THE

MATHEMATICAL THEORY
OF
NON-UNIFORM GASES
A N A C C O U N T O F THE KINETIC THEORY
O F VISCOSITY, THERMAL C O N D U C T I O N A N D
D I F F U S I O N IN GASES

SYDNEY CHAPMAN, F.R.S.

Geophysical Institute, College, Alaska
National Center for Atmospheric Research, Boulder, Colorado
AND
T. G. COWLING, F.R.S.
Professor of Applied Mathematics
Leeds University

THIRD EDITION
P R E P A R E D IN C O - O P E R A T I O N WITH
D. B U R N E T T

CAMBRIDGE
UNIVERSITY PRESS
Published by the Press Syndicate of the University of Cambridge
The Pitt Building, Trumpington Street, Cambridge CB2 IRP
40 West 20th Street, New York, NY 10011-4211 USA
10 Stamford Road, Oakleigh, Melbourne 3166, Australia

Copyright Cambridge University Press 1939, 1932

© Cambridge University Press 1970
Introduction © Cambridge University Press 1990

First published 1939

Second edition 1932
Third edition 1970
Reissued as a paperback with a Foreword by Carlo CerclgnanI
in the Cambridge Mathematical Library Series 1990
Reprinted 1993

ISBN 0 321 40844 X paperback

Transferred to digital printing 1999

 CONTENTS

Foreword page vii

Preface xiii

Note regarding references xiv

Chapter and section titles XV

List of diagrams XX

List of symbols xxi

Introduction I

Chapters I - I Q 10-406

Historical Summary 407

Name index 4"

Subject index 415

References to numerical data for particular gases

(simple and mixed) 4*3

M
 FOREWORD

The atomic theory of matter asserts that material bodies are made up of small
particles. This theory was founded in ancient times by Democritus and
expressed in poetic form by Lucretius. This view was challenged by the
opposite theory, according to which matter is a continuous expanse. As
quantitative science developed, the study of nature brought to light many
properties of bodies which appear to depend on the magnitude and motions
of their ultimate constituents, and the question of the existence of these tiny,
invisible, and immutable particles became conspicuous among scientific
enquiries.
As early as 1738 Daniel Bernoulli advanced the idea that gases are formed
of elastic molecules rushing hither and thither at large speeds, colliding and
rebounding according to the laws of elementary mechanics. The new idea,
with respect to the Greek philosophers, was that the mechanical effect of the
impact of these moving molecules, when they strike against a solid, is what
is commonly called the pressure of the gas. In fact, if we were guided solely
by the atomic hypothesis, we might suppose that pressure would be produced
by the repulsions of the molecules. Although Bernoulli's scheme was able to
account for the elementary properties of gases (compressibility, tendency to
expand, rise of temperature in a compression and fall in an expansion, trend
toward uniformity), no definite opinion could be formed until it was investi-
gated quantitatively. The actual development of the kinetic theory of gases
was, accordingly, accomplished much later, in the nineteenth century.
Although the rules generating the dynamics of systems made up of molecules
are easy to describe, the phenomena associated with this dynamics are not so
simple, especially because of the large number of particles: there are about
2X7X IO'9 molecules in a cubic centimeter of a gas at atmospheric pressure
and a temperature of 0 °C.
Taking into account the enormous number of particles to be considered, it
would of course be a perfectly hopeless task to attempt to describe the state
of the gas by specifying the so-called microscopic state, i.e. the position and
velocity of every individual particle, and we must have recourse to statistics.
This is possible because in practice all that our observation can detect is
changes in the macroscopic state of the gas, described by quantities such as
density, velocity, temperature, stresses, heat flow, which are related to the
suitable averages of quantities depending on the microscopic state.
J. P. Joule appears to have been the first to estimate the average velocity
of a molecule of hydrogen. Only with R. Clausius, however, the kinetic theory
of gases entered a mature stage, with the introduction of the concept of mean
free-path (1858). In the same year, on the basis of this concept, J. C. Maxwell
developed a preliminary theory of transport processes and gave an heuristic
derivation of the velocity distribution function that bears his name. However,
[vii]
viii FOREWORD
he almost immediately realized that the mean free-path method was inadequate
as a foundation for kinetic theory and in 1866 developed a much more accurate
method, based on the transfer equations, and discovered the particularly simple
properties of a model, according to which the molecules interact at distance
with a force inversely proportional to the fifth power of the distance (nowadays
these are commonly called Maxwellian molecules). In the same paper he gave
a better justification of his formula for the velocity distribution function for
a gas in equilibrium.
With his transfer equations, Maxwell had come very close to an evolution
equation for the distribution, but this step must be credited to L. Boltzmann.
The equation under consideration is usually called the Boltzmann equation
and sometimes the Maxwell-Boltzmann equation (to acknowledge the impor-
tant role played by Maxwell in its discovery).
In the same paper, where he gives an heuristic derivation of his equation,
Boltzmann deduced an important consequence from it, which later came to
be known as the //-theorem. This theorem attempts to explain the irreversibil-
ity of natural processes in a gas, by showing how molecular collisions tend to
increase entropy. The theory was attacked by several physicists and
mathematicians in the 1890s, because it appeared to produce paradoxical
results. However, a few years after Boltzmann's suicide in 1906, the existence
of atoms was definitely established by experiments such as those on Brownian
motion and the Boltzmann equation became a practical tool for investigating
the properties of dilute gases.
In 1912 the great mathematician David Hilbert indicated how to obtain
approximate solutions of the Boltzmann equation by a series expansion in a
parameter, inversely proportional to the gas density. The paper is also repro-
duced as Chapter XXII of his treatise entitled Grundzige einer allgemeinen
Theorie der linearen Integralgleichungen. The reasons for this are clearly stated
in the preface of the book ('Neu hinzugefugt habe ich zum Schluss ein Kapitel
iiber kinetische Gastheorie. [...] erblicke ich in der Gastheorie die glazendste
Anwendung der die Auflosung der Integralgleichungen betreffenden
Theoreme').
In 1917, S. Chapman and D. Enskog simultaneously and independently
obtained approximate solutions of the Boltzmann equation, valid for a
sufficiently dense gas. The results were identical as far as practical applications
were concerned, but the methods differed widely in spirit and detail. Enskog
presented a systematic technique generalizing Hilbert's idea, while Chapman
simply extended a method previously indicated by Maxwell to obtain transport
coefficients. Enskog's method was adopted by S. Chapman and T. G. Cowling
when writing The Mathematical Theory of Non-uniform Gases and thus came
to be known as the Chapman-Enskog method.
This is a reissue of the third edition of that book, which was the standard
reference on kinetic theory for many years. In fact after the work of Chapman
and Enskog, and their natural developments described in this book, no essential
 FOREWORD ix
progress in solving the Boltzmann equation came for many years. Rather the
ideas of kinetic theory found their way into other fields, such as radiative
transfer, the theory of ionized gases, the theory of neutron transport and the
study of quantum effects in gases. Some of these developments can be found
in Chapters 17 and 18.
In order to appreciate the opportunity afforded by this reissue, we must
enter into a detailed description of what was the kinetic theory of gases at the
time of the first edition and how it has developed. In this way, it will be clear
that the subsequent developments have not diminished the importance of the
present treatise.
The fundamental task of statistical mechanics is to deduce the macroscopic
observable properties of a substance from a knowledge of the forces of
interaction and the internal structure of its molecules. For the equilibrium
states this problem can be considered to have been solved in principle; in fact
the method of Gibbs ensembles provides a starting point for both qualitative
understanding and quantitative approximations to equilibrium behaviour. The
study of nonequilibrium states is, of course, much more difficult; here the
simultaneous consideration of matter in all its phases - gas, liquid and solid
- cannot yet be attempted and we have to use different kinetic theories, some
more reliable than others, to deal with the great variety of nonequilibrium
phenomena occurring in different systems.
A notable exception is provided by the case of gases, particularly monatomic
gases, for which Boltzmann's equation holds. For gases, in fact, it is possible
to obtain results that are still not available for general systems, i.e. the
description of the thermomechanical properties of gases in the pressure and
temperature ranges for which the description suggested by continuum
mechanics also holds. This is the object of the approximations associated with
the names Maxwell, Hilbert, Chapman, Enskog and Burnett, as well as of the
systematic treatment presented in this volume. In these approaches, out of all
the distribution functions / which could be assigned to given values of the
velocity, density and temperature, a single one is chosen. The precise method
by which this is done is rather subtle and is described in Chapters 7 and 8.
There exists, of course, an exact set of equations which the basic continuum
variables, i.e. density, bulk velocity (as opposed to molecular velocity) and
temperature, satisfy, i.e., the full conservation equations. They are a con-
sequence of the Boltzmann equation but do not form a closed system, because
of the appearance of additional variables, i.e. stresses and heat flow. The same
situation occurs, of course, in ordinary continuum mechanics, where the system
is closed by adding further relations known as 'constitutive equations'. In the
method described in this book, one starts by assuming a special form for /
depending only on the basic variables (and their gradients); then the explicit
form of f is determined and, as a consequence, the stresses and heat flow are
evaluated in terms of the basic variables, thereby closing the system of
conservation equations. There are various degrees of approximation possible
X FOREWORD
within this scheme, yielding the Euler equations, the Navier-Stokes equations,
the Burnett equations, etc. Of course, to any degree of approximation, these
solutions approximate to only one part of the manifold of solutions of the
Boltzmann equation; but this part turns out to be the one needed to describe
the behaviour of the gas at ordinary temperatures and pressures. A byproduct
of the calculations is the possibility of evaluating the transport coefficients
(viscosity, heat conductivity, diffusivity,...) in terms of the molecular param-
eters. The calculations are by no means simple and are presented in detail in
Chapters 9 and 10. These results are also compared with experiment (Chapters
12, 13 and 14).
In 1949, H. Grad wrote a paper which became widely known because it
contained a systematic method of solving the Boltzmann equation by expanding
the solution into a series of orthogonal polynomials. Although the solutions
which could be obtained by means of Grad's 13-moment equations (see section
15.6) were more general than the 'normal solutions' which could be obtained
by the Chapman-Enskog method, they failed to be sufficiently general to
cover the new applications of the Boltzmann equation to the study of upper
atmosphere flight. In the late 1950s and in the 1960s, under the impact of the
problems related to space research, the main interest was in the direction of
finding approximate solutions of the Boltzmann equation in regions having a
thickness of the order of a mean free-path. These new solutions were, of
course, beyond the reach of the methods described in this book. In fact, at
the time when the book was written, the next step was to go beyond the
Navier-Stokes level in the Chapman-Enskog expansion. This leads to the
so-called Burnett equations briefly described in Chapter 15 of this book. These
equations, generally speaking, are not so good in describing departures from
the Navier-Stokes model, because their corrections are usually of the same
order of magnitude as the difference between the normal solutions and the
solutions of interest in practical problems. However, as pointed out by several
Russian authors in the early 1970s, there are certain flows, driven by tem-
perature gradients, where the Burnett terms are of importance. For this reason
as well for his historical interest, the chapter on the Burnett equations still
retains some importance.
Let us now briefly comment on the chapters of the book, which have not
been mentioned so far in this foreword. Chapters 1-6 are of an introductory
nature; they describe the heavy apparatus that anybody dealing with the kinetic
theory of gases must know, as well as the results which can be obtained by
simpler, but less accurate tools. Chapter 11 describes a classical model for
polyatomic gases, the rough sphere molecule; this model, although not so
accurate when compared with experiments, retains an important role from a
conceptual point of view, because it offers a simple example of what one should
expect from a model describing a polyatomic molecule. Chapter 16 describes
the kinetic theory of dense gases; although much has been done in this field,
the discussion by Chapman and Cowling is still useful today.
 FOREWORD zl
Where is kinetic theory going today? The main recent developments are in
the direction of developing a rigorous mathematical theory: existence and
uniqueness of the solutions to initial and boundary value problems and their
asymptotic trends, but also rigorous justification of the approximate methods
of solution. Among these is the method described in this book. It is unfair,
however, to criticise, in the light of the standards and achievement of today,
the approach described in this book, as is sometimes done. In addition to still
being a good description of an important part of the kinetic theory of gases,
this book has played the important role of transmitting the solved and unsolved
problems of kinetic theory to generations of students and scholars. Thus it is
not only useful, but also historically important.
Carlo Cercignani
Milano
 EXTRACT FROM
P R E F A C E TO F I R S T E D I T I O N

In this book an account is given of the mathematical theory of gaseous

viscosity, thermal conduction, and diffusion. This subject is complete in
itself, and possesses its own technique; hence no apology is needed for
separating it from related subjects such as statistical mechanics.
The accurate theory originated with Maxwell and Boltzmann, who
established the fundamental equations of the subject. The general solution
of these equations wasfirstgiven more than forty years later, when within
about a year (1916-1917) Chapman and Enskog independently obtained
solutions by methods differing widely in spirit and detail, but giving iden-
tical results. Although Chapman's treatment of the general theory was
fully effective, its development was intuitive rather than systematic and
deductive; the work of Enskog showed more regard for mathematical
form and elegance. His treatment is the one chosen for presentation here,
but with some differences, including the relatively minor one of vector and
tensor notation.* A more important change is the use of expansions of
Sonine polynomials, following Burnett (1935). We have also attempted
to expound the theory more simply than is done in Enskog's dissertation,
where the argument is sometimes difficult to follow.
The later chapters describe more recent work, on dense gases, on the
quantum theory of collisions (so far as it affects the theory of the transport
phenomena in gases), and on the theory of conduction and diffusion in
ionized gases, in the presence of electric and magnetic fields.
Although most of the book is addressed to the mathematician and
theoretical physicist, an effort has been made to serve the needs of labora-
tory workers in chemistry and physics by collecting and stating, as clearly
as possible, the chief formulae derived from the theory, and discussing
them in relation to the best available data.
S.C.
1939 T. G. C.

