Transition Regime,” Phys. Fluids 11(3), 497-506 (1968). 27. Halbritter, L., “Torque on a Rotating Ellipsoid in a Rarefied Gas,” Z. Naturforsh. 29a, 1717-1722 (1974). 28. Smirnov, L.P. and Chekalov, V.V., “Slow Rotation of a Sphere in a Bounded Rarefied Gas Volume,” Izv. AN SSSR, M. Zh. G. (4), 117-124 (1978) 29. Chandrasekhar, S., Hydrodynamic and Hydromagnetic Stability (Dover, New York, 1961). 30. Landau, L.D. and Lifshitz, E.M., Mechanics (Pergamon Press, New York, 1960). 31. Landau, L.D. and Lifshitz, E.M., Fluid Mechanics (Addison-Wesley, Reading, Massachusetts, 1960). 32. Loyalka, S.K., “Motion of a Sphere in a Gas: Numerical Solution of the Linearized Boltzmann Equation,” Phys. Fluids A 4(5), 1049-1056 (1992).

Chapter 12 BOUNDARY SLIP PHENOMENA IN A BINARY GAS MIXTURE

1.

THE FIRST-ORDER CHAPMAN-ENSKOG APPROXIMATION FOR A BINARY GAS MIXTURE.

The non-uniform state of a binary gas mixture is described by the distribution functions:

fi where Ci

0 1 f i r , t 1 < i Ci , r , t ; i 1, 2 ,

mi

0 f i r , t

12

2kT

(12-1)

0 Vi and fi r , t is given by:

§ · mi ni r , t ¨ ¨ 2S kT r , t ¸¸ © ¹

32

§ m v u r , t 2 i i ¨ exp ¨ 2kT r , t ©

· ¸. ¸ ¹

(12-2)

The corrections, < i Ci , r , t , may be expressed in terms of the first-order Chapman-Enskog solutions [1], A i , Di , and Bi , in the following manner: 1

< i1 Ci , r , t Ai1 d12

x1

w ln T wr

n1n2 m2 m1 nU

Di d12 2Bi : 1

ln p ,

1

wu , wr

(12-3)

(12-4)

290

Analytical Methods for Problems of Molecular Transport

where ni is the number density of the i -th constituent, mi is the molecular mass, p p1 p2 is the hydrostatic pressure, xi ni n , n n1 n2 , and U n1m1 n2 m2 . In Eq. (12-4), an external force field is assumed to be absent. The first-order Chapman-Enskog solutions for the thermal conductivity, 1 1 A1 and A2 , are given by [1]:

A1 1

1

A2

a1S3 2 C12 C1 1

1

a1S3 2 C22

12

kT D12 § m1 · § 2m2 · C1 ¸ , x1 ¨© 2kT ¸¹ ¨© m ¹

(12-5)

12

k D § m · § 2m · C2 T 12 ¨ 2 ¸ ¨ 1 C2 ¸ , x2 © 2kT ¹ © m ¹

(12-6)

where a1 and a1 are the transport coefficients, kT DT D12 is the thermal diffusion ratio, D12 is the diffusion coefficient, and m x1m1 x2 m2 . In the 1 same manner, the first-order Chapman-Enskog solutions for diffusion, D1 1 and D2 , are given by [1]: 12

D1

D12 § m1 · § 2m2 · C1 ¸ , x1 ¨© 2kT ¸¹ ¨© m ¹

D2

D12 § m2 · § 2m1 · C2 ¸ . ¨ ¸ ¨ x2 © 2kT ¹ © m ¹

1

1

(12-7)

12

(12-8)

and the first-order Chapman-Enskog solutions for viscosity, B1 and B2 , are given in terms of their respective components by [1]: 1

B1ij

b1 C1i C1 j 13 C12G ij ,

B2 ij

b1 C2i C2 j 13 C22G ij .

1

1

1

(12-9)

(12-10)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

2.

291

THE TRANSPORT COEFFICIENTS FOR A BINARY GAS MIXTURE.

The first-order Chapman-Enskog solutions for the unknown transport coefficients a1 , a1 , d 0 , b1 , and b1 , are derived from the following algebraic systems of equations [1]: 12

154

a11a1 a11a1 D 1

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

,

(12-11)

12

a00 d 0

G0

3 2

,

(12-12)

12

2kT · n ¨ ¸ © m0 ¹

b11b1 b11b1

b11b1 b11b1

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

154

a11a1 a11a1 D1

1 §

5 2

E 1 5 2

E1

,

(12-13)

n2 , n2

(12-14)

n1 , n2

(12-15)

The thermal conductivity, diffusion, and viscosity coefficients are then found to be [1]:

N

12 ª § 2kT ·1 2 º § 2kT · nk « x1 ¨ ¸ a1 x2 ¨ ¸ a1 » , «¬ © m1 ¹ »¼ © m2 ¹ 5 4

(12-16)

12

D12

1 xx 2 1 2

§ 2kT · ¨ ¸ © m0 ¹

d0 ,

(12-17)

and:

P

p x1b1 x2b1 ; p

nkT .

(12-18)

292

Analytical Methods for Problems of Molecular Transport

If one introduces the following standard notations of Chapman and Cowling [1]:

A

C

1 5

:12 2,2

1,3

: 1,2 , B 5:12 15 121,1 , 1,1 :12 :12

1

2 5

:121,2 1,1

:12

1 8

, and E

kT

, 1,1 M 1M 2 :12

then the known algebraic coefficients in Eqs. (12-11)-(12-15), a11 , a11 , a11 , a11 , a00 , b11 , b11 , b11 , and b11 , may be expressed in the following manner: a00

x1 x2

a11

x12

5 2

kT , E

(12-19)

kT

5 x1 x2

> P1 @1

u ª 14 6 M 12 ¬

a11

x22

5 2

kT

> P2 @1

kT 1 M1 E

5M 22

5 x1 x2

(12-20a)

M 22 B

2M 1M 2 Aº , ¼

kT 1 M2 E

(12-20b)

u ª 14 6 M 22 5M 12 M12 B 2M 1M 2 Aº , ¬ ¼

a11

b11

a11

x12

5 2

5 x1 x2

kT

> P1 @1

kT M1 M 2 E

kT 1 M1 E

5 x1 x2

12

2 3

ª¬ 114 B 2 Aº¼ ,

M1 M 2 A ,

(12-21)

(12-22a)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

b11

x22

b11

b11

5 2

kT

> P2 @1

5 x1 x2

5 x1 x2

kT E

kT 1 M2 E

2 3

M 2 M1 A ,

293 (12-22b)

23 A .

(12-23)

> P1 @1

is given by Eq. (5-62) with the substitutions m m1 and The coefficients a11 and b11 have been derived from the coefficients a11 and b11 by changing the indices associated with the gas constituents; > P1 @1 o > P2 @1 , x1 o x2 , x2 o x1 , M 1 o M 2 , and M 2 o M1 . Then, the transport coefficients may be presented in the form: where

V

V1 .

12

a1

52

m1 § 2kT · ¨ ¸ k © m1 ¹

>N1 @1 p > ' a @1

(12-24)

^

`

u x1Q1 x2 ª¬ P1M1M 2 11 4 B 8 A 2 P1 P2 º¼ ,

12

a1

m2 § 2kT · ¨ ¸ k © m2 ¹ 2 5

>N 2 @1 p > ' a @1

(12-25)

^

`

u x2Q2 x1 ª¬ P2 M 1M 2 11 4 B 8 A 2 P1 P2 º¼ ,

12

§ 2kT · ¸ © m0 ¹

E nx1 x2 1 , kT

d0

3 2¨

b1

1 ª x1 R1 x2 p > ' b @1 ¬«

b1

(12-26)

1 2

1 E > P2 @1 23 A º , ¼»

1 ª x2 R2 x1 p > ' b @1 «¬

1 2

1 E > P1 @1 32 A º . »¼

Here, the following notations have been incorporated:

(12-27)

(12-28)

294

Analytical Methods for Problems of Molecular Transport

> ' a @1

x12Q1 x22Q2 x1 x2 Q12 ,

> 'b @1

x12 R1 > P1 @1 x22 R2 > P2 @1 x1 x2 R12 ,

(12-30)

(12-31)

(12-32)

1

(12-29)

1

Q1

P1 6M 22 5M 12 4 M12 B 8M 1 M 2 A ,

Q2

P2 6 M12 5M 22 4 M 22 B 8M 1 M 2 A ,

Q12

3 M1 M 2

P1

M1 E > P1 @1 , P2

2

5 4 B 4M1 M 2 A 11 4 B 2P1P2 ,

1

1

M 2 E > P2 @1 ,

(12-33)

(12-34)

R1

2 3

M1 M 2 A ,

(12-35)

R2

2 3

M 2 M1 A ,

(12-36)

R12

3.

1 2

1

1

1

E > P1 @1 > P2 @1 34 A EM1M 2 .

(12-37)

THE SECOND-ORDER CHAPMAN-ENSKOG APPROXIMATION FOR A BINARY GAS MIXTURE.

The second-order Chapman-Enskog solutions for the functions, A i , Di , and Bi , can be expressed as:

Ai

1 2 ar1Ci ª S3 2 Ci2 ar 2 S3 2 Ci2 º kT Di , ¬ ¼

(12-38)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

295

12

Di

ª º D12 § mi · i 1 m j 1 zi S3 2 Ci2 » , ¨ ¸ Ci « 1 2 xi © 2kT0 ¹ m ¬ ¼

Bi

Ci Ci Bi Ci2

(12-39)

and: $

$

,

br1 Ci Ci 1 br*2 S5 2 Ci2 1

(12-40) 12

where ar*2 ar2 ar1 , br*2 br2 br1 , zi xi 2kT mi d r1 D12 , the upper and lower signs, r , refer to the first and second constituents of the gas mixture, respectively, and i z j . The transport coefficients of the thermal conductivity, diffusion and viscosity, ar1 , ar2 , d r1 , d 0 , br1 , and br2 for the second-order Chapman-Enskog approximation can be calculated from the three linear algebraic systems of equations given by: 2

¦ a p a pq

D q ; p, q z 0 ,

(12-41)

E q ; p, q z 0 ,

(12-42)

Gq ,

(12-43)

p 2

2

¦ bp bpq

p 2

and: 1

¦ d p a pq

p 1

where D r1 , E r1 , and G 0 are given by Eqs. (12-11)-(12-15) in Section 12-2 while the remaining coefficients, D r2 E r2 G r1 0 . The coefficient, a00 , of the algebraic system for diffusion, Eq. (12-43), is specified by Eq. (12-19) while the other coefficients in this system, a01 a10 and a0 1a10 , are expressible in the form: a01

x1 x2

a01

x1 x2

5 2

kT CM 2 M 11 2 , E

(12-44)

kT CM 1M 21 2 . E

(12-45)

and: 5 2

296

Analytical Methods for Problems of Molecular Transport

All three of the systems given in Eqs. (12-41)-(12-43) involve the coefficients, a11 , a11 , b11 , and b11 which are given by Eqs. (12-20)-(1223). For p 1, 2 and q r2 , explicit analytical expressions for a pq and bpq may be found if one uses the analytical representations of the appropriate bracket integrals given in [2,3]. These coefficients, a pq and bpq , are then found to be [2,3]:

: 2,3 · kT § ¨ 7 2 1 2,2 ¸ > P1 @1 ¨© :1 ¸¹ 1,2 M kT ª 35 2 2 2 2 :12 x1 x2 2 4 5 M M « 16 12M 1 5M 2 21 1 2 8 M 1 E «¬ :121,1 x12 85

a12

§ : 1,3 : 1,4 M 22 ¨ 194 121,1 12 121,1 ¨ : :12 12 ©

· § : 2,2 : 2,3 ¸ M 1M 2 ¨ 7 121,1 2 121,1 ¸ ¨ : :12 12 ¹ ©

·º ¸» , ¸» ¹¼

: 2,3 · kT § ¨ 7 2 2 2,2 ¸ > P2 @1 ¨© : 2 ¸¹ 1,2 M kT ª 35 2 2 2 2 :12 x1 x2 1 4 5 M M « 16 12M 2 5M 1 21 2 1 8 M 2 E «¬ :121,1

a1 2

x22 85

§ : 1,3 : 1,4 M 12 ¨ 194 121,1 12 121,1 ¨ : :12 12 ©

x1 x2 M 23 2 M11 2

a12

1,4 1 :12

2

:121,1

7

:12 2,2 :121,1

kT E

kT E 2

· § :12 2,2 :12 2,3 7 2 M M ¸ 1 2¨ ¸ ¨ : 1,1 :121,1 12 ¹ ©

1,2 (1,3) § 189 :12 19 :12 ¨ 595 8 4 ¨ 16 :121,1 :121,1 © : 1,4 : 2,2 : 2,3 · 12 121,1 7 121,1 2 121,1 ¸ , :12 :12 :12 ¸¹

x1 x2 M13 2 M 21 2

a1 2

(12-46a)

1,2 (1,3) § 595 189 :12 19 :12 ¨ 16 8 4 ¨ :121,1 :121,1 ©

:12 2,3 ·

¸,

:121,1 ¸¹

(12-46b)

·º ¸» , ¸» ¹¼

(12-47a)

(12-47b)

297

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

x12 85

a22

x1 x2

: 2,3 : 2,4 kT § 77 ¨ 4 7 1 2,2 1 2,2 > P1 @1 ¨© :1 :1

kT ª 35 4 2 2 4 « 64 40 M1 168M1 M 2 35M 2 M1E ¬

78 M 22 84M 12 35M 22

1,2 12 1,1 12

:121,4

x22 85

x1 x2

72 M 14

1,1 12

:

2,2

:12 2,3

2,4 3,3 º 3 :12 2 2 :12 2 M M 2 M M », 1 2 1 2 :121,1 :121,1 :121,1 »¼

: 2,3 : 2,4 kT § 77 ¨ 4 7 2 2,2 2 2,2 > P2 @1 ¨© :2 :2

· ¸ ¸ ¹

kT ª 35 4 2 2 4 « 40 M 2 168M 1 M 2 35M 1 M 2 E ¬ 64

:

: 1,5

78 M 12 84M 22 35M 12 1 M 12 8

(12-48a)

:121,3

14 M 24 121,1 72 M 1M 2 4 M 12 7 M 22 121,1 :12 :121,1 :12

14 M 1M 23

a2 2

::

18 M 22 108M 12 133M 22 72 M 24

· ¸ ¸ ¹

108M 22

:121,2

:

133M 12

:121,4

1,1 12

(12-48b)

:121,3

:

: 1,5

1,1 12

:

2,2

14 M 14 121,1 72 M 1M 2 4 M 22 7 M12 121,1 1,1 :12 :12 :12

14 M 2 M 13

:12 2,3

2,4 3,3 º 3 :12 2 2 :12 2 M M 2 M M », 2 1 1 2 :121,1 :121,1 :121,1 »¼

298

Analytical Methods for Problems of Molecular Transport

x1 x2 M 13 2 M 23 2

a2 2

241 8

14

b12

b1 2

:121,3

kT E

: 1,4

1,2 § 833 :12 ¨ 8505 8 ¨ 64 :121,1 ©

: 1,5

: 2,2

72 121,1 14 121,1 772 121,1 :121,1 :12 :12 :12

(12-49)

:12 2,3

:121,1 2,4 :12 3,3 · 2 2 ¸, :121,1 :121,1 :121,1 ¸¹

: 2,3 · kT § kT M 2 ¨ 7 2 1 2,2 ¸ x1 x2 23 ¨ ¸ E M1 > P1 @1 © :1 ¹ § : (1,2) : (2,2) : (2,3) u ¨ 352 M 1 7 M 1 121,1 214 M 2 121,1 23 M 2 121,1 ¨ :12 :12 :12 ©

x12 85

x22 85

: 2,3 kT § ¨ 7 2 2 2,2 > P2 @1 ¨© :2

· kT M 1 ¸ x1 x2 23 ¸ E M2 ¹

§ : (1,2) u ¨ 352 M 2 7 M 2 121,1 ¨ :12 ©

21 M1 4

§ : 1,2 ¨ 352 7 121,1 ¨ :12 ©

21 4

: kT § 35 M1 ¨ 2 7 121,1 E ¨© :12

21 4

b1 2

x1 x2 23 M 2

b12

x1 x2 23

kT E

· ¸, ¸ ¹

1,2

:12(2,2) :121,1

32 M 1

:12(2,3) ·

¸,

2,3 · 3 :12 ¸, 2 :121,1 :121,1 ¸¹

:121,1

3 2

(12-50b)

:121,1 ¸¹

:12 2,2

:12 2,2

(12-50a)

:12 2,3 ·

¸,

:121,1 ¸¹

(12-51a)

(12-51b)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

: 2,3 : 2,4 kT § 301 ¨ 12 7 1 2,2 1 2,2 > P1 @1 ¨© :1 :1

x12 85

b22

· ¸ ¸ ¹

kT ª 1 2 2 « 4 M1 140M 1 245M 2 M1E ¬ § : 1,2 : 1,3 · M 1M 22 ¨ 49 121,1 8 121,1 ¸ ¨ :12 :12 ¸¹ © x1 x2

2 3

18 M 2 154 M 12 147 M 22 21 M 23 2

b2 2

:12 2,2

:

1,1 12

2,4 3,3 º 3 :12 2 :12 3 M M M 3 », 2 1 2 2 :121,1 »¼ :121,1 :121,1

: 2 2,3 : 2 2,4 kT § 301 7 ¨ > P2 @1 ¨© 12 : 2 2,2 : 2 2,2

kT ª 1 2 2 « M 2 140 M 2 245M1 M2E ¬ 4 § : 1,2 : 1,3 · M 2 M 12 ¨ 49 121,1 8 121,1 ¸ ¨ :12 :12 ¸¹ ©

1 8

2 3

M 1 154 M 22

212 M 13

(12-52a)

:12 2,3

x22 85

x1 x2

147 M 12

· ¸ ¸ ¹

(12-52b)

:12 2,2

:

1,1 12

:12 2,3

2,4 3,3 º 2 :12 3 :12 3 M M M 3 », 1 2 1 2 :121,1 »¼ :121,1 :121,1

§ :121,2 :121,3 49 8 ¨ 385 ¨ 4 :121,1 :121,1 © :12 2,2 21 :12 2,3 3 :12 2,4 :12 3,3 · 3 301 ¸. 8 2 2 :121,1 :121,1 :121,1 :121,1 ¸¹

b2 2

299

x1 x2 23 M 1M 2

kT E

(12-53)

300

Analytical Methods for Problems of Molecular Transport

The coefficients, a pq and bpq , when p, q 0 , have been specified from the expressions for when p, q ! 0 by changing the indices connected with the gas constituents in the manner shown previously, i.e. > P1 @1 o > P2 @1 , x1 o x2 , x2 o x1 , M 1 o M 2 , and M 2 o M 1 . The same procedure applied to the expressions for a1 2 and b1 2 yield expressions for a12 and b12 . The remaining coefficients, a pq and bpq , can be found from the symmetry conditions, a pq aqp and bpq bqp . All of the coefficients described above depend upon the : -integrals, many of which have been previously tabulated in [2]. In Appendix F, a full set of these : -integrals, sufficient for use with the second-order approximation, have been included. Among these are values for the 1,4 1,5 3,3 previously unreported, : , : , and : . All of the : -integral values included in Appendix F have been fully recomputed and an annotated Mathematica® program for use in further calculations has been provided.

4.

ANALYTICAL METHODS OF SOLUTION FOR PLANAR BOUNDARY VALUE PROBLEMS INVOLVING BINARY GAS MIXTURES.

A semi-infinite expanse of a binary gas mixture is considered which is bounded by a planar surface lying in the y z plane and located at x 0 . Far from the surface, the gas mixture is maintained at a constant mean-mass velocity gradient normal to the surface, h wu wx f , and constant tangential gradients of the partial concentrations, d12 wx1 wy f , and temperature (in the y -direction), g w ln T wy f . The total pressure is assumed to be constant. For these conditions in a binary gas mixture, the mean-mass velocity far from the surface is expressible in the following asymptotic form: u x

u0 , xh

(12-54)

where u0 is a constant velocity that arises owing to the influence of the surface on the gas mixture. The asymptotic behavior of a binary gas mixture may only be found via solution of the Boltzmann equations with appropriate boundary conditions at the wall. For this geometry, the distribution function for the i -th constituent may be expressed as:

fi r

m º 0 ª iy < i ci , r ) ir ci , x » , f i «1 i xhv ¬ kT ¼

(12-55)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

301

12

where ci mi 2kT v i is the dimensionless velocity, < i ci , r is the Chapman-Enskog solution, ) ir ci , x is a function that allows for the influence of the wall, and the signs, and , refer to the reflected and incident molecules, respectively. The functions, ) ir ci , x , can be found from the following linearized Boltzmann equations: 0 w) i vix f i wx

ni2 I i ) i ni n j I ij ) i ) j ,

(12-56)

where i 1, 2 with i z j , and I i and I ij are the standard notations [1]. These linearized collision integrals are given by

ni2 I i ) i

0 0 ³³ fi vi fi v ª¬) i vi ) i v

) i vci ) i vc º¼ giD i T , gi d : dv ,

ni n j I ij ) i ) j

³³ f f ª¬) 0

i

0

j

i

) j

) ic ) cj º¼ gijD ij T , gij d : dv j .

(12-57)

(12-58)

Using the Maxwellian boundary model, Eq. (6-4), the boundary conditions for the functions, ) ir ci , x , can be written in the form:

) i ci ,0 D iW ª¬ Aiy g Diy d12 º¼ 2 2 D iW Bixy h 1 D iW ) i cix , ciy , ciz ,0 .

(12-59)

Let the scalar product be introduced by: 2

M1 ci , x ,M2 ci , x ¦ ³ M1 fi0 M2 dvi .

(12-60)

i 1

Then, using both the commutative property of the standard bracket integrals and the conservation of momentum present during molecular encounters, the scalar product of Eqs. (12-56) with mi viy gives:

302

Analytical Methods for Problems of Molecular Transport

w n 2 ) i , mi viy mi vix viy ,) i ci , x n 2 mi viy ,) i wx n12 ª¬)1 , m1v1 y º¼ n1 n2 ª¬)1 ) 2 , m1v1 y m2 v2 y º¼ 1 12

n22 ª¬) 2 , m2 v2 y º¼

^

2

`

^

` (12-61)

0.

where the bracket integral notations of Eq. (12-61) are defined in [1] as:

n 2 ^ F , G` n12 > F1 , G1 @1 n1n2 > F1 F2 , G1 G2 @12 n22 > F2 , G2 @2 ,

> Fi , Gi @i ³ Gi Ii Fi dvi , > F1 G2 , H1 K 2 @12 ³ F1I12 H1 K 2 dv1 ³ G2 I 21 H1 K 2 dv 2 , with F , G , Fi , Gi , H i , and Ki representing arbitrary functions of the molecular velocities of the subscripted constituents, and I i and I ij having been previously defined in Eqs. (12-57) and (12-58), respectively. In the dimensionless variables, ci , this moment solution may be written in the form:

c c

) i ci , x const .

ix iy ,

(12-62)

If one takes into account the asymptotic behaviors of the functions, 12 ) i ci , x 2 mi 2kT u0 ciy , then this function in the two-moment approach may be chosen in the form:

) ir ci , x M i1 2 a r x ciy .

(12-63)

This form of the correction to the distribution function means that Eq. (1262) may be expressed in the form:

c c

) i ci , x 0 ,

ix iy ,

(12-64)

which is the exact analytical solution of Eq. (12-56). The Maxwellian method [4] involves the use of the analytical solution expressed by Eq. (12-64) at the surface where x 0 , i.e.:

303

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

c c

) i ci ,0 0 ,

ix iy ,

(12-65)

where ) i ci ,0 K cix ) i ci ,0 K cix ) i ci ,0 and K cix is the Heaviside step function that is defined as:

1 ; x ! 0, ¯0 ; x 0.

K x ®

For this method, the boundary conditions are satisfied only in an integral sense and not in an explicit sense. To get some useful results using this method, additional assumptions must be made. The Maxwellian analysis is constructed around the assumption that, in Eq. (12-63), a r f a 0 a . Then, the constant, a , may be calculated if one uses the analytical solution given by Eq. (12-65). The Maxwellian method may be classified as a onemoment approach. Another analytical method, first proposed by Loyalka [5], involves using the two general moment solutions of the Boltzmann equations, Eq. (12-56). The first of these moment solutions, given by Eq. (12-65), is the same as that used in the Maxwellian method. The second moment solution may be derived from Eqs. (12-56) if one uses a property of the Chapman-Enskog solution for viscosity. For the particular planar geometry being considered here, the Chapman-Enskog corrections may be found from the following expressions: 0 f1 c1x c1 y

0 f 2 c2 x c2 y

n12 I1 B1xy n1n2 I12 B1xy B2 xy ,

(12-66)

n22 I 2 B2 xy n1n2 I 21 B1xy B2 xy .

(12-67)

Now, the scalar product of Eq. (12-56) with Bixy , taken in accordance with Eq. (12-60), yields:

w vix Bixy ,) i ci , x wx

^ ` n ^) c , x , B ` ³ ) c , x ª n I B n n I B B º dv ¬ ¼ ³ ) c , x ª n I B n n I B B º dv . ¬ ¼ 1

1

2

2

n 2 Bixy ,) i ci , x 2 1 1

2 2 2

1 xy

2 xy

1 2 12 1

2 21

2

i

1 xy

1 xy

i

2 xy

2 xy

ixy

1

2

304

Analytical Methods for Problems of Molecular Transport

Here, the commutative property of the bracket integrals has been used. From this, taking into account Eqs. (12-64), (12-66) and (12-67), one obtains:

w vix Bixy ,) i ci , x wx

) c , x , c c i

i

ix iy

0,

which yields:

M

1 2 2 cix ciy Bi i

c ,) c , x 2 i

i

i

const .

(12-68)

The Loyalka method is a generalization of the Maxwellian method and may be classified as a two-moment approach because two-moment solutions to the Boltzmann equations are used in its formulation. Specifically, for the Loyalka method, the two solutions used are the following exact moment solutions:

c c

) i ci , x 0 ,

ix iy ,

M

1 2 2 cix ciy Bi i

(12-69)

c , ª¬) c ,0 ) c , f º¼ 2 i

i

i

i

i

0.