* The notation used in this book for three-dimensional Cartesian tensors was devised
jointly by E. A. Milne and S. Chapman in 1916, and has since been used by them in many
branches of applied mathematics.

(»•]
 P R E F A C E TO T H I R D E D I T I O N

Until now, this book has appeared in substantially its 1939 form, apart
from certain corrections and the addition, in 1952, of a series of notes
indicating advances made in the intervening years. A more radical revision
has been made in the present edition.
Chapter 11 has been wholly rewritten, and discusses general molecular
models with internal energy. The discussion is primarily classical, but in
a form readily adaptable to a quantum generalization. This generalization
is made in Chapter 17, which also discusses (in rather more detail than
before) quantum effects on the transport properties of hydrogen and
helium at low temperatures. The theory is applied to additional molecular
models in Chapter 10, and these are compared with experiment in Chapters
12-14; the discussion in these chapters aims for the maximum simplicity
consistent with reasonable accuracy. Chapter 16 now includes a short
summary of the BBGKY theory of a dense gas, with comments on its diffi-
culties. A new Chapter 18 discusses mixtures of several gases. Chapter 19
(the old Chapter 18) discusses phenomena in ionized gases, on which an
enormous amount of work has been done in recent years. This chapter has
been much extended, even though attention is confined to aspects related
to the transport phenomena. Finally, in Chapter 6 and elsewhere, a more
detailed account is given of approximate theories, especially those that
illuminate some feature of the general theory.
To accommodate the new material, some cuts have been necessary.
These include the earlier approximate discussion of the electron-gas in
a metal, and the Appendices A and B. The Historical Summary, and the
discussion of the Lorentz approximation have been curtailed. The discussion
of certain other topics has been modified, especially in the light of the work
of Kihara, of Waldmann, of Grad and of Hirschfelder, Curtiss and Bird.
A few minor changes of notation have been made; these are set out at the
end of the list of symbols on pp. xxv and xxvi.
The third edition has been prepared throughout with the co-operation
of Professor D. Burnett. We are deeply indebted to him for numerous
valuable improvements, and for his continuous attention to details that
might otherwise have been overlooked. He has given unstinted assistance
over a long period.
Our thanks are due to many others for their interest and encouragement
Special mention should be made of Professors Waldmann and Mason for
their helpful interest. Our thanks are also due, as earlier, to the officials
of the Cambridge University Press for their willing and expert help both
before and during the printing of this edition.
s. c.
1969 T. G. C.
{"<' 1
 NOTE R E G A R D I N G REFERENCES

The chapter-sections are numbered decimally.

The equations in each section are numbered consecutively and
are preceded by the section number, (3.41. 1), (3.41. a)
References to equations are also preceded by the section
number and where a series of numbers occur they are elided
(3.41,2,3,...) or (3.41,1-16).
References to periodicals give first (in italic type) the name of the
periodical, next (in Clarendon type) the volume-number, then
the number of the page or pages referred to, and finally the
date in parenthesis.

[Xiv]
 CHAPTER A N D SECTION TITLES

Introduction
1. The molecular hypothesis (i)—2. The kinetic theory of heat (i)—3. The three states of
matter (i)—4. The theory of gases (2)—5. Statistical mechanics (3)—6. The interpretation
of kinetic-theory results (6)—7. The interpretation of same macroscopic concepts (7)—
8. Quantum theory (8).

Chapter 1. Vectors and tensors

1.1. Vectors (10)—1.11. Sums and products of vectors (11)—1.2. Functions of position
(12)—1.21. Volume elements and spherical surface elements (13)—1.3. Dyadica and
tensors (14)—1.31. Products of vectors or tensors with tensors (16)—1.32. Theorems on
dyadics (17)—1.33. Dyadics involving differential operators (18).
Some results on Integration
1.4. Integrals involving exponentials (19)—1.41. Transformation of multiple integrals
(20)—1.411. Jacobians (20)—1.42. Integrals involving vectors or tensors (21)—1.421. An
integral theorem (22).
1.5. Skew tensors (23).

Chapter 2. Properties of a gas: definitions and theorems

2.1. Velocities, and functions of velocity (25)—2.2. Density and mean motion (a6)—
2.21. The distribution of molecular velocities (27)—2.22. Mean values of functions of the
molecular velocities (28)—23. Flow of molecular properties (29)—2.31. Pressure and the
pressure tensor (32)—2.32. The hydrostatic pressure (34)—2.33. Intermolecular forces
and the pressure (35)—2.34. Molecular velocities: numerical values (36)—2.4. Heat (36)—
2.41. Temperature (37)-2.42. The equation of state (38)-2.43. Specific heats (39)—
2.431. The kinetic-theory temperature and thermodynamic tempcrature(4i)—2.44. Specific
heats: numerical values (42)—2.45. Conduction of heat (43)—2.5. Gas-mixtures (44).

Chapter 3. The equations of Boltzmann and Maxwell

3.1. Bottzmann's equation derived (46)-3.11. The equation of change of molecular pro-
perties (47)—3.12. £*/expressed in terms of the peculiar velocity (48)—3.13. Transforma-
tion of )<j>9fdc (48)—3.2. Molecular properties conserved after encounter; summational
invariants (49)—3.21. Special forms of the equation of change of molecular properties
(50)—3.3. Molecular encounters (5a)—3.4. The dynamics of a binary encounter (53)—
3.41. Equations of momentum and of energy for an encounter (53)—3.42. The geometry of
an encounter (55)—3.43. The apse-line and the change of relative velocity (55)—3.44.
Special types of interaction (57)—3.5. The statistics of molecular encounters (58)—3.51.
An expression for A(5 (60)—3.52. The calculation of ijtit (61)—3.53. Alternative expres-
sions for nA^; proof of equality (64)—3.54. Transformations of some integrala (64)—
3.6. The limiting range of molecular influence (65).

[xv]
xvi CHAPTER A N D SECTION TITLES

Chapter 4. Boltzmann's H-ihtonm and the Maxwelllan vclodty-

dtatribution
4.1. Bolttmann't //-theorem: the uniform ateady Mate (67)-4.11. Propertiei of the
Maxwellian atate (70)—4.12. Maxwell't original treatment of velocity-dittribution (7a)—
4.13. The ttetdy itate in « amooth veitel (73)-4.14. The ateady ttttc in the pretence of
external forcet (78)—4.2. The //-theorem and entropy (78)-4.21. The //-theorem and
reveraibility (70)-4.3. The //-theorem for gat-mixturea: equipartition of kinetic energy
of peculiar motion (80)—4.4. Integral theoremi: /(F). [F, G].(F, GU8a)—4.41. Inequalitiea
concerning the bracket expreationa [F, G], {F, G) (84).

Chapter S. The free path, the collision-frequency and persistence of

velocities
5.1. Smooth rigid elaatic apherical molecule* (86)—5.2. The frequency of collitioni (86)—
5.21. The mean free path (87)—5.22. Numerical valuet (88)—S3. The dittribution of
relative velocity, and of energy, in collitiont (80)—5.4. Dependence of collition-frequency
and mean free path on apeed (00)—5.41. Probability of a free path of a given length (03)—
5.5. The pertittence of velocitiet after collition (¢3)—5.51. The mean peniitence-ratio(oj).

Chapter 6. Elementary theories of the transport phenomena

6.1. The trantport phenomena (¢7)-6.2. Viacotity (¢7)—6M. Viacoaity at low pretturea
(08)—6.3. Thermal conduction (too)—631. Temperature-drop at a wall (101)—6.4.
Diffution (toi)—6.3. Defecti of free-path theories (103)—6.6. Collition interval theoriea
(104)-6.61. Relaxation timet (10O—6.62. Relaxation and diffution (106)—6.63. Gaa
mixturea (108).

Chapter 7. The non-uniform state for a simple gaa

7.1. The method of tolution of Bolttmann't equation (no)—7.11. The tubdivition of £(/):
the firtt approximation/'*' (in)—7.12. The complete formal eolution (m)—7.13. The
condition! of aolubility (113)—7.14. The tubdivition of ¢/(114)—7.15. The parametric
expreation of Enakog'a method of tolution (118)—7.2. The arbitrary parameter! i n / (no)
—7.3. The tecond approximation to/(iai)—7.31. The function • ' " Qai)—7.4. Thermal
conductivity dap—7.41. Viacotity (ta6)—7.3. Sonine polynomiala (ia7)—7.51. The
formal evaluation of A and A (ia8)—7.52. The formal evaluation of B and M (130).

Chapter 8. The non-uniform state for a binary gas-mixture

8.1. Boltxmann'a equation, and the equation of tranafer. for a binary mixture (13a)—
8 A The method of aohition (134)—821. The tubdivition of 8/(136)—83. The tecond
approximation to/(138)—8-31. The functiont »'", A. D, B (139)-8.4. Diffution and
thermal diffution (140)—8.41. Thermal conduction (14a)—-8.42. Viacoaity (144)—8.3. The
fourfirttgaa-cotmcientt (145)—831. The coemcienta of conduction, diffution, and thermal
diffution (14¾)—8*52. The coefficient of viacoaity (147).

Chapter 9. Viscosity, thermal conduction, and diffusion: general

expressions
9.1. The evaluation of fa"1, a'") and [»'"• b'«'l (uo)—93. Velocity tranaformationa
(140)—93. The expreationa fS(Vl) » , . S(*l) ^ . 1 , . and M«l> V&u S(Vt) y f c t l . i (»*>)
—931. The integrala //,,(¾)and L,J.x) Osi)—932. HxJLx) and InJiX) at functiont oft
andt(i53)—9J3. The evaluation ofI5(y*) ^ . . $(1¾ y j , , and [3(<rflyf't?i.S(yi)<g.'g.],.
 CHAPTER AND SECTION TITLES xvii

(155)-9.4- The evaluation of [5(¾¾ ^,,5(^,) ^ . 1 , , and [Sfif?) ^Vu S(«T)Ssfot1is(is6)

—9.5. The evaluation of fS(«?) V,. 5(^1) ^.1. and fS(?,) Sffy,. 5«ff) 'g.'ff.l. 057)—
9.6. Table of formuloe (is8)—9.7. Viscosity and thermal conduction in a simple gas
('59)—9.71. Kihara's approximations (160)—9.8. The determinant elements a^, &„ for a
gas-mixture (162)—9.81. The coefficient of diffusion/),,-, first an J second approximations
fDnli. fDi.l. (161)—9.82. The thermal conductivity for a gas-mixture: first approxima-
tion fAl, (164)-9.83. Thermal diffusion (16O—9.84. The coefficient of viscosity for 1
gas-mixture; first approximation Mi (165)—9.85. Kihara's approximations for a gas-
mixture (166).

Chapter 10. Viscosity, thermal conduction, and diffusion; theoretical

formulae for special molecular models
10.1. The functions O(r) (167).
10.2. Rigid elastic spherical molecule* without field* of force (168)—10.21. Viscosity
and conduction for a simple gas (168)—10.22. Gas-mixtures; [Dlt]t, [£>lt]t> Mi, [kr]i>
frfi O69)
10.3. Molecule* that are centre* of force (160)—10.31. Inverse power-law force (170)
—10.32. Viscosity and conduction for a simple gas (17a)—10.33. Maxwellian molecules
(173)—10.331. Eigenvalue theory (175)—10.34. The inverse-square law of interaction
(176).
Molecules possessing both attractive and repulsive fields
10.4. The Lennard-Jones model (179)-10.41. Weak attractive fields (181)—10.42.
Attractive forces not weak; the 12.6 model (183)—10.43. The exp;6 and other models(186).
10.5. The Lorentx approximation (188)—10.51. Interaction proportional to r" (100)
—10.52. Deduction of the Lorentz results from the general formulae dot)—10.53. Quasi-
Lorentzian gas (193)
10.6. A mixture of mechanically similar molecules (194)—10.61. Mixtures of isotopic
molecules (195)-

Chapter 11. Molecules with internal energy

11.1. Communicable internal energy (197)—11.2. Liouville's theorem (198)—11.21. The
generalized Boltzmann equation (199)—11.22. The evaluation of ijjit (zoo)—11.23.
Smoothed distributions (aoi)—11.24. The equations of transfer (aoa)—1IJ. The uniform
steady state at rest (203)—11.31. Boltzmann's closed chains (205)—11.32. More general
steady states (205)—11.33. Properties of the uniform steady state (ao6)—11.34. Equi-
partition of energy (208)—11.4. Non-uniform gases (ao8)—11.41. The second approxima-
tion to/, (ai 1)—11.5. Thermal conduction in a simple gas (211)-11.51. Viscosity: volume
viscosity (214)—11.511. Volume viscosity and relaxation (215)—11.52. Diffusion (ai6)—
11.6. Rough spheres (217)—11.61. Transport coefficients for rough spheres (219)—11.62.
Defects of the model (220)—11.7. Spherocylinders (221)—11.71. Loaded spheres (aaa)—
11.8. Nearly smooth molecules: Eucken's formula (aa2)—11.81. The Mason-Monchick
theory (224).