(12-70)

These two analytical solutions give useful results if one takes into account two assumptions concerning the corrections to the distribution functions. Specifically, one must introduce two constants, a r f a and a 0 aw , where a z aw , which may be calculated from the two analytical solutions given in Eqs. (12-69) and (12-70).

5.

THE SLIP COEFFICIENTS FOR A BINARY GAS MIXTURE.

The slip phenomena in gas mixtures are of fundamental significance when it comes to the specification of appropriate boundary conditions for flows in the slip-flow regime. In this regime, the boundary condition for the mean mass velocity component tangential to a body surface may be presented in the form [6-13]:

uW rS cm

OP w ln T * Q* PnW cTsl cDsl D12 d12 W , P wrW

(12-71)

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

305

where rS is the radius vector of a given point on the body surface, n indicates a direction normal to the surface and points into the gas at rS while W indicates a direction tangential to the body surface at the point rS . Also, * cm , cTsl , and cDsl are the velocity-slip, thermal-creep, and diffusion-slip coefficients, respectively and the following notations have been used: 12

P § 2kT · ¨ ¸ S p© m ¹

OP

8 5

Q*

P ON , U OP

ON

64 75

,

12

N § m · ¨ ¸ T, S p © 2kT ¹

in which OP and ON are the mean free paths calculated from the viscosity and thermal conductivity coefficients, P and N , respectively, and PnW is the pressure tensor component. The first term in Eq. (12-71) yields a correction to continuum expressions which is proportional to the Knudsen number, Kn , and which, when Kn o 0 , may be neglected. However, the other terms in Eq. (12-71), which are associated with the thermal-creep and diffusion-slip phenomena, may be significant even if Kn o 0 because the influence of these phenomena depends only upon the value of the tangential component of the external non-uniformities in the gas mixture. In the slip-flow regime, for the bulk flow of the gas mixture, the NavierStokes equations can be used to develop the same description as the Boltzmann equations if one introduces the appropriate slip boundary condition given by Eq. (12-71). For this regime, the use of the NavierStokes equations together with the slip boundary condition is preferable to the use of the full Boltzmann equations because it allows one to reduce the number of independent variables from ten (for the Boltzmann equations for a binary gas mixture) to four (in the general case with the Navier-Stokes equations). For stationary boundary value problems and for some specific geometries the number of these independent variables may be additionally reduced. It should be noted that the slip conditions themselves must be determined by a solution of the planar boundary value problem involving the Boltzmann equations. These slip coefficients can then be used with suitable accuracy to formulate transport problems for arbitrary geometries.

306

Analytical Methods for Problems of Molecular Transport

The slip-flow phenomena are basic to understanding cross-effects in gas mixtures (where they can be the most important factors), the motions of aerosol particles suspended in non-uniform media, and the flow of gas mixtures through capillary tubes, all of which have numerous technological applications. The slip effects are also of importance in studies of many natural phenomena such as aerosol mechanics and flow in porous media, which also play a significant role in numerous other technological applications such as chemical and physical vapor deposition, nanofabrication, and most low-pressure applications. In some cases, such as physical vapor transport experiments under micro-gravity conditions, for example, these phenomena may become the dominant transport factors due to the production of side-wall gas creep which may suppress natural convection [14-20]. In spite of considerable work in this field [21-63], there is no completely accurate analysis employing realistic models of the intermolecular interactions that would have reliable accuracy in all applications. The half-range moment method [34,35,51] described in Chapter 8, was previously believed to be able to yield accurate results for slip problems. Unfortunately, this method involves a very complicated analysis (even in low-moment approaches) and the accuracy of the method is not predictable. The relatively simple Loyalka’s method described in this chapter results in a simpler analysis and good accuracy [5]. To calculate the slip coefficients for binary gas mixtures, Loyalka’s method is among the most useful. First, it yields relatively simple expressions and, second, it has no difficulties associated with the use of general intermolecular and gas-surface interaction laws. Moreover, the use of the two general conservation laws as exact moment solutions provides a suitable accuracy to the results. Next in this section, Loyalka’s method is used to derive analytical expressions for the slip coefficients of a binary gas mixture. This analysis employs the second-order Chapman-Enskog solutions for the transport parameters. One uses the same statement of the planar boundary value problem as was described in Section 12.4. The velocity, u0 , given by Eq. (12-54) is presented in the form:

u0

cm OP h cDsl D12 d12 cTslQ * g .

(12-72)

Additionally, this asymptotic velocity may be calculated by simple integration, i.e.:

U u0

m v

12 r i iy , M i a

f ciy ,

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

307

where a r f a const . This integration yields: 12

u0

§ 2kT · ¸ © m0 ¹

1 2¨

(12-73)

a.

Within the framework of the Loyalka method, one has two unknown constants, a r f a and a 0 aw , which can be found from the two exact moment solutions given in Eqs. (12-69) and (12-70). Substitution of the boundary condition given by Eq. (12-59) into Eqs. (12-69) and (12-70) yields the analytical expressions for the constants, a and aw . Then, Eqs. (12-72) and (12-73) yield the following expressions for the slip coefficients:

cm

cDsl

5 8

pM 1 2S

P

2

ª

i 1

¬

4b

º

¦ 2 D iW xi br1 « K1 S Mr11 2 K 2Z1i » ,

2

ª

§

i 1

¬«

©

¦ 2D iW « M i1 2 K1 ¨ 1

i 1

mj

· 14 zi ¸ m ¹

§ ·º i 1 m j br1 K 2 ¨ 1 12 ziZ2i ¸ » , m © ¹ ¼»

* cTsl

3 a*

ª 2 « ¦ D iW xi ar1 «¬ i 1

1 2

K1H1i br1M i1 2 K 2H 2i

where:

x1M 1 x2 M 2 , a*

x1a1M 11 2 x2 a1M 21 2 ,

2

Z1i 1 br*2 174 br*2 , Z2i 1 72 br*2 ,

H1i 1 14 ar*2 , and H 2i 1 72 br*2 72 ar*2br*2 .

(12-75)

12 º § m · kT D12 ¨ 0 ¸ cDsl » , © 2kT ¹ »¼

M

(12-74)

¼

i

(12-76)

308

Analytical Methods for Problems of Molecular Transport

Here, also, the following notations have been used:

K1

2 D1W x1b1 2 D 2W x2b1 K

K2

1 4

D1W x1M 11 2 D 2W x2 M 21 2

2

,

x1b1 x2b1 1 .

(12-77)

(12-78)

Equations (12-74)-(12-78) are suitable for the computation of results corresponding to arbitrary models of the intermolecular interaction provided that appropriate values of the transport coefficients, ar1 , br1 , etc, are available. Now, consider the limiting case of a simple gas. The limiting formulas may be derived from Eqs. (12-74)-(12-78) if one makes the substitutions: x1 1 , m1 m2 m , V 1 V 2 V , and D1W D 2W DW . The factors K1 and K 2 are then given by:

K1

1 4

2 2 DW DW1 and K 2

1 1 b . 4 1

(12-79)

The limiting expressions for the slip coefficients are the same as those given by Eqs. (9-62) and (9-63) and the diffusion-slip coefficient is found to be cDsl 0 .

6.

DISCUSSION OF THE SLIP COEFFICIENT RESULTS.

The expressions for the slip coefficients that have been obtained in the preceding sections of this chapter are in simple algebraic forms readily usable in computations. Such computations are straightforward provided that values of the coefficients, ar1 , br1 , etc., are available. These values depend upon the intermolecular interaction (potential) model that is used to compute the : -integrals. While many of the necessary : -integrals are tabulated in the classical texts [2,3], there are several additional : -integrals that have been newly computed for inclusion in this text. The various : integrals, both old and new, are given in Appendix F. Tables 12-1 and 12-2 contain values of the transport coefficients, P , N ,

and D12 , and of the slip coefficients, cm , cDsl , cTsl , and cTsl , that have been computed for a selection of binary gas mixtures using different intermolecular potential models. In Table 12-1 the rigid-sphere potential

309

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

Table 12-1. Values of the transport coefficients ( P , N , and D12 ) and of the slip coefficients

( cm , cDsl , cTsl , and cTsl ) for a selection of binary gas mixtures obtained using the rigidsphere potential model and the first- and second-order Chapman-Enskog approximations ( p =1 atm, T0 =293 K). The units of the viscosity, P , are gm cmí1 secí1, the units of the thermal conductivity, N , are cal cmí1 secí1 Kí1, and the units of the diffusion coefficient, D12 , are cm2 secí1. The slip coefficients are all dimensionless. The thermal conductivity coefficients have been calculated under the assumption that the molecules are effectively monatomic. For all of these calculated values, the assumption has been made that the tangential momentum accommodation coefficient for each gas species is equal to unity, i.e. D1W = D 2W =1. The column labeled ‘Order’ refers to the order of the Chapman-Enskog approximation that was used in calculations.

P

N

( u 105) 4.5642 4.6736

D12

cm

cDsl

cTsl

cTsl

1 2

( u 104) 1.6874 1.7125

0.6379 0.6857

1.1189 1.0914

-0.1775 -0.1920

1.1102 0.9786

1.1200 0.9961

N2-Ar

1 2

1.9262 1.9545

4.1837 4.2788

0.1659 0.1690

1.1206 1.0936

0.1170 0.1265

1.1120 0.9825

1.1029 0.9822

x1 =0.5

N2-O2

1 2

1.8202 1.8471

4.5041 4.6063

0.1752 0.1783

1.1167 1.0893

0.0363 0.0395

1.1235 0.9924

1.1211 0.9979

N2-CO2 x1 =0.5

1 2

1.5254 1.5493

3.1895 3.2666

0.1269 0.1296

1.1253 1.0982

0.2211 0.2374

1.0976 0.9625

1.1106 0.9821

Mixture

Order

N2-H2

x1 =0.99 x1 =0.5

model was used while in Table 12-2 the Lennard-Jones (6-12) potential model was used. For all mixtures and models the results obtained using the first-order Chapman-Enskog approximation are compared with those obtained using the second-order Chapman-Enskog approximation. In all of the cases shown in Tables 12-1 and 12-2 molecules have been assumed to be diffusely reflected at the relevant surfaces such that D1W D 2W 1 . Table 123 summarizes values of the parameters that are associated with each of the intermolecular potential models used. A variety of additional models are also available in the literature for which results have not been computed here but which may also be of interest to the reader [64]. This source contains tabulated values of the transport collision integrals for a variety of modern potential models and provides substantial information about the various intermolecular collision parameters included in these models. Some approximate methods using the various alternative potential models described in this reference, and which may be used to calculate the various transport coefficients for the second- and thirdorder Chapman-Enskog approximations, may be found in the literature [6569] but, since these analyses generally are not complete, they tend to be of limited use. It is clear from Tables 12-1 and 12-2 that there exists a significant dependence of the slip coefficients upon the intermolecular potential (within the context of the Loyalka method) even when the calculations are limited to

310

Analytical Methods for Problems of Molecular Transport

Table 12-2. Values of the transport coefficients ( P , N , and D12 ) and of the slip

coefficients ( cm , cDsl , cTsl , and cTsl ) for a selection of binary gas mixtures obtained using the Lennard-Jones (6-12) potential model and the first- and second-order Chapman-Enskog approximations ( p =1 atm, T0 =293 K). The units of the viscosity, P , are gm cmí1 secí1, the units of the thermal conductivity, N , are cal cmí1 secí1 Kí1, and the units of the diffusion coefficient, D12 , are cm2 secí1. The thermal conductivity coefficients have been calculated under the assumption that the molecules are effectively monatomic. The slip coefficients are all dimensionless. For all of these calculated values, the assumption has been made that the tangential momentum accommodation coefficient for each gas species is equal to unity, i.e. D1W = D 2W =1. The column labeled ‘Order’ refers to the order of the Chapman-Enskog approximation that was used in calculations. All of the necessary : -integrals have been calculated numerically using the appropriate reduced temperatures and without using the typical interpolation procedure.

P

N

( u 105) 4.7246 4.7576

D12

cm

cDsl

cTsl

cTsl

1 2

( u 104) 1.7375 1.7446

0.7361 0.7564

1.1189 1.1021

-0.1896 -0.1959

1.1072 1.0342

1.1228 1.0518

N2-Ar

1 2

1.9872 1.9932

4.3334 4.3543

0.1870 0.1877

1.1208 1.1064

0.1307 0.1331

1.1138 1.0514

1.1091 1.0489

x1 =0.5

N2-O2

1 2

1.8825 1.8893

4.6598 4.6858

0.1982 0.1990

1.1168 1.1009

0.0380 0.0390

1.1238 1.0548

1.1216 1.0549

N2-CO2 x1 =0.5

1 2

1.6006 1.6037

3.3829 3.3940

0.1477 0.1481

1.1252 1.1141

0.2589 0.2621

1.1051 1.0560

1.1303 1.0815

Mixture

Order

N2-H2

x1 =0.99

x1 =0.5

the first-order Chapman-Enskog solutions whereas, in a simple gas, such a dependence is observed only for calculations employing the second- or higher-order Chapman-Enskog solutions. As is evident from Eqs. (12-74)(12-76), the thermal-creep and diffusion-slip coefficients are significantly more sensitive to the order of the Chapman-Enskog approximation than the velocity-slip coefficient. This is undoubtedly due to the dependence of the thermal-creep and diffusion-slip coefficients on cross-effects; a supposition that is confirmed by direct calculation of the slip coefficients for a selection of gas mixtures. For example, with the rigid-sphere model for a mixture of N2 and O2, the relative differences between the first-order Chapman-Enskog derived values with respect to the second-order Chapman-Enskog derived values are 13.5 % for cT* sl and 8.1 % for cD sl . For the same mixture and the rigid-sphere model the relative difference observed for cm is only 2.5 %. The same situation occurs in a simple gas as well where the relative differences observed for cT* sl and cm are 13 % and 2.5 %, respectively. Of course, for a simple gas, one must remember that there is no diffusion-slip effect to observe. The relative differences of the slip coefficients that one observes when using the generally more realistic Lennard-Jones (6-12) potential model are somewhat less than those observed for the rigid-sphere model but the same

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

311

Table 12-3. Relevant parameters for two of the most commonly used intermolecular potential models; the rigid-sphere model [1] and the Lennard-Jones (6-12) potential model [2]. Gas He H2 N2 Ar O2 CO2

Rigid-sphere model V (Å) 2.193 2.745 3.784 3.659 3.636 4.643

Lennard-Jones (6-12) potential model V (Å) H k (K) 2.576 10.22 2.94 35.65 3.72 85.65 3.418 124.0 3.487 100.5 3.95 201.5

trend with respect to the sensitivity is observed. In a mixture of N2 and H2, * the relative differences determined for cTsl and cDsl (with respect to the second-order approximation values) are 6.7 % and 3.2 %, respectively while the relative difference determined for cm is only 1.5 %; again indicating that the velocity-slip coefficient is not as sensitive to the order of the approximation used. Further, if one compares the relative differences between the values of these coefficients using the different potential models with the same order * approximation for the same mixture of N2 and CO2 one finds that cTsl and cDsl vary from the Lennard-Jones (6-12) model to the rigid-sphere model by 9.1 % and 9.4 %, respectively; clearly showing that the more realistic model likely produces substantially more realistic results. From this discussion it is clear that whenever possible it is desirable to use the higher-order approximations and most realistic potential models as significant improvements in the theory of the slip phenomena may be realized under these circumstances. One slip phenomenon of interest is the thermal transpiration effect which is observed in capillary tubes. Table 12-4 compares the experimental and theoretical values of the slip factor, csl cTsl cDsl kT D12 Q , which governs this effect [48]. The experimental values given in Table 12-4 are slightly different from those reported in [48] owing to the use of the secondorder Chapman-Enskog approximations in determining the viscosity and velocity-slip coefficients which are used to calculate the slip factor for experimentally obtained pressure difference data. The statement of the problem for the thermal transpiration effect and the method solution are given in Problem 12.5. The specific values of the tangential momentum accommodation coefficients, D1W and D 2W , shown in Table 12-4 were chosen so as to produce the best possible agreement between the measured and theoretical slip factors for the selected binary gas mixtures. Table 12-4 shows that excellent agreement between theory and experiment is obtained if these values are assumed. This further demonstrates the substantial improvement in the accuracy of theoretically calculated slip phenomena

312

Analytical Methods for Problems of Molecular Transport

Table 12-4. The tangential momentum accommodation coefficients, D iW , in a binary gas mixture chosen for calculation of the slip coefficients and slip factors. These specific D iW were chosen so as to produce the best agreement between the measured and theoretical slip factors for the selected binary gas mixtures. Gas mixture

D1W

D 2W

>csl @exp

>csl @th

Error (%) †

He:Ar (4:1) 0.8 0.8 1.1863 1.1708 He:Ar (2:1) 0.8 0.8 1.1874 1.1559 He:Ar (1:1) 0.8 0.8 1.1043 1.1146 He:Ar (1:2) 0.8 0.8 1.0903 1.0714 He:CO2 (2:1) 0.8 0.8 1.3846 1.3141 0.8 0.8 1.2997 1.2720 He:CO2 (1:1) Ar:CO2 (2:1) 0.8 0.8 0.9593 1.0098 Ar:CO2 (1:1) 0.8 0.8 0.9573 1.0143 † Relative error has been calculated with respect to the experimental values of i.e. csl csl csl u 100% .

> @exp > @th > @exp

1.3 2.7 0.9 1.7 5.1 2.1 5.3 6.0 the slip,

values that can be obtained when using higher-order Chapman-Enskog approximations; in this case the second-order approximation. The above discussion has shown that there are strong dependencies of * cTsl and cDsl on the potential model used and on the order of the ChapmanEnskog approximation used. However, cm does not show the same degree of sensitivity as the other slip coefficients. There exists a direct way to assess the accuracy of the analysis reported here by comparing the values of the slip coefficients that it predicts with the very accurate numerical values reported by Takata and Aoki in [62]. The relative deviation of the values calculated from Eqs. (12-75) and (12-76) from the values reported in [62] is less than 2 % over a wide range of the molecular mass ratio. From this, one may conclude that the overall degree of accuracy of the slip coefficients reported here is, almost certainly, less than 2 % as well.

PROBLEMS 12.1. Using the first-order Chapman-Enskog approximation, determine the diffusion-slip coefficient for a binary gas mixture by means of the Maxwell method. Use a planar geometry in which the pressure and temperature are assumed to be constant. Solution: The distribution function for the i -th constituent may be expressed in the form:

fi

r

fi

0

12 ª º · D12 § mi · § i 1 m j d12 ¸ M i1 2 a r x ciy » , «1 ciy ¨ 1 2 ¨ ¸ xi © 2kT ¹ © m «¬ »¼ ¹

313

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

where d12 wx1 wy f , and x , and y are the coordinates normal and tangential to the planar surface, respectively. The Maxwell method is constructed on the assumptions that a r f a and a 0 a f a . The constant, a , is found from the moment solution given by Eq. (12-65):

c c

) i ci ,0 0 .

ix iy ,

The boundary conditions (one for each constituent) are given by:

)

i

ci ,0

12 ª i 1 m j D12 § mi · º 2D iW d12 « 1 » ciy 1 D iW M i1 2 aciy , ¨ ¸ m xi © 2kT ¹ »¼ «¬

where D iW is the tangential momentum accommodation coefficient for the i th constituent. The constant, a , can then be found by simple integration. Thus, Eq. (12-73) yields: cDsl

D

1W

m2 D 2W m1 K 0 ,

where:

K0

m1m2

m D1W x1 m1 D 2W x2 m2

.

12.2. For a non-uniform binary gas mixture, determine the mean velocities of the constituent gases and the mean mass velocity of the mixture. Consider the specific case in which there is no net number flow (transport) of molecules and in which the pressure and temperature of the gas mixture are assumed to be uniform. Solution: The distribution function of the i -th constituent is given by:

fi

m · 0 § f i ¨1 i u v i Di d12 ¸ , © kT ¹

where u is the mean mass velocity of the mixture and in which Di and d12 are expressed as: 12

Di

ª º D12 § mi · i 1 m j 1 zi S3 2 Ci2 » and d12 ¨ ¸ Ci « 1 2 xi © 2kT0 ¹ m ¬ ¼

n 1n1 .

314

Analytical Methods for Problems of Molecular Transport

The mean mass velocity of the i -th constituent can be calculated by integrating in the following manner:

viE

m · 0 § ni1 ³ viE fi dv i ni1 ³ f i ¨1 i uD viD DiD d12 D ¸ viE dv i © kT ¹ i 1 m j D12 uE 1 d12 E . m xi

This yields:

v1

u

m2 D12 d12 , m x1

(P-1)

v2

u

m1 D12 d12 . m x2

(P-2)

and:

For the specific case under consideration in which there is no net number flow of molecules, the velocities, v1 and v 2 , may be found from the equations:

v1 v 2

D12 n n1n2 n1 ,

n1 v1 n2 v 2

0,

(P-3)

(P-4)

which yield:

v1

D12 n11n1 and v 2

D12 n21n1 .

From these, the mean mass velocity is then determined to be:

u

U1 v1 U 2 v 2 U1 U 2

m2 m1

U

D12n1 .

(P-5)

12.3. For a non-uniform binary gas mixture, determine the number flux vector (current) of the first constituent assuming that the first constituent is

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

315

diffusing through the second and that there is no net number flow (current) of the second constituent. Assume the pressure and temperature of the gas mixture to be uniform. Solution: From the statement of the problem, one has that v 2 0 . Since the gas mixture conditions in this problem are similar to those in Problem 12.2, one then has, from Eq. (P-2), that:

u

m1 D12 d12 . m x2

Using this in Eq. (P-1) yields: n1 v1

n n2 D12n1 .

12.4. For a non-uniform binary gas mixture, determine the mean velocities of the constituent gases. In this more general problem, assume that the partial densities of the constituents, and the pressure and temperature of the gas mixture are all non-uniform. Solution: The distribution function of the i -th constituent is given by:

fi

m wu · 0 § f i ¨1 i u v i Ai ln T Di d12 2Bi : ¸ , wr ¹ © kT

(P-6)

where d12 is specified by Eq. (12-4). The same integration as in Problem 12.2 yields:

v1

u

m2 D12 d12 kT ln T , m x1

(P-7)

v2

u

m1 D12 d12 kT ln T , m x2

(P-8)

and:

where kT is the thermal diffusion ratio and u is the mean mass velocity which is unknown until the exact conditions governing the diffusion are specified. Note: Terms containing Sonine polynomials which occur in the functions, A i and Di , should vanish after the integration procedure and integrations over the function Bi should be identically zero.

316

Analytical Methods for Problems of Molecular Transport

12.5. For a non-uniform binary gas mixture, determine the relationship between the mean velocity of the gas mixture and its mean mass velocity. Solution: Let vi be the mean velocity of the i -th constituent of the gas mixture in the laboratory frame of reference. This velocity may be expressed in the form:

vi

w Wi

u Ui ,

(P-9)

where w and u are the mean and mean mass velocities of the gas mixture, respectively, and Wi and Ui are the diffusion velocities of the i -th constituent relative to two different frames of reference moving with the velocities, w and u , respectively. If one takes into account Eqs. (P-7) and (P-8), the relative diffusion velocities, Wi and Ui , may be found from the two following algebraic systems of equations:

W1 W2

v1 v 2

n1W1 n2 W2

U1 U 2

D12 d12 kT ln T , x1 x2

0,

v1 v 2

U1U1 U 2 U 2

D12 d12 kT ln T , x1 x2

0.

Upon solving these systems for Wi and Ui , the results may be substituted into Eq. (P-9) to yield:

u

w

m1 m2 D12 d12 kT ln T . m

(P-10)

12.6. Calculate the diffusion coefficient for a H2-N2 binary gas mixture at Use the first-order Chapman-Enskog T 293 K and p 1 atm. approximation. Assume that the molecules act as rigid spheres. Solution: Using Eqs. (12-13), (12-17) and (12-19) one obtains: D12 1

3 16

kT

. 1,1 M 1M 2 nm0 :12

The : -integral for rigid-sphere molecules is given by [3]:

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture 12

§ kT · * ¸ © 2S m ¹

:12 l , r SV 122 ¨

1 2

l ª 1 1 º 1 « », r 1 1 2 l 1 » «¬ ¼

317 (P-11)

where m* m1m2 m1 m2 . Then, one can obtain the following analytical expression for the diffusion coefficient. 1

D12

12

3 8

kT § kT · ¨ ¸ pV 122 © 2S m* ¹

(P-12)

.

For

the specific mixture under consideration, one may use 3.2645 u 108 cm from [1] and, from [70], p 1.01325 u 106 dyne cmí2, mH 1.00794 gm molí1, mN 14.00674 gm molí1, and k 1.38066 u 1016 erg Kí1 and N A 6.02214 u 1023 molí1 for Boltzmann’s constant and Avogadro’s number, respectively. Then, Eq. (P-12) yields 1 D12 0.6379 cm2 secí1.

V 12

12.7. For the second-order Chapman-Enskog approximation, obtain an analytical expression for the diffusion coefficient that relates it to the expression for the diffusion coefficient obtained in Problem 12.6 using the first-order Chapman-Enskog approximation. Solution: The transport coefficient, d 0 , is calculated from the following algebraic system of equations:

d 1a11 d 0 a01 d1a11

0, 12

2kT · n ¨ ¸ , © m0 ¹

d 1a10 d 0 a00 d1a10

3 2

d 1a11 d 0 a01 d1a11

0,

1 §

where the coefficients, a00 , a11 , a11 , and a11 a11 are given in Section 12.2 while the coefficients, a01 a10 and a01 a10 , are given by Eqs. (1244) and (12-45), respectively. Then, taking into account Eqs. (12-13) and (12-17), one can represent the diffusion coefficient obtained using the second-order approximation in the form: D12 2

D12 1

1 , 1 '

318

Analytical Methods for Problems of Molecular Transport

where the following notation has been introduced:

a0 1 '

a a10 a01 11 a10 a11 a a a00 11 11 a11 a11

a10 a11

a11 a10

2 a01 a11 a021a11 2a01a10 a11

a00 a11a11 a121

.

12.8. For the first-order Chapman-Enskog approximation, derive an analytical expression for the viscosity coefficient for a binary gas mixture. Solution: The viscosity coefficient is given by P p (x1b1 x2b1 ) . The transport coefficients, b1 and b1 , are specified from the following algebraic system of equations found in Section 12.2 (Eqs. (12-14) and (12-15)):

b11b1 b11b1

E 1

5 2

n2 , and b11b1 b11b1 n2

E1

5 2

n1 , n2

The solution of these equations yields:

> P @1

5 2

kT

x12b11 2 x1 x2b11 x22b11 . b11b11 b121

12.9. For the first-order Chapman-Enskog approximation, derive an analytical expression for the thermal conductivity coefficient for a binary gas mixture. Solution: The thermal conductivity coefficient is given by Eq. (12-16):

N

12 ª § 2kT ·1 2 º § 2kT · 54 nk « x1 ¨ a x ¸ 1 2¨ ¸ a1 » , «¬ © m1 ¹ »¼ © m2 ¹

The transport coefficients, a1 and a1 , are specified from the algebraic system of equations found in Section 12.2 (Eqs. (12-11) and (12-12)): 12

a11a1 a11a1 D 1

154

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

, and:

319

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture 12

a11a1 a11a1 D1

154

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

,

The solution of these equations yields:

>N @1

2

75 8

k T

x12 m11a11 2 x1 x2 m1m2

1 2

a11 x22 m21a11

a11a11 a121

.

12.10. For the second-order Chapman-Enskog approximation, derive an analytical expression for the viscosity coefficient for a binary gas mixture. Solution: The viscosity coefficient is given by P p (x1b1 x2b1 ) . The transport coefficients, b1 and b1 , are specified from the following algebraic system of equations:

b2 2 b2 b1 2b1 b1 2b1 b2 2b2

0,

5 2

b21b2 b11b1 b11b1 b21b2

n2 , n2

n1 , n2

b21b2 b11b1 b11b1 b21b2

5 2

b22 b2 b12b1 b12b1 b22b2

0.

The solution of this system of equations is given by:

b1

' b1

> 'b @2

and b1

' b1

> 'b @2

,

where the following notations have been introduced:

> 'b @2

b2 2 b21 b21 b22

b1 2 b11 b11 b12

b1 2 b11 b11 b12

b2 2 b21 , b21 b22

320

Analytical Methods for Problems of Molecular Transport 5 2

' b1

n1 n 'cb1 52 22 'ccb1 , 2 n n n1 n 'cb1 52 22 'ccb1 , 2 n n

' b1

52

'cb1

b2 2 b21 b22

b1 2 b11 b12

b2 2 b21 , 'ccb1 b22

'cb1

b2 2 b21 b22

b1 2 b11 b12

b2 2 b21 , and 'ccb1 b22

b2 2 b21 b22

b1 2 b11 b12 b2 2 b21 b22

b2 2 b21 , b22 b1 2 b11 b12

b2 2 b21 . b22

Using these notations, the viscosity coefficient is found to be:

> P @2

5 2

kT

x12 'cb1 x1 x2 'ccb1 'cb1 x22 'ccb1

> 'b @2

.

12.11. For the second-order Chapman-Enskog approximation, derive an analytical expression for the thermal conductivity coefficient for a binary gas mixture. Solution: The thermal conductivity coefficient is given by Eq. (12-16):

N

12 ª § 2kT ·1 2 º § 2kT · « nk x1 ¨ ¸ a1 x2 ¨ ¸ a1 » , «¬ © m1 ¹ »¼ © m2 ¹ 5 4

(12-16)

The transport coefficients, a1 and a1 , are specified from the following algebraic system of equations:

a2 2 a2 a1 2 a1 a1 2 a1 a2 2 a2

0,

a21a2 a11a1 a11a1 a21a2

154

12

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

,

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture 12

a21a2 a11a1 a11a1 a21a2

154

a22 a2 a12 a1 a12 a1 a22 a2

0.

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

,

The solution of this system of equations is given by:

a1

' a1

> ' a @2

' a1

and a1

,

> ' a @2

where the following notations have been introduced:

a2 2 a21 a21 a22

> ' a @2

a1 2 a11 a11 a12

a1 2 a11 a11 a12

a2 2 a21 , a21 a22

12

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

12

'ca1 154

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

15 4

n2 § 2kT · ¨ ¸ n 2 © m2 ¹

' a1

154

' a1

15 4

'ca1

a2 2 a21 a22

a1 2 a11 a12

a2 2 a21 , 'cca1 a22

'ca1

a2 2 a21 a22

a1 2 a11 a12

a2 2 a21 , and 'cca1 a22

12

n1 § 2kT · ¨ ¸ n 2 © m1 ¹

'cca1 ,

12

'ca1

a2 2 a21 a22

'cca1 ,

a1 2 a11 a12

a2 2 a21 a22

a2 2 a21 , a22

a1 2 a11 a12

a2 2 a21 . a22

321

322

Analytical Methods for Problems of Molecular Transport

Table 12-5. A comparison of diffusion-slip coefficient values. Values determined numerically [62] are compared with corresponding values determined using the Maxwell method for selected values of the mass ratio, Z , for the specific case when D1W D 2W 1 and x1 x2 12 .

Z m2 m1

2

cDsl , Eq. (P-13) cDsl , numerical G (%)

0.3235

0.5333

0.5694

0.5974

0.1637

0.2514

0.2634

0.2619

97.6

112

116

128

4

5

10

G (%) = cDsl ,num cDsl , Max cDsl ,num u 100 . Using these notations, the thermal conductivity coefficient is found to be:

>N @2

2

75 8

k T

x12 m11'ca1 x1 x2 m1m2

1 2

'cc 'c x m

> ' a @2

a1

a1

2 2

1 cc 2 ' a1

.

12.12. Evaluate the accuracy of the Maxwell method for determining the diffusion-slip coefficient by comparing it with the numerical results [62] for different values of the mass ratio, Z m2 m1 , Consider the specific case when D1W D 2W 1 and x1 x2 12 . Solution: For the specific case under consideration, the diffusion-slip coefficient given in Problem 12.1 may be expressed in the form:

cDsl

4Z Z 1

Z

2

1 Z 1

.

(P-13)

The numerical values and values computed from Eq. (P-13) can then be assembled in tabular form for ease of comparison. This has been done and is given in Table 12-5. As one can see from Table 12-5, with relative errors on the order of 100%, the Maxwell method does not appear, at least for the specific case under consideration in this problem, to be an effective technique for determination of the diffusion-slip coefficient. However, the current case is quite limited and one should not make the mistake of inferring that the accuracy of the Maxwell method overall is represented by the current problem. Rather, one should understand from the current problem only that the Maxwell method can lead to unacceptably inaccurate results in some circumstances.

12.13. Using the first-order Chapman-Enskog approximation and the Maxwell method, determine the slip-flow coefficient for a binary gas mixture. Consider the case in which the temperature and partial pressures of the gas constituents are uniform.

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

323

Solution: The distribution function of the i -th constituent in the gas mixture is given by:

fi r

m º 0 ª iy 2cix ciy br1h ) ir ci , x » , f i «1 i xhv ¬ kT ¼

where one should use b1 if i 1 and b1 if i model yields:

2 . The Maxwellian boundary

) i ci ,0 2 2 D iW cix ciy br1h 1 D iW ) i cix , ciy , ciz ,0 . The correction to the distribution function, ) i ci , x , can be chosen in the form ) ir ci , x M i1 2 a r x ciy . The Maxwell method is constructed on the assumption that a r f a 0 a . Then this constant can be calculated from the exact analytical solution to the Boltzmann equations given by:

c c

) ir ci ,0

ix iy ,

0.

Taking into account Eqs. (12-72) and (12-73), one can then obtain:

cm

pM 1 2

P

5 8

S

2 D1W x1b1 2 D 2W x2b1 . D1W x1M11 2 D 2W x2 M 21 2

12.14. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O where O is the mean free path of the gas molecules. The two vessels are maintained at the same constant temperature, T c T cc T , and are filled at the beginning t 0 with a binary gas mixture having slightly different mixture fractions in each vessel such that x1c 0 x1 0 and x2c 0 x2 0 in the first vessel while in the second vessel x1cc 0 x1 0 G x1 0 and x2cc 0 x2 0 G x1 0 . When the capillary is opened a pressure difference, G p pcc pc , will develop as a function of time. Determine the time dependence of the relative density difference G x1 t x1cc t x1c t and the time dependence of the pressure difference, G p t . Also, determine the maximum value of the pressure difference that develops. Solution: The basic equations of balance for the number of molecules of each constituent gas are given by:

324

Analytical Methods for Problems of Molecular Transport

V c

dnic dt

J iz and V cc

dnicc dt

J iz ,

where J iz S R 2 ni viz is the net molecular current (referred to as the molecular number flux in some texts) of the i -th constituent through the capillary. If one takes into account Eq. (P-9) from Problem 12.5, the basic equations can be written in the form:

V c

V cc

dnic dt

S R 2 ni wz Wiz ,

dnicc S R 2 ni wz Wiz , dt

(P-14)

(P-15)

where wz is the mean velocity of the gas mixture and Wiz is the diffusion velocity of the i -th constituent relative to a frame of reference moving with velocity, w . For the number density difference, G ni t nicc t nic t , one can obtain:

S R2

d G ni dt

V

ni wz Wiz ,

(P-16)

where V V cV cc V c V cc . The number density difference may then be expressed in the form:

G ni

· § pxi · § xi ¨ G p G xi ¸ n . ¸ © kT ¹ © p ¹

G¨

(P-17)

Then, Eq. (P-16) yields: · d § x1 ¨ G p G x1 ¸ dt © p ¹

S R2

· d § x2 ¨ G p G x1 ¸ dt © p ¹

S R2

V

V

x1 wz W1z ,

(P-18)

x2 wz W2 z .

(P-19)

The average values introduced here are slightly different from those that would exist in the equilibrium state because the pressure difference that

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

325

develops is much less than the initial pressure. The mean molecular velocity, w , is related to the mean mass velocity, u , by (see Problem 12.5): § m m1 · w u¨ 2 ¸ D12d12 , © m ¹

(P-20)

and the diffusion velocities are then defined by:

n1W1

n2 W2

nD12d12 ,

(P-21)

where the vector, d12 , is specified in Eq. (12-4). In the particular case under consideration in this problem, the z -component of this vector can be expressed in the form:

d12 z

§ § m m1 · G p · L1 ¨ G x1 x1 x2 ¨ 2 ¸. ¸ © m ¹ p ¹ ©

(P-22)

Note that in the slip-flow regime, the mean mass velocity in the capillary may be specified from the boundary value problem of:

u 0 and P 2u p with:

u z R cm OP

wu z cDsl D12 d12 z and p r , L p r ,0 G p . wr

The solution of this boundary value problem is found to be:

uz r

1 wp ª R 2 1 4cm Kn r 2 º cDsl D12 d12 . z ¬ ¼ wz 4P

from which one can obtain: R

uz

2 u z r rdr R 2 ³0

R 2 1 4cm Kn 8P L

G p cDsl D12 d12 z .

(P-23)

Now, taking into account Eqs. (P-20)-(P-23), Eqs. (P-18) and (P-19) can be written in the form:

326

Analytical Methods for Problems of Molecular Transport

d G x1 dt

Z1xG x1 Z1 pG p ,

(P-24)

d Gp dt

Z xG x1 Z pG p ,

(P-25)

where the following notations have been employed:

§ m m1 · D12 , a ax1 x2 ¨ 2 ¸ © m ¹ p

S R2

Z1x

aD12 , Z1 p

Zx

apc12 D12 , c12

Zp

ª pR 2 1 4cm Kn º § m m1 · c12 x1 x2 ¨ 2 a« D ». 12 ¸ 8P © m ¹ «¬ »¼

VL

§ m m1 · cDsl ¨ 2 ¸ , and © m ¹

The characteristic equation for this homogeneous linear system of differential equations is the quadratic equation:

Z1x s Z1 p Z x Z p s

0,

which has the following roots:

s1,2

1 2

ª « Z Z r Z p p 1x « «¬

2 º · 4Z xZ1 p » . ¸ 2 » ¸ Z p ¹ »¼

§ Z1x ¨1 ¨ Zp ©

For the different terms in this expression one can make the following determinations regarding the orders of various quantities, specifically: 12

§ 2kT ·

P UV O , D12 V O , V ¨ ¸ , and: © m ¹

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

327

Z p R 2 nkT R 2 2 Kn 2 1 . 2 2 Z1x UV O O The last of these implies that:

Z Z1x Z 1 , x 1 , and 1 p 1 . Zp Zp Zp Now, taking into account these determinations of order, it follows that s1 Z1x and s2 Z p . Thus, since both of the roots of the characteristic equation have been shown to be negative, it further follows that each term in the complete solution to the homogeneous system of equations will decay exponentially to zero as t o f . The solution of the system is found to be:

G x1 t G x1 0 exp Z1x t and:

G p t K exp Z1x t exp Z p t where:

K

8P c12 D12 § Z1x G x1 0 2 ¨1 R 1 4cm Kn ¨© Z p

1

· ¸ . ¸ ¹

The maximum pressure difference is found to be:

'p max where c12

G x1 0

8P c12 D12 § Z1x § Z p · · ln ¨ ¨1 ¸¸ . R 1 4cm Kn ©¨ Z p © Z1x ¹ ¹¸ 2

cDsl (m2 m1 )/m .

12.15. Two vessels having volumes, V c and V cc , are joined by a long capillary where R and L are the capillary radius and length, respectively, L R , and R O (which corresponds to the slip-flow regime) where O is the mean free path of the gas molecules. The two vessels are maintained at different temperatures, T c and T cc with T c ! T cc . Assume that the two vessels initially t 0 contain identical gas mixtures, each with the same pressure and mole ratios, i.e. and x1c (0) x1cc(0) q (1 q)

328

Analytical Methods for Problems of Molecular Transport

x2c (0) x2cc (0) q (1 q) , where q is the initial pressure ratio of the gas constituents. When the capillary is opened a pressure difference, G p t pcc pc , will develop as a function of time. Determine the steadystate pressure difference that will develop in this system, 'pst G p f . Solution: For this problem one can use the same basic system of differential equations that was employed in Problem 12.14 in which the mean velocity, wz , and the diffusion velocities, Wiz , must be modified by the inclusion of additional terms to account for the non-uniformity of the temperature in the system. This basic system of differential equations is given by: · d § x1 ¨ G p G x1 ¸ dt © p ¹

S R2

· d § x2 ¨ G p G x1 ¸ dt © p ¹

S R2

V

V

x1 wz W1z ,

x2 wz W2 z .

where G p t pcc pc and G x1 x1cc x1c . The diffusion velocities, Wiz , are given by:

n1W1z

GT º ª nD12 « d12 z kT , where: TL »¼ ¬

n2W2 z

d12 z

§ § m m1 · G p · L1 ¨ G x1 x1 x2 ¨ 2 ¸, ¸ © m ¹ p ¹ ©

and in which kT is the thermal diffusion ratio. The mean velocity, wz , can be expressed in the form:

wz

R 2 1 4cm Kn 8P L

G p cDsl D12 d12 z

GT º § m m1 · ª Q D12 « d12 z kT , 2 TL ¨© m ¸¹ TL »¼ ¬

* * cTsl

GT

(P-26)

* where cm , cTsl , and cDsl are the velocity-slip, thermal-creep, and diffusionslip coefficients, respectively. Then, the basic system of equations can be written in the form:

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

329

d G x1 dt

Z1xG x1 Z1 pG p Z1T G T ,

(P-27)

d Gp dt

Z xG x1 Z pG p ZT G T ,

(P-28)

where the following notations have been employed:

§ m m1 · D12 , a ax1 x2 ¨ 2 ¸ © m ¹ p

S R2

Z1x

aD12 , Z1 p

Z1T

Zp

ª pR 2 1 4cm Kn º § m m1 · c12 x1 x2 ¨ 2 a« D », 12 ¸ 8P © m ¹ «¬ »¼

ZT

ap T * T c12 Q , and c12 T

akT D12 , Zx T

apc12 D12 , c12

VL

§ m m1 · cDsl ¨ 2 ¸, © m ¹

§ m m1 · kT D12 * ¨ 2 cTsl ¸ * © m ¹ Q

The last terms of Eqs. (P-27) and (P-28) can be regarded as the driving forces which will determine the ultimate steady-state (or stationary state) of this process. The characteristic equation of the homogeneous system has two negative roots as was shown in Problem 12.14. This again indicates that the solutions to the homogeneous system decay exponentially to zero as t o f . However, the steady-state solution for the non-homogeneous system, where the contribution of the particular solutions dominates the behavior of the system, can be specified from the following equations:

Z1xG x1 f Z1 pG p f Z1T G T ,

Z xG x1 f Z pG p f ZT G T .

330

Analytical Methods for Problems of Molecular Transport

G p f , is then found to be:

The steady-state pressure difference, 'pst

'pst

Z § ZZ G T T ¨ 1 x 1T Z p © Z1xZT

· § Z xZ1 p ¸ ¨¨1 ¹ © Z1xZ p

1

· ¸ . ¸ ¹

An evaluation of the two terms found in the expression for Z p shows that the first term is the dominant term, i.e.: 1

R2 p R 2 nkT R 2 nkT R2 § m2 m1 · 2 1 , ¨ ¸ 2 2 8P c12 D12 x1 x2 © m ¹ P D12 mnV O O where the following estimations of order have been used: 12

§ 2kT ·

P mnV O , D12 V O , and V ¨ ¸ . © m ¹ This means that:

Zp

ap

R 2 1 4cm Kn 8P

.

Further, the order of the ratio, Z xZ1 p Z1xZ p , is determined to be:

Z xZ1 p D12 P V 2 O 2 nm O 2 2 1 . Z1xZ p pR 2 nkTR 2 R Now, neglecting the small terms, one can obtain:

'pst

8PQ *csl , T R 2 1 4cm Kn

GT

where the following notation has been introduced:

csl

c12 c12 T

kT D12

Q

*

* cTsl cDsl

kT D12

Q*

.

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

331

REFERENCES 1. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 2. Hirschfelder, J.O., Curtiss, C.F., and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954). 3. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 4. Maxwell, J.C., in The Scientific Papers of J.C. Maxwell, Vol. 1, ed. by Niven, W.D. (Dover, Mineola, NY, 2003). 5. Loyalka, S.K., “Approximate Method in the Kinetic Theory,” Phys. Fluids 14(11), 22912294 (1971). 6. Kennard, E.H., Kinetic Theory of Gases (McGraw-Hill, New York, 1938). 7. Williams, M.M.R. and Loyalka, S.K., Aerosol Science, Theory and Practice (Pergamon Press, New York, 1991). 8. Kogan, M.N., Rarefied Gas Dynamics (Plenum Press, NY, 1969). 9. Cercignani, C., Theory and Application of the Boltzmann Equation (Scottish Academic Press, Edinburg, UK, 1975). 10. Hidy, G.M. and Brock, J.R., The Dynamics of Aerocolloidal Systems, Vol. 1 (Pergamon Press, New York, 1970). 11. Derjaguin, B.V. and Yalamov, Yu.I., “The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and their Experimental Testing,” in Topics in Current Aerosol Research, vol. 3, part 2, edited by Hidy, G.M. (Pergamon Press, Oxford, 1972). 12. Loeb, L.B., The Kinetic Theory of Gases, 3rd edition (Dover, New York, 1961). 13. Present, R.D., Kinetic Theory of Gases (McGraw-Hill, New York, 1958). 14. Rosner, D.E., “Side-wall Gas ‘Creep’ and ‘Thermal Stress Convection’ in Microgravity Experiments on Film Growth by Vapor Transport,” Phys. Fluids A 1(11), 1761-1763 (1989). 15. Gupta, R.N., Scott, C.D., and Moss, J.N., Slip-Boundary Equations for Multicomponent Non-equilibrium Air Flow (NASA TP2455, Nov. 1985). 16. Hess, D.W. and Jensen, K.F., Microelectronic-Processing Chemical Engineering Aspects (American Chemical Society, Washington DC, 1989). 17. Wolf, S. and Tauber, R.N., Silicon Processing for the VLSI ERA, Vol. 1. Process Technology (Lattice Press, Sunset Beach, CA, 1986). 18. Ikegawa, M. and Kobayashi, J., “Deposition Profile Simulation Using the Direct Simulation Monte Carlo Method,” J. Electrochemical Soc. 136(10), 2982-2986 (1989). 19. Coronell, D.G. and Jensen, K.F., “Analysis of Transition Regime Flows in Low-pressure Chemical Vapor Deposition Reactors Using the Direct Monte Carlo Method,” J. Electrochemical Soc. 139(8), 2264-2273 (1992). 20. Coronell, D.G. and Jensen, K.F., “Simulation of Rarefied Gas Transport and Profile Evolution in Nonplanar Substrate Chemical Vapor Deposition,” J. Electrochemical Soc. 141(9), 2545-2551 (1994). 21. Kramers, H.A. and Kistemaker, J., “On the Slip of a Diffusing Gas Mixture Along a Wall,” Physica (Amsterdam) 10(8), 699-713 (1943). 22. Kucherov, R.Ya. and Richenglas, L.E., “Slip and Temperature Jump at the Boundary of a Gaseous Mixture,” Zh. Eks. Teor. Fis. (Russian) 36(6), 1758-1761 (1959). 23. Kucherov, R.Ya., “Diffusion Slip and Convective Diffusion of a Gas in Capillaries,” Zh. Tekh. Fiz. (Russian) 27(9), 2158-2161 (1957).

332

Analytical Methods for Problems of Molecular Transport

24. Schmitt, K.H., “Untersuchungen an Schwebstoffteilchen in Diffundierendem Wasserdampf,” Z. Naturforschg. 16a, 144-149 (1961). 25. Brock, J.R., “The State of a Binary Gas Mixture Near a Catalytic Surface,” J. Catalysis 2, 248-250 (1963). 26. Brock, J.R., “Forces on Aerosols in Gas Mixtures,” J. Colloid Sci. 18(6), 498-501 (1963). 27. Zhdanov, V.M., “The Theory of Slip on the Boundary of a Gaseous Mixture,” Zh. Tekh. Fiz. (Russian) 37(1), 192-197 (1967). 28. Lang, H. and Eger, K., “Gas Diffusion in Narrow Capillaries,” Z. Physik Chem. 68, 130148 (1969). 29. Lang, H. and Müller, W.J.C., “Slip Effects in Mixtures of Monoatomic Gases for General Surface Accommodation,” Z. Naturforschg. 30a, 885-867 (1975). 30. Schmitt, K.H. and Waldmann, L., “Untersuchungen an Schwebstoffteilchen in Diffundierenden Gasen,” Z. Naturforschg. 15a(10), 843-851 (1960). 31. Waldmann, L. and Schmitt, K.H., “Über das bei der Gasdiffusion Auftretende Druckgefälle,” Z. Naturforschg. 16a, 1343-1354 (1961). 32. Waldmann, L., “On the Motion of Spherical Particles in Nonhomogeneous Gases,” in Rarefied Gas Dynamics, ed. Talbot, L. (Academic Press, New York, 1961), pp. 323-344. 33. Waldmann, L. and Schmitt, K.H., “Thermophoresis and Diffusiophoresis of Aerosols,” Chapter VI in Aerosol Science, ed. Davies, C.N. (Academic Press, New York, 1966), pp.137-162. 34. Breton, J.P., “Interdiffusion of Gases Through Porous Media – Effect of Molecular Interactions,” Phys. Fluids 12(10), 2019-2026 (1969). 35. Breton, J.P., “The Diffusion Equation in Discontinuous Systems,” Physica 50, 365-379 (1970). 36. Yalamov, Yu.I., Ivchenko, I.N., and Derjaguin, B.V., “Calculation of the Surface Diffusion Slip Velocity of a Gas Mixture,” in Rarefied Gas Dynamics, Vol. 1, eds. Trilling, L. and Wachman, H.Y. (Academic Press, New York, 1969), pp. 295-300. 37. Ivchenko, I.N. and Yalamov, Yu.I., “Diffusion Slip of a Binary Gas Mixture,” Mekh. Zhid. G. (Russian) 4, 22-26 (1971). 38. Loyalka, S.K. and Ferziger, J.H., “Model Dependence of the Slip Coefficient,” Phys. Fluids 10(8), 1833-1839 (1967). 39. Loyalka, S.K. and Ferziger, J.H., “Model Dependence of the Temperature Slip Coefficient,” Phys. Fluids 11(8), 1668-1671 (1968). 40. Loyalka, S.K., “Momentum and Temperature Slip Coefficient with Arbitrary Accommodation at the Surface,” J. Chem. Phys. 48(12), 5432-5436 (1968). 41. Loyalka, S.K., “Velocity Slip Coefficient and the Diffusion Slip Velocity for a Multicomponent Gas Mixture,” Phys. Fluids 14(12), 2529-2604 (1971). 42. Loyalka, S.K., “Kinetic Theory of Thermal Transpiration and Mechanocaloric Effect,” J. Chem. Phys. 55(9), 4497-4503 (1971). 43. Lang, H. and Loyalka, S.K., “Diffusion Slip Velocity. Theory and Experiment,” Z. Naturforschg. 27a, 1307-1319 (1972). 44. Loyalka, S.K., “Temperature Jump in a Gas Mixture,” Phys. Fluids 17(5), 897-899 (1974). 45. Loyalka, S.K., “Kinetic Theory of Thermal Transpiration and Mechanocaloric Effect. II,” J. Chem. Phys. 63(9), 4054-4060 (1975). 46. Loyalka, S.K. and Storvick, T.S., “Kinetic Theory of Thermal Transpiration and Mechanocaloric Effect. III,” J. Chem. Phys. 71(1), 339-350 (1979). 47. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “Slip Coefficients for Binary Gas Mixtures,” JVST A 15(4), 2375-2381 (1997).