Chapter 12. Viscosity: comparison of theory with experiment

12.1. Formulae for n for different molecular models (aa6)—12.11. The dependence of
viscosity on the density (227)—12JZ. Viscosities and equivalent molecular diameters (aao).
The dependence of the viscosity on the temperature
12.3. Rigid elastic spheres (229)—12.31. Point-centres offeree (230)—12J2. Sutherland's
formula (23a)—12J3. The Lennard-Jones la, 6 model (235)—12J4. The exp; 6 and polar-
gas models (tyj).
*vH« CHAPTER A N D SECTION TITLES
C M miihirea
12.4. The viscosity of a gas-mixture (238)—12.41. The variation of the viscosity with
composition (no)—12.42. Variation of the viscosity with temperature (242)—12.43.
Approximate formulae (»43).
12.5. Volume viscosity (144).

Chapter 13. Thermal conductivity: comparison of theory with

experiment
13.1. Summary of the formulae (247)—13.2. Thermal conductivities of gases at 0 °C. (248)
—13.3. Monatomic gases (230)—13.31. Non-polar gases (251)—13.32. Polar gases (252)—
13.4. Monatomic gas-mixtures (253)—13.41. Mixtures of gases with internal energy (234)
—13.42. Approximate formulae for mixtures (255).

Chapter 14. Diffusion: comparison of theory with experiment

14.1. Causes of diffusion (287)-14.2. The first approximation to £),. (2<8)—14.21. The
second approximation to D,t (a;q)—14.3. The variation of D„ with the pressure and
concentration-ratio (160)—14.31. Comparison with experiment for different concentration-
ratios (260)—1442. Molecular radii calculated from D„ (262)—14.4. The dependence of
D„ on the temperature; the intermolecular force-law (264)—14.5. The coefficient of
self-diffusion D„ (26O—14.51. Mutual diffusion of isotopes and like molecules (267).
14.6. Thermal diffusion, and the thermo-diffusion effect (268)—14.7. The thermal
diffusion factor an (271)—14.71. The sign and composition-dependence of [«II1I (272)—
14.711. Experiment and the sign of an (274)—14.72. a,, and the intermolecular force-law:
•sotopic thermal diffusion (275)-14.73. a„ and the intermolecular force-law: unlike
molecules (376).

Chapter 15. The third approximation to the velocity-distribution

function
15.1. Successive approximations t o / (280)—15.2. The integral equation for /"' (282)—
153. The third approximation to the thermal flux and the stress tensor (284)—15.4. The
terms in q"> (286)—15.41. The terms in p"> (288)-15.42. The velocity of diffusion (290)
—15.5. The orders of magnitude of<7'" and p"1 (201)—15.51. The range of validity of the
third approximation (292)—15.6. The method of Crad (293)—15.61. The Mott-Smith
approach (29s)—15.62. Numerical solutions (29¾).

Chapter 16. Dense gases

16.1. Collitional transfer of molecular properties (207)—16.2. The probability of collision
(208)—16.21. The factor X (ao8)—16.3. Boltamann's equation: d.fldt hop)—16.31. The
equation for/'" (301)—16.32. The second approximation tod.fldt (302)—16.33. The value
of f" (303)-1644. The mean values of pCC and ipCC (704)—16.4. The collisional
transfer of molecular properties (304)— 16.41. The viscosity of a dense gas (306)—16.42.
The thermal conductivity of a dense gas (307)—16.5. Comparison with experiment (308)—
16.6. Extension to mixed dense gases (311).
16.7. The BBKGY equations (in)—16.71. The equations of transfer (314)—16.72.
The uniform steady state (3»s)—16.73. The transport phenomena (316).
16.8. The evaluation of certain integrals (319).
 CHAPTER A N D SECTION TITLES xix

Chapter 17. Quantum theory and the transport phenomena

T h e q u a n t u m theory of molecular collisions
17.1. T h e wave fields of molecules (322)—17.2. Interaction of two molecular streams
(323)—17.3. T h e distribution of molecular deflections (323)—17.31. T h e collision-
probability and mean free path (326)—17.32. T h e phase-angles in (328)—17.4. Comparison
with experiment for helium (329)—17.41. Hydrogen at low temperatures (332)—17.5.
Degeneracy for Fermi-Dirac particles (333)—17.51. Degeneracy for Bose—Einstein
particles (33s)—17.52. Transport phenomena in a degenerate gas (336).
Internal energy
17.6. Quantized internal energies (336)—17.61. Encounter probabilities (337)—17.62.
T h e Boltzmann equation (138)—17.63. T h e uniform steady state (no)—17.64. Internal
energy and the transport phenomena (34').

Chapter 18. Multiple gas mixtures

18.1. Mixtures of several constituents (343)—18.2. T h e second approximation (343)—
18.3. Diffusion ( 3 4 4 ) - 1 8 . 3 1 . Heat conduction (346)-18.32. Viscosity (348)—18.4.
Expressions for the gas coefficients (348)—18.41. T h e diffusion coefficients (348)—18.42.
Thermal conductivity (349)—18.43. Thermal diffusion (351)—18.44. T h e viscosity (352).
Approximate values In special cases
18.5. Isotopic mixtures (352)—18.51. One gas present as a trace (354)—18.52. Ternary
mixture, including electrons (354).

Chapter 19. Electromagnetic phenomena in ionized gases

19.1. Convection currents and conduction currents (358)—19.11. T h e electric current in
a binary mixture (3H0)—19.12. Electrical conductivity in a slightly ionized gas (360)—
19.13. Electrical conduction in a multiple mixture (361).
Magnetic fields
19.2. Boltzmann's equation for an ionized gas in the presence of a magnetic field (361)—
19.3. T h e motion of a charged particle in a magnetic field (363)—19.31. Approximate
theory of diffusion in a magnetic field (364)—19.32. Approximate theory of heat conduction
and viscosity (368)—19.4. Boltzmann's equation: the second approximation to f for an
ionized gas (370)—19.41. Direct and transverse diffusion (373)—19.42. T h e coefficients of
diffusion (373)—19.43. Thermal conduction (376)—19.44. T h e stress tensor in a magnetic
field (378)—19.45. Transport phenomena in a Lorentzian gas in a magnetic field (378)—
19.5. Alternating electric fields (379).
P h e n o m e n a i n strong electric fields
19.6. Electrons with large energies (382)—19.61. T h e steady state in a strong electric field
(383)—19.62. Inelastic collisions (389)—19.63. T h e steady state in a magnetic field (391)—
19.64. Ionization and recombination (392)—19.65. Strongly ionized gases (395)—19.66.
Runaway effects (398).
T h e Fokker—Planck Approach
19.7. T h e Landau and Fokkcr-Planck equations (400)—19.71. T h e superpotentials (402).
19.8. Collislonleas plasmas (403).
 DIAGRAMS

i The passage of molecules across a surface-element page 30

2 Change of relative velocity in a molecular encounter 54
3 Geometry of a molecular encounter, direct (3 a) or inverse (36) 56
4 Maxwell's diagram of molecular orbits 57
5 The collision of two smooth rigid elastic spherical molecules 58
6 Probability of a molecular encounter 59
7 Graphs of *-** and **«-** 71
8 and 9 Curves illustrating the variation of log^ii', logiS^J, A,B
and c with log(Ar/elt) for the 12, 6 and exp; 6 models 185-6
10 Comparison of theoretical and experimental viscosities, on the
12, 6 model, for He, Ne, H„ A, N, and CO, 236
11 The viscosity of a mixture (H„ HC1): variation with composi-
tion, at different temperatures 243
12 The variation of Du with composition for H,-N, and He-A
mixtures 261
13 Comparison of the calculated and the experimental viscosities
of He* and He4 at very low temperatures 331
14 The distribution function for electronic speeds: the steady
state in a strong electric field 389

[XX]
 LIST OF SYMBOLS
Clarendon type is used for vectors, roman clarendon type for unit
vectors, and sans serif type for tensors.
The bracket symbols [ , ] and { , } are defined on page 83.
In general, symbols which occur only in a few consecutive pages are not
included in this list. Greek symbols are placed at the end of the list.
The italicfiguresindicate the pages on which the symbols are introduced.
ap, 128, 145 a,,, 128,146 a',,, a"„, 162 A{y), 171 A(<V), 123
m
J*»\ J*%\ 129,146 j*'i \j*'fi\i46 cfP\i28 ap\ ap\ 145
A, 123 Alt At, 139,21t Av A„ 143 A„ 344 A„ 347
A
> 163 A*, 350
f>,55 bp, 130,147 b„, 130,148 b'n, b"PV 162 B(V),i24 BvBt,2ii
»^,»^,131,148 b"»,/jo b<»>, bp\ 147 Bt, B»/jp, 2//
B„ 344 B, 163 Brt, 350
c e
c,c,2S c0, ^,27,44 ce<39 cp, 40 0,,44 i> t> ci. cJ. 53
c„ ct, c'v c't, S4 4, c"r, 222 (c'r)„ (cj)„ 207 Cv, Cp, 40 C, C, C, 27
C',28 C„44 C„4S %V,i22 «?„«'„«'„«'„/.?* 0,163 c*>35'
f
p. '45 D,„ 102 Du, 103 DT, 141 DM, 224 DT„ 34s
Dj, log, ,/.. 47 ®i/i. ^ 1 /„ 134 9?, ®t>, 134
346 ®f. ®.
'Ei'46 8t '*7
Jt'11* Dl'*8
dr,i3 dc, 2$ dk, 14 dc, 62 d*',59
Dl>"6
da„ igg
d
Br'"
e
BC ' 3
d„ 210,343
8
Be' 25
d
BP; '4''"
D„ D* 139' 211 W.344
WY20 <*it. <*n< '3$
E
«„ et, 176 e„ 3S8 e, e', J9 < E> 37 B, 359 e, 8,19
e„ /i°, 207 E, 163

f(c,r,t),f(C,r,t),28 Uc„r, 1),44 / w \ / n \ / w "° WJP,'34

f»\clt ...e„rlt... r„ t), 312 f, 101 F, 46 P„ 47
8iv8tvg'it<g'tv G<53 git<gn<g'it<g'n<g>g'>G>54 f<t'>f>f'<'5°
G0,14g G9, iso ®„, 90, IJO g (gravity), 227
(xxij
XXii LIST OF SYMBOLS

#.67 #it(x).'5' Hi(X),'S6 H

(Hamiltonian), 108
U.,WP,m K203 H,36i
1(F), /,(F). W), 82 UK), 83 /,, 206
),324 3,2 J (Jacobian), 20 /(//1),/^,/^///
/,(/1/), JMft), Wtf\ /,1(/,/1), / n / ^ . /J* W 20* j , 358
k
>37 *T, '4' *7*34S K„K„ *„,*/* k,/*57
/,/i,W 1^),92 /(CJ./oo /(<:,), J*J Mx)./5/ ^i(x)./56
m,26 m„44 w0, Wj,«,, S3 A*. J* Mx,Mt,s3
11,27,44 »„44 N>4* Nn,Nlt,87 N„207 Nx,Nt,...,2i6 ^,312

P,3S /»o.*>9 Pxx,Px* 34 P„ P „ P „ , / 6 4 P;IQ8 pn,32

P,P,i69 Pt,P.,Fi\i99 p.jj p., 45 p«\ p">,//5,/J<S
01,0,,0,,,/64 q,43 q*>,q", 115,13* Q",'98 Q"„Q„(^\iQ9
Q,4°S
r, 12 R„ R,, R„, R;„ /65 R, 3g
s, 127 s (index), 172 sv st, su, 240 t'n, 262 Sj*(x), 127
S,Stt,i8i s,i27 8,,8,,/65 S. 4'
t, 12 T, 127 T, 37
U, 15 u, 08 u'u, 103 u, v, to, 25 «0» vt> wo< 27 V, V, W, 27
MO./70 v(r(S),3i3 \,38
W+,7' ^^),^^),167
x„ x t , 138 x„ 343
Z„ 206,339

GREEK SYMBOLS
a,, 128,146 a^g,b),6o ax(g,b),6i »it(g,x),323 *i(g,X),326
a
**'£, 337 i* (thermal diffusion), 142 a„, 275

fiv /jo. '47

7,4'
 LIST OF SYMBOLS xxiti

<*,, 146 A, 164 A, 283 A$5, 47 A^„ 48

v 12
Aj^i.A.^,, 60 &*,345 >
e e
> 5S u.lSo

0< '3 *. 3S9

i(=-J-*)>324

Kl2, 170 K, IJ2 K'n, l8o

\,ioo Wi, .-,159 K,iog X',X",222 ^,,^,,255 K>< 349

fi.98 Mi 159 P»io8 f>xu<fte,243 1^,352

v, 170 v', 180

ro
i*(ci)> 94 «"1!. 95 w
> 214> 3°6
p, 27, 44 p„ 44
ff
&1, o* G"u. 57 <*> 88 u, ""n /^7

r„ 88 rfa), 92 r, 104, 215 T„, 706 T„ 251

», v0, 0,,,,, 170 uov 177

9,13 ¢, t{c),..., 28 $,29 1(,,,44 flL'SS V*,n4 <!>,«>, <t>J» /J*

X (deflection), 55 x (dense gas), 298

f,S7 ^,^,^,49 r*\SO Y(76 ^,203

OftW. '55 Q M rS8 w„ w,, 206

The notation r, is used to denote the product r(r— 1)... (r-q+1), e.g.
on pages 127 and 173.
Differences in notation from earlier editions are noted in the following list,
which gives the new equivalents of earlier combinations.