Chapter 12. Boundary Slip Phenomena in a Binary Gas Mixture

333

48. Huang, C.M., Tompson, R.V., Ghosh, T.K., Ivchenko, I.N., and Loyalka, S.K., “Measurements of Thermal Creep in Binary Gas Mixtures,” Phys. Fluids 11(6), 16621671 (1999). 49. Ivchenko, I.N., Loyalka, S.K. and Tompson, R.V., “Boundary Slip Phenomena in a Binary Gas Mixture,” ZAMP 53(1), 58-72 (2002). 50. Yalamov, Yu.I., Shkanov, Yu. and Savkov, S.A., “Boundary Conditions of Sliding of a Binary Gas Mixture and Their Use in the Dynamics of Aerosols. 1. Flow of a Gas Mixture Along a Solid Flat Wall,” Inzh. Fizich. Zh. (Russian) 66(4), 421-426 (1994). 51. Savkov, S.A., Slip Boundary Conditions for Non-Uniform Binary Gas Mixtures and their Application to Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 52. Volkov, I.V. and Galkin, V.S., “Analysis of the Slip Coefficients and Temperature Jump in a Binary Gas Mixture,” Mekh. Zhid. G. (Russian) 6, 152-159 (1990) 53. Siewert, C.E. and Sharipov, F., “Model Equations in Rarefied Gas Dynamics: ViscousSlip and Thermal-Slip Coefficients,” Phys. Fluids 14(12), 4123-4129 (2002). 54. Siewert, C.E., “Viscous-Slip, Thermal-Slip, and Temperature-Jump Coefficients as Defined by the Linearized Boltzmann Equation and the Cercignani-Lampis Boundary Condition,” Phys. Fluids 15(6), 1696-1701 (2003). 55. Siewert, C.E., “The Linearized Boltzmann Equation: A Concise and Accurate Solution of the Temperature-Jump Problem,” J. Quantitative Spectroscopy and Radiative Transfer 77(4), 417-432 (2003). 56. Takata, A. and Aoki, K., “Two Surface Problems of a Multicomponent Mixture of Vapors and Non-condensable Gases in the Continuum Limit in the Light of Kinetic Theory,” Phys. Fluids 11(9), 2743-2756 (1999). 57. Sone, Y., Takata, A., and Golse, F, “Notes on the Boundary Conditions for FluidDynamics Equations on the Interface of a Gas and Its Condensed Phase,” Phys. Fluids 13(1), 324-334 (2001). 58. Takata, A. and Aoki, K., “The Ghost Effect in the Continuum Limit for a Vapor-Gas Mixture Around Condensed Phases: Asymptotic Analysis of the Boltzmann Equation,” Transport Theory and Stat. Phys. 30(2&3), 205-237 (2001). 59. Aoki, K., Bardos, C., and Takata, A., “Knudsen Layer for Gas Mixtures,” J. Statistical Phys. 112(3&4), 625-655 (2003). 60. Takata, A. and Aoki, K., “The Ghost Effect in the Continuum Limit for a Vapor-Gas Mixture Around Condensed Phases: Asymptotic Analysis of the Boltzmann Equation,” Transport Theory and Stat. Phys. 31(3), 289-290 (2002). 61. Takata, A., Yasuda, S., Kosuge, S., and Aoki, K., “Numerical Analysis of Thermal-Slip and Diffusion-Slip Flows of a Binary Mixture of Hard-Sphere Molecular Gases,” Phys. Fluids 15(12), 3745-3766 (2003). 62. Takata, A. and Aoki, K., “Numerical Analysis of the Shear Flow of a Binary Mixture of Hard-Sphere Gases Over a Plane Wall,” Phys. Fluids 16(6), 1989-2003 (2004). 63. Takata, A., “Kinetic Theory Analysis of the Two-Surface Problem of a Vapor-Vapor Mixture in the Continuum Limit,” Phys. Fluids 16(7), 2182-2198 (2004). 64. Maitland, G.C., Rigby, M., Smith, E.B., and Wakeham, W.A., Intermolecular Forces: Their Origin and Determination (Oxford University Press, New York, 1981, reprinted and corrected edition, 1987). 65. Mason, E.A., “Transport Properties of Gases Obeying a Modified Buckingham (exp-six) Potential,” J. Chem. Phys. 22(2), 169-186 (1954). 66. Mason, E.A., “Higher Approximations for the Transport Properties of Binary Gases Mixtures. I. General Formulas,” J. Chem. Phys. 27(1), 75-84 (1957).

334

Analytical Methods for Problems of Molecular Transport

67. Mason, E.A., “Higher Approximations for the Transport Properties of Binary Gases Mixtures. II. Applications,” J. Chem. Phys. 27(3), 782-790 (1957). 68. Mason, E.A. and Saxena, S.C., “Approximate Formula for the Thermal Conductivity of Gas Mixtures,” Phys. Fluids 1(5), 361-369 (1958). 69. Mason, E.A. and Monchick, L., “Heat Conductivity of Polyatomic and Polar Gases,” J. Chem. Phys. 36(6), 1622-1639 (1962). 70.CRC Handbook of Chemistry and Physics, 71 edition, ed. Lide, D.R. (CRC Press, Boca Raton, 1990).

Appendix A BRACKET INTEGRALS FOR THE PLANAR GEOMETRY

1.

BRACKET INTEGRALS INVOLVING TWO SONINE POLYNOMIALS. In this section, bracket integrals of the form: ars

ªcS r c 2 , cS s c 2 º , 32 ¬ 32 ¼

are discussed. Consider, for example, the bracket integral a11 that is contained in the first-order Chapman-Enskog solution for thermal conductivity. This integral can be written as:

a11

ªcS 1 c 2 , cS 1 c 2 º 3 ª cx S 1 c 2 , cx S 1 c 2 º 32 32 ¬ 32 ¼ ¬ 32 ¼

^³ D T , g d : ³ ³ gf f ª¬c S c S c cc S cc cc S cc º dv ` dv . ¼

3n 2 ³ cx S3 2 c 2 1

c1 x

1 32

2 1

0

0

1

1 x 32

2

1 1x 3 2

1 x 32

2

(A-1)

2

1

1

Using the momentum conservation equation, one obtains:

a11

3 ª¬ cx c 2 , cx c 2 º¼ .

(A-2)

336

Analytical Methods for Problems of Molecular Transport

The differential scattering cross section for rigid-sphere molecules can be expressed in the form:

D T , g d :

bdbd H

b

db dT d H dT

1 4

V 2 sin T dT d H ,

(A-3)

where V is the diameter of a molecule. and Let the variables of integration be changed from v and v1 to G g v1 v (the center of mass and relative molecular velocities, respectively) by the relations:

1 g , v G 2

1 g , G 2

v1

1 2

(A-4a)

m v 2 v12

(A-4b)

m G 2 14 g 2 ,

(A-4c)

and: dvdv1

dg , dG

(A-4d)

and g are dimensional variables. where G After changing the variables of integration, one can make the following substitutions: 12

G

§ m· ¨ kT ¸ G , © ¹

g

1§ 2¨

c

12

m· ¸ g , © kT ¹

1 2

(A-5a)

G g

,

(A-5b)

(A-5c)

337

Appendix A. Bracket Integrals for the Planar Geometry

c1

dG

1 2

G g

§ kT · ¨ ¸ © m¹

,

(A-5d)

dG ,

(A-5e)

32

and:

§ kT · dg 8 ¨ ¸ © m¹

32

(A-5f)

dg .

It is convenient to introduce the spherical coordinates g ,D , E for g . The relationships between these spherical coordinates and the coordinates in Cartesian velocity space, g x , g y , g z , are given by:

gx

g cos D ,

g sin D cos E ,

gy

g sin D sin E .

gz

Substituting the dimensionless variables into a11 , and neglecting all vanishing integrals, one can obtain:

a11

3 4

12S

§ kT · ¸ © m¹

V 2S 3 ¨

f

2S

³ sin T dT 0

u ³ exp G y2 dG y f

f

³

f

³ 0

f

dH

³ exp Gx dGx 2

f

f

exp Gz2 dGz ³ g 3 exp g 2 dg 0

2S

S

u³ sin D dD 0

(A-6)

³ dE 0

^

u Gx4 Gx2G y2 Gx2Gz2 Gx2 g 2 2Gx2 g x2

g

2 x

g cx2

`

2G y2 g x g y g x g y g cx g cy 2Gz2 g x g z g x g z g cx g cz , where g cx , g cy , and g cz are given by: g cx

g ª¬ cos D cos T sin D sin T cos H º¼ ,

g cy

g ª¬sin D cos E cos T cos D cos E sin T cos H sin E sin T sin H º¼ ,

(A-7a)

(A-7b)

338

Analytical Methods for Problems of Molecular Transport

g cz

g ª¬sin D sin E cos T cos D sin E sin T cos H . cos E sin T sin H º¼

(A-7c)

After integration over all values of G and neglecting vanishing integrals one can express the bracket integral, a11 , in the following form:

a11

3 4

2

V S

3 2

12f

§ kT · ¨ m¸ © ¹

S

u ³ sin D dD 0

³g

3

0

2S

2S

0

0

S

exp g 2 dg ³ sin T dT

³ d H ³ d E ^ 54 g

2

0

12 g 4

u ª¬ cos 2 D sin 2 T sin 2 D sin 2 T cos 2 H º¼

(A-8)

`

g 4 cos 2 D sin 2 T 8 3

12f

§ kT · ¸ © m¹

SV 2 ¨

7 2 ³ g exp g dg 0

12

§ kT · 8 SV 2 ¨ ¸ © m¹

.

Finally, in summary, the bracket integral, a11 , is given by: 12

ªcS 1 c 2 , cS 1 c 2 º 8 SV 2 § kT · , (A-9) ¨ m¸ 32 ¬ 32 ¼ © ¹ where one must remember that molecules have been assumed to be rigid spheres. a11

2.

BRACKET INTEGRALS CONTAINING SEVERAL COMPONENTS OF MOLECULAR VELOCITY.

In this section, the bracket integral associated with the first-order Chapman-Enskog solution for viscosity:

b11c

ª¬cx c y , cx c y º¼ ,

is considered. Using the dimensionless variables from Section A.1 and omitting vanishing integrals, one can obtain the following expression for this bracket integral:

339

Appendix A. Bracket Integrals for the Planar Geometry 12

b11c

1 4

ª¬cx c y , cx c y º¼ 2S

S

u³ sin T dT 0

2S

u ³ dE 0

f

³ 0

f

³ exp

§ kT · ¸ © m¹

V 2S 3 ¨ f 0

f

dG ³ exp

Gx2

0

G y2

x

f

S

d H ³ g 3 exp g 2 dg ³ sin D dD (A-10)

dG

y

f

^

u ³ exp Gz2 dGz g x g y g x g y g cx g cy f

` .

After integration over the components of G , and using the relations for g cx and g cy given by Eqs. (A-7a) and (A-7b), one can obtain:

b11c

1 4

12S

§ kT · ¸ © m¹

V 2S 3 2 ¨ f

2S

³ sin T dT

³ dH

0

0

S

u³ g 3 exp g 2 dg ³ sin D dD 0

2S

0

³ dE g

4

(A-11)

0

^

`

u sin 2 D cos 2 D cos 2 E sin T ª¬1 cos 2 H º¼ . The remaining simple integrations then yield: 12

b11c

¬ªcx c y , cx c y ¼º

4 5

§ kT ·

V2 S ¨ ¸ © m¹

.

(A-12)

Molecules have been assumed to be rigid spheres. Analytical expressions for other bracket integrals contained in higher-order Chapman-Enskog solutions and a more general calculation scheme for arbitrary intermolecular potentials may be found in [1-3].

3.

BRACKET INTEGRALS CONTAINING TWO DISCONTINUOUS FUNCTIONS.

In this section a method of calculating bracket integrals that contain the discontinuous sign functions is considered. As an example, the following bracket integral will be evaluated for the case of rigid-sphere molecules:

340

Analytical Methods for Problems of Molecular Transport

ª¬ c y sign cx , c y sign cx º¼

^

1 0 f 0 dv1 n 2 ³ c y sign cx ³ D T , g d : ³ gf

(A-13)

u ª c y sign cx c1 y sign c1 x ¬

`

ccy sign ccx c1c y sign c1cx º dv . ¼ This integral may be represented as the sum of four terms: 1 2 3 4 I I I I .

I

(A-14)

The first term does not contain the discontinuous functions and, therefore, may be calculated by the standard method described in Section A.1. Consider the second integral in detail. The integrand does not contain any collision parameters and, therefore, ³ D T , g d : can be calculated independently of the other integrations yielding:

³ D T , g d :

SV 2 .

Now, let the variables of integration be changed from v, v1 to , g v v . Then, after introducing the dimensionless variables G and G 1 g , this integral becomes:

I

2

2

V S

2

2S

f

u ³ dE 0

f

12f

§ kT · ¨ ¸ © m¹

S

3 2 ³ g exp g dg ³ sin D dD 0

f

0

³ exp Gx dGx ³ exp Gy dGy 2

f

2

(A-15)

f

^

u ³ exp Gz2 dGz G y2 g 2 sin 2 D cos 2 E f

u sign ª¬Gx g cos D º¼ ª¬Gx g cos D º¼

` .

After integration over all values of G y , Gz , and E this expression becomes:

341

Appendix A. Bracket Integrals for the Planar Geometry 12 f

§ kT · ¸ © m¹

2 I

V2¨ f

S

3 2 ³ g exp g dg ³ sin D dD 0

0

u ³ exp Gx2 dGx 1 g 2 sin 2 D

(A-16)

f

u sign ª¬Gx g cos D º¼ ª¬Gx g cos D º¼ . Note that the method of calculation used in these integrations has been previously described in [4,5]. Here, this method has been generalized by the introduction of appropriate recurrence relations. Using the new variables of integration, x cos D and z Gx g , one can reduce Eq. (A-16) to the form: 2 I

12 1

§ kT · ¸ © m¹

V2¨

f

³ dx

³

1

f

f

dz ³ g 4 exp g 2 ¬ª1 z 2 ¼º dg 0

u ª1 g 2 1 x 2 º sign > z x @> z x @ ¬ ¼ 12 1

§ kT · V ¨ ¸ © m¹ 2

(A-17)

f

³ dx ³ dz sign > z x @> z x@

1

f

u 83 S ª 1 z 2 «¬

5 2

52 1 x 2 1 z 2

7 2

º . »¼

It is convenient to use the following notations:

J m 1 2 1

1

f

2³ dx ³ 0

2

J m 1 2

1

sign > z x @> z x @

1 z 2

f

2

f

2³ 1 x dx ³ 0

m 1 2

sign > z x @> z x @

1 z 2

f

(A-18a)

dz ,

m 1 2

dz .

(A-18b)

After transformation, these integrals may be written as:

J m 1 2 1

1 x ° f ½° m 1 2 m 1 2 2 ³ dx ® ³ 1 z 2 dz 2 ³ 1 z 2 dz ¾ , x 0 ¯° f ¿°

(A-19a)

342

Analytical Methods for Problems of Molecular Transport

J m 1 2 and J m 1 2 . 1

Table A-1. Values of the integrals,

2

J m 1 2

J m 1 2

1

m

2

4 3 2

3 2 2 3 2 2 15 11 2 58 45 2

1

J 3 2

4 3 2 2

J 3 2

16 3

2

J 5 2

4 3

J 5 2

8 9

1

1

J 7 2 1

3

1

1

J m 1 2 2

2

28 23 2 8 7 2

2 15

J 9 2

4

2

J 7 2 2

8 45

J 9 2

4 105

2

13 35

2³ 1 x 2 dx 0

(A-19b)

f x m 1 2 m 1 2 ° °½ u® ³ 1 z2 dz 2 ³ 1 z 2 dz ¾ . x ¯° f ¿°

Then, using the following relationship [6]:

³

dz

1 z 2

z m 1 2

2 m 1 1 z 2

m 1 2

2m 2 2m 1

³

dz

1 z 2

m 1 2

, (A-20)

recurrence formulas for m t 2 may be obtained:

8

1 § ¨ 1 m 3 2 2 m 1 2 m 3 © 2

· 2m 2 1 J m 1 2 , ¸ ¹ 2m 1

J m 1 2

J m 1 2

8 § 25 2 m 2 25 2 m 1 · 2 m 2 2 J . ¨ ¸ 2m 1 © 2m 3 2m 5 ¹ 2m 1 m 1 2

1

2

(A-21a)

(A-21b)

Table A-1 gives the expressions for these integrals for several values of m . Using the integrals from Table A-1, one obtains: 2 I

12

3 8

§ kT · ¸ © m¹

SV 2 ¨

J 1 52

12

§ kT · SV 2 ¨ ¸ © m¹

1 6

5 2

J 7 2 2

3 2 2 .

(A-22)

343

Appendix A. Bracket Integrals for the Planar Geometry 3 Now, consider the third integral, I , which may be written as:

3 I

12f

§ kT · 14 V 2S 3 ¨ ¸ © m¹ 2S

f

u ³ dE 0

f

S

3 2 ³ g exp g dg ³ sin D dD 0

f

0

³ exp Gx dGx ³ exp Gy dGy 2

f

2

f

S

u ³ exp Gz2 dGz ³ sin T dT f

0

2S

³ dH 0

u ª¬G y g sin D cos E º¼ u ª¬G y g sin D cos E cos T (A-23) cos D cos E sin T cos H sin E sin T sin H º¼ sign Gx g cos D

u sign Gx g ª¬cos D cos T sin D sin T cos H º¼ . First, the variable of integration should be changed from T to T . Then, after integration with respect to E , G y , and Gz , the integrand may be written as:

^

`

S 2 1 g 2 ª sin 2 D cos T sin D cos D sin T cos H º ¬

¼

u sign Gx g cos D

u sign Gx g ª¬cos D cos T sin D sin T cos H º¼ . Here, it is convenient to introduce new collision angles to simplify integration over the collision parameters. Let the variables of integration be changed from g ,T , H to g c, D c, H c . The relationships between these variables are shown in Fig. A-1. From the spherical triangle, the following expressions may be obtained:

cos T cos D cos D c sin D sin D c cos H c ,

(A-24a)

cos D c cos D cos T sin D sin T cos H .

(A-24b)

Neglecting vanishing integrals, one obtains:

344

Analytical Methods for Problems of Molecular Transport

Figure A-1. The relationship between various collision angles.

I

3

12f

kT · V S ¨ ¸ © m¹ 1 4

1 §

2

f

³g 0

3

S

exp g 2 dg ³ sin D dD

S

S

0

0

0

u ³ exp Gx2 dGx ³ sin D c dD c³ d H c f

(A-25)

u sign ª¬Gx g cos D º¼ ª¬Gx g cos D c º¼ . At this point, changing the variables of integration to ( z Gx g , x cos (D ) , y cos (D c) ) and then performing the integrations with respect to g and H c , Eq. (A-25) becomes:

I

3

12 1

§ kT · SV ¨ ¸ © m¹ 2

163 f

u ³ dz 1 z

2

1

³ dx ³ dy

1

5 2

1

(A-26)

sign > z x @> z y @ .

f

For other bracket integrals in planar boundary value transport problems the integrand may contain the following expressions:

345

Appendix A. Bracket Integrals for the Planar Geometry

1 z 2

m 1 2

sign > z x @> z y @ ,

and:

1 x 1 z 2

2

m 1 2

sign > z x @> z y @ .

For these terms, let the following notations be introduced:

I m 1 2 1

1 1

I m 1 2 2

1

f

³ dx ³ dy ³

1

1

sign > z x @> z y @

1 z 2

f

1

f

1

f

³ 1 x dx ³ dy ³ 2

1

m 1 2

sign > z x @> z y @

1 z 2

(A-27a)

dz ,

m 1 2

dz .

(A-27b)

To facilitate calculations of these integrals it is convenient to obtain recurrence formulas that may be derived by using the relation given by Eq. (A-20). 1 First, consider the integral, I m 1 2 , which may be represented in the form:

I m 1 2 1

x f sign > z x @> z y @ ° dx dz ³ ® ³ dy ³ m 1 2 1 ° 1 f 1 z2 ¯ 1

½ sign > z x @> z y @ ° ³ dy ³ dz ¾ . m 1 2 x f ° 1 z2 ¿ 1

f

In the first term y d x and therefore:

1 ; f z y , x z f , sign > z x @> z y @ ® yzx . ¯ 1 ; In the second term y t x and therefore:

1 ; f z x , y z f , sign > z x @> z y @ ® xz y . ¯ 1 ;

(A-28)

346

Analytical Methods for Problems of Molecular Transport

1 Using these representations for sign > z x @> z y @ , the integral I m 1 2 may be expressed in the following form:

1

I m 1 2

1 f x x dz dz ° dx dy dy 2 ® ³ ³ ³ ³ ³ m 1 2 m 1 2 2 1 1 y 1 z ° 1 f 1 z 2 ¯ 1

½ dz ° . 2 ³ dy ³ m 1 2 ¾ 2 x x 1 z ° ¿

(A-29)

y

1

Now, by means of Eq. (A-20), the following recurrence formula may be obtained:

2m 2 1 211 2 m I m 1 2 2m 1 2m 1 2m 3

I m 1 2 1

1

8

dx

2m 1 2m 3 ³1 1 x 2 m3 2

(A-30)

;

mt2 .

1 From this general expression the basic integral, I 3 2 , becomes:

I 3 2 1

1

8 1 2 4³

dx

1

12

1 x 2

.

(A-31)

In the same way, the following recurrence formula may be obtained for 2 I m 1 2 :

I m 1 2 2

2m 2 2 211 2 m I m 1 2 2m 1 3 2m 1 2m 3

1 x dx 2m 1 2m 3 ³ 1 x 1 x dx ; m t 2 , 4 ³ 2m 1 1 x 1

2

1

1

with the basic integral:

2

1

4 2m 2

2

m 1 2

2

m 3 2

(A-32)

347

Appendix A. Bracket Integrals for the Planar Geometry

2

I3 2

16 3

1 x dx 2 4³ 1 x 2

1

1

2

1

.

12

(A-33)

Expressions for several of these integrals are given in Table A-2. Using these table integrals yields: 3 I

12

§ kT · 163 SV 2 ¨ ¸ © m¹

I 5 2 1

12

§ kT · 12 S V 2 ¨ ¸ © m¹

2 2 .

(A-34)

The integral given by Eq. (A-23) does not alter if the variable of integration, T , is changed from T to T S . This integral then becomes 4 equal to I , and: 3 I

4 I .

1 The integral, I , after simple integration may be written as:

I

1

12

§ kT · SV ¨ ¸ © m¹ 2

7 3

(A-35)

.

Thus, for ª¬ c y sign cx , c y sign cx º¼ , the final expression is given by:

I

1 2 3 I I 2 I

12

§ kT · 16 S V 2 ¨ ¸ © m¹

1 8 2 .

(A-36)

It is necessary to note that this bracket integral is different from that used in the Gross-Ziering theory. The relationship between these integrals is: 12

>) ,< @

§ m · >) ,< @ nS 3 2 ¨ ¸ © 2kT ¹

(A-37)

,

where >) ,< @ is the Gross-Ziering notation. obtains:

ª¬ c y sign cx , c y sign cx º¼

1 12

1 8 2 SO

1

Using this notation, one

.

This expression is identical to that obtained previously in [7].

(A-38)

348

Analytical Methods for Problems of Molecular Transport

Table A-2. Values of the integrals,

1 2 I m 1 2 and I m 1 2 .

I m 1 2

I m 1 2 1

m

I 3 2 1

1

8 1 2 1

2

4 ³ 1 x 2 1

I 3 2 2

1 2

8 3

I5 2

2

2 2

1 2 1

4 ³ 1 x 2

dx

1

I 5 2 2

1

16 3

8 9

I 7 2 1

I 9 2 1

4

4.

4 15 2 105

16 9 2

192 115 2

2

3 2

2

4 3 2 1

43 ³ 1 x 2 1

3

1 2

1 x dx

I 7 2 2

4 45

I 9 2

2 105

2

1 x dx

32 17 2 128 73 2

BRACKET INTEGRALS CONTAINING ONE DISCONTINUOUS FUNCTION.

In the case of the planar geometry for stationary transport problems, a moment system can be obtained by means of the Maxwellian integral equation given by Eq. (11.52). For these problems it is convenient to employ the discontinuous distribution functions which might satisfy the boundary conditions at the wall surface. In the full velocity space these distribution functions for the linearized transport problems can be presented by the following general expression:

f

^

0 f 1 12 ª¬) 2 c, x )1 c, x º¼

`

12 ª¬) 2 c, x )1 c, x º¼ sign cx ,

(A-39)

where ) 2 c, x and )1 c, x are the corrections to the distribution function for which cx ! 0 and cx 0 , respectively, and x is a normal coordinate to the wall. If this distribution function is employed for closing the full velocity space moment system, the right-hand-sides of moment equations contain both ordinary bracket integrals and others which can be expressed in the form:

ª¬< c sign cx , Q c º¼ ,

(A-40)

349

Appendix A. Bracket Integrals for the Planar Geometry

where < c and Q c are functions of the molecular velocity components. Consider, as an example, evaluation of the bracket integral defined by:

I

ªsign cx , c 2 cx º . ¬ ¼

(A-41)

This integral occurs in the planar heat transport problem (see Section 11.9) and can be obtained from Eq. (11.16) by substitution of r 1 and cr 0 . Integrating with respect to the variables, T , H , and g , and introducing new variables of integration via the following relationships:

Gx

xg , G y

yg , Gz

zg ,

12 1

f

cos D t ,

one can obtain:

I

§ 2kT0 · 96V 2S 1 ¨ ¸ © m ¹

f

f

³ dt ³ dx ³ dy ³ F t , x, y, z dz ,

1

t

f

(A-42)

f

where:

F

x t 2 13 1 x 2 y 2 z 2

5

,

and molecules have been assumed to be rigid spheres. In order to facilitate the integration procedure in Eq. (A-42) with respect to variables y and z , it is convenient to introduce polar coordinates by means of the relationships:

y

U cos \ , z U sin \ .

(A-43)

Then, after integrations with respect to \ and U , this integral becomes:

I

12 1

§ 2kT0 · 24V ¨ ¸ © m ¹ 2

³ t

1

2

f

13 dt ³ 1 x 2 t

4

xdx .

(A-44)

The integrations with respect to x and t offer no difficulty. On integrating over all values of these variables one can obtain: 12

I

§ 2kT0 · 2 23 ¨ ¸ V . © m ¹

The same value of this integral has been obtained in [8].