Old C-i ki%dk, kydk n^nJ^FtG) «io> "»

New c gctltde', gct^d n*{F,G} xltxt

xxlv LIST OF SYMBOLS
Old B, 6,,6, nDvnDt af»,a$» 2*08
New 26,28,,26, DltDt n^a^Ma^ gffl

Old p„ p,
New M\vlt Afjp,
The symbols R„ R,, R„, Rj, used here are unrelated to those of earlier
editions.
 INTRODUCTION

1. The molecular hypothesis

The purpose of this book is to elucidate some of the observed properties of
the natural objects called gases. The method used is a mathematical one.
The foundation on which our work is based is the molecular hypothesis of
matter. This postulates that matter is not continuous and indefinitely
divisible, but is composed of a finite number of small bodies called mole-
cules. These in any particular case may be all of one kind, or of several kinds:
the number of kinds is usually far less than the number of molecules. Free
atoms, ions and electrons are considered merely as special types of molecule.
The individual molecules are too small to be seen individually even with
the most powerful ultra-microscope.
The joint labours of experimental and theoretical physicists have sug-
gested certain hypotheses regarding the structure and interaction of
molecules: the precise details, however, are known for only a few kinds of
molecule. The mathematician has therefore to consider ideal systems, chosen
as illustrating the particular features of actual gas-molecules that are to be
studied, and to work out their properties as accurately as possible. The
difficulty of this undertaking imposes limitations on the models that can be
used. For example, if the systems are not spherically symmetrical, the
investigation of their interactions includes the solution of some difficult
dynamical problems: the mass-distribution and field of force of a molecule
are therefore usually taken to be spherically symmetrical. As this book
shows, the investigations even then are very complicated; the complexity is
enormously enhanced when the condition of spherical symmetry is relaxed
in the least degree. 1'he special models of molecules that are considered in
this book are described in 3.3 and in Chapter 11.

2. The kinetic theory of heat

The molecular hypothesis is of great importance in chemistry as well as in
physics. For some purposes, particularly in chemistry and crystallography,
the molecules can be considered statically; but usually it is essential to take
account of the molecular motions. These are not individually visible, but
there is evidence that they are extremely rapid. An important extension
of the molecular hypothesis is the theory (called the kinetic theory of heat)
that the molecules move more or less rapidly, the hotter or colder the body
of which they form part; and that the heat energy of the body is in reality
mechanical energy, kinetic and potential, of the unseen molecular motions,
relative to the body as a whole. The heat energy is thus taken to include the
translatory kinetic energy of the molecules, relative to axes moving with the
Ii1
2 INTRODUCTION [3
element of the body of which at the time these molecules form part; it
includes also kinetic energy of rotation, and kinetic and potential energy of
vibration, if the molecular constitution permits of these motions.
Since heat energy is regarded as hidden mechanical energy, it must be
expressible in terms of mechanical units. Joule, in fact, showed that the
ordinary measure of a given amount of heat energy is proportional to the
amount of mechanical energy that can be converted, for example by friction,
into the given quantity of heat. The ratio
measure of heat energy in heat units
measure of the same energy in mechanical units
is therefore called Joule's' Mechanical equivalent of heat'—usually denoted
by J.
3. The three states of matter
The molecular hypothesis and the kinetic theory of heat are applicable to
matter in general. The three states of matter—solid, liquid, and gaseous—
are distinguished merely by the degree of proximity and the intensity of the
motions of the molecules. In a solid the molecules are supposed to be packed
closely, each hemmed in by its neighbours so that only by a rare chance can
it slip between them and get into a new set. If the solid is heated, the motions
of the molecules become more violent, and their impacts in general produce
a slight thermal expansion of the body. At a certain point, depending on the
pressure to which the body is subjected, the motions are sufficiently intense
for the molecules, though still close-packed, to be able to pass from one set of
neighbours to another set: the liquid state has then been attained. Further
application of heat will ultimately lead to a state in which the molecules
break the bonds of their mutual attractions, so that they expand to fill any
volume available to them; the matter has then attained the gaseous state.
At certain pressures and temperatures two states of matter (liquid and gas,
solid and liquid, or solid and gas) can coexist in equilibrium; all three states
can coexist at a particular pressure and temperature.

4. The theory of gases

In a solid or liquid the mutual forces between pairs of neighbouring mole-
cules are considerable, strong enough, in fact, to hold the mass of molecules
together, at least for a time, even if all external pressure is removed. A static
picture of a solid is obtained if the molecules are imagined to be rigid bodies
in contact: a molecule can be supposed to possess a size, equal to the size of
such a rigid body.
The density of a gas is ordinarily low compared with that of the same
substance in the liquid or solid form. The molecules in a gas are therefore
separated by distances large compared with their sizes, and they move
hither and thither, influencing each other only slightly except when two or
51 INTRODUCTION 3
more happen to approach closely, when they will sensibly deflect each other's
paths. In this case the molecules are said to have encountered one another;
expressed otherwise, an encounter has occurred. Obviously an encounter
is a less definite event than a collision between two rigid bodies; definiteness
can be imparted to the conception of an encounter only by specifying a
minimum deflection which must result from the approach of two molecules,
if the event is to qualify for the name encounter.
When the molecules are regarded as rigid bodies not surrounded by fields
of force, their motion between successive impacts is quite free from any
mutual influences: each molecule is said to traverse a free path between its
successive collisions. The average or mean free path will be greater or less,
the rarer or denser the gas.
The conception of the free path loses some if its definiteness when the
molecules, though still rigid, are surrounded by fields of force. The loss of
definiteness is greater still if the molecules are non-rigid. The conception
can, however, be applied to gases composed of such molecules, by giving to
encounters, in the manner described above, the definiteness that attaches to
collisions.
Collisions or encounters in a gas of low density are mainly between pairs
of molecules, whereas in a solid or liquid each molecule is usually near or in
contact with several neighbours. The legitimate neglect of all but binary
encounters in a gas is one of the important simplifications that have enabled
the theory of gases to attain its present high development.

5. Statistical mechanics
In ordinary mechanics our aim is usually to determine the events that follow
from prescribed initial conditions. Our approach to the theory of a gas must
be different from this, for two reasons. Firstly, we never know the detailed
initial conditions, that is, the situation and state of motion of every molecule
at a prescribed initial instant; secondly, even if we did, our powers are quite
unequal to the task of following the subsequent motions of all the many
molecules that compose the gas. Hence we do not even attempt to consider
the fate of the individual molecules, but interest ourselves only in statistical
properties—such as the mean number, momentum or energy of the mole-
cules within an element of volume, averaged over a short time interval, or the
average distribution of linear velocities or other motions among these
molecules.
It is not only necessary, for mathematical reasons, to restrict our aims in
this way: it is also physically adequate, because experiments on a mass of gas
measure only such 'averaged' properties of the gas. Thus our aim is to
find out how, for example, the distribution of the 'averaged' or 'mass'
motion of a gas, supposed known at one instant, will vary with the time; or
again, how a non-uniform mixture of two sets of molecules of different kinds
will vary, by the process known as diffusion.
4 INTRODUCTION P
In such attempts, we consider not only the dynamics of the molecular
encounters, but also the statistics of the encounters. In this we must use
probability assumptions, such, for example, as that the molecules are in
general distributed 'at random', or evenly, throughout a small volume, and
moreover, that this is true also for the molecules having velocities in a
certain range.
The pioneers in the development of the kinetic theory of gases employed
such probability considerations intuitively. Their work laid the foundations
of a now very extensive branch of theoretical physics, known as statistical
mechanics, which deals with systems much more general than gases. This
applies probability methods to mechanical problems, and as regards its
underlying principles it shares some of the obscurities that attach to the
theory of probability itself. These philosophical difficulties were glimpsed
already by the founders of the subject, and have been partly though not
completely clarified by subsequent discussion.
In one aspect, the theory of probability is merely a definite mathematical
theory of arrangements. The simplest problem in that subject is to find in
how many different ways m different objects can be set out in n rows (m > n),
account being taken of the order of the objects in the rows. A great variety
of problems of this and more complicated types can be solved, in a completely
definite way.
One such problem throws some light on the uniformity of density in a gas.
Consider all possible arrangements of m molecules in a certain volume, sup-
posed divided into n cells of equal extent, m being very large compared with n.
The number of different arrangements, if regard is paid only to the presence,
and not to the order or disposition, of individual molecules in each cell, is nm.
Among these arrangements there will be many in which the total numbers of
molecules in the respective cells i to n have the same particular set of values
<tt, a „ . . . , «„, where of course
«i + «a+ ••• + a B = m-
It is not difficult to show that, when m/n is large, the great majority of the
nm arrangements correspond to distributions for which every number a, to
aK differs by a very small fraction from the average number m/n per cell.
Hence, if we regard the original nm arrangements as all equally probable (on
the ground, for example, that all the cells are equal in volume, and that there
is no reason why any particular molecule should be placed in one cell rather
than in another),* we are led to conclude that in any arbitrarily chosen mass
of gas the density of the molecules will almost certainly be very nearly
uniform throughout the volume.
It needs little consideration to recognize that this somewhat vague
statement is very different from the original results about the arrangements
* This, of course, implies that the volume of the molecules is negligible: if the volume
of one cell is altetdy largely occupied by molecules, another molecule may be supposed
less likely to find a place in this cell than in a relatively empty one.
5] INTRODUCTION 5
of the molecules; it is less precise (though it can be expressed in the form of
inequalities with narrow limits): moreover, it depends on an assumption as
to a priori probability. Every statement about probability depends, in a
similar way, on some assumption as to a priori probability, and is less definite
than the results of the arrangement theory.
Similar considerations as to arrangements can be applied to the distribu-
tion of a given total amount of translatory kinetic energy between the
molecules of a gas when the mass-centre of the whole set is at rest. Here it is
assumed that all velocities of a given molecule are a priori equally probable.
The result obtained is that the velocities of the molecules are almost
certainly distributed in a manner agreeing very nearly with a formula first
inferred (from intuitive and unjustifiable probability considerations) by
Maxwell. The a priori assumption cannot be verified: but it can be shown,
using a purely dynamical theorem due to Liouville, that as the state of the
gas varies with the passage of time, the 'arrangements' which are found
initially to be most abundant, as regards both space and velocity-distribution,
will always remain most abundant. Hence it is concluded that the uniform
density and the Maxwellian velocity-distribution will always be the most
probable, though a particular mass of gas may, very rarely (with a degree of
improbability that can be estimated), pass through a state which departs to
some extent from these usual or normal conditions.
These results of statistical mechanics, and others of a like kind, illustrate
the use made of probability in the kinetic theory of gases. The results
obtained in this theory are usually stated in a quite definite form, but the
validity of the conclusions cannot be rated higher than that of the argu-
ments leading thereto. Since in these arguments we appeal to probability,
the results of the kinetic theory remain only probable. But the study of
statistical mechanics suggests that statements of probability about systems
containing a very large number of independent units, such as molecules,
usually have a degree of probability so high as to be equivalent, for all
practical purposes, to certainty: results which statistical mechanics asserts
to be extremely probable are usually taken as rigorously true in experimental
work and in thermodynamic theory. Hence though in theory we cannot
exclude the rare possibility of a fleeting departure from the most probable
states, in practice there need be no question whether the results of kinetic
theory will agree with those of experiment.
By statistical mechanics we are led to certain conclusions about the
equilibrium states of systems, independent of the mode whereby these
equilibrium states are attained; but statistical mechanics does not show how,
or at what rate, a system will attain an equilibrium state. This can be
determined only if we know certain details about the molecules or other
units composing the system, details which, for the purposes of statistical
mechanics, can be ignored.
It is the province of a detailed kinetic theory to study the problems of
6 INTRODUCTION [6
non-equilibrium states, and such investigations occupy the greater part of
this book. The probability methods of the kinetic theory are also, however,
applied in the earlier chapters (3 and 4) to determine the equilibrium states;
the results thus obtained are merely special cases of much more general
results of statistical mechanics.

6. The interpretation of kinetic-theory results

The methods of the kinetic theory are successful in giving results of practical
interest, although the molecular models chosen are not believed to corre-
spond at all closely with actual molecules. By comparing results obtained for
different models, we are able to gain some idea as to how far any particular
kind of result depends on this or that feature of the molecular model. It
appears that the assumption that the centres of molecules approach each
other more closely, the greater their speed of mutual approach, leads to
quantitative results for various properties of gases more in accordance with
those actually observed than the assumption that the molecules are rigid.
Thus a molecule surrounded by a field of force is a better model for quanti-
tative treatment, if not for simple illustrative discussions, than a rigid
molecule.
In actual gases, at moderate temperatures, in all but a very small fraction
of the molecular encounters the least distance between the centres of the
molecules is still distinctly greater than would correspond to an overlapping
of the normal detailed structures of the molecules. These structures are
therefore not of immediate concern in the kinetic theory of gases; they
determine the exterior fields of force, which form the outworks of the
molecule, and it is only the nature of the outworks that is here important.
It can be adequately specified, for our purpose, by a formula expressing the
approximate rate of variation of the force-intensity with distance from the
centre of the molecule, over the range of distance outwards from that
corresponding to close encounters. At smaller distances the field might have
any value without affecting the kinetic-theory calculations; the actual
structure of the molecule within this minimum distance can be ignored,
and the molecule may without detriment be regarded as a point-centre of
force.
The restriction of kinetic-theory calculations to molecules that are
spherically symmetrical is also not of such importance as might appear from
the practical certainty that many actual molecules are not at all spherically
symmetrical. This is because the molecules in general rotate, and at
encounters they may be oriented relative to each other in any manner: the
detailed consequences of a particular encounter depend on the orientations,
but such consequences averaged over a large number of encounters are
probably not very different from the corresponding averaged results of
the encounters of a set of spherically symmetrical molecules, the force
71 INTRODUCTION 7
between pairs of which, at any distance, is equal to the average, over all
orientations, of the force between pairs of the actual molecules whose centres
are at that distance apart. Such averaging of the consequences of encounters
is of the essence of kinetic-theory calculations, so that many of the results
obtained in this book should be correct qualitatively, and not far from correct
quantitatively, for gases whose molecules are non-spherical. The chief
exceptions are in problems involving the total heat energy, since the actual
molecules may possess an average amount of internal energy different from
that of the spherically symmetric models.