(A-45)

350

Analytical Methods for Problems of Molecular Transport

REFERENCES 1. Chapman, S. and Cowling, T.G., The Mathematical Theory of Non-Uniform Gases (Cambridge University Press, Cambridge, U.K., 1990). 2. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 3. Hirschfelder, J.O., Curtiss, C.F., and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954). 4. Rolduguin, V.I., Application of the Non-Equilibrium Thermodynamics Method in Boundary Problems of the Kinetic Theory of Gases (M.S. Thesis, Moscow, 1979). 5. Savkov, S.V., The Slip Boundary Conditions of Non-Uniform Binary Gas Mixture and Application of them in Aerosol Dynamics (M.S. Thesis, Moscow, 1987). 6. "Mathematical Tables," in Handbook of Chemistry and Physics, 71st edition, edited by Lide, D.R. (CRC Press, Boston, 1990), #283, p.A-38. 7. Derjaguin, B.V. and Yalamov, Yu.I., "The Theory of Thermophoresis and Diffusiophoresis of Aerosol Particles and Their Experimental Testing," In Topics in Current Aerosol Research, vol. 3, part 2, edited by Hidy, G.M. (Pergamon Press, Oxford, 1972). 8. Ivchenko, I.N., "Generalization of the Lees Method in Boundary Problems of Transfer," J. Colloid and Interface Sci. 135(1), 16-19 (1990).

Appendix B BRACKET INTEGRALS FOR CURVILINEAR GEOMETRIES

1.

THE SPECIAL FUNCTION OF THE FIRST KIND FOR THE SPHERICAL GEOMETRY.

Here, the special functions introduced in Section 11.4 will be

1 investigated in detail. For spherical transport problems, the function I1 > @ r is determined by Eq. (11-21) and, taking into account Eqs. (11-17) and (1118), the following expression may be written: 1

2S

I1 > @ r 72S 2 ³ dt ³ d E

1

1

0

f

³

f

f

f

dy ³ dz ³ F t , E , y, z , x dx , f

(B-1)

x0

where the notations introduced in Eq. (11-17) have been used. It is very difficult to use Eq. (B-1) to obtain numerical values for

1 I1 > @ r , however some analytical simplification is readily achievable. The integration with respect to x offers no difficulty if the following recurrence formula is used [1]:

³

dx

a x 2

ª

m 1

1 x 2ma a x 2

m

2m 1 dx 2ma ³ a x 2

m

(B-2)

º

r r 1 2m « x m 1 dx » ¦ ³ 2 « r m m «¬ 2a r 1 4a mr 2r a x 2 4a a x 2 »¼»

.

352

Analytical Methods for Problems of Molecular Transport

To integrate with respect to y and z , the variables of integration are changed from y and z to y c and z c which are defined by: 12

yc

y 1 t2

zc

z 1 t2

12

cos E ,

(B-3a)

sin E .

(B-3b)

After this substitution it is convenient to introduce the following polar coordinates:

U cos M c ,

yc

zc

U sin M c .

Then, the substitution, M M c E , permits one to integrate over all values of E . Having performed this transformation and integration, one obtains the following analytical expression: I1 > @ r

1

1

2S

f

1

0

0

144

S

>1@ >1@ >1@ >1@ >1@ >1@ ³ dt ³ dM ³ U f1,1 \ 1,1 f1,2\ 1,2 f1,3\ 1,3 d U .

(B-4)

>@ Here, the quantities f1,i> @ and \ 1,i may be written as: 1

f1,1> @ 1

>@ f1,2 1

t 2 13 ,

(B-5a)

ªt 1 t 2 1 2 cos M U 2 t 1 t 2 U º r 2 1 1 2 «¬ »¼ 12 t 2 ª 1 t 2 1 t 2 cos M U º , «¬ »¼

>@ f1,3 1

>1@ \ 1,1

1

t 1 t2

1 8

12

a x 2 0

,

(B-5b)

cos M U t 1 t 2 ,

4

(B-5c)

(B-5d)

353

Appendix B. Bracket Integrals for Curvilinear Geometries

>i @

Table B-1. Numerical values of the functions, I1 r . Note that for i 1, 2 , the apparent values of these functions (evaluated numerically) appear to be the same although the analytical formulations of these functions appear to be different. As of this time these functions have not been proven to be identical.

r

i I > @ r

r

i I > @ r

r

i I > @ r

1.00 1.10 1.20 1.30 1.40

9.982Eí1 8.267Eí1 6.944Eí1 5.917Eí1 5.093Eí1

1.50 1.75 2.00 2.50 3.00

4.440Eí1 3.261Eí1 2.489Eí1 1.591Eí1 1.116Eí1

3.50 4.00 5.00 6.00 10.0

8.143Eí2 6.052Eí2 3.982Eí2 2.791Eí2 1.049Eí2

>1@ \ 1,2

35 128

a 4 a x02

>1@ \ 1,3

35 128

7 48

a

a 9 2

2

1

35 3 a a x02 192

a x02

3

1 8

a

1

x0

t U r2 1

12

12

2

a

U 2 2U 1 t 2

x02

S arctan x0 a 1 2

1 2

a

4

(B-5e)

,

,

(B-5f)

(B-5g)

cos M 2 t 2 ,

.

(B-5h)

The numerical values of the functions, I1 > @ r and I1 > @ r (which will be discussed in Section B.2), are given in Table B-1 and the dependence of

i I1 > @ r on the radial coordinate is shown in Fig. B-1. A numerical analysis

i shows that the special functions, I1 > @ r , are essentially equal to each other for i 1, 2 with the relative deviations of the numerical values being less than 1% although equality has not been definitively established.

1

2.

2

THE SPECIAL FUNCTION OF THE SECOND KIND FOR THE SPHERICAL GEOMETRY.

The second special function related to spherical boundary value transport problems can be expressed in the form: 1

2 I1 > @ r 1440S 2 8 3S

1

2S

³ dt ³ d E

1

0

f

³

f

f

f

dy ³ dz ³ F) dx , f

x0

(B-6)

354

Analytical Methods for Problems of Molecular Transport

1.0

Normalized Special Function Value

0.8

0.6

0.4

0.2

0.0 0

2

4

6

8

10

Dimensionless Radial Coordinate

Figure B-1. Dependence of

I1 > @ r on the radial coordinate, r .

i

where the notations of Section 11.4 are used. By means of the transformations employed in Section B.1, this integral can be expressed in the form: 1 1 1

2 I1 > @ r 20 8 3S I1 > @ r 2880S 1 8 3S 1

2S

f

1

0

0

^

`

2 > 2@ > 2@ > 2@ > 2@ > 2@ u ³ dt ³ dM ³ U d U f1,1> @\ 1,1 f1,2 \ 1,2 f1,3 \ 1,3 ,

(B-7)

> @ may be written as: where the quantities f1,i> @ and \ 1,i 2

f1,1> @

15 1 t 2

> @ f1,2

1 t 5

2

2

t

2

2

12

2t

2

13 ª 1 t 2 «¬

12

cos M U º , »¼

13 95 t 1 t 2 a 1

u ª 1 t 2 cos 2 M U 2 2 1 t 2 «¬

12

cos M U º , »¼

(B-8a)

(B-8b)

355

Appendix B. Bracket Integrals for Curvilinear Geometries > @ f1,3 2

2t U 2 ª 1 t 2 ¬ 5

> 2@ \ 1,1

a x

> 2@ \ 1,2

>@ >@ x0\ 1,2 \ 1,3 ,

> 2@ \ 1,3

1 10

2 0

12

2t cos M U cos M t º¼ ,

2t 2t 4 133 t 2 53 4t 1 t 2 2

2

2

4 3

(B-8c)

1 3

,

1

(B-8d)

1

(B-8e)

x0 a 1 a x02

5

.

(B-8f)

Here, the quantities a and x0 are the same as those given in Eqs. (B-5g) and (B-5h), respectively. Numerical values of this special function are given in Table B-1.

3.

THE SPECIAL FUNCTION OF THE FIRST KIND FOR THE CYLINDRICAL GEOMETRY.

This function is defined by Eqs. (11-18) and (11-21). Since the lower limit of integration with respect to x is independent of the variable, z , it is convenient to perform the first integration over all values of z and then to integrate with respect to x . Having performed these integrations, one can obtain a triple integral which can be numerically evaluated. To integrate over all values of y , first change the variable of integration to y c y (1 t 2 )1 2 cos E and then break this integral into two parts by means of the relationship: f

³

dy c

f

0

f

³

dy c ³ dy c .

f

0

After these very simple transformations, one can obtain: 1

2S

f

1

0

0

^

>1@ >1@

S 1 ³ dt ³ d E ³ dy I 2> @ r 315 16

1

>1@ >1@

(B-9) >1@

>1@

>1@

>1@

`

u f 2,1\ 2,1 f 2,2\ 2,2 g 2,1) 2,1 g 2,2) 2,2 ,

356

Analytical Methods for Problems of Molecular Transport

where the following notations are introduced:

t

>@ f 2,1

>@ g 2,1

>@ f 2,2

t 1 t 2

>@ g 2,2

t 1 t 2

>1@ \ 2,1

c x

1

1

1

>1@ \ 2,2

1

1 7

13 ,

cos E ª 1 t 2 «¬

12

cos E ª 1 t 2 «¬

7 2

x0 c x02

>1@ ) 2,2

d x 2 0

1 35

12

cos E y º , »¼

(B-10b)

12

cos E y º , »¼

(B-10c)

,

(B-10d)

1 2

6c

>1@ ) 2,1

(B-10a)

12

2 0

1 35

2

7 2

x0 d x02

2

ª16c 4 8c 3 c x 2 0 «¬

c x02

2

5c

1

1

c x02

3

º . »¼

,

(B-10f)

1 2

6d

(B-10e)

2

ª16d 4 8d 3 d x 2 0 «¬

d x02

2

5d

1

1

d x02

3

º . »¼

(B-10g)

357

Appendix B. Bracket Integrals for Curvilinear Geometries

>i @

Table B-2. Numerical values of the functions, I 2 r . Note that again, as in Table B-1, the values of these functions appear to be the same but that analytical equality has not been proven. Here, the values for the different functions have been given separately in order to demonstrate their similarity.

r

I 2> @ r

2 I 2> @ r

r

I 2> @ r

2 I 2> @ r

1.00 1.10 1.20 1.30 1.40 1.50 1.75 2.00 2.50 3.00 3.50 4.00 5.00 6.00 7.00

0.9999 0.9091 0.8333 0.7692 0.7142 0.6666 0.5714 0.4999 0.3999 0.3333 0.2857 0.2499 0.2000 0.1666 0.1426

1.0000 0.9090 0.8333 0.7691 0.7142 0.6666 0.5714 0.5000 0.4000 0.3333 0.2857 0.2500 0.1999 0.1666 0.1430

8.00 9.00 10.00 12.00 15.00 17.00 20.00 25.00 30.00 50.00 60.00 70.00 80.00 90.00 100.00

0.1249 0.1111 0.1000 0.0833 0.0665 0.0588 0.0500 0.0399 0.0334 0.0200 0.0165 0.0151 0.0123 0.0112 0.0110

0.1250 0.1110 0.0999 0.0833 0.0667 0.0588 0.0499 0.0400 0.0332 0.0199 0.0167 0.0132 0.0123 0.0110 0.0090

1

1

In Eqs. (B-10a)-(B-10g) the quantities x0 , c , and d are defined as:

x0

12

t y r2 1

c 1 ª y 1 t2 «¬

,

12

(B-11a)

2

cos E º , »¼

(B-11b)

and:

d 1 ª 1 t2 «¬

12

2

cos E y º . »¼

(B-11c)

Numerical values of the functions, I 2> @ r and I 2> @ r , are given in

1 Table B-2 and the dependence of I 2> @ r on the radial coordinate is shown z 1 in Fig. B-2. Moreover, the numerical values of ³ I 2> @ r dr for various 1 values of z are given in Table B-3.

1

2

358

Analytical Methods for Problems of Molecular Transport

Normalized Special Function Value

1

0.8

0.6

0.4

0.2

0 1

10

100

Dimensionless Radial Coordinate

Figure B-2. Dependence of

4.

I 2> @ r on the radial coordinate, r .

1

THE SPECIAL FUNCTION OF THE SECOND KIND FOR THE CYLINDRICAL GEOMETRY.

The special function, I 2> @ r , is determined by Eqs. (11-19) and (11.22). Performing the same transformations as in Section B.3, one can obtain:

2

1 1 1

2 I 2> @ r 20 8 3S I 2> @ r 315 S 1 8 3S 4 1

2S

f

(B-12)

u ³ dt ³ d E ³ dy 1

0

0

^

`

> 2@ > 2@ > 2@ > 2@ > 2 @ > 2@ > 2 @ > 2@ u f 2,1 \ 2,1 f 2,2 \ 2,2 g 2,1 ) 2,1 g 2,2 ) 2,2 ,

where the following functions are employed:

359

Appendix B. Bracket Integrals for Curvilinear Geometries Table B-3. Numerical values of the integrals, \ i

z

z ³1 I 2 >i@ r dr .

z

\1

\2

z

\1

\2

1.00 1.05 1.10 1.20 1.30 1.40 1.50

0.0000 0.0488 0.0935 0.1823 0.2624 0.3365 0.4055

0.0000 0.0488 0.0935 0.1823 0.2624 0.3365 0.4055

1.75 2.00 3.00 4.00 5.00 7.00 10.00

0.5596 0.6931 1.0984 1.3860 1.6090 1.9453 2.3019

0.5596 0.6931 1.0983 1.3860 1.6090 1.9454 2.3021

> @ f 2,1 2

ª t ® t 2 13 1 t 2 «sin 2 E ¬ ¯

8c 1 §¨ 1 t 2 ©

> @ f 2,2 2

12

1 t 2t 2

2

12

2

º °½ cos E y ·¸ cos 2 E » ¾ , ¹ ¼ ¿°

13 §¨ 1 t 2 ©

x0 c 1t 1 t 2 §¨ 1 t 2 ©

12

12

(B-13a)

cos E y ·¸ cos E ¹ 2

cos E y ·¸ cos 2 E ¹

(B-13b)

x0t t 2 13 ,

> @ g 2,1 2

ª t ® t 2 13 1 t 2 «sin 2 E ¬ ¯

1 §

8d ¨ 1 t ©

> @ g 2,2 2

12

1 t 2t 2

2

2

12

2 º °½ cos E y ·¸ cos 2 E » ¾ , ¹ ¼ ¿°

13 §¨ 1 t 2 ©

x0 d 1t 1 t 2 §¨ 1 t 2 ©

x0t t 2 13 ,

12

12

(B-13c)

cos E y ·¸ cos E ¹ 2

cos E y ·¸ cos 2 E ¹

(B-13d)

360

Analytical Methods for Problems of Molecular Transport > 2@ \ 2,1

>@ \ 2,2 ,

> 2@ \ 2,2

c x

> 2@ ) 2,1

>@ ) 2,2 ,

> 2@ ) 2,2

d x

1

2 0

(B-13e) 1

c x

>1@ \ 2,1

2 0

9 2

,

(B-13f)

1

2 0

(B-13g) 1

>1@ ) 2,1

d x 2 0

9 2

.

(B-13h)

The quantities shown in these formulas are identical to those given in Eqs.

2 (B-10a)-(B-11c). Numerical values of I 2> @ r are given in Table B-2 and z > 2@ numerical values of ³ I 2 r dr are given in Table B-3. 1

5.

APPROXIMATE EXPRESSIONS FOR THE SPECIAL FUNCTIONS.

Using the general properties of the Boltzmann equation [2], one can obtain very simple approximate expressions for the bracket integrals considered in Sections B.1-B.4. These approximate expressions will be derived in this section by employing the first-order approximation to the Chapman-Enskog solution given in Eq. (5-39). In order to use the first-order approximation of the Chapman-Enskog solution, the bracket integrals must be transformed into the more convenient form:

ª c k sign cr cr , c 2 cr º ¬ ¼

1 ªc k sign cr cr , cr S3 2 c 2 º ¬ ¼

³ c k sign cr cr I cr S3 2 c 2 dv , k 1

(B-14)

0, 2 .

Now, using Eq. (5-39), these bracket integrals can be approximated by:

ª c k sign cr cr , c 2 cr º ¬ ¼ 1 1 0 v S c 2 f c k sign cr cr dv , 2 1 ³ r 3 2 n a1

1 where a1 is given by Eq. (5-40).

(B-15)

Appendix B. Bracket Integrals for Curvilinear Geometries

361

It is very important to note that the integrals in the right-hand-side of Eq. (B-15) are ordinary integrals in the velocity space. The values of these integrals depends on the problem geometry. For the spherical geometry, after integration, one can obtain the following expressions: I1 > @ r

4 5

2 I1 > @ r

64 5

1

2r 2 ,

(B-16)

1

2 8 3S r 2 .

(B-17)

The appropriate expressions for the cylindrical geometry can be written as: I 2> @ r

4 5

2 I 2> @ r

64 5

1

2r 1 ,

(B-18)

1

2 8 3S r 1 .

(B-19)

The accuracy of this analysis can be easily determined by means of a comparison with the appropriate numerical values. The relative errors for

1

2 I i > @ r and I i > @ r , for r 1 , are 13% and 4%, respectively.

REFERENCES 1. "Mathematical Tables," in Handbook of Chemistry and Physics, 71st edition, edited by Lide, D.R. (CRC Press, Boston, 1990), #67, p.A-24. 2. Loyalka, S.K., "Approximate Method in the Kinetic Theory," Phys. Fluids 14(11), 22912294 (1971).

Appendix C BRACKET INTEGRALS FOR POLYNOMIAL EXPANSION METHOD

1.

CALCULATION OF THE BRACKET INTEGRALS OF THE FIRST KIND.

To complete the calculations of the polynomial expansion method in the four-moment approach for the spherical geometry, one must know values of the following bracket integrals:

D k>1n@

ª H k cr S0 n cT2 cI2 , cr c 2 º , ¬ ¼

D k> 2n@

ª c 2 H k cr S0 n cT2 cI2 , cr c 2 º , ¬ ¼

where H k cr and S0 cT2 cI2 are, respectively, Hermite and Sonine polynomials of the molecular velocity components. First, consider the 1 method of calculating the bracket integrals, D k> n@ , when molecules are assumed to be rigid spheres. To obtain general analytical expressions for these integrals, one must use the generating functions for the Hermite and Sonine polynomials, H x, t and S0 x, s , which are given by: n

H x, t exp t 2 2 xt , and:

x cr ,

(C-1)

364

Analytical Methods for Problems of Molecular Transport

1 § xs · exp ¨ ¸ , 1 s © 1 s ¹

S 0 x, s

x cT2 cI2 .

(C-2)

Then, the Hermite and Sonine polynomials may be expressed as:

Hk x

wk H x, t wt k t

(C-3)

, 0

and: n S0 x

1 wn S 0 x, s n ws n s

.

(C-4)

0

Having used Eqs. (C-3) and (C-4), one can obtain: >i @

D kn

2 1 w k w n exp t [ >i @ t , s n wt k ws n 1 s

,

(C-5)

t ,s 0

where:

ª

§ ©

s

·¹

º

[ >i@ t , s «c 2 i 1 exp ¨ 2cr t cT2 cI2 ¸ , cr c 2 » . 1 s ¬

(C-6)

¼

The bracket integral of Eq. (C-6) can be calculated in the same manner as the original one but it has some special properties that can be taken into account to facilitate the calculations. The analysis is simplified if the following new variables of integration are introduced:

t

,

(C-7)

Grc

Gr

GTc

GT

s g sin D cos E , 2s

(C-8)

GIc

GI

s g sin D sin E , 2s

(C-9)

2

and:

Appendix C. Bracket Integrals for Polynomial Expansion Method

365

where Gr , GT , GI are the Cartesian coordinates of the dimensionless velocity, G , and g ,D , E are the spherical coordinates of the dimensionless velocity, g . Then, having performed six integrations with respect to the variables, T , H , E , Grc , GTc , and GIc , and having employed Leibnitz’ formula [1] for the k -th derivative with respect to t , one can obtain: >1@ D kn

12

16 3

§ 2kT · ¸ © m ¹

S 1 2V 2 ¨

1 wn 1 2 s ³ sin D dD n n ws 0 S

f § s ª º· sin 2 D » ¸ u³ g 3 exp ¨ g 2 «1 ¬ 2s ¼¹ © 0

u< k> @ s,D , g dg 1

s 0

(C-10)

,

where:

< k>1@ s,D , g

3 2

F k 0, g

s g 3 f 2 D 2s

1 k 1 k F 0, g g 2 f1 D , 2

F t , g exp 12 t 2 2 gt cos D ,

f1 D

3 2

cos 2 D 12 ,

and:

f 2 D sin 2 D cos D . k To calculate the k -th derivative, F 0, g , one should use the recurrence relation that is given by:

F k 1 0, g 2 g cos D F k 0, g k F

k 1

0, g

;

k t0 ,

(C-11)

366

Analytical Methods for Problems of Molecular Transport

Table C-1. Analytical values of the bracket integral, method.

>1@ , D1n

for the polynomial expansion

12

D1>1n@

anV 2 2S 2kT m

n

1

2

3

4

an

16 15

16 105

4 315

8 3465

with:

F 0 0, g 1 .

(C-12)

>@ For the particular case when the index, k 1 , the bracket integrals, D1n , may be expressed in the form: 1

D1>1n@

12

§ 2kT · 163 2S V 2 ¨ ¸ © m ¹

1

1 1 M n x dx , n ³0

(C-13)

where:

Mn1 x

3 2

9 1 x x 4nQ

n n 1 n2 x 2 12 4Q 3 4nQ 3 n n 1 Q 3 2

2

4

n 1

4n n 1 Q 4

n n 1 n 2 Q 4

n 3

n 2

.

Here, the following notations are introduced:

Q 3 m

Q 4 m

3 4 5 3 m 1 23 m

4 5 6 4 m 1 24 m

x 2m ;

x2m ;

m t1 ,

m t1 .

>@ Specific values of D1n are given in Table C-1. 1

(C-14)

(C-15)

367

Appendix C. Bracket Integrals for Polynomial Expansion Method

2.

ANALYTICAL EXPRESSIONS FOR THE BRACKET INTEGRALS OF THE SECOND KIND.

The method described in the preceding section may be used to calculate 2 the integrals, D k> n@ , as well. After some fairly complicated algebraic transformations, one can obtain: 12

> 2@ D kn

§ kT · 2SV 2 ¨ ¸ © m¹

8 3

f

1 wn 1 2 s ³ sin D dD n n ws 0 S

§ s ª º· 2 u ³ g exp ¨ g 2 «1 sin 2 D » ¸< k> @ s,D , g dg ¬ 2s ¼¹ © 0 s

(C-16)

3

, 0

where:

°

< k> 2@ s,D , g F k 0, g ® f1 D g 3 cos D °¯

2 ª º½ · s § 3s 2 4s 2 3 3 13s 22 s 8 » ° ¨1 ¸ D f 2 D « 32 g 5 sin g ¾ 2 2 ¸ 4 2s¨ « »° 2 s 2 s © ¹ ¬ ¼¿

1 2

kF

k 1

§ 3s 2 4s ·º ª 10 7 s 2 2 ¸ » f1 D D g g 4 ¨1 sin 2 ¨ ¸ 2 s » s 2 « ¯° ¬ © ¹¼

0, g ®°« 12

½° 3s k 2 g 4 cos D f 2 D ¾ k k 1 F 0, g 2 s °¿

s ° °½ u ® g 3 cos D f1 D 34 g 3 f 2 D ¾ 2 s ¯° ¿° 1 k 3 12 k k 1 k 2 g 2 F 0, g f1 D , 2

f1 D

3 2

cos 2 D 12 , and f 2 D sin 2 D cos D .

> @ Here, for the specific case of k 1 , the integrals, D1n , may be expressed in the form: 2

368

Analytical Methods for Problems of Molecular Transport

Table C-2. Analytical values of the bracket integral, method.

> 2@ , D1n

for the polynomial expansion

12

D1>n2@ bnV 2 2S 2kT m n

1

2

3

4

bn

163

46 15

38 105

19 630

> 2@

D1n

12

4 3

§ 2kT · 2S V ¨ ¸ © m ¹ 2

1

1 2 M n x dx , n ³0

where:

Mn 2 x

^88 96 x Q n 168 116 x Q n n 1 110 36 x Q n n 1 n 2 24 x Q ` 3 1 x x ^96Q n 552 48 x Q n n 1 816 12 x Q n n 1 n 2 474 72 x Q n n 1 n 2 n 3 96 27 x Q ` ,

3 2

x 2 12

n

2

n 1

2

4

4

n2

2

4

n 3

2

4

2

n

2

n 1

2

5

5

n2

2

5

n 3

2

5

n4

2

5

and:

Q 5 m

5 6 7 5 m 1 25 m

x2m ;

m t1 .

> @ Some values of D1n are given in Table C-2. 2

REFERENCES 1. Bronshtein, I.N. and Semendyayev, K.A., A Guidebook to Mathematics (Springer Verlag, New York, N.Y., 1973).

Appendix D THE VARIATIONAL PRINCIPLE FOR PLANAR PROBLEMS

1.

SOME DEFINITIONS AND PROPERTIES FOR INTEGRAL OPERATORS.

In the Kinetic Theory of Gases, the following form of an integral operator, Q , is frequently used:

Q) c, r

³ K c, cc ) cc, r dcc ,

(D-1)

where the integration extends over all values of the molecular velocity. For planar transport problems, the appropriate operator is defined by:

Q) c, x

³ K c, cc ) cc, x dcc .

(D-2)

To begin with, a fundamental definition for this integral operator is introduced. The adjoint operator, Q , in the space defined by the scalar product given by Eqs. (9-16) and (9-17) is specified by:

ª¬< , Q) º¼

ª ) , Q < º . ¬ ¼

(D-3)

The operator, Q , is termed a self-adjoint operator if:

Q

Q .

(D-4)

370

Analytical Methods for Problems of Molecular Transport

For a self-adjoint operator, Eq. (D-3) becomes: ª¬< , Q) º¼

ª¬) , Q< º¼ .

(D-5)

It has been proven in [1,2] that the condition for the operator, Q , to be selfadjoint is that:

K c, cc

K cc, c .

(D-6)

Another way of stating Eq. (D-6) is to say that the function, K c, cc , in Eqs. (D-1) and (D-2) is symmetric in the variables, c and cc . From the symmetry of the kernel, K c, cc , in Eq. (9-14), one can easily conclude that the operator, H , is self-adjoint. In the formulation of the variational

principle, the adjoint operator, H L , where L is defined by Eq. (9-13), is

employed to facilitate certain transformations. The adjoint operator, H L , may be derived from the following relationships:

ª¬< , H L) º¼

ª¬ H < , L) º¼

ª ) , L H < º ¬ ¼

ª ) , L H < º , ¬ ¼

(D-7)

which yield:

H L 2.

L H

L H .

(D-8)

THE VARIATIONAL PRINCIPLE.