7. The interpretation of some macroscopic concepts

Our aim is to explain things that are seen and directly measurable by means
of imagined things that are not seen and not directly measurable. The general
lines along which we are to proceed have already been indicated, in describing
the molecular hypothesis and the kinetic theory of heat. There remain,
however, further points on which there is room for freedom of interpretation.
The criterion by which our choice is to be judged is whether the relations
found between the quantities we identify with measurable macroscopic
quantities do or do not approximate to the observed relations between those
macroscopic quantities. Success in this test affords ground for a reasonable
expectation that any hitherto unknown macroscopic relation suggested by
the kinetic theory on its own basis of interpretation will be confirmed on
experimental trial.
It should be emphasized that only approximate agreement is to be
expected between observed macroscopic relations and those inferred from
kinetic theory, because some divergence between the two sets of relations
may reasonably be attributed to the imperfect representation of the actual
molecules by the 'model' molecules with which the mathematician works.
The kinetic-theory interpretations of some typical macroscopic properties
are briefly summarized here.
The combined masses of the molecules of a set are taken as giving the
macroscopically observed mass of the set.
The heat energy of a small portion of matter is identified with the trans-
latory kinetic energy of the molecular motions relative to the element as a
whole, together with the total of such other forms of molecular energy as are
interchangeable with translatory kinetic energy at encounters. Thus in
diatomic and polyatomic molecules the relative motion of the atomic nuclei
may contribute kinetic and potential energy to the heat energy; but energy
like the kinetic and potential energy of electrons in an inert gas-molecule,
which is normally unaffected by encounters, is neglected. Correspondingly,
certain of our molecular models, such as the smooth rigid spherical model,
make no provision for a possible interchange of translatory and rotatory
kinetic energy: we can ignore the energy of rotation in discussing these.
8 INTRODUCTION 18
The pressure of a gas on a bounding surface is identified with the mean
time-rate of communication of momentum to the surface, per unit area, by
molecular impacts; the momentum is imparted in a more or less dis-
continuous manner, but the individual impulses are so small, frequent,
and numerous as to simulate a continuous pressure. In addition to this
momentum-pressure there is a much smaller stress due to the action at a
distance between the molecules of the gas and the wall; this is normally not
taken into account in the present book.
The mean translatory kinetic energy per molecule, relative to the general
motion of the gas, is taken to be proportional to the thermodynamic tem-
perature, the constant of proportionality depending on the units of energy
and temperature, but not on the gas. Such an identification is permissible (on
the understanding that it is to be justified by its results), so long as we are
concerned with the phenomena of a single portion of gas; but the question
arises whether this identification will be valid for different portions of gas,
composed of molecules of different kinds. So long as our discussion is con-
fined purely to the phenomena of gases, the question seems to depend on
whether, when we mix two different gases, to which, according to this
definition, we ascribe equal temperatures, the same temperature will
characterize the mixture, as is observed to be the case when we mix actual
rare gases whose thermodynamic temperatures are equal. The kinetic theory
is able to give a fairly satisfactory affirmative answer to this question (4.3),
to this extent justifying its procedure as regards temperature definition.
But questions as to the thermal equilibrium of two different gases with
another body (say a diathermanous wall between the compartments of a
vessel containing the two gases) are outside the scope of the kinetic theory of
gases: they lie within the domain of statistical mechanics, which considers
assemblies much more general than gases.

8. Quantum theory
A wise conservatism, rather than reasons valid a priori, prompted the
pioneers of the kinetic theory to attribute to their imagined molecules the
same rules of behaviour—or, in technical language, the same mechanical
laws of motion—as those that characterize the objects of our ordinary
experience. Their rigid spherical molecules were idealizations of ordinary
billiard balls, while their point-centres of force were suggested by planets
viewed, from the large astronomical standpoint, as point-centres of gravita-
tional attraction.
The consequences of this assumed behaviour of molecules correspond
closely in general to the observed behaviour of gases; this supports the
view that molecules do behave in the supposed way. It is not a matter for
surprise, however, that the kinetic-theory consequences of the assumption
do not fit the whole range of observed facts. The discrepancies are of a
81 INTRODUCTION 9
nature to be explained by attributing to the molecules rules or laws of
behaviour (including statistical laws) that deviate from the classical laws in
a way suggested by the study of many other phenomena of matter, parti-
cularly spectroscopic phenomena. The new laws and the body of science
dealing with them, known as the quantum laws and quantum theory, are
only briefly touched on in this book (Chapter 17), whose main aim is to
record the basic development of the kinetic theory of gases, founded on
classical mechanics.

2 CKT
 1
VECTORS A N D TENSORS

1.1. Vectors
The notation and calculus of vectors, and also of three-dimensional Cartesian
tensors, are largely used in this book. In this chapter we summarize the vector
and tensor notation and calculus that we adopt.
Any physical quantity possessing both magnitude and direction is called
a vector quantity, or, briefly, a vector. Such quantities will be denoted by
symbols in heavy (Clarendon) type, in various founts, as, for example,*
a, A, V, w, n.
The (positive) magnitude of a vector denoted by a Clarendon symbol will
usually be denoted by the same symbol in the corresponding ordinary type,
e.g. for a, A, V, u> by a, A, V, <o (of course in the case of a unit vector no
such magnitude symbol is needed).
The component of a vector A along a direction inclined to A at an angle
d{o<,d^n)it defined to be A cos 0; this may be positive or negative. Any
vector is completely specified when its components in three mutually
perpendicular directions are given. When these directions are those of the
axes Ox, Oy, Oz of a Cartesian system, f the components are called the
rectangular Cartesian components relative to these axes. They may be
denoted by adding the suffices x, y, z to the symbol denoting the magnitude
of the vector (e.g. ax, ay, a, denote the x, y, z components of a), or by special
symbols (as in 1.2 for r and C, and as in 1.33 for c). The magnitude of a
vector is given in terms of its rectangular components by an equation of the
f0rm
a*-a* + al + al (1.1,1)
Let a be a vector whose components relative to the axes Ox, Oy, Oz are
* For the convenience of the reader certain special convention! regarding «uch types will
be made in thi* book, a> follow*:
(i) vector* whoae magnitude ia unity (or, briefly, unit victors), will be denoted by
ordinary tmall upright letter* in Clarendon type, namely
e,h,i,j,k, 1, n;
(ii) script Clarendon capital*, such a*

will denote certain non-dimensional vectors associated with vectors represented by the
corresponding Clarendon italic capitals, namely
C.G*
f Throughout this book all Cartesian axe* of reference are understood to be mutually
perpendicular (or orthogonal) and right-handed.
[10]
1.11] VECTORS AND TENSORS 11
ax, ay, a„ and let Ox', Oy', Oz' be a second set of orthogonal axes whose direc-
tion cosines relative to the first set are (/„«„»,), (/„ ma, na), (/,, mt, nt).
Then the components ay, ay, a^ of a relative to the second set of axes are
g,VCn y
«*- = / 1 a I +w l «„ + »!«, (I.I, 2)
and two similar equations. Similarly

and so on. These equations take a simpler form if in place of /1( /,, /„ Wj, m„...,
we write t^, txy, f„., tyx., tvy The nine symbols tair, where a. andfimay
stand for x ory or z, define a matrix which we call the transformation matrix;
the typical element taf. of this matrix is the cosine of the angle between the
axes OOL, Ofi'. In this notation, the equations of transformation may be
written „ . .
a

«. = S W - (i .1,5)

1.11. Sums and products of vectors

The sum of two vectors is defined as the vector whose components are the
sums of the corresponding components of the vectors. Thus the rule for the
addition of vectors is the same as the parallelogram law for the composition of
forces or velocities.
Let a, b be two vectors inclined at an angle 0 ( ^ if). Then ab cos 6 is a
scalar quantity (i.e. a quantity possessing magnitude but not direction).
It is called the scalar product of a and b, and is denoted by a. b. In terms of
the components of a and b,
a.b = axbx + avbv + a,bt = b.a. (i-",')
From this it follows that
(a+b).(c+d) = a.c + b.c+a.d+b.d,
of which the following are important special cases
(a + b).(a + b) = a1 + 2a.b + bt,
(a-b).(a-b) = ai-za.b + bt,
(a + 6).(a-6) = a8-6».
The vector product of the vectors a, b is defined to be the vector of magni-
tude ab sin 0, perpendicular to both a and b, and in the direction of trans-
lation of a right-handed screw, rotated in the sense from a to b, through the
angle 0(<n) between a and b. It is here denoted by a A b. Its Cartesian
componensare a^_a^ a,bx-axb„ ajt^-a^ (1.11,2)
12 VECTORS AND TENSORS |1.2
Using these expressions, it may readily be proved that
OA(OAC) -(a.c)6-(a.6)c. (1.11,3)
In connection with vector products it is of interest to distinguish a special
class of vectors, associated with rotation about an axis: typical vectors of
this class are the angular velocity of a body, and the moment of a force.
The direction of such a 'rotation-vector' is taken to be along the axis, in
the direction of translation of aright-handedscrew rotated in the sense of
the quantity considered. Thus the sign of a rotation-vector depends on a
convention as to the relation between the positive directions of translation
along, and rotation about, a given axis, and would be reversed if this con-
vention were altered. Since the same convention is used in the definition of a
vector product, the vector product of two ordinary vectors is an example of
a rotation-vector: the vector product of an ordinary vector and a rotation-
vector, in whose definition the convention is used twice, does not have its
sign altered if the convention is changed, and so is an ordinary vector.
In mechanical equations rotation-vectors can be equated only to other
rotation-vectors, and not to vectors of other types.

1.2. Functions of position

Any point in space may be specified either by the' potitkm-veetor' r giving its
displacement from some origin O, or by its Cartesian coordinates x, y, z (the
components of r), referred to a set of rectangular axes with 0 as origin. For
brevity, the phrase 'at the point r at time V will usually be contracted to
'atr.e.
A function <j> of position may be denoted by <f>(r) or 4{x,y, z), if scalar; if
it is a vector function, the functional symbol will be printed in heavy type,
as <f>(r), and its Cartesian components will be denoted by <f>J.r), <j>Jir), <f>,(r),
or, more briefly, by $„ ¢,, ¢,.
The equations of transformation of the operator whose components are
8/8x, 8]8y, 8\8z, from one set of axes Ox, Oy, Ox to another set Ox', Oy', Oz',
are the same as for a vector: for, in the notation of 1.1,
8_ 8x 8_ 8y^ 8_ 8z_ 8_
8x' " 8x' 8x+ 8x' 8y+ 8x' 8z
.8 8 8
~lidx + mi
8y+n*8z'
Thus the operator in question may be treated as a vector; it will be denoted
by 8/dr or V.
The result of the operation of 8/dr on a scalar function <f>(r) is called the
gradient of the function; it is a vector with components d<f>jdx, 8ft dy, 8$/8z.
1.21] VECTORS AND TENSORS 13

When <j>{r) is a function of the magnitude r alone, it is readily seen that

d<j> rd<f> . .

in particular, s-«2r. (1-2.2)

The scalar product of d/cV and a vector function <f>(r), i.e. 8/dr.<f> or
V .<f>, is called the divergence of the vector (sometimes written as div^);
it is, of course, invariant for a change of axes. Clearly

Similarly, if C is a vector whose x, y, z components are U, V, W, and

<f>(C) is any vector function of C,

where 0X, <f>u, </>t are the x,y, z components of 4>(C). Likewise if cS(C) is any
scalar function of C, an associated vector is

with c o m p o n e n t s ' , — . , ^ . In particular, if 0(C) = C* = U*+V*+W*,

it is readily seen that .^,
^=2C; (,.,,6)

more generally, if 0(C) = F(C*), where F is any function, it is easy to verify

that
8$ dF{C*) -8F . .

Again, if A is any vector independent of C, it is easy to verify that

^{C.A)-A. (1.2,8)

1.21. Volume elements and spherical surface elements

An element of volume enclosing the point r or (x,y, z) will be denoted by
the symbol dr. This must be distinguished from dr, which denotes the small
vector joining r to an adjacent point, and from dr, which denotes a small
increment in the length r. If Cartesian coordinates are employed, it is con-
venient to take dr as the parallelepiped dxdydz; using polar coordinates
r, 0,9, we take dr = r* sin 0drddd% and so on.
14 VECTORS AND TENSORS [1.3
If k denotes a unit vector, then the point whose position vector, relative
to an origin O, is k, lies on a sphere of unit radius (or 'unit sphere') with
centre O. Thus dk must be interpreted not as an element of volume, but as
an element of the surface of the unit sphere, or, what is equivalent, as the
element of solid angle subtended by this element of surface at O; the
element dk will be supposed to include the point k. The element may be of
any form; if k is specified by its polar angles 0, <p, it is appropriate to take
dk = sinddddt?.