In the linearized planar transport problems studied to date, it has generally been the case that the main quantity of physical interest can be expressed as:

I

ª ) , p º ¬ ¼

ª p , ) º , ¬ ¼

(D-9)

where the following notation has been introduced:

)

H) .

(D-10)

Appendix D. The Variational Principle for Planar Problems

371

and ) is the correction to the distribution function which satisfies the following integral form of the Boltzmann equation:

) c, x L) c, x p c, x .

(D-11)

Applying the operator, H , to Eq. (D-11), one obtains:

) c, x H L) c, x p c, x L) c, x p c, x ,

(D-12)

where L H LH 1 and H 1 is the inverse of H . Now, a functional may be constructed of the form:

F ) ,)

ª ) , p º ª ) ,) L) p º , ¬ ¼ ¬ ¼

(D-13)

where ) is a trial function. If this trial function is equal to the real correction to the distribution function, ) , then the integral, I , is found from:

F ) ,) .

I

(D-14)

Now, consider a trial function of the form:

) ) G) .

(D-15)

For this trial function, the functional of Eq. (D-13) becomes:

F ) G) ,) G)

ª p ¬

, ) º O ªG) º , ¼ ¬ ¼

ª p , ) º ª G) , ) L) p º ¬ ¼ ¬ ¼

(D-16)

2

which implies that:

GF

2 O ªG) º . ¬ ¼

(D-17)

Since the first variation is equal to zero, the functional is stationary when ) c, x ) c, x and ) c, x ) c, x . Now, the basic integral of

372

Analytical Methods for Problems of Molecular Transport

interest, I , can be expressed in terms of the stationary value of the functional by the following simple expression:

I

F ) ,)

Fst ) ,) ,

(D-18)

and, hence, to find an approximate value for I , one need only employ a simple trial function satisfying the stationary condition [3-10]. It should be noted in this context that a non-self-adjoint operator such as the Boltzmann operator cannot lead to an extremum (minimum) but, rather, must yield a saddle point because it has been proven in [4] that the second variation, G 2 F , may have either sign.

REFERENCES 1. Morse, P.M. and Feshbach, H., Methods of Theoretical Physics (McGraw-Hill, NY, 1953). 2. Courant, R. and Hilbert, D., Methods of Mathematical Physics, Vol. 1 (Interscience, NY, 1953). 3. Kahan, T., Rideau, G., and Roussopolous, P., Les Methodés d'Approximation Variationelles Dans la Theorié des Collisions Atomiques et Dans la Physique des Piles Nucléaires (Gauthier-Villars, Paris, 1956). 4. Pomraning, G.C. and Clark, M.Jr., "The Variational Method Applied to the Monoenergetic Boltzmann Equation. Part I," Nucl. Sci. Engineering 16, 147-154 (1963). 5. Cercignani, C. and Pagani, C.D., "Variational Approach to Boundary-Value Problems in Kinetic Theory," Phys. Fluids 9(6), 1167-1173 (1966). 6. Loyalka, S.K., "Momentum and Temperature-Slip Coefficients with Arbitrary Accommodation at the Surface," J. Chem. Phys. 48(12), 5432-5436 (1968). 7. Lang, H., "Ein Variationsprinzip für die Linealisierte, Stationäre Boltzmann-Gleichung der Kinetischen Gastheorie," Acta Mechanica 5, 163-188 (1968). 8. Loyalka, S.K., "Linearized Couette Flow and Heat Transfer Between Two Parallel Plates," in Rarefied Gas Dynamics, edited by Trilling, L. and Wachman, H.Y. (Academic Press, New York) Proceedings of the 6th International Symposium on Rarefied Gas Dynamics (1969). pp. 195-203. 9. Loyalka, S.K., "Slip Problems for a Simple Gas," Z. Naturforsh. 26a(6), 964-972 (1971). 10. Loyalka, S.K., "Slip in the Thermal Creep Flow," Phys. Fluids 14(1), 21-24 (1971).

Appendix E SOME DEFINITE INTEGRALS

1.

SOME FREQUENTLY ENCOUNTERED INTEGRALS. One commonly encountered integral is of the form (see Table E-1): f

in

³x

n

exp x 2 dx .

0

For even and odd values of n , respectively, this may be rewritten as: f

³x

2n

0

f

³x 0

exp E x 2 dx

2 n 1

2n 1 2n 3 5 3 1

S

2n 1

E 2 n 1

exp E x 2 dx

1 2

n

E n 1

;

nt0 .

(E-1)

;

n t1 ,

(E-2)

374

Analytical Methods for Problems of Molecular Transport

Table E-1. Values of the commonly encountered n

0 1 2

in

5

in

6

10 945 64

in

2.

1 4

15 16

1

n

2

1 2

S

n

in integrals.

1

S

60

3 8

8 105 32

3 12 10395 128

4

1 2

S 7

11

S

3

S 9

S

12

13

S

360

SOME INTEGRALS ENCOUNTERED IN BOUNDARY PROBLEMS. 2 ³ cx exp c dc

1 2

S ,

(E-3)

exp c 2 dc S ,

(E-4)

³ cx c

2

2 2 ³ cx exp c dc

1 4

S3 2 ,

(E-5)

2 2 ³ cx c y exp c dc

1 4

S ,

(E-6)

2 2 2 ³ cx c y exp c dc

1 8

S3 2 ,

(E-7)

5 8

S3 2 ,

(E-8)

1 4

2 2

³ cx c

exp c 2 dc

³ cx c

2

32 exp c 2 dc

S ,

(E-9)

375

Appendix E. Some Definite Integrals 2 2 2 ³ cx c c 32 exp c dc

3 2

S ,

(E-10)

S3 2 ,

(E-11)

³ cx c 2

2

23 exp c 2 dc

³ cx c

2

52 exp c 2 dc 14 S ,

1 4

(E-12)

2 2 2 ³ cx c c 52 exp c dc

1 2

S ,

(E-13)

³ cx c 2

2

52 exp c 2 dc 0 ,

(E-14)

2 1 2 2 ³ cx c y S3 2 c exp c dc

18 S ,

(E-15)

2 2 1 2 2 ³ cx c y S3 2 c exp c dc

18 S 3 2 ,

(E-16)

S3 2 c 2 exp c 2 dc 85 S 3 2 ,

(E-17)

2 2

³ cx c

1

3 1 2 2 ³ cx S3 2 c exp c dc

14 S ,

(E-18)

1 2 2 ³ cx ª¬ S3 2 c º¼

2

5 8

13 8

exp c 2 dc

S3 2 ,

(E-19)

2

1 2 3 2 ³ cx ª¬ S3 2 c º¼ exp c dc

S ,

(E-20)

376

Analytical Methods for Problems of Molecular Transport 3 2 2 2 ³ cx S3 2 c exp c dc

161 S ,

(E-21)

2 2 3 1 2 2 ³ cx S3 2 c S3 2 c exp c dc

23 32 S ,

(E-22)

2 2 2 ³ cx 2 c exp c dc

18 S 3 2 ,

(E-23)

S ,

(E-24)

2 2 2 ³ cx c 2 c exp c dc

1 2 2 2 2 ³ cx 2 c S3 2 c exp c dc

5 8

2 2 2 2 2 ³ cx 2 c S3 2 c exp c dc

0 .

S3 2 ,

(E-25)

(E-26)

Here, the following notation has been used:

f

³ dc

3.

³ dcx 0

f

³

f

f

dc y

³ dcz

.

f

SOME INTEGRALS CONNECTED WITH THE SECOND-ORDER CHAPMAN-ENSKOG SOLUTION. 1 2 2 2 2 2 ³ cx ª¬ S3 2 c a2 S3 2 c º¼ exp c dc

1 4

S 1 34 a2 2 ,

(E-27)

377

Appendix E. Some Definite Integrals

³ cx c

2

7 8

S 1

³ cx c

2

1 2 2 32 ª S3 2 c 2 a2 S3 2 c 2 º exp c 2 dc ¬ ¼ 13 28

2

a2

,

ª S 1 c 2 a2 2 S 2 c 2 º exp c 2 dc 32 ¬ 32 ¼

1 2

1 4

2

S 1 a2

,

1 2 2 2 2 2 2 ³ cx c y ª¬ S3 2 c a2 S3 2 c º¼ exp c dc

18

2 2

³ cx c

1 4

2

S 1 a2

,

ª S 1 c 2 a2 2 S 2 c 2 º exp c 2 dc 5 S 3 2 , 32 8 ¬ 32 ¼

2 2 2 2 º 3 ª 1 2 c S c a S c exp c 2 dc 2 x 32 ³ ¬ 32 ¼

§ 23 2 S ¨1 26 a2 ©

433 208

2 ª a2 2 º ¸· , ¬ ¼ ¹

1 2 2 2 2 2 1 2 2 2 2 ³ cx c y ª¬ S3 2 c a2 S3 2 c º¼ ª¬1 b2 S5 2 c º¼ exp c dc

1 8

S

32

1

7 2 b 2 2

7 2

2 2

a2 b2

2 2 1 2 º 3 2ª c c 1 b S c exp c 2 dc x y 2 5 2 ³ ¬ ¼

(E-30)

(E-31)

,

(E-32)

(E-33)

1 4

(E-29)

13 8

(E-28)

2 ª º S «1 b2 2 174 b2 2 » , ¬ ¼

(E-34)

378

Analytical Methods for Problems of Molecular Transport 2 1

3 2

³ cx c y ª¬1 b2

4.

S5 2 c 2 º exp c 2 dc ¼

1 4

S 1 12 b2 2 .

(E-35)

SOME INTEGRALS CONNECTED WITH NONLINEAR TRANSPORT PROBLEMS. f

³ exp cx dcx 2

1 2

S 1 2 1 erf u x ,

(E-36)

(E-37)

ux

f

³ cx exp cx dcx 2

1 2

exp u x2 ,

1 2

ªu x exp u x2 1 S 1 2 1 erf u x º , 2 ¬ ¼

1 2

exp u x2 1 u x2 ,

ux

f

³ cx exp cx dcx 2

2

ux

f

³ cx exp cx dcx 3

2

ux

where: erf x

2

S1 2

x

2 ³ exp t dt 0

is the error function.

,

(E-38)

(E-39)

Appendix F OMEGA-INTEGRALS FOR SECOND-ORDER APPROXIMATION

All of the coefficients associated with the Chapman-Enskog algebraic systems of equations given in Eqs. (12-11)-(12-13) for the first-order Chapman-Enskog approximation and given in Eqs. (12-41)-(12-43) for the second-order Chapman-Enskog approximation are expressible in terms of the : -integrals which are dependent upon the specific model of the intermolecular potential that one is using. For a simple gas, these integrals have been described in Section 5.9. For gas mixtures, however, the : integrals must be slightly generalized. In this case, the : -integrals are defined by:

:12 l , r

ª :12 l ,r º : l , r å , ¬ ¼ r .s.

where [:12(l ,r ) ]r .s. is the : -integral for rigid-sphere molecules which is defined in [1] as: 12

l ,r º

ª: ¬ 12 ¼ r .s.

§ kT · ¨ * ¸ © 2S m12 ¹

r 1 ª«1 1 1 l º» SV 2 , 12 2 « 2 l 1 » ¬ ¼

where: * m12

m1m2 and V 12 m1 m2

1 2

V 1 V 2 .

380

Analytical Methods for Problems of Molecular Transport

The functions, : , are known as the reduced : -integrals (called the reduced collision integrals in some texts) and are given in terms of the reduced temperature, T * , by [2]: l ,r å

: l ,r å T *

ª r 1 T * ¬«

f r 2 1

º ¼»

³Q 0

l å

§ E* · E * exp ¨ * ¸ E * © T ¹

r 1

dE * ,

in which the following notations have been used:

Q

l å

E *

ª 1 1 l º » 2 «1 2 l 1 » «¬ ¼

f

³ 1 cos F b db l

*

*

,

0

·¸

* * b*2 M r * ¨ S 2b ³ 1 *2 ¨ r E* r0* ©

f§

F

1

¸ ¹

1 2

dr * , r *2

where r * r V 12 is the reduced intermolecular distance, b* b V 12 is the reduced impact parameter, M * M H is the reduced intermolecular potential * 2 energy, T * kT H is the reduced temperature, E * 12 m12 g H is the reduced relative kinetic energy, H is the energy parameter of the potential model being used, and r0* is determined from the largest, positive, real root of the expression:

·¸

§ b*2 M * r0* ¨1 ¨ r0*2 E* ©

¸ ¹

0.

In Table F-1 below, we have presented values of the : å -integrals computed for selected values of the reduced temperature, T * , using the Lennard-Jones (6-12) potential model. The values tabulated here are those needed for the first- and second-order Chapman-Enskog approximations. All listed values have been independently recalculated from those tabulated in [3] using a program developed by Maitland. [2] and subsequently revised by us to include increased numerical quadratures in the variables, b* and E* . Additionally, we note that a simple, concise, and very transparent program can be constructed in Mathematica® which will also permit good graphical display of the results at several stages in the calculations including

381

Appendix F. Omega-Integrals for Second-Order Approximation l

plots of the scattering angle, F (b * , E * ) , the functions, Q ( E * ) , and the l ,r resulting : -integrals, : (T * ) . We have constructed this Mathematica® program and have included it below as Table F-2 for the convenience of the reader. Given with this program is relevant output which includes a complete alternate set of : -integral values calculated using the Lennard-Jones (6-12) potential model which are needed for the first- and second-order Chapman-Enskog approximations. These values are analogous to those given in Table F-1, which were calculated with a modified version of an earlier Fortran program, and are shown computed for the same set of values of the reduced temperature, T * , that were specified in the earlier Fortran program. We note that while much of the nomenclature in the Mathematica® program is self-explanatory, the program has been annotated with a number of comments to assist the reader. We are suggesting that the reader may use this Mathematica® program to explore results for different intermolecular potential models. The reader may also wish to explore ways in which the program may be improved with respect to its accuracy and speed. The major time-consuming parts of this program are the computation of the scattering angle, F (b* , E * ) , and, in particular, the determination of r0* (b* , E * ) . For potential functions where non-polynomial equations are involved, one might wish to explore the use of the Mathematica® command ‘FindRoot’ (with a suitable starting value r0* | b* ).

382

Analytical Methods for Problems of Molecular Transport

Table F-1(a). Values of some : -integrals computed for selected values of the reduced * temperature, T , using the Lennard-Jones (6-12) potential model. The values tabulated here are the complete set of : -integrals needed for the first- and second-order ChapmanEnskog approximations. All listed values have been independently recalculated from those tabulated in [3] using a Fortran program developed by [2] and subsequently modified by us to give improved accuracy. Note that the last column of the table on page 541 of [2] appears to 2,5 3,3 be mislabeled. Instead of being labeled : it should be labeled : .

T*

: 1,1

0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00 2.10 2.20 2.30 2.40 2.50 2.60

2.6505 2.4699 2.3157 2.1826 2.0671 1.9661 1.8775 1.7992 1.7298 1.6679 1.6126 1.5628 1.5178 1.4771 1.4400 1.4062 1.3752 1.3468 1.3206 1.2964 1.2740 1.2531 1.2337 1.2157 1.1988 1.1829 1.1681 1.1541 1.1409 1.1285 1.1167 1.1056 1.0950 1.0850 1.0755 1.0577 1.0416 1.0268 1.0132 1.0007 0.9891

: 1,2

2.2586 2.0816 1.9345 1.8113 1.7077 1.6198 1.5448 1.4803 1.4244 1.3756 1.3328 1.2949 1.2613 1.2312 1.2042 1.1799 1.1578 1.1376 1.1192 1.1024 1.0868 1.0724 1.0591 1.0467 1.0352 1.0244 1.0143 1.0048 0.9959 0.9874 0.9795 0.9719 0.9648 0.9580 0.9515 0.9394 0.9284 0.9183 0.9090 0.9003 0.8923

: 1,3

1.9674 1.7998 1.6659 1.5578 1.4697 1.3971 1.3365 1.2855 1.2419 1.2045 1.1720 1.1435 1.1184 1.0961 1.0761 1.0582 1.0419 1.0271 1.0136 1.0012 0.9897 0.9791 0.9693 0.9601 0.9516 0.9435 0.9360 0.9288 0.9221 0.9158 0.9097 0.9040 0.8985 0.8933 0.8883 0.8790 0.8704 0.8625 0.8551 0.8482 0.8418

: 1,4

1.7426 1.5908 1.4740 1.3828 1.3103 1.2516 1.2034 1.1632 1.1292 1.1001 1.0749 1.0528 1.0333 1.0160 1.0005 0.9865 0.9737 0.9621 0.9515 0.9416 0.9325 0.9240 0.9161 0.9087 0.9018 0.8952 0.8891 0.8832 0.8777 0.8724 0.8674 0.8626 0.8580 0.8536 0.8494 0.8415 0.8341 0.8273 0.8209 0.8149 0.8092

: 1,5

1.5692 1.4363 1.3372 1.2614 1.2021 1.1546 1.1157 1.0834 1.0560 1.0325 1.0121 0.9942 0.9784 0.9642 0.9514 0.9398 0.9292 0.9195 0.9105 0.9022 0.8945 0.8873 0.8805 0.8741 0.8681 0.8624 0.8570 0.8519 0.8470 0.8424 0.8380 0.8337 0.8296 0.8257 0.8219 0.8148 0.8082 0.8019 0.7961 0.7906 0.7854

: 2,2

2.8438 2.6796 2.5337 2.4029 2.2854 2.1798 2.0849 1.9995 1.9226 1.8533 1.7907 1.7341 1.6826 1.6358 1.5932 1.5542 1.5184 1.4856 1.4553 1.4274 1.4015 1.3776 1.3553 1.3346 1.3152 1.2971 1.2802 1.2643 1.2494 1.2353 1.2221 1.2095 1.1977 1.1865 1.1758 1.1561 1.1383 1.1220 1.1071 1.0934 1.0809

383

Appendix F. Omega-Integrals for Second-Order Approximation

T*

: 1,1

2.70 2.80 2.90 3.00 3.10 3.20 3.30 3.40 3.50 3.60 3.70 3.80 3.90 4.00 4.10 4.20 4.30 4.40 4.50 4.60 4.70 4.80 4.90 5.00 6.00 7.00 8.00 9.00 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100 200 300 400

0.9783 0.9683 0.9589 0.9501 0.9418 0.9340 0.9267 0.9197 0.9131 0.9069 0.9009 0.8953 0.8898 0.8847 0.8797 0.8749 0.8704 0.8660 0.8618 0.8577 0.8538 0.8500 0.8464 0.8429 0.8129 0.7898 0.7712 0.7556 0.7423 0.6641 0.6235 0.5963 0.5760 0.5599 0.5465 0.5352 0.5254 0.5168 0.4630 0.4339 0.4142

: 1,2

0.8848 0.8777 0.8711 0.8649 0.8590 0.8534 0.8481 0.8430 0.8382 0.8336 0.8292 0.8250 0.8210 0.8171 0.8134 0.8098 0.8063 0.8030 0.7997 0.7966 0.7936 0.7906 0.7878 0.7850 0.7610 0.7419 0.7261 0.7126 0.7008 0.6295 0.5913 0.5655 0.5461 0.5308 0.5181 0.5073 0.4979 0.4897 0.4383 0.4106 0.3919

: 1,3

0.8357 0.8300 0.8245 0.8194 0.8145 0.8099 0.8054 0.8012 0.7971 0.7932 0.7895 0.7859 0.7824 0.7791 0.7758 0.7727 0.7697 0.7667 0.7639 0.7611 0.7584 0.7558 0.7533 0.7508 0.7292 0.7117 0.6970 0.6844 0.6733 0.6052 0.5684 0.5435 0.5248 0.5099 0.4977 0.4872 0.4782 0.4702 0.4207 0.3940 0.3761

: 1,4

0.8039 0.7988 0.7940 0.7894 0.7850 0.7808 0.7768 0.7730 0.7693 0.7657 0.7623 0.7590 0.7558 0.7527 0.7497 0.7468 0.7440 0.7413 0.7386 0.7361 0.7336 0.7311 0.7288 0.7264 0.7060 0.6894 0.6753 0.6632 0.6526 0.5865 0.5507 0.5265 0.5083 0.4938 0.4819 0.4718 0.4630 0.4552 0.4071 0.3812 0.3641

: 1,5

0.7804 0.7757 0.7713 0.7670 0.7629 0.7590 0.7552 0.7516 0.7481 0.7448 0.7416 0.7384 0.7354 0.7325 0.7296 0.7269 0.7242 0.7216 0.7191 0.7166 0.7142 0.7119 0.7096 0.7074 0.6877 0.6716 0.6580 0.6462 0.6359 0.5714 0.5364 0.5127 0.4949 0.4808 0.4691 0.4592 0.4506 0.4431 0.3962 0.3710 0.3545

: 2,2

1.0692 1.0584 1.0483 1.0389 1.0301 1.0218 1.0141 1.0067 0.9998 0.9932 0.9870 0.9810 0.9754 0.9700 0.9649 0.9599 0.9552 0.9507 0.9463 0.9422 0.9381 0.9343 0.9305 0.9269 0.8963 0.8728 0.8539 0.8381 0.8245 0.7437 0.7008 0.6717 0.6498 0.6323 0.6178 0.6055 0.5947 0.5852 0.5256 0.4931 0.4710

384

Analytical Methods for Problems of Molecular Transport

Table F-1(b).

T*

: 2,3

0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 1.20 1.25 1.30 1.35 1.40 1.45 1.50 1.55 1.60 1.65 1.70 1.75 1.80 1.85 1.90 1.95 2.00 2.10 2.20 2.30 2.40 2.50 2.60 2.70 2.80 2.90 3.00 3.10 3.20

2.5812 2.4096 2.2577 2.1241 2.0070 1.9046 1.8150 1.7364 1.6674 1.6065 1.5526 1.5048 1.4620 1.4238 1.3893 1.3582 1.3301 1.3044 1.2810 1.2596 1.2399 1.2218 1.2050 1.1895 1.1750 1.1616 1.1490 1.1373 1.1262 1.1159 1.1061 1.0969 1.0882 1.0800 1.0722 1.0577 1.0446 1.0326 1.0216 1.0115 1.0022 0.9935 0.9854 0.9778 0.9707 0.9640 0.9577

: 2,4

2.3626 2.1835 2.0298 1.8990 1.7881 1.6940 1.6138 1.5452 1.4861 1.4349 1.3903 1.3512 1.3166 1.2860 1.2586 1.2340 1.2119 1.1919 1.1736 1.1570 1.1417 1.1276 1.1147 1.1026 1.0915 1.0810 1.0713 1.0622 1.0536 1.0456 1.0380 1.0308 1.0240 1.0175 1.0114 0.9999 0.9895 0.9799 0.9711 0.9629 0.9553 0.9482 0.9416 0.9353 0.9294 0.9238 0.9185

: 2,5

2.1704 1.9898 1.8407 1.7185 1.6181 1.5350 1.4658 1.4075 1.3581 1.3157 1.2791 1.2471 1.2191 1.1944 1.1723 1.1526 1.1348 1.1186 1.1040 1.0906 1.0783 1.0669 1.0564 1.0466 1.0376 1.0291 1.0211 1.0136 1.0066 0.9999 0.9936 0.9876 0.9820 0.9765 0.9714 0.9618 0.9529 0.9447 0.9372 0.9301 0.9235 0.9173 0.9114 0.9059 0.9007 0.8957 0.8910

: 2,6

2.0009 1.8256 1.6868 1.5764 1.4880 1.4162 1.3572 1.3081 1.2667 1.2314 1.2011 1.1747 1.1515 1.1310 1.1128 1.0964 1.0816 1.0682 1.0560 1.0447 1.0344 1.0248 1.0159 1.0077 0.9999 0.9927 0.9858 0.9794 0.9733 0.9675 0.9621 0.9569 0.9519 0.9471 0.9426 0.9341 0.9262 0.9189 0.9121 0.9057 0.8997 0.8941 0.8887 0.8837 0.8788 0.8742 0.8699

: 3,3

2.3989 2.2249 2.0799 1.9572 1.8523 1.7618 1.6833 1.6147 1.5544 1.5011 1.4538 1.4115 1.3737 1.3396 1.3089 1.2809 1.2555 1.2323 1.2110 1.1914 1.1733 1.1567 1.1412 1.1268 1.1134 1.1008 1.0891 1.0781 1.0678 1.0580 1.0488 1.0401 1.0319 1.0241 1.0167 1.0029 0.9904 0.9790 0.9684 0.9587 0.9497 0.9414 0.9336 0.9263 0.9194 0.9130 0.9069

: 4,4 2.5712 2.3939 2.2395 2.1052 1.9885 1.8871 1.7987 1.7214 1.6537 1.5940 1.5412 1.4943 1.4525 1.4150 1.3814 1.3510 1.3234 1.2984 1.2755 1.2546 1.2354 1.2176 1.2013 1.1861 1.1721 1.1589 1.1467 1.1353 1.1245 1.1145 1.1050 1.0960 1.0876 1.0796 1.0720 1.0579 1.0452 1.0335 1.0229 1.0131 1.0040 0.9956 0.9877 0.9804 0.9735 0.9670 0.9609

385

Appendix F. Omega-Integrals for Second-Order Approximation

T*

: 2,3

3.30 3.40 3.50 3.60 3.70 3.80 3.90 4.00 4.10 4.20 4.30 4.40 4.50 4.60 4.70 4.80 4.90 5.00 6.00 7.00 8.00 9.00 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100 200 300 400

0.9518 0.9461 0.9407 0.9356 0.9308 0.9261 0.9216 0.9174 0.9133 0.9093 0.9056 0.9019 0.8984 0.8950 0.8917 0.8885 0.8854 0.8824 0.8566 0.8363 0.8195 0.8052 0.7927 0.7165 0.6751 0.6469 0.6256 0.6086 0.5945 0.5825 0.5720 0.5628 0.5050 0.4735 0.4521

: 2,4

0.9135 0.9087 0.9041 0.8997 0.8955 0.8915 0.8876 0.8839 0.8803 0.8769 0.8736 0.8703 0.8672 0.8642 0.8612 0.8584 0.8556 0.8529 0.8295 0.8106 0.7948 0.7813 0.7694 0.6956 0.6552 0.6276 0.6068 0.5901 0.5764 0.5646 0.5544 0.5454 0.4891 0.4584 0.4373

: 2,5

0.8865 0.8821 0.8780 0.8740 0.8702 0.8666 0.8630 0.8596 0.8563 0.8531 0.8500 0.8470 0.8441 0.8413 0.8386 0.8359 0.8333 0.8308 0.8086 0.7906 0.7754 0.7623 0.7508 0.6786 0.6389 0.6118 0.5914 0.5751 0.5616 0.5501 0.5401 0.5312 0.4761 0.4461 0.4250

: 2,6

0.8657 0.8616 0.8578 0.8540 0.8505 0.8470 0.8437 0.8404 0.8373 0.8343 0.8314 0.8285 0.8258 0.8231 0.8205 0.8179 0.8154 0.8130 0.7916 0.7742 0.7594 0.7465 0.7353 0.6643 0.6253 0.5986 0.5785 0.5625 0.5492 0.5379 0.5280 0.5194 0.4653 0.4357 0.4144

: 3,3

0.9011 0.8956 0.8904 0.8855 0.8808 0.8763 0.8720 0.8678 0.8639 0.8600 0.8564 0.8528 0.8494 0.8461 0.8429 0.8398 0.8368 0.8339 0.8090 0.7893 0.7731 0.7593 0.7474 0.6743 0.6349 0.6081 0.5879 0.5718 0.5585 0.5471 0.5372 0.5285 0.4740 0.4444 0.4245

: 4,4 0.9551 0.9496 0.9444 0.9395 0.9348 0.9302 0.9259 0.9218 0.9178 0.9140 0.9103 0.9068 0.9033 0.9000 0.8968 0.8937 0.8907 0.8878 0.8627 0.8428 0.8264 0.8124 0.8001 0.7247 0.6834 0.6551 0.6338 0.6167 0.6025 0.5904 0.5799 0.5706 0.5122 0.4804 0.4585

386

Analytical Methods for Problems of Molecular Transport

Table F-2. A Mathematica® program that may be used to compute values of the : * integrals for selected values of the reduced temperature, T . This program is currently configured to use the Lennard-Jones (6-12) potential model and the same set of reduced temperatures reported in Table F-1, but can be readily adapted to other potential models and alternative sets of reduced temperatures.