1.3. Dyadics and tensors

Any two vectors a, b determine, relative to the set of axes chosen, a matrix,
each of whose components is the product of one component of a with one of b,
namely , , ,,
axox, axov, axo,,
aybx, ayb¥, avb„ (1.3,1)
a,bx, a,by, a,b,.
Such a matrix gives the ordered components, relative to the given axes, of an
entity called a dyadic, which will be denoted by ab.* It is to be noted that the
dyadic ba differs from ab unless the vectors a, b are parallel. The order of
the suffixes in the matrix may be remembered by aid of the symbol a^^,
indicating how the suffices succeeding x are disposed in (1.3, 1). 1
The components of the dyadic ab relative to a second set of axes Ox', Oy',
Oz' are given, in the notation of 1.1, by
a«-V = G>rV)(2**V)
r >
= ZS« r V y «-V- (*-3.»)
r »
Any 3 x 3 matrix (related to a set of axes Ox, Oy, Oz) of the type

('•3. 3)
W.i

of which the general term may be denoted by toaf, is said to constitute the
array of components (relative to those axes) of a second-order tensor (which
will be denoted by the symbol w), provided that the components, wa.F say,
relative to any other set of axes Ox', Oy', Oz', are such that

r *
This set of equations of transformation is the same as the set (1.3, 2) for the
components of a dyadic, so that every dyadic is a tensor.
* Thii symbol mutt be carefully distinguished from a.b. The insertion of the dot changes
the symbol for the dyadic to that for the scalar product of two vectors.
1.3] VECTORS AND TENSORS IS

The matrix of components of a dyadic or tensor must be carefully distin-

guished from the determinant which might be formed from the matrix; the
matrix is an ordered set of numbers, and the determinant is a certain sum of
products of these numbers.
The sum of two tensors is defined as the tensor whose components are
equal to the sums of the corresponding components of the two tensors.
The product of a tensor and a scalar magnitude k is defined as the tensor
whose components are each k times the corresponding components of the
original tensor.
If the rows and columns of the matrix (1.3, 3) are interchanged, a new
tensor is derived, which is known as the tensor conjugate to w, and denoted
by w. When this is identical with w, w is said to be symmetrical. If w is not
symmetrical, a symmetrical tensor denoted by w can be derived from it,
whose components are the means of the corresponding components of w
and w, so that = ,,— , . , .
= w = J(w + w). (1.3,5)
The components of w are
»xx. J(w*» + «Vx). 1(W» + W„),
i K x + ^xv). «W> i(«V. + «W).
!(»« + «>»). i ( » * + «"».). «W
The simplest symmetrical tensor is the unit tensor U, whose components
relative to any set of orthogonal axes are given by
u
xx = umi = u- = '> Ux» = uvx " etc- = o; (1.3,6)
it is easy to show that they are unaltered by transformation of orthogonal
axes.
The sum of the diagonal terms of the dy adic a b is ax bx+ay by + a, b, or a. b,
which is invariant for change of axes. Thus the sum to^+Wyy + w^ of the
diagonal terms of any tensor w will also be an invariant; it is known as the
divergence of the tensor. If the divergence of a tensor vanishes, it is said to
be non-divergent.
From any tensor w a non-divergent tensor, denoted by # , can be derived,
by subtraction of one-third of the divergence from each of the diagonal
terms: thus 0 , / . ,., , ,
* = w - \{v>„+»„„+wJ U. (1.3, 7)
The components of & are
1(2»« -tvm -w„), w^, to„,
«W JteWwr-Wxr-wJ, *"„,
w«. «V i(2to„-tc„-wm).
The symbols 0 and ~ may both be placed above a tensor symbol, as in w,
which in accordance with (1.3, 7) signifies
16 VECTORS AND TENSORS [1.31
o
Clearly the components of w are
K * « a - » W -*»»). 1 K » + »„*). K»»+ «>«).
1K.+".,). l(*w»,-Wa-w«). i(«».+w«).
iKc+»*.). iK„+»„,), i(2w»-«'xx-«' w )-
If h, i, j are three mutually perpendicular unit vectors (1.1, footnote),
hh + ii+jj = U, (1.3,9)
as is evident if the elements of the tensors are written out in full. Hence also

hh+a+jj = U=o. (1.3,10)

1.31. Products of vectors or tensors with tensors

The product w. a of the tensor w and a vector a is defined as the vector whose
components are given by ( w . f l ) a = £ ^ . (l.31, l}
0
The product a. w (which is in general not equal to w. a) is similarly defined
by the relation . * _,

Clearly w.a = a.w, U.a = a . U = a , (1-31,2)

and if p is any symmetrical tensor, p. a = a. p.
The simple product w. w' of two tensors w, w' is defined as the tensor with
components (w.W) M -^ - ^
S 0-31,3)
r
The double, or scalar product w: w' is defined as the scalar equal to the
divergence of w. w'; thus
w:w'«Z2>« / ,w*, = w':w, 0-3^4)
« t
that is, it is equal to the sum of the products of corresponding components of
w and w'. In particular, w: w is the sum of the squares of the components
of w, and U: U = 3.
From these definitions it follows that each of the above products satisfies
the distributive law of ordinary algebra; but the commutative law is not in
general satisfied, since, except in the case of the double product of two
tensors, the terms of the product cannot be interchanged without altering
the value of the expression.
An important particular case of (1.31,4) is
U:Yt = K„+wn + «>„ (1-31.5)
1.32] VECTORS AND TENSORS 17

which gives the divergence of w. Thus $t: U or U: # is zero (by the definition
of #). Also (i .3, 7) may be written
& = w-iU(U:w), («-3».6)
and so #:&' = # : { w ' - JU(U:w')}
= *:w'-J(#:U)(U:W)
= $/: w',
whence, by symmetry, &: &' = &: w' = w: $/'. (1.31, 7)
Again, it follows from (1.31, 4) that
w:w' = E2>^«>^ = w:w', (I-3 1 . 8 )
«0
and so w: w' = Jw: W + Jw: w'
= Jw: w' + Jw: w'
= w: w',
whence, by symmetry,
w:w'= w:w'= w: w'. (1.31,9)

1.32. Theorems on dyadics

Since dyadics form a special class of tensors, the above notations and results
for tensors also apply to dyadics.
If ab is symmetrical, ab = ba, and a must be a scalar multiple of b;
further, whatever a and b,
ab — ba, ab = i(ab + ba) = ba. (1.32,1)
0 0
The notation ab may be illustrated by the special case CC, where C is a
vector with amplitude C and components (U, V, W); the components of this
tensor are l / i - J C , UV, UW, j
VU, V*-\C*, VW, (1.32,2)
WU, WV, W* -|C».
The product of a dyadic ab by a vector d has a specially simple form; for
{(ab).d}a = 2 ( « * M , = 2 ¾ ½ = *.(b.d),
0 0
and so (ab).d= a(b.d). (1.32,3)
Similarly d.(ab) = (d.a)b. (132,4)
The scalar product of the vector w. a by a vector b is equal to the double
product of the tensors w, ab; for
(w.a).6 = 2 ( w . a ) A = S S » « ^ ^ . = E S > « X a 6 ) * = w:a
*-
« «A «0
(132.5)
Similarly 6.(0. w) _ 6a: w. (1.32,6)
18 VECTORS AND TENSORS [1.33
When w is itself of the form cd, it follows that
ab:cd = a.(b.cd) = a.{(b.c)d} = (a.d){b.c), (1.32,7)
whence also ab.cd = ac: bd. (1.32, 8)
From these results and (1.31, 6, 7) it follows that
0 0 0

-C&.CtCt-iCWJ-.CC,)
= (C,.C,)»-JCiCi. (1.32,9)
Again, if w is independent of C,

^ ( w : C C ) = ^&:CC) = 2W.C. (1.32,10)

1.33. Dyadics Involving differential operators

One of the vectors in a dyadic may be a vector differential operator such as
d/dr ( s V). If, for example, c is a vector with components u, v, to, the
8
components of =- c (m Vc) are
or
du dv 8v>'
dx' 'dx' 8x'
du dv dvo
dy' dy' d^' ("•33. 0
du dv dvo
dz' dz' to\

and the components of «- c (a Vc) are

1/ du_dv dto\ i/dv du\ i/dw du\

l\2dx dy~dz)' 2\dx+dy)' 2\dx+dz)'
i/du dv\
2\dy+dx)' 3\2dy dx~dz)' 2\dy+dz)' (133.2)
1 /dv du>\ 1/ *?_*f_M
2\dz+dx)' 2la* + 0y/' 3^2a* _ a* ayj";
If (as in 2.2) c0 denotes the velocity of a medium, the tensor ~-c0(aVct)

will be called the velocity-gradient tensor. Its symmmetncal part Vc„ and its
_o_
non-divergent symmetrical part Vc0 will be called respectively the rate-qf-
1.4] VECTORS AND TENSORS 19

strain and the rate-of-shear tensors; for these we use the notation
o
esVc0, 8 = Vc 0 . (1-33.3)
When the operator d/dr appears in the product of two tensors, or of a vector
and a tensor, attention must be paid to the order in which the terms occur, so
that in each case the terms on which the operator acts may be made clear. For
example, when a dyadic ab is multiplied by d/dr, both a and b being func-
tions of r, the operator, being supposed to act on the components of the
tensor, should be written before it; thus

or V.ab = (a.V)b + b(V.a). (1.33,4)

If, on the other hand, in the product w. a or a. w, the tensor w is of the form
a

=- b, these products should be written as follows

a.w = a.j£& = (a.V)6, w . a = a . w = a.(Vb). (1.33, 5)

Similarly in obtaining the double product of ab and Vc, (1.32, 7) should be

put in the form . _ ,. _. , ,.
r
ab:\c = a.(b.V)c. (1.33,6)
As further examples of the notation,
V{a.(Vb)} = (Va).(V6) + («.V)(Vfr), (,.33, 7)
and, if T is a scalar function of r,
V(a.vr) = (Va).vr+(o.V)vr. (1-33.8)
We may also note here the form of the components of the product V. p,
where p is a tensor function of position; the x-component is given by
«»•*-(£•').-%+£+%•• <••"•"
SOME RESULTS ON INTEGRATION

1.4. Integrals involving exponentials

Consider the integral f* e_aCt ^ ^

In this write s = aC*\ then it becomes equal to

j a -j(r+i) J" e-sfr-»ds = l a - * * " r(—\ (1.4, 1)

20 VECTORS AND TENSORS (1.41
if r > — i. In particular, if r is an integer,

'r^(?JC-^.i.U...—a-** 1 1 , (1-4. 2)
/; 2 2 2 2 2

or j " e-*CdC - *«-»<'+» ( ^ ) I, (1.4,3)

according as r is even or odd.

1.41. Transformation of multiple Integrals

Consider the multiple integral
jjj...F{u1,ut,...,un)du1dut...duK
extended over any range of values of the variables u. As is well known from
integration theory, if the variables of integration are changed to a set
vl%vt,...,vn, then the integral is transformed to
fff...P(vuvt, ...,»„) |J| </©,</»,... dvn,
where ^(¢,,11,,..,,11,)8^11,,11, «,), and J denotes the Jacobian
determinant . . -
du, cut oun
ST,' 8vi' '"' dvi
X » * - . »u »))
«(«>i. —* ^ ^

\to~n' SJn eS~m

The new integral extends over the range of values of the variables v that
corresponds to the range of values of the original variables u.
A special case of this result was used in 1.21, when it was pointed out that
the volume dr, whose expression in Cartesian coordinates is dxdydx, is
taken to be equal to r* sin 0drd0d<p in terms of polar coordinates r, 0, 9,
where * = r cos 9 sin 0, y = r sin 9 sin 0, z = r cos 0. It may easily be verified
that in this case „ .

and so \J\drd0d<? « r*sinOdrdOdf.

1.411. Jacobians
The general Jacobian of 1.41 may conveniently be denoted by d(u)/B{v),
regarding ult u,,..., «„ and t>„ t»„ ..., vn as components of vectors M and 0 in
ft-dimensional spaces. Alternatively, if each set is divided into two groups,
1.42] VECTORS AND TENSORS 21
(«i.« O . («m+i>«m+s> •••>«») and (vv ...,vr), (»r+„ ...,»„), regarded as
components of vectors of dimensions m and n — tn or r and n — r, namely
u', u" and »', t>", the Jacobian may be denoted by 8(u', u")/d(v', v"). This
notation can obviously be extended to the case of division of the n com-
ponents into more than two groups. For example, consider the case when
n = 6, m = 3, r = 3, so that«', «", »', v" are all three-dimensional vectors.
If we write out in full the determinants in the equations
B(U' + \LU",U") _8(u',u")
' d(vT,v") B{v',v"y I M " . »)
8(u',u' + k'u')_8(u',u") ttA„*

where k, k' are any constants, the truth of these equations is readily seen.
The notation adopted here is convenient and suggestive.

1.42. Integrals involving vectors or tensors

Let C be a vector with components U, V, W, and let dC denote an element
of volume in a space in which C denotes the vector of displacement from an
origin. Consider integrals (supposed convergent) of the type
\KC)dC,
taken over the whole of the C-space.
If cS is a function of odd degree in U or V or W, the part of the integral for
which 4> is positive cancels the part for which <f> is negative, and the integral
vanishes.
If $(C) = U*F(C), then by symmetry
fU*F{C)dC = fV*F{C)dC = $W*F{C)dC
= ij(U*+ V*+ W*)F(C)dC = if&F(C)dC. (1.42,1)
Thus jF(C)CCdC=iUjF(C)C*dC (1-42,2)
(the integrals involving the non-diagonal terms of CC vanish, since these
terms are odd functions of U, V, or W). Hence also
fF(C)CCdC-o, (1-42.3)
and if A is any constant vector
\F\C) {A. C) CdC = A.\F{C) CCdC
= iA-UjF\C)C*dC
= \A\F(C)C*dC. (1.42,4)
Again, let ¢{(0) = U*F(C). Using polar coordinates C, 0, 9 such that
22 VECTORS AND TENSORS (1.421
U - Ccos^, V = Csin0 cos<p, W = Csind sin?, it is found that
\F\C) U*dC = jjjF\C) C*coi*6.C* sin 6dCddd9
- IfjfFXQ&.&sinOdCdOdf
= lfF(C)CdC, (,.42,5)

since I cos*dsin0dd = - I sinddd.