Evaluating Omega-Integrals: INI, SKL, RVT 2005-2006 We have used the Lennard-Jones (6-12) potential model as an example. The results of this program compare well with those in: 1. Hirschfelder, J.O., Curtiss, C.F. and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954). In the program we refer to this reference as HCB. Take special note of pages 557, 558, and 1132 in this text. 2. Maitland, G.C., Rigby, M., Smith, E.B. and Wakeham, W.A., Intermolecular Forces (Oxford University Press, Oxford, 1981). In the program we refer to this reference as MRSW. Take special note of pages 538-541 in this text. The program consists of three parts: 1. Calculation of the scattering angle, chi (F), and construction of an interpolation approximation to it. 2. Calculation of the Q-integrals using chi, and again interpolation approximations to these. 3. Calculation of the omega-integrals. In the following, we have not focused on high accuracy. Our purpose has been to indicate how a simple program for calculating the omega-integrals can be constructed. We encourage the reader to explore the program. We have not included several comparative tests that were carried out during construction of the program. We thank our students Ryan M. Meyer, Zebadiah Smith, and Earl Lynn Tipton for reviewing and helping us with improvements in parts of the program. Needs["Graphics`Graphics`"]; Needs["Graphics`MultipleListPlot`"]; Needs["Graphics`Legend`"]; Off[NIntegrate::ncvb]; Off[NIntegrate::slwcon];

Appendix F. Omega-Integrals for Second-Order Approximation

387

Off[General::spell]; Off[General::spell1]; (* Define the potential: In the following we have chosen the Lennard-Jones (6-12) potential. *) Clear[ustar, rstar]; ustar[rstar_]:=4(rstar^-12 - rstar^-6); (* Construct the chi function, specialized form for the Lennard-Jones (6-12) potential. In the calculation of chi we have used a standard transformation, pstar=1/(rstar*2), which reduces the order of the polynomial in the integrand, and also changes the interval of integration to a finite interval. For calculating pstar0 we have use the Matrix Eigenvalue method of finding zeros of a polynomial. One can also use the NSolve or FindRoot functions depending upon the potential. We have not accounted for the orbiting condition in the scattering angle chi explicitly; we have smoothed over it through the choice of a coarse computational grid on bstar and estar. A more careful computation would be needed for precise work. For using the NSolve function, one can have a construct of the type: pstar0[bstar_,estar_]:=Min[Select[pstar/.NSolve[(1 – bstar^2 * pstar - 4(pstar^6 * pstar^3)/estar)m0,pstar]//N, (Im[#]m0&&Re[#]>0) &]]; *) Clear[pstar0, bstar, estar, chi, pstar00]; pstar0[bstar_,estar_]:=Min[Select[Eigenvalues[{{0, 0, 0, 0, 0, estar/4}, {1, 0, 0, 0, 0, -estar/4 * bstar^2}, {0, 1, 0, 0, 0, 0}, {0, 0, 1, 0, 0, 1}, {0, 0, 0, 1, 0, 0}, {0, 0, 0, 0, 1, 0}}]//N, (Im[#]m0&&Re[#]>0) &]]; chi[bstar_,estar_]:=Re[Pi – bstar * With[{pstar00=pstar0[bstar,estar]}, NIntegrate[Evaluate[1/Sqrt[pstar(1 - bstar^2 * pstar - 4(pstar^6 - pstar^3)/estar)]], {pstar, 0, pstar00}, MaxRecursiono30, AccuracyGoalo8, SingularityDeptho10]]]; (* We avoid the almost 0 imaginary part that drifts in because of numerics. *) (* Construct and test chiapprox which is based on interpolation. For higher precision or higher range on tstar, the reduced temperature, the limits and the grids below would need to be modified. *)

388

Analytical Methods for Problems of Molecular Transport

Clear[bstarlowlim, bstaruplim, estarlowlim, estaruplim, bstarlist, estarlist, reschi, chiapprox]; bstarlowlim=10^-5; bstaruplim=5; estarlowlim=10^-5; estaruplim=1200; bstarlist=bstarlowlim + Range[0.0,bstaruplim,0.025]; estarlist=estarlowlim + Range[0.0,estaruplim,0.025]; reschi=Outer[chi,bstarlist,estarlist]; (* The above evaluates chi on bstar, estar grid. *) chiapprox=ListInterpolation[reschi,{{First[bstarlist], Last[bstarlist]}, {First[estarlist], Last[estarlist]}}]; (* The above constructs the interpolation function for chi. *) Print[chiapprox]; InterpolatingFunction[{{0.00001,5.00001},{0.00001,1200.}},]

(* Plot (1-Cos[chi[bstar,estar]]), compare with HCB see page 557. *) Clear[chiapproxplotsHCB]; chiapproxplotsHCB=Map[Plot[1.0 Cos[Re[chiapprox[Sqrt[bstarsq], #]]], {bstarsq, o1000, PlotRangeo{{10^-10, 6}, 10^-10, 6}, PlotPointso {0, 2}}, Tickso{{0, 1, 2, 3, 4, 5, 6}, {0, 0.5, 1.0, 1.5, 2.0}}, FrameoTrue, RotateLabeloTrue, FrameLabelo{"b*2", "1-CosF"}, oIdentity] &, {0.4, 0.8, 1.0, 10.0}]; DisplayFunctiono Print["Plots of 1-CosFapprox vs b*2 for E*={0.4,0.8,1.0 and 10.0}", "\n "]; Show[GraphicsArray[Partition[chiapproxplotsHCB, 4]], DisplayFunctiono$DisplayFunction];

389

6

2 1.75 1.5 1.25 1 0.75 0.5 0.25

1Cos¢

1Cos¢

Appendix F. Omega-Integrals for Second-Order Approximation

1

2

3 4 b 2

5

6

2 1.75 1.5 1.25 1 0.75 0.5 0.25 1

2

3 4 b 2

5

6

(* For the construction of qstar[[L]][estar], we first evaluate the integrals on estar grid for given L (b in the textbook notation), and then construct an interpolation function for each L. We call the interpolation functions qstarapprox2 which are elements of the list qstarapproxlist. Lhigh is the user assigned integer, dictated by the orders of omega-integrals needed. *) Clear[qstarapprox1, qstarapprox2, qstarapproxlist, Lhigh]; qstarapprox1[L_,estar_]:=2(1-((1+(-1)^L)/(2(1+L))))^-1 * NIntegrate[(1 - Cos[chiapprox[bstar,estar]]^L) * bstar, {bstar, First[bstarlist], Last[bstarlist]}, MaxRecursion5, AccuracyGoal4]; qstarapprox2[L_]:=Interpolation[Map[{#, qstarapprox1[L,#]} &,estarlist]]; Lhigh=4; qstarapproxlist=Map[qstarapprox2[#] &,Range[1,Lhigh]]; Print[qstarapproxlist]; {InterpolatingFunction[{{0.00001,1200.}},],InterpolatingFunc tion[{{0.00001,1200.}},],InterpolatingFunction[{{0.00001,120 0.}},],InterpolatingFunction[{{0.00001,1200.}},]}

(* Tabulate and Plot qstar (L, estar). page 558. *)

Compare with HCB

Print[TableForm[Table[Map[(Flatten[{N[#], Table[qstarapproxlist[[L]][#], {L, Lhigh}]}]) &, {10^-4, 10^-3, 10^-2, 10^-1, 1, 2, 3, 4, 5, 10, 100, 1000}]], TableHeadings{None, {" E* ", " Q*(1, E*)", "Q*(2, E*)", "Q*(3, E*)", " Q*(4, E*)"}}]]; Clear[qstarplot1]; qstarplot1=Plot[Evaluate[Chop[Table[ qstarapproxlist[[L]][estar], {L, 1, Lhigh}], 10^-8]], {estar, 0.1, 4}, PlotPoints1000, PlotRangeAll, Ticks{{0, 1, 2, 3, 4}, Automatic}, FrameTrue, RotateLabelTrue, FrameLabel{" E* ", " Q*(b, E*)"}, AxesOrigin{0, 0}, PlotStyleMap[AbsoluteDashing[{5, #^2.5}] &,{1, 2, 3, 4}], PlotLegendMap[ToString, N[{1, 2, 3, 4}]], LegendPosition{1, -0.3}, LegendSize{0.8, 0.8}, LegendShadowNone, LegendLabelStyleForm["b", FontSize14]];

390

Analytical Methods for Problems of Molecular Transport

E 0.0001 0.001 0.01 0.1 1. 2. 3. 4. 5. 10. 100. 1000.

Q +1,E / 34.8393 33.1847 19.8657 5.38281 2.40659 1.46589 1.18881 1.06699 0.997701 0.855383 0.594234 0.41265

Q +2,E / 27.3578 26.2679 17.4243 6.12706 2.80343 2.14097 1.61842 1.37593 1.24426 1.01136 0.7044 0.495646

Q +3,E / 29.8057 28.5725 18.5817 6.23423 2.81556 1.94271 1.52617 1.30831 1.18465 0.95951 0.662703 0.465145

Q +4,E / 26.2607 25.2987 17.4597 6.57795 3.02483 2.37249 1.84504 1.54584 1.37286 1.06798 0.73363 0.519093

Q +b,E /

6

b

5

1.

4

2. 3.

3 4. 2

1 1

2 E

3

4

(* Compare again with HCB page 558, who have given a LogLinear Plot of Q*(1,E*), Q*(2,E*), and Q*(4,E*). *) Clear[qstarplot2]; qstarplot2=LogLinearPlot[Evaluate[Chop[Map[ qstarapproxlist[[#]][estar] &, {1, 2, 4}], 10^-8]], {estar, 0.1, estaruplim}, PlotPoints5000, PlotRangeAll, * FrameTrue, RotateLabelTrue, FrameLabel{" E ", " Q*(b, E*)"}, AxesOrigin{0.1, 0}, AspectRatio1.2, PlotStyleMap[AbsoluteDashing[{5, #}] &,{1, 5, 15}], Epilog{Text[" Q*(b, E*) for b = 1,2,4", {1.0, 5.0}], Text[" b = 1", {0.1, 2.0}], Text[" b = 2", {0.5, 1.5}], Text[" b = 4", {1.5, 1.0}]}];

391

Appendix F. Omega-Integrals for Second-Order Approximation

6

b

Q + ,E / f o r

5

b

1,2, 4

Q +b,E /

4

3

b

2

1

b

2

b

1

0.1

1

10 E

4

1 00

1000

(* Evaluate the omega-integrals using qstarapprox2[[L]][estar]. Compare with HCB pages 11261127 and MRSW page 541. We have set AccuracyGoal high to get reasonably accurate results for small Tstar. We have also evaluated the tail part of the integral approximately as per MRSW. *) Clear[omegastar]; Off[NIntegrate::ploss]; omegastar[L_,s_,tstar_]:=(NIntegrate[Chop[ qstarapproxlist[[L]][estar]] * Exp[-estar/tstar] * estar^(s+1), {estar, First[estarlist], Last[estarlist]}, MinRecursion10, MaxRecursion30, AccuracyGoal8]/(Factorial[(s+1)] * tstar^(s+2)) + Chop[qstarapproxlist[[L]][Last[estarlist]]] * NIntegrate[Exp[-estar/tstar] * estar^(s+1), {estar, Last[estarlist], Infinity}]/(Factorial[(s+1)] * tstar^(s+2))); Print["Table of Omega Integrals ", "\n "]; Do[Print["b = ", L];

392

Analytical Methods for Problems of Molecular Transport

Print[TableForm[Chop[Map[{#, omegastar[L, 1, #], omegastar[L, 2, #], omegastar[L, 3, #], omegastar[L, 4, #], omegastar[L, 5, #], omegastar[L, 6, #]} &, {0.1, 0.2, 0.3, 0.5, 0.7, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 50.0, 100.0}]], TableHeadings{None, {"T* ", "r=1 ", "r=2 ", "r=3 ", "r=4 ", "r=5 ", "r=6 "}}]], {L, 1, Lhigh}]; Table of Omega Integrals b =

1

T 0.1 0.2 0.3 0.5 0.7 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 50. 100.

b = T 0.1 0.2 0.3 0.5 0.7 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 50. 100.

r1 4.03052 3.14477 2.6577 2.06944 1.73095 1.44051 1.07552 0.950098 0.884639 0.842821 0.812858 0.789791 0.771157 0.755583 0.742235 0.575969 0.516794

r2 3.57351 2.74038 2.26061 1.70836 1.42466 1.20433 0.951483 0.864833 0.817083 0.78494 0.760923 0.741824 0.726007 0.712527 0.700797 0.546111 0.48969

r3 3.25852 2.44475 1.96826 1.47 1.24203 1.07614 0.888318 0.819374 0.779017 0.75077 0.729126 0.711621 0.696951 0.684343 0.673301 0.524777 0.470324

r4 3.01939 2.20647 1.74315 1.31043 1.12922 1.00046 0.849385 0.78935 0.752666 0.726388 0.705976 0.689324 0.675289 0.663177 0.65254 0.50829 0.455435

r5 2.82695 2.00606 1.5696 1.20216 1.05596 0.951348 0.821901 0.766952 0.732417 0.70733 0.687689 0.67159 0.657978 0.646207 0.635852 0.494927 0.443509

r6 2.6648 1.83615 1.43693 1.12668 1.0054 0.916617 0.800763 0.749051 0.715933 0.691661 0.672565 0.656868 0.643571 0.632057 0.621921 0.483736 0.433762

r1 4.56849 3.63283 3.1838 2.64092 2.27737 1.91273 1.37254 1.17317 1.07194 1.01019 0.967913 0.936631 0.912183 0.892297 0.875624 0.683549 0.616226

r2 4.05719 3.24077 2.83 2.2811 1.92119 1.59293 1.1759 1.0389 0.969964 0.926819 0.896217 0.872757 0.853822 0.837976 0.82437 0.649791 0.585166

r3 3.719 2.98294 2.57354 2.00572 1.66732 1.38953 1.07222 0.97065 0.917304 0.882326 0.85655 0.836199 0.819399 0.805097 0.79265 0.625553 0.562809

r4 3.47506 2.78939 2.35902 1.78801 1.48643 1.2588 1.01133 0.929309 0.883842 0.852839 0.829375 0.810506 0.79473 0.781177 0.769302 0.606706 0.545531

r5 3.2892 2.62689 2.16912 1.61841 1.35839 1.17244 0.971338 0.900594 0.859519 0.830714 0.808538 0.79051 0.775326 0.762215 0.750686 0.591349 0.531643

r6 3.14222 2.47914 2.00066 1.48839 1.26698 1.11283 0.942531 0.878744 0.840357 0.812914 0.791551 0.774064 0.759271 0.746461 0.735172 0.578434 0.520264

2

Appendix F. Omega-Integrals for Second-Order Approximation

b =

3

T 0.1 0.2 0.3 0.5 0.7 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 50. 100.

b = T 0.1 0.2 0.3 0.5 0.7 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 50. 100.

393

r1 4.6529 3.64653 3.12014 2.50274 2.13222 1.7864 1.29264 1.10901 1.01433 0.955952 0.915696 0.885775 0.862323 0.843213 0.827177 0.643025 0.579091

r2 4.12477 3.19922 2.70053 2.11889 1.78412 1.49032 1.11282 0.984202 0.918331 0.876791 0.847243 0.824573 0.806281 0.790986 0.777862 0.610888 0.54966

r3 3.76508 2.88425 2.40109 1.8529 1.5546 1.30887 1.01663 0.919351 0.867753 0.833868 0.808919 0.789249 0.773037 0.759258 0.747282 0.587878 0.528516

r4 3.49582 2.63981 2.17056 1.65819 1.39593 1.19164 0.95883 0.879455 0.835342 0.805309 0.782632 0.764437 0.749251 0.736227 0.724831 0.570016 0.512193

r5 3.28207 2.43999 1.98643 1.51165 1.28309 1.1125 0.920299 0.851593 0.811756 0.783903 0.762519 0.745173 0.73059 0.718018 0.706977 0.555479 0.499083

r6 3.10483 2.27233 1.83608 1.39949 1.20089 1.05667 0.892353 0.830375 0.793201 0.766717 0.746158 0.729366 0.715185 0.702922 0.692127 0.543265 0.488348

r1 4.9391 3.92571 3.43691 2.86481 2.48993 2.10675 1.50388 1.26756 1.14572 1.07167 1.02153 0.984951 0.956765 0.934145 0.915416 0.71244 0.643407

r2 4.38923 3.49768 3.05538 2.49314 2.12111 1.76422 1.27437 1.10635 1.023 0.972207 0.937115 0.910811 0.889963 0.872772 0.858177 0.677832 0.611512

r3 4.02342 3.215 2.78624 2.21284 1.85296 1.53592 1.14761 1.02299 0.960207 0.920578 0.892193 0.870234 0.852368 0.83732 0.824323 0.65302 0.588504

r4 3.75748 3.00503 2.56878 1.98837 1.65362 1.38128 1.07189 0.973386 0.921683 0.887715 0.862615 0.842741 0.826292 0.812258 0.800019 0.633697 0.570684

r5 3.55326 2.83367 2.38056 1.80744 1.5054 1.27473 1.02261 0.940009 0.894636 0.863818 0.840527 0.821803 0.806141 0.792677 0.780869 0.617927 0.556337

r6 3.39054 2.68347 2.21396 1.6625 1.3948 1.19933 0.987894 0.915372 0.873884 0.844987 0.822804 0.804792 0.789623 0.776523 0.764997 0.604644 0.544568

4

REFERENCES 1. Ferziger, J.H. and Kaper, H.G., Mathematical Theory of Transport Processes in Gases (North-Holland, Amsterdam-London, 1972). 2. Maitland, G.C., Rigby, M., Smith, E.B., and Wakeham, W.A., Intermolecular Forces: Their Origin and Determination (Oxford University Press, New York, 1981, reprinted and corrected edition, 1987). 3. Hirschfelder, J.O., Curtiss, C.F., and Bird, R.B., Molecular Theory of Gases and Liquids (John Wiley and Sons, New York, 1954).

Author Index

In this author index, bold page entries signify a citation in the references. Unbolded page entries signify a citation by name in the text. –A– Allen, M.D.; 237 Anselm, A.I.; 37 Aoki, K.; 238, 312, 333

–B– Bakanov, S.P.; 178, 238 Barantsev, R.G.; 89 Bardos, C.; 333 Barrett, J.C.; 89 Bassanini, P.; 237, 266, 287 Becker, M.; 89 Bird, G.A.; 28, 139 Bird, R.B.; 73, 331, 350, 393 Boffi, V.; 238 Bogoliubov, N.N.; 37 Breton, J.P.; 332 Brock, J.R.; 89, 140, 222, 237, 238, 331, 332 Bronshtein, I.N.; 368 Brush, S.G.; 51 Buckley, R.L.; 217, 219, 237

–C– Cadle, R.D.; 237 Campbell, P.H.; 89 Cercignani, C.; 51, 89, 237, 238, 266, 287, 331, 372 Chambré, P.L.; 139

Chandrasekhar, S.; 287 Chapman, S.; 14, 28, 51, 67, 73, 292, 331, 350 Chekalov, V.V.; 272, 287 Cheng, R.K.; 140 Cipolla, J.W.Jr.; 209 Clark, M.Jr.; 372 Coronell, D.G.; 331 Courant, R.; 372 Cowling, T.G.; 14, 28, 51, 67, 73, 292, 331, 350 Curtiss, C.F.; 73, 331, 350, 393

–D– Davies, C.N.; 237, 332 de Boer, J.; 37 Derjaguin, B.V.; 89, 178, 237, 238, 331, 332, 350 Dirac, P.A.M.; 89 Dwyer, H.A.; 238

–E– Eger, K.; 332 Epstein, P.S.; 237

–F– Ferziger, J.H.; 14, 37, 51, 73, 208, 331, 332, 350, 393 Feshbach, H.; 209, 372 Fiebig, M.; 89

396

Analytical Methods for Problems of Molecular Transport

Fisher, S.S.; 238 Flügge, S.; 286 Ford, G.W.; 51, 73 Fuchs, N.A.; 140

–G– Galkin, V.S.; 333 Ghosh, T.K.; 333 Golse, F.; 333 Goodman, F.O.; 89 Gora, E.K.; 37 Grad, H.; 286 Grosch, C.E.; 287 Gross, E.P.; 89, 155, 159, 178, 266, 268, 287 Gupta, R.N.; 331

–H– Halbritter, L.; 272, 287 Heineman, M.; 139 Hess, D.W.; 331 Hickey, K.A.; 188, 208, 209 Hidy, G.M.; 89, 179, 237, 238, 331, 350 Hilbert, D.; 372 Hirschfelder, J.O.; 73, 331, 350, 393 Huang, C.M.; 333 Huang, K.; 14, 51 Hurlbut, F.C.; 287

–I– Ikegawa, M.; 331 Ishihara, Y.; 238 Ivchenko, I.N.; 83, 84, 89, 90, 106, 139, 140, 178, 189, 196, 208, 209, 238, 261, 266, 287, 332, 333, 350

–J– Jackson, E.A.; 89, 178, 266, 287 Jacobsen, S.; 222, 237 Jeans, J.; 28 Jensen, K.F.; 331 Jew, H.; 89

–K– Kahan, T.; 372 Kaper, H.G.; 14, 37, 51, 73, 331, 350, 393 Keller, J.B.; 88 Kelly, G.E.; 89 Keng, E.Y.H.; 237 Kennard, E.H.; 51, 139, 331 Kistemaker, J.; 331 Klein, M.J.; 51 Knudsen, M.; 51

Kobayashi, J.; 331 Kogan, M.N.; 14, 73, 89, 139, 208, 243, 286, 287, 331 Kosuge, S.; 333 Kramers, H.A.; 331 Krook, M.; 243, 287 Kucherov, R.Ya.; 331 Kumar, K.; 286 Kušþer, I.; 89

–L– La Mer, V.K.; 237 Landau, L.D.; 14, 28, 37, 90, 237, 287 Lang, H.; 89, 179, 332, 372 Lea, K.C.; 237 Lees, L.; 242, 286 Levich, V.G.; 14 Lide, D.R.; 334, 350, 361 Lifshitz, E.M.; 14, 28, 37, 90, 237, 287 Liu, C.Y.; 242, 286, 287 Liu, V.C.; 89 Loeb, L.B.; 51, 331 Loyalka, S.K.; 83, 84, 89, 90, 149, 178, 179, 183, 188, 189, 196, 197, 198, 199, 208, 209, 217, 218, 219, 221, 222, 237, 238, 266, 273, 287, 303, 331, 332, 333, 361, 372

–M– Maitland, G.C.; 73, 333, 380, 393 Malinauskas, A.P.; 140 Mason, B.J.; 140 Mason, E.A.; 140, 333, 334 Maxwell, J.C.; 141, 178, 331 Millikan, R.A.; 79, 89 Monchick, L.; 334 Morse, P.M.; 209, 372 Moss, J.N.; 331 Mott-Smith, H.M.; 242, 287 Muckenfuss, C.; 287 Müller, W.J.C.; 332 Muntz, E.P.; 89

–N– Niven, W.D.; 331

–O– Orr, C.Jr.; 237

–P– Pagani, C.D.; 237, 266, 287, 372 Pang, S.C.; 89 Patterson, G.N.; 139 Phillips, W.F.; 237, 287

397

Author Index Pitaevskii, L.P.; 37 Poddoskin, A.B.; 238 Pomraning, G.C.; 372 Porodnov, B.T.; 159, 178 Present, R.D.; 331

–R– Raabe, O.G.; 237 Rabinovich, Ya.I.; 237 Ranz, W.E.; 237 Richenglas, L.E.; 331 Rideau, G.; 372 Rigby, M.A.; 73, 333, 393 Rolduguin, V.I.; 90, 178, 350 Rose, M.H.; 88 Rosenblatt, P.; 237 Rosner, D.E.; 331 Roussopolous, P.; 372

–S– Salwen, H.; 287 Savkov, S.A.; 178, 263, 287, 333 Savkov, S.V.; 90, 350 Saxena, S.C.; 334 Saxton, R.L.; 237 Schaaf, S.A.; 139 Schadt, C.F.; 237 Schefer, R.W.; 140 Schmitt, K.H.; 140, 237, 332 Scott, C.D.; 331 Semendyayev, K.A.; 368 Sengers, J.V.; 89 Sharipov, F.; 333 Shizgal, B.; 89 Shkanov, Yu.; 333 Siewert, C.E.; 333 Sigimura, T.; 287 Smirnov, L.P.; 272, 287 Smith, E.B.; 73, 333, 393 Smoluchowski, M.; 140 Sone, Y.; 238, 333 Springer, G.S.; 266, 287

Storozhilova, A.I.; 237 Storvick, T.S.; 332 Suetin, P.E.; 159, 178 Szymanski, Zd.; 139

–T– Takata, A.; 312, 333 Talbot, L.; 106, 140, 332 Tauber, R.N.; 331 Teagan, W.P.; 266, 287 Tompson, R.V.; 83, 84, 90, 189, 196, 208, 209, 287, 332, 333 Trilling, L.; 332, 372

–U– Uhlenbeck, G.E.; 37, 51, 73, 139, 155, 178, 266, 287

–V– Volkov, I.V.; 333

–W– Wachman, H.Y.; 89, 332, 372 Wakeham, W.A.; 73, 333, 393 Waldmann, L.; 97, 140, 237, 332 Wang-Chang, C.S.; 139, 140, 155, 178, 266, 287 Weaver, D.P.; 89 Williams, M.M.R.; 140, 237, 331 Willis, D.R.; 140 Wolf, S.; 331

–Y– Yalamov, Yu.I.; 89, 106, 140, 178, 238, 331, 332, 333, 350 Yamamoto, K.; 238 Yasuda, S.; 333 Yushkanov, A.A.; 238

–Z– Zhdanov, V.M.; 332 Ziering, S.; 89, 139, 155, 159, 178, 266, 268, 287

Subject Index

–A– accommodation coefficients; 77 additive quantities; 1 adjoint operator; 369 aerosol particle; 211 external parameters; 212 anti-symmetric unit tensor; 276 apse-line; 25 arbitrary intermolecular potential, first-order approximation; 67 second-order approximation; 68 auxiliary parameter; 92 auxiliary system of ordinary differential equations; 92 average gas speed; 236 Avogadro’s number; 317 azimuthal symmetry in the sphere torque problem; 273

–B– binary encounters; 31 binary gas mixtures; 289 average molecular mass; 290 Chapman-Enskog solutions, first-order, diffusion; 290 thermal conductivity; 290 viscosity; 290 second-order, diffusion; 294 thermal conductivity; 294 viscosity; 294

collision interval; 43 diffusion coefficient; 290 diffusion coefficient for a H2-N2 mixture; 316 diffusion coefficients for different order approximations; 317 diffusion-slip coefficient by Maxwell method; 312 distribution functions; 289 flow through a capillary, constant temperature with unequal initial mixtures; 323 temperature gradient with equal initial mixtures; 328 hydrostatic pressure; 290 mass density; 290 mean free path; 42 mean mass velocity; 313 mean velocities of constituents; 313, 315 mole fractions of constituents; 290 molecular masses of constituents; 290 number densities of constituents; 290 number flux vector (current); 314 planar problems; 300, 306 diffusion-slip coefficient; 312 limiting case of a simple gas; 308 Loyalka method; 303 Maxwell method; 302 slip coefficients; 307

400

Analytical Methods for Problems of Molecular Transport

relationship between mean velocity and mean mass velocity; 316 slip coefficient; 304 slip-flow coefficient using the firstorder Chapman-Enskog approximation; 322 tangential momentum accommodation coefficients; 312 thermal conductivity coefficien,t first-order Chapman-Enskog approximation; 318 thermal conductivity coefficient, second-order Chapman-Enskog approximation; 320 thermal diffusion ratio; 290 transport coefficients; 291, 309, 310 two vessels connected by a slot; 47 uniform steady-state; 42 viscosity coefficient, first-order Chapman-Enskog approximation; 318 second-order Chapman-Enskog approximation; 319 Boltzmann collision integral; 16, 34 Boltzmann constant; 10, 317 Boltzmann equation; 15 condition of applicability; 35 difficulties in solution; 239 dimensionless form; 53 infinite, unsteady, non-uniform gas; 53 integral form; 182 Boltzmann H-theorem; 39 boundary conditions, condensable gas; 79 non-condensable gas; 77 ordinary moment form; 168 reflected molecules of the same liquid vapor; 86 small Knudsen number; 82 solid impenetrable wall; 86 boundary dispersion kernel; See dispersion kernel boundary values; 77 bracket integrals, analytical expressions; 247 arbitrary intermolecular potentials; 67 commutative property; 151 curvilinear geometry; 351

cylindrical geometry, special function approximations; 361 special function of the first kind; 355 special function of the second kind; 358 independent of intermolecular potential; 271 linearized problems; 248 method of calculation; 159 notations; 302 of Gross-Ziering; 347 planar geometry; 251 first-order Chapman-Enskog solution for viscosity; 338 with one discontinuous sign function; 348 with Sonine polynomials; 335 with two discontinuous sign functions; 339 special functions, approximate expressions; 360 spherical geometry, polynomial expansion method; 363, 367 special function approximations; 361 special function of the first kind; 351

–C– capillary flow between vessels, isothermal gas mixture; 130 isothermal simple gas; 130 non-isothermal gas mixture; 127 non-isothermal simple gas; 129 Cauchy problem; 109 center of mass; 336 central force field; 25 Chapman-Enskog correction to the distribution function; 142 Chapman-Enskog method; 91 Chapman-Enskog solutions, diffusion in a gas mixture; 290 first-order; 289 polynomial expansion, thermal conductivity; 71 viscosity; 72 second coefficients; 70

401

Subject Index second-order; 68, 294 thermal conductivity; 71 viscosity; 72 thermal conductivity in a gas mixture; 290 viscosity in a gas mixture; 290 Chapman-Enskog theory in the limit of a simple gas; 60 characteristic, density; 243 in linearized problems; 247 dimension; 75 mean velocity; 243 system; 92 temperature; 243 in linearized problems; 247 characteristics as molecular trajectories; 92 classical mechanics; 1 Clausius-Clapeyron equation; 87, 88 closure of moment systems; 19 coaxial cylinders, heat flux between; 133, 226 temperature distribution between; 226 collision interval; 43 collision operator; 34, 71 equal to zero; 71 self-adjointness; 151 collisional invariants; 241 heat conduction from a sphere; 254 communicable energy; 9 concentric spheres, heat flux between in a binary gas mixture; 225 heat flux between in a monatomic gas; 134 temperature distribution between in a binary gas mixture; 225 condensation coefficient; 79 condensation on a spherical droplet; 106 cone of influence; 94, 242 cylindrical geometry; 244, 245 sphere; 93 spherical geometry; 243 conservation of energy; 71 conservation of momentum; 71 conservation of tangential momentum; 146 continuity equation; 241

continuum equations; 19 in the slip-flow regime; 211 continuum medium; 18 continuum regime; 76 coordinates in velocity space, cylindrical vs. Cartesian; 246 spherical vs. Cartesian; 246 correction terms to the distribution, integral equations for first- and second-orders; 55 Couette flow, Gross-Ziering method; 171 Loyalka method; 171 Maxwell method; 169 Cunningham slip correction factor; 217 current density from a metal surface with a work function; 49 curvilinear coordinates; 239, 240 curvilinear geometry; 271

–D– de Broglie wavelength; 12 definite integrals; 373 in boundary problems; 374 in non-linear transport problems; 378 in the second-order Chapman-Enskog solution; 376 degree of rarefaction; 75 degrees of freedom; 212 density; 2 density of a drop; 108 diameter of a molecule; 336 differential scattering cross section for rigid-sphere molecules; 336 diffusely reflected molecules; 46 diffusion coefficient, binary gas mixture; 291 different order approximations; 317 H2-N2 gas mixture; 316 diffusion velocities; 316 diffusion-pressure effect; 131 diffusion-slip coefficient, accuracy of Maxwell method; 322 binary gas mixture; 305, 312 dimensionless radial vector; 241 dimensionless thermal force; 221 NaCl aerosols in Ar; 222 Dirac delta function; 81, 162 discontinuous Maxwellian distribution function,

402

Analytical Methods for Problems of Molecular Transport

curvilinear geometries; 242 planar geometry; 242 disk in a rarefied gas, thermophoretic force on; 120 dispersion kernel; 77, 80, 81, 85 dispersion operator; 252 distribution function; 3, 11 Chapman-Enskog correction; 142 cylindrical coordinates; 44 discontinuous in velocity space; 94, 142 first-order approximation; 55 free-molecular in the full velocity space; 94 full corrections to; 81 in a linear expansion; 54 integral equation for the second-order correction; 56 kinetic energy of translation in a Maxwellian gas; 45 Maxwellian; 41 moment of; 4 of incident molecules; 78 of reflected molecules; 78 rebounding molecules; 87 rebounding molecules with evaporation/ condensation; 88 second-order approximation; 56 second-order conditions; 56 slip-flow problem; 142 surface of discontinuity; 94 thermal-creep problem; 142 drag force, in the free-molecular regime; 217 on a sphere in a monatomic gas; 112 drag ratio; 217

–E– effective molecular diameter; 67 effusion; 47 electron gas, flux density; 48 energy accommodation coefficient; 79, 87, 88 energy conservation; 21 energy parameter of a potential model; 380 energy transport equation; 241 equation of state for a perfect gas; 10 error function; 378 evaporation coefficient; 79

expansion in powers of a small parameter; 53 expansion into a vacuum, from a spherical surface; 138 from an infinite planar surface; 137 exponential integrals; 188 external force; 15 external non-uniformity of a gas; 193

–F– first variation; 371 first-order Chapman-Enskog approximation, slip-flow coefficient for a binary gas mixture; 322 thermal conductivity coefficient for a binary gas mixture; 318 viscosity coefficient for a binary gas mixture; 318 flow between two parallel surfaces; 236 flow in a cylindrical tube; 230, 231 flow through a capillary, binary gas mixtures, driving forces; 329 pressure difference; 323 relative density; 323 steady-state pressure difference; 328 diffusion velocities; 325 mean mass velocity; 325 net mass transfer in a closed loop; 233 flux density vector; 6 flux of the energy; 79 flux of the number of molecules; 79 force, particle in a uniform flow; 96 round disk in a uniform flow; 116 fourth moment equation for heat conduction from a sphere; 254 free-molecular regime; 76, 91 full-range moment equations, apparent viscosity coefficient; 172 Couette flow; 172 mean velocity; 172 pressure tensor component; 172 two-moment approach; 172 functional of a trial function; 371

–G– gamma function; 62 gas flow,

403

Subject Index along a plane wall; 82 number density around a sphere in the free-molecular regime; 100 number density at large distance from a small slot; 123 number density at the exit of a round hole; 122 past a sphere in the free-molecular regime; 99 generating functions, Hermite polynomials; 363 Sonine polynomials; 363 geometry of an encounter; 21 Gross-Ziering bracket integrals and their relationship to Chapman-Enskog bracket integrals; 159 Gross-Ziering method; 155 apparent viscosity coefficient; 171 Couette flow; 171 mean velocity profile; 171 pressure tensor component; 171 growth rate of a drop; 109

–H– half-range moment method; 141, 155 four-moment approximation; 155 mean velocity of the gas; 161 polynomials in velocity space; 156 thermal-creep problem; 166 velocity-slip problem; 155 half-space integral notation; 376 heat conduction between two parallel plates; 265 ‘smoothed’ distribution function method; 268 half-range moment method; 268 polynomial expansion method; 268 heat conduction from a sphere; 254, 271 free-molecular heat flux; 258 polynomial expansion method; 264 surface heat flux; 258 heat flux, between coaxial cylinders; 226 Chapman-Enskog solution as a molecular property; 280 domain of applicability of continuum formula; 285 domain of applicability of freemolecular formula; 285 four-moment approach; 279

between concentric spheres; 225 ‘smoothed’ distribution function method; 278 domain of applicability of continuum formula; 286 domain of applicability of freemolecular formula; 286 four-moment approach; 277 from a sphere; 224 per unit length, inner cylinder; 280, 281 Heaviside step function; 303 Hermite polynomials; 261 generating function; 363 Hilbert space; 145 hydrostatic pressure; 8

–I– impenetrable surface condition; 7 infinite gas in an unsteady, non-uniform state; 53 integral boundary conditions for heat and tangential momentum transport problems; 253 integral boundary model; 252 integral operators in Kinetic Theory; 369 planar problems; 369 integral properties of a gas; 181 integral theorem; 57 intermolecular potential models, Lennard-Jones (6-12); 67 parameters of; 67, 311 invariant quantities; 18 isothermal-creep coefficient; 83, 164 isothermal-slip coefficient in the sphere drag problem; 216

–J– Jacobian; 42, 111

–K– kinematic viscosity; 84, 146, 213 kinetic energy; 9 peculiar motion; 9 volume element; 9 Kinetic Theory of Gases; 5, 11 regimes of; 76 Knudsen iteration; 76 Knudsen layer; 181 boundary conditions for NavierStokes equations; 82 Knudsen number; 75

404

Analytical Methods for Problems of Molecular Transport

–L– Lagrange system; 92, 110 latent heat of evaporation/condensation; 87, 88 Legendre polynomials; 104 Leibnitz’ formula; 365 Lennard-Jones (6-12) potential; 202 linearized boundary condition; 96 linearized transport problems; 181 Liouville theorem; 15, 34 local Maxwellian distribution function; 55 Loyalka method; 149 analytical formulas, pressure jump, monatomic condensable gas over liquid; 176 monatomic non-condensable gas; 174 temperature-jump, monatomic condensable gas over liquid; 176 monatomic non-condensable gas; 174 apparent viscosity coefficient; 171 boundary-jump effects; 175 Couette flow; 171 generalization of the Maxwell method; 150 mean velocity profile; 171 pressure tensor component; 171 slip coefficients; 151 slip velocities; 151 temperature-jump problem; 152

–M– macroscopic quantities; 11 free-molecular regime; 98 mass density; 2 mass of the Earth’s atmosphere; 50 Maxwell integral transport equations; 240 heat conduction between plates; 266 Maxwell method; 141 apparent thermal conductivity coefficient; 173 apparent viscosity coefficient; 169 Couette flow; 169 form of the distribution function; 145 main deficiency; 149 mean velocity far from the wall; 146 mean velocity profile; 169

pressure tensor component; 169 slip factors; 147 slip-flow problem; 142 temperature distribution between parallel plates; 173 temperature-jump problem; 147 thermal-creep problem; 142 Maxwellian boundary model; 77, 183, 274 dispersion kernel; 85 heat conduction between two parallel plates; 267 heat conduction from a sphere; 256 number density of fictitious gas; 78 reflected distribution function; 85 temperature of fictitious gas; 78 Maxwellian distribution function; 41 as a first approximation; 55 local; 55 Maxwellian gas, distribution function of kinetic energy of translation; 45 number of molecules impinging upon a unit surface element in a given angular range; 45 number of molecules impinging upon a unit surface element in a given velocity range; 45 Maxwellian velocity distribution; 40 mean distance between collisions; 43 mean free path; 36, 43, 65 from thermal conductivity coefficient; 305 from viscosity coefficient; 305 in N2 at STP; 44 Knudsen number; 75 other representation; 146 simple gas; 44 two-dimensional, rigid-sphere gas; 37 mean heat velocity; 13 mean time between collisions; 43 mean value of a molecular property; 110 mean velocity; 4, 11, 167 equilibrium gas; 46 fictitious value; 82 molecular beam; 46 of a gas near a moving wall; 82 profile; 200 real; 82

405

Subject Index thermal-creep in a tube; 232 uniform flow past a sphere; 113 variational method; 187 mechanical equilibrium condition, thermal-creep in a cyindrical tube; 231 methods of closure, polynomial expansions in velocity space; 241 special choice for the distribution function; 242 metric coefficients; 240 cylindrical geometry; 240 spherical geometry; 240 microscopic boundary conditions; 252 mobile operator; 18 model of diffuse reflection; 79 modified two-stream Maxwellians; 243 molecular encounters; 19 summational invariants of; 55 molecular flux, in a given angular range; 45 in a given velocity range; 45 molecular interaction; 19 molecular property; 4, 16 conserved during encounters; 18 fluxes; 5, 11 mean value of; 4, 98 rate of variation; 34 moment equations; 16, 19 features of boundary conditions; 162 moment of the collision integral; 34 moment of the collision operator; 256 moments of discontinuous distribution functions; 245 momentum conservation; 20 monatomic gas; 18 mutual potential energy; 25

–N– Navier-Stokes equations; 76 as moment equations in the Maxwellian analysis; 143 near free-molecular regime; 76 net flux; 6 net mass transport, in a cylindrical tube; 230 thermal-creep in a cyindrical tube; 231 Newton’s equations of motion; 27 non-dimensionalization; 53 non-stationary gas flows; 109

normal coordinate; 182 number density; 2, 3, 11 saturated vapor; 107 simple gas in a rotating cylinder; 50 number flow of molecules; 6 numerical quadratures; 380

–O– omega-integrals, first-order approximation; 379 with arbitrary intermolecular potential; 67 gas mixtures; 379 reduced; 68, 380 rigid-sphere molecules; 68, 379 second-order approximation; 379 tables of; 382

–P– parallel plates, frictional force on due to horizontal motion; 125 gas flow between; 126 gas temperature between; 131 parameter of non-uniformity; 54 parameters of an encounter; 22 peculiar velocity, mean value of; 4 perfect gas; 10 phase space; 4 phase volume element; 27 planar boundary transport problems; 181 Planck constant; 13 plate in a steady gas flow, drag when angled to flow; 118 drag when parallel to flow; 117 point-mass hypothesis; 12 polar coordinates; 261, 349, 352 polynomial expansion method; 260 cylindrical geometry; 262 planar geometry; 262 spherical geometry; 260 potential flow field for a sphere; 114 pressure; 6, 11 pressure difference due to thermal-creep in a tube; 231 pressure distribution; 8 pressure tensor; 6 general expression; 66

–R– radial momentum equation; 241 rarefied gases; 1

406

Analytical Methods for Problems of Molecular Transport

reduced heat flux; 266 experiment vs. theory; 270 ratio; 265, 269 reduced impact parameter; 380 reduced intermolecular distance; 380 reduced intermolecular potential energy; 380 reduced omega-integrals; 380 reduced relative kinetic energy; 380 reduced temperature; 68, 202, 285, 380 reduced torque on a rotating sphere; 277 reflected distribution function, condensible gas; 80 correction to; 81 reflection, diffuse; 77 operator; 184 specular; 77 relative fluctuation of an additive quantity; 12 relative molecular velocity; 336 relative motion of two molecules; 25 resistance force on a moving disk; 227 Reynolds number; 213 Richardson thermionic emission formula; 49 rigid-sphere gas, model; 19 molecular mean free path; 36 temperature independence; 203 two-dimensional; 37 root mean square velocity of a molecular beam; 46 rotating gas, equivalence to an external force field; 50 number density distribution; 50 round disk in a gas, force on from different temperature sides; 119 force on in a uniform flow; 116 number density behind in a flow; 121, 124

–S– saturated vapor density at a surface; 80 scalar product; 185, 301 scattering cross section; 29 differential; 29 integral; 30

rigid spheres; 30 total; 30 scattering solid angle; 31 second variation; 372 second-order Chapman-Enskog solutions; 306 thermal conductivity coefficient for a binary gas mixture; 320 viscosity coefficient for a binary gas mixture; 319 second-order correction, as a summational collision invariant; 56 homogeneous term; 56 non-homogeneous term; 57 self-adjoint operator; 369 sign function; 156 simple gas, in a rotating cylinder; 50 steady-state number density variation in a force field; 49 two vessels connected by a slot; 47 uniform steady-state; 39, 41 slip coefficient; 165, 201 binary gas mixture; 322 isothermal; 83 Lennard-Jones (6-12) potential; 202 second-order Maxwell method; 203 variational method; 187 slip factor; 311 experiment vs. theory; 312 slip problem geometry; 142 slip velocity; 82 slip-flow regime; 76, 211 average gas speed between two parallel surfaces; 236 boundary conditions; 219 calculation error in thermal conductivity coefficient; 225 drag force on parallel round disks; 227 flow in a cylindrical tube, isothermal pressure difference; 231 net mass transport; 230 velocity distribution; 230 flow in capillary tubes; 233 heat flux, between coaxial cylinders; 226 between concentric spheres; 225 from a sphere; 224

407

Subject Index temperature distribution, around a sphere; 224 between coaxial cylinders; 226 between concentric spheres; 225 thermal conductivity coefficient, around a heated sphere; 224 calculation error in heated wire in a cylinder method; 227 heated wire in a cylinder method; 227 thermal-creep in a cylindrical tube; 231, 232 torque, on a sphere; 234 on an infinite cylinder; 235 slip-flow results; 188 half-range moment method; 189 ILT boundary model; 189 Maxwellian boundary model; 189 Loyalka method; 189 variational method; 189 smooth molecules; 20 smoothed distribution function method; 259 collision operator; 259 corrections to density and temperature; 259 Sonine polynomials; 62, 261 generating function; 363 orthonormal conditions; 261 special functions; 252 cylindrical geometry; 361 spherical geometry; 263 specific heat per unit mass at constant volume in a monatomic gas; 68 speed ratio; 117 sphere drag problem; 95, 211, 214 dimensional analysis; 213 experiment vs. theory; 218, 219 external parameters; 212 satisfied conditions; 212 sphere in a monatomic gas, drag force; 112 heat flux; 115 temperature distribution; 115 temperature of; 111 spherical coordinates; 337 standard collision operator; 56, 145 standard temperature and pressure; 1

stationary functional; 371 STP; 1 summational invariants; 55 supersaturation; 107 surface element of a sphere; 276 surface tension coefficient; 108 system of moment equations, method of closure; 241

–T– tangential momentum, accommodation coefficient; 78, 117, 312 conservation of; 146 temperature; 9, 11 of a gas near a wall; 84 of a particle; 104 of a sphere in a monatomic gas; 111 of a surface; 79 real; 84 temperature defect at the wall, Loyalka method; 206 Maxwell method, first-order; 205 second-order; 205 temperature distribution, around a sphere; 224 between coaxial cylinders; 226 between concentric spheres; 225 inside a particle; 104 temperature-jump; 83, 84, 202, 220 Lennard-Jones (6-12) potential; 202 Loyalka method and general boundary conditions; 206 Maxwell method and general boundary conditions; 204 Maxwell method, second-order; 204 thermal conduction equation, boundary conditions; 220 first-order Chapman-Enskog solution; 62 rigid-sphere molecules; 63 thermal conductivity; 212 binary gas mixture; 291, 318, 320 first-order approximation; 68 first-order Chapman-Enskog theory; 147 heat conduction from a sphere; 258 heat loss from a hot sphere method; 224, 225

408

Analytical Methods for Problems of Molecular Transport

heated wire in a cylinder method; 226 calculation error; 227 monotomic gas of rigid spheres; 66 of a particle; 104 second-order approximation; 69 rigid-sphere molecules; 71 thermal diffusion ratio; 315, 328 thermal flux; 11 thermal flux vector; 10 thermal force; 106 dimensionless; 221 one component; 103 thermal force problem; 211, 218 dimensional analysis; 213 experiment vs. theory; 222, 223 external parameters; 212 geometry; 102 satisfied conditions; 212 third moment equation; 213 thermal transpiration effect; 311 thermal-creep; 84, 220 binary gas mixture; 305 geometry; 85 in a cylindrical tube, mean velocity variation; 232 pressure difference; 231 influence of boundary models; 168 Lennard-Jones (6-12) potential; 202 second-order Maxwell method; 203 velocity; 84 thermal-creep results; 196 mean velocity profile half-range moment method; 199 thermal-creep coefficient, half-range moment method, ILT boundary model; 196 integral boundary model; 196 Loyalka method; 196 variational method; 196 velocity defect, half-range moment method, ILT boundary model; 196 integral boundary model; 196 Loyalka method; 196 variational method; 196 thermal-slip; See thermal-creep thermophoresis; 101, 218 corrections to; 223 time duration of an encounter; 35 torque problem; 272

torque, on a rotating body, flat disk; 135 infinite cylinder; 136 sphere; 137 on a sphere; 234, 273 arbitrary potential model; 284 in CO2; 284 in N2; 284 Lennard-Jones (6-12) potential; 284 Maxwellian boundary model; 274 moment system; 274 net; 276 on an infinite cylinder; 235 on two coaxial cylinders; 283 on two concentric spheres; 281 per unit length; 136 coaxial cylinders; 284 slip-flow regime; 235 transport coefficients; 70 functions for calculating second-order; 70 functions for Lennard-Jones (6-12) potential; 70 monatomic gas; 68 transport equation; 17 trial function; 371 two vessels connected by a slot, gas mixture density ratio; 47 kinetic equilibrium; 47 two vessels connected by a tube, pressure difference due to thermalcreep; 231, 232 velocity distribution in the tube; 231 two-dimensional gas; 28 two-sided Maxwellian distribution function; 242 in linearized problems; 247

–U– uniform flow past a sphere, mean velocity distribution; 113 potential flow field; 114 pressure; 114 temperature distribution; 114 uniform steady-state of a gas; 41

–V– variational method; 181 planar geometry; 181

409

Subject Index slip-flow problem; 183 thermal-creep coefficient; 195 thermal-creep problem; 192 velocity defect; 195 variational parameter; 186 variational principle; 370 velocity defect; 164, 168 at the wall; 190 influence of boundary models; 168 Maxwell method, slip-flow problem; 204 thermal-creep problem; 204 slip-flow problem, integral representation of the distribution function; 207, 208 thermal-creep problem, integral representation of the distribution function; 208 Loyalka method; 208 variational method; 187 velocity distribution, due to thermal-creep; 231 in a cylindrical tube; 230 Maxwellian; 40 velocity,

actual; 3 mean; 2 peculiar; 3 relative; 3 velocity-slip coefficient for a binary gas mixture; 305 viscosity; 212 binary gas mixture; 291, 318, 319 first-order approximation; 68 first-order Chapman-Enskog theory; 147 monotomic gas of rigid spheres; 66 second-order approximation; 69 rigid-sphere molecules; 73 viscosity equation, first-order Chapman-Enskog solution; 64 rigid-sphere molecules; 65 volume element; 4

–W– work function; 49

–Z– zone of influence for a cylinder; 133

Our partners will collect data and use cookies for ad personalization and measurement. Learn how we and our ad partner Google, collect and use data. Agree & close