Jo SJo
Similarly it may be proved that
jF\Q U*V*dC- A / ^ Q CM- (1 42. 6)
1.421. An integral theorem
Let w be any tensor independent of C. Then thefiveintegrals

(i) jF\C)CQCC:w)dC, (ii) fF{C)CC{CC:v*)dC,

(iii) fF{C)(X((X:*)dC, (iv) $fF{C)(CC:CC)dC,

(v) ^jF{C)C*dC,
represent identical tensors, if F{C) is any function of C such that the
integrals converge.
For, if (ii) is subtracted from (i), the result is

\\F{C)UC\CC:vi)dC

or JU(w .fFXQCHXdC),
which vanishes, by (1.42,3). Similarly the result of subtracting (ii) from
(iii) is
ijF(QCCC\U:<M)dC,
which likewise vanishes. Thus the equality of (i), (ii) and (iii) is established.
Again, by (1.31,7,9) the integral (i) is equal to

jF(QCC(CC:w)dC.
A typical diagonal element of this tensor is

jF\C)U\CC:w)dC.
Neglecting terms in the integrand which involve functions odd in U, V or W,
this may be written
JF{C) u\ um^ + K 4 W + w*w„) dc,
1.5) VECTORS AND TENSORS 23
or, using (1.42, 5, 6),

= fA4F(C)OdC.
Similarly the typical non-diagonal element

jF(C)UV(CC:%dC

reduces to the form 2JF(C) U217¾ dC,

which by (1.42, 6) is equal to
&WzujF{C)C*dC.
Thus the integral (i) is equal to

&WfF(C)C*dC,
i.e. to (v). Finally, the equality of the integrals (iv) and (v) follows from
(1.32, 9). Thus the theorem is proved.

By a similar argument, if W is a symmetric non-divergent tensor,

[F(C)CC(CC: W)»</C = ^ - W7W J* F(C)C*dC. (1.421,1)

1.5. Skew tensors

Let w be a tensor with conjugate w. If w = — w, the tensor w is said to be
antisymmetric, or skew. In this case

so that w has only three independent components.

If w 4= — w, the skew tensor ft is defined by
# = i(w-w), (1.5,2)
so that itjxfi = \{w^ -wtl) = -8,,,.
Since w = $(w + w), clearly w = w + w. 0-5>3)
The two tensors on the right of (1.5, 3) are the symmetric and skew parts of w.
The tensor w has components

0-5,4)
i(u>a-w„), i K „ -«>,„), 0.
24 VECTORS AND TENSORS 11.5
Thus it can be specified by the three components &„, &„, 6^, which are

These can be shown to be the components of a vector & which is of the

rotational type (I.I). This is called the vector of the tensor w or w.
In the special case when w is a dyadic AB, the vector is clearly \Af\B\ in
the case of a differential dyadic, such as Ve, the vector is J V A c, or \ curl c.
It is easy to show that w. a — a A &, so that
w.a»w.a+aA& (1.5,5)
An example occurring later (cf. 15.3) is
(Vc,).Vr - W9.VT+iVTA curie,
-e.Vr+JVrAcurlco.
 2
PROPERTIES OF A G A S : DEFINITIONS
AND THEOREMS

2.1. Velocities, and functions of velocity

The linear velocity of the mass-centre of a molecule will be denoted
vectorially by the letter c, and its components relative to Cartesian
coordinates by («, v, to); its magnitude c will be called the molecular speed.
The velocity-vector c may be regarded as the position-vector, or vector of
displacement from an origin, of a point in a velocity-space or velocity-domain:
this point is called the velocity-point of the molecule.
This representation of a velocity by a point in an auxiliary velocity-space
suggests the analogy used in 1.2 between scalar or vector functions of r
and similar functions of c. Thus a scalar function of velocity, 0(c) or
<j>{u, v, to), has an associated gradient function 8<j>/dc or d<f>/du, d<j>jdv, d<f>/dw;
similarly with a vector function of velocity, $(c), there is associated a scalar
function
» AJ**JA»JA<
dc'v 8u dv dtc
corresponding to the divergence. Again, a triple integration with respect to
it, v, w corresponds to a volume integration in the velocity-space; this will be
denoted by
j...dc,
the symbol dc denoting an element of volume (of any shape) in the velocity-
space, surrounding the point c. Unless the contrary is expressly stated, such
an integration is supposed to extend over the whole velocity-space.
The phrase 'velocities in a range dc about the value c' will be contracted
to 'velocities in the range e, dc', or, more briefly, to 'velocities in the range
dc'. Similarly 'during a time-interval dt including the instant t' will be
contracted to 'during a time t, dt' or 'during a time dt'. Likewise 'the
volume-element dr containing the point r' will be contracted to' the volume-
element r, dr' or, more briefly, to 'the volume-element dr'.
The position and velocity of a molecule can both together be represented
by a point in a space of six dimensions, whose coordinates in that space are
the three components of r and the three components of c.
If the molecule is not merely a mass-point, but is of finite size, it will in
general possess rotatory motion; and if it is not rigid, it may also possess
vibratory or other internal motion. The condition of such a molecule may be
represented by a point in a space of n dimensions, where n is the number of
[25]
26 PROPERTIES OF A GAS: [2.2
independent positional and velocity (or momentum) variables needed to
specify the configuration and motion of the molecule.
For example, if the molecule is rigid, it has six positional variables
(three of location and three of orientation), and six velocity variables (three
translatory and three rotatory); in this case n = 12. Some of these may be
unimportant in particular cases: thus if the molecule be spherically sym-
metrical, its three variables of orientation will have no dynamical interest.
If it is also smooth, its angular velocity will be unalterable by collisions, and
its three angular-velocity variables are also without further interest; in this
case n a 6 as for a point-molecule. But if the molecule, though spherical, be
rough, its three angular-velocity variables are important, affecting and being
affected by collisions; in this case « = 9.
If a gas is composed of molecules of more than one kind, we may associate
a separate velocity-domain with each kind; each domain will have the
appropriate number of dimensions for that kind of molecule, corresponding
to the number of independent variables of position and velocity for such
molecules.

2.2. Density and mean motion

In a non-uniform, variable, continuous medium, the density at the point r at
time t is defined as the limit of the mean density (mass/volume) in a small
volume dr surrounding the point r, as the dimensions of dr diminish
indefinitely, the limit being supposed independent of the shape of dr. This
definition cannot usefully be applied to a medium like a gas, composed of
discrete molecules, especially when these are separated by distances large
compared with molecular dimensions; it leads to a value of the density which
varies rapidly from point to point, and with passage of time, and does not
correspond to any ordinary measurable quantity. Hence some other defini-
tion is necessary.
Consider in the first instance a 'simple' gas, that is, a gas composed of
molecules all of which are alike. Let the mass of any molecule be m. Let dr
denote a small volume surrounding the point r, which is large enough to
contain a great number of molecules, while still possessing dimensions small
compared with the scale of variation of such macroscopic quantities as the
pressure, temperature, or mass-velocity of the gas. (For example, a cube of
edge one-hundredth of a millimetre contains about 2-687 x I o l ° molecules in
a gas 'at standard temperature and pressure'.*) Let the mass contained in
dr be averaged over a time t, dt which is long compared with the average time
that would be taken by a molecule to cross dr if undeflected, yet short com-
pared with the scale of time-variation of the macroscopic properties of the
gas. (For example, since the mean speed of molecules in a gas at S.T.P. is
* Thit phrase is commonly abbreviated t o ' at S.T.P.'; it signifies at a temperature of 0 °C„
and a pressure of 760 nun. of mercury (1-013 x 10" dynes/cm.*).
2.21] DEFINITIONS AND THEOREMS 27
several hundred metres per second, a molecule would move one-hundredth
of a millimetre in less than io~ 7 of a second.) Then the averaged value of the
mass contained by dr will be proportional only to its volume, and will not
depend on its shape. It will be denoted by pdr; p is termed the mass-density
or the density of the gas at r, t.
Similarly, the number of molecules in dr averaged over dt is proportional
to dr. It will be denoted by ndr; n is called the number-density of the mole-
cules. The quantities p and n are connected by the relation
p = nm.
Both p and n are functions of position and time; when it is desired to indicate
this, they may be denoted by p(r, t), n{r, t).
The mean molecular velocity at r, t in a simple gas, denoted by c 0 , is defined
by the vector equation . . . „
(ndr)c0 = 2lc,
where the summation on the right is extended over the ndr molecules in the
small volume r, dr, both ndr and Ec being averaged over a small time-
element t, dt. Similarly we obtain any other mean value; for example, the
mean speed is (Lc)/(ndr). In particular, the mean momentum of a mole-
cule at r, t is equal to mc0; like c 0 itself, it is, in general, a function of r and t.
The translational motions of the individual molecules in dr may be
specified either by their 'actual' velocities c (i.e. their velocities relative to
some standard frame of reference) or by their velocities C relative to axes
moving with some velocity c', so that C = c — c'. Generally c 0 , the mean
velocity of the gas at the point, will be adopted as the velocity c'; C is then
written C, and is termed the peculiar velocity of the molecule; also C is the
peculiar speed. The mean peculiar velocity of molecules at r, t is c 0 — c 0 , that
is, zero. The components of C, c' and C, c 0 are denoted respectively by
(V, V, W), («', v',»') and (U, V, W), («0, v0, tv0).

2.21. The distribution of molecular velocities

The distribution of velocities among the large number ndr of molecules in
dr can be represented by the distribution of their velocity-points c in the
velocity-space. Owing to the continual changes of velocity by molecular
encounters, and to the appearance and disappearance of velocity-points, as
molecules pass into or out of dr, the distribution of velocity-points will vary
with time. As, however, the number ndr of velocity-points is very large, it
will be assumed that, just as there is a number-density of molecules in the
actual space occupied by the gas, so there is a statistically definite number-
density of the n dr velocity-points in the velocity-space. This number-density
is supposed to be proportional to the volume of the element dr, but not to
depend on its shape. It will in general be a function of r and of t as well as of
position in the velocity-space; it will therefore be denoted by f(c,r,t)dr.
The definition implies that the probable number of molecules which, at the
28 PROPERTIES OF A GAS: [2.22
time t, are situated in the volume element r, dr, and have velocities lying in
the range
6 e, dc, is equal
n to ,, ,, • ,
f(c,r,t)dcdr.
This does not mean that the given element dr actually contains this number
of molecules having velocities in the range c, dc at the time t; this is the
average number of such molecules when thefluctuationswhich occur in a
short time dt are, as it were, smoothed out. The function/(c, r, t) or, briefly,/
is termed the velocity-distribution function. Its definition involves probability
concepts; any result in which it appears will be a result as to the probable, or
average, behaviour of the gas.
The distribution of velocity-points is clearly unaffected if the origin in the
velocity-space is changed to the point c'. In this case the position-vector of
a velocity-point is changed to c — e' or C'\ the volume-element dc is now
denoted by dC, and contains the same number of molecules, i.e.
f(C' + c',r,t)dC'dr.
This will often be written as/(C, r, t) dC'dr, with an appropriate change in
the nature of the function/. The change of variable from c to C can be made
in any integral without altering its value. Usually c' is taken to be c0, so that
C becomes C, the peculiar velocity, and/becomes f(C,r,t).
The whole number of molecules in the element dr is obtained by inte-
grating/^ </r or fdC'dr throughout the whole velocity-space; this number
is, by hypothesis, ndr. Hence
n - //(c, r, t)dc - / / ( C + c\ r, t) dC.
Clearly the function / is never negative, and it must tend to zero as c
or C becomes infinite. It is assumed to be finite and continuous for all
values of t.
The function /gives the number-density of points in the six-dimensional
space of 2. i, in which the coordinates of a point are the components of c and r
for a molecule. Similarly, if the molecule also possesses rotational or vibratory
motion, account can be taken of this by employing a space of more dimen-
sions, as in 2. i; the number-density/of representative points in this space is
the generalized velocity-distribution function. The function /then repre-
sents the distribution of density and of translatory, rotatory and internal
motions.

2.22. Mean values of functions of the molecular velocities

Let <j>(c) be any function of the molecular velocity c. The function <f> may be
a scalar, vector or tensor; for example, it may be c itself, or C, or a com-
bination of components such as uv* or uW, or again a function of the speed
c or C". It may also be a function of position and time, and will therefore be
denoted by $(c, r, t) or, in terms of C , by $(C' + c', r, t) or, with an appro-
2.3J DEFINITIONS AND THEOREMS 29
priate change in the function <f>, by 0(C, r, t). Any such function (f> we
call a molecular property.
Let 2 0 denote the time-average during t, dt of the sum of the values of 0
for the ndr molecules in r, dr, and write
2^ = n$dr. (2.22,1)
Then ^ is the mean value* of 0 for the molecules at (or near) the point r.
It is a function of r, t, even if 0 itself does not explicitly involve position and
time. It can be expressed in terms of the velocity-distribution function/;
for each of the f(c,r,t)dcdr molecules in dr, whose velocities are in the
range c, de, contributes 0(c, r, t) to 20, and their aggregate contribution is
<}>fdcdr. By integration over the whole velocity-space, we obtain
20 = drftfdc,
whence n$ = j<f>fdc = f<f>fdC
In particular, since c0, by definition, is the mean molecular velocity at r, t,
nc0 = jcfdc; (2.22, 2)
clearly C = o, V = P= W = 0. (2.22, 3)

2.3. Flow of molecular properties

Consider the passage of molecules across a small element of surface dS,
moving in the gas with any velocity c'. The surface element is supposed to
have a positive and a negative side. Let n be a unit vector drawn normal to
the element in the direction from the negative to the positive side. The
passage of a molecule across dS is regarded as positive or negative according
as the molecule crosses to the positive or negative side of dS. The velocity
C of a molecule relative to dS is equal to e - c', or, if C is the peculiar velocity
of the molecule, to c0 + C—c'.
Consider the molecules whose peculiar velocities lie in the range C, dC.f
If one such molecule crosses the element dS in a time dt so short that we
may ignore the possibility of the molecule encountering another during dt,
then at the beginning of dt the molecule must lie somewhere inside the
cylinder on dS as base, I with generators specified in length and direction by
* The significance of the bar placed over 4> is here totally different from that of the bar
placed over the symbol for a tensor, in 1.3. When, as msy happen, 41 denotes a tensor, or
when a single bar occurs over a dyadic (and therefore tentorial) expression (as in (a.31,3)
for example), it is necessary to know in which of the two possible senses it is used. In the
remainder of this book a single bar placed over the symbol for a tensor will indicate that
the mean value is to be taken. The double bar, on the other hand, always has the same
meaning as in 1.3.
f For brevity, the phrase 'molecules whose peculiar velocities lie in the range C, dC
is contracted to 'molecules C, dC'.
X Here, as frequently, it is convenient to regard molecules as mass-points. Thus the exact
position of a molecule can be specified, and also the exact time at which it crosses the
element dS.
30 PROPERTIES OF A GAS: [2.3
— C'dt (see Fig. i). Thus, if dr denotes the volume of this cylinder, the
number of molecules C, dC crossing dS during dt \sfdCdr.
Now dr " ± C" cos 6 dt dS, where 0 is the angle between C and n, the sign
+ or — being chosen so as to make the expression for dr positive. But 0 is
acute or obtuse (and so cos 0 is positive or negative), according as theflowis
positive or negative. Thus the flow is expressed, both in magnitude and
8lgn
*by f(C)dC.C'cosddtdS.
But C"cos0 is the component of C normal to dS, which is equal to C'.n;
we denote it by C„.# Then theflowof molecules C, dC across dS in time dt is
C'J(C)dCdtdS. (2.3,1)

Fig. 1
The netflowof molecules across dS during time dt is found by summing over
all velocity groups, i.e. by integrating over the whole range of C, which gives
dSdtjC^fiC)dC = dSdtnCi. (2.3, 2)
The number of molecules crossing from the negative to the positive side is
similarly .
dSdt \ C'J(C)dC.
Jc.-xt
The molecules that cross the element carry with them their energy,
momentum, and so on. The net rate of transport of such quantities across
dS can be found by methods similar to those just used. Thus let <f>(C) denote
any scalar molecular property; each of the molecules C, dC that cross
dS carries an amount ¢((7) of <j> with it. Hence the contribution of the group
to theflowof <}> across dS during dt is, by (2.3, 1),
4>(C).C'J{<C)dCdtdS,
• The suffix n in thii symbol, and in p, and C„ (2.31), has reference to n, the normal to
dS; it has no relation to it the number-density.
2.3] DEFINITIONS AND THEOREMS 31
and the total net flow of <f> across dS during dt is
dSdtjC'n <f>{C)f(C)dC = dSdtriC^C). (2.3, 3)
The expression (2.3, 2) for the flow of molecules across dS is a special case of
this result, corresponding to (j>(C) = 1.
The rate of flow of <f> across dS per unit area is obtained by dividing
(2-3. 3) b y dSdt, giving
»c;#c). (2.3,3')
Since C'n = C. n, this is the component along n of the vector
nC'<j>(C).
Now C = C+c0-c', so that
HC^(CJ = nC^(C) + n(co-c')?(CJ- (2.3.4)
Hence the component of this vector in any direction n represents the rate of
flow of the property <j>(C) per unit area across a surface normal to this
direction, and moving with the velocity c'.
The number-flow is given by the vector n(c0 — c'), since in this special case
0(C) = 1, C<j>(C) = C = o. This vector is the product of the number-
density and the mean velocity of the molecules relative to the surface
element. If c' = c0 the number-flow is zero, whatever the orientation of the
surface.
This enables us to interpret the second term on .the right-hand side of
(2.3, 4). Its component normal to dS represents the contribution to the rate
of flow of <f>{C) due to the net number-flow of the molecules across dS, each
carrying, on the average, the quantity <j>(C) of ¢. The first term, on the other
hand, is independent of the number-flow, and its component normal to dS
represents the rate of flow of 0 when the number-flow vanishes, i.e. when
the element shares the mean motion of the gas at the point.
The vector nC<p(C) may conveniently be termed the 'flux-vector' for the
property (j>. The rate of flow of 0 across unit area of a surface which moves
with the gas is the component of the flux-vector normal to the surface. If,
however, the surface is in motion relative to the gas, the rate of flow is
increased by the normal component of the relative velocity, multiplied by
n<f>(C). In general, when the flow of some molecular property across a
surface is considered, it will be assumed that the surface is moving with the
gas.
In the case of a vector property <f>(C) of the molecular velocities it is con-
venient to consider the flux-vector of each component of <f>, which will be a
scalar quantity, as in the preceding discussion. Thus the flux-vector of the
component 0 a , where a stands for any one of *, y, and z, is nC<f>a(C).
These results may be generalized so as to apply to molecules that are free
to rotate, or that possess other internal degrees of freedom; <f> may then
depend on the variables 'specifying the orientation, angular velocity, and
32 PROPERTIES OF A GAS: [2.31
internal state, as well as on the translational velocity. Theflowwill again be
represented by an expression of the form (2.3, 4), where averages are now
to be taken over all values of the velocities and also of the other variables
specifying the motion.
2.31. Pressure and the pressure tensor
The case in which 0(C) is equal to some component of the molecular
momentum mc is of great importance, because of its connection with the
pressure distribution.
At the boundary of the containing vessel, every molecule that impinges
and rebounds exerts an impulse equal to the difference between its momenta
before and after impact. When such impacts are sufficiently numerous and
sufficiently uniform in distribution, they simulate a continuous force on the
boundary, equal in magnitude and direction to the rate at which momentum
is being communicated to the surface by impacts. The force per unit area of
the surface is called the pressure (or 'boundary pressure') on the surface at
the point. The surface clearly exerts an equal and opposite pressure on the
gas. The pressure is a vector, whose direction is not necessarily normal to
the surface at the point considered.
Suppose that dS is an element of the surface of the containing vessel,
moving with the velocity c', and let the internal face be taken as the negative
face. Let the direction of the outward normal to dS be that of the unit vector
n, and let pn denote the pressure on the wall at this point. Then by the
definition of /»„, the momentum communicated to dS in the time dt is
pndSdt.
As in 2.3, we may show that the total momentum of the molecules im-
pinging on the element dS in time dt is equal, before impact, to
dSdti C;«c/(C)rfC,
where the suffix ( + ) signifies that the integration is extended only over that
part of the velocity-range for which C'H, the n-component of the velocity of
a molecule relative to dS, is positive (since only molecules for which C'n is
positive can impinge on the surface). Similarly the total momentum of the
molecules rebounding from dS during dt is

dSdti (-C'n)mcf(C)dC,

the suffix ( - ) signifying that the range of integration is over all values of C
for which C'n is negative: the minus sign before C'n is introduced because
C'n enters into the integrand through the expression for the number of
molecules C, dC leaving dS during dt, and this number is essentially positive.
Thus the total momentum communicated to dS during dt, which is the
difference between the momentum of the impinging molecules and that of
2.31] DEFINITIONS AND THEOREMS 33
those rebounding from the surface, is equal to

dSdt[j C'nmcf(C)dC- j (-C;)mcf{C)dc\

=dSdt\C'nmcf{C)dC
= dSdt.nmC^c.
Hence pn = nmC^c = pC^e. (2.31, 1)
The velocity c' of the wall will not in general equal the mean velocity c0
of the neighbouring gas. Experiment shows that the behaviour of a gas in
the neighbourhood of a wall may be rather complicated; some molecules
do not immediately rebound off the wall, but enter it or adhere to it for
a time before they return to the gas. If the gas is neither condensing upon
nor evaporating from the surface, the total number of impinging molecules,
namely, .
dSdt\ C'J{C)dC,

must equal the number rebounding, which is

dSdtj_(-CM(C)dC;

hence dSdtjC^f(C)dC = 0
or C;=o.
That is, the mean velocity relative to the wall, for the molecules in its neigh-
bourhood, has no component normal to the wall; the gas may, however, have
a mean motion relative to the wall, in a direction parallel to the surface.
Using this result and (2.22, 3), we have

C'„c = Cfc0 + C) = C'nc0 + C'nC = C'nC

= (n:C')C = {n.(V-c')}C
= {n.(C+c 0 -V)}C
= (n.C)C+{n.(c D -c')}C
= (n.C)C=C n C.
Hence from (2.31, i), using (1.32,3), we obtain the following alterna-
tive forms for pn:
pn = pC^C = P(n.C)C = n .pCC
= n.p = p.n, (2-3 1 . 2 )
34 PROPERTIES OF A OAS: {2.32
where p is the symmetrical tensor defined by the equation
'pU*, pUv, pDw;
p = p C C = pVU, PV\ pVW, (2.31,3)
fWU, pWV, pW, .
This tensor depends only on the distribution of the peculiar velocities;
its components are given by
P*x=pU\ Pn=pV\ p„=pWi, (2.3:,4)
Py. = P*,-pVW> P«=Pa = pWU, pxy=pvx = pUV. (2.31, S )
The pressure distribution at any point P within the gas is defined as follows.
As in 2.3, let dS be any surface element containing P, and let n be its unit
positive normal vector. Let dS share the mean motion of the gas at P, so that
c' = c0,C = C = o, and therefore C'n = 0. T\\enpn, the pressure across dS,
towards its positive side, is defined as the rate of flow of molecular momentum
mc across dS, per unit area, in the positive direction. This is given by (2.3, 3'),
if Mistaken to be the vector function mc. Consequently pn = nC'nmc = pC'nc,
as in (2.31, 1); since C'n = o, this is equivalent to (2.31, 2), which therefore
holds in the interior as well as at the boundary of the gas. In the interior,
however, n may have any direction.
The distribution of pressure across planes in all directions through P is
therefore determined by the pressure tensor p.
When n is x, the unit vector in the direction of Ox, Pn**Px** P x . of
which the components are pzx, pIy, pa; the other components of p are com-
ponents of the similarly defined vectors/>„, p,. Thus the components of p are
the components of the pressures across surfaces parallel to the three co-
ordinate planes.
The above results are valid whether the molecules possess only energy of
translation, or have internal energy of rotation, vibration, or any other form.
2.32. The hydrostatic pressure
The normal component of the pressure on a surface normal to the unit
vector n is -—-, —s
n.pn = n.pCnC = pCl. (2.32,1)
Thus the normal component of the pressure on any surface is essentially
positive; that is, the normal force exerted on any surface by the gas is always
a pressure, and never a traction.
The sum of the normal pressures across three planes through any point P,
parallel to the coordinate planes, is
/ ^ + / ^ + / ^ = / ^ + V*+ »")
= />C». (232, 2)
2.33] DEFINITIONS AND THEOREMS 35
Thus the mean of the normal pressures across any three orthogonal planes is
JpC*. This is called the mean hydrostatic pressure, or the pressure at P; we
denote it by p. By (..31, 5) /> = J p : U . (2.32,3)
If the non-diagonal elements of the tensor p vanish, and the diagonal
elements are equal,
n then .. . . .
P =Pn= Puu = P—
and p =/>U.
In this case (cf. (1.31, 2))
/>„ = p . n = / > U . n = / > n ,
so that the pressure on any surface element through the point r is normal
to the surface: its magnitude is independent of the orientation of the surface,
and equal to the hydrostatic pressure. These are the conditions satisfied
by the pressure in hydrostatic problems; hence such a pressure system, in
which p is a scalar multiple of U, is said to be hydrostatic.

2.33. Intermolecular forces and the pressure

In the above discussion the whole of the pressure of the gas on the walls of
the containing vessel, or across a hypothetical internal surface, was tacitly
assumed to be due to the transfer of momentum. In actual gases, intermole-
cular forces also contribute to these pressures. Since at distances large com-
pared with the molecular diameters these forces are usually attractive, they
add an attractive component to the total pressure. This component is
relatively small for gases at ordinary temperatures: but in liquids or solids
the importance of intermolecular forces may equal-or exceed that of the
momentum transfer.
Attractions between the gas molecules operate to reduce the pressure on
the walls in the following way. The average resultant force exerted on a
molecule in the interior of the gas by the other molecules of the gas is in
general zero, because only the adjacent molecules exert any appreciable
attraction, and so, unless there is a very steep density gradient at the point,
the attractions are approximately equal in all directions; but at the walls the
gas lies on one side only, so that there is a resultant attraction inwards, which
is roughly proportional to the number of attracting molecules in the neigh-
bourhood, i.e. to the density of the gas. Consequently the momentum im-
parted to the wall by each molecule impinging thereon will be smaller than
if there were no attractive force, by an amount proportional to the density p.
The rate at which molecules strike the wall is also proportional to p; hence
the correction to the momentum pressure is proportional to /?*, whereas the
momentum pressure itself varies as p. Thus the correction becomes of
greater importance as the density increases; it is small for gases at ordinary
pressures, and at temperatures well above their critical temperatures.*
* For a detailed discussion of the effects of these factors on the equation of state of a gas,
see R. H. Fowler's Statistical Mechanics, chapters 8 and 9 (1928, 1936).
36 PROPERTIES OF A OAS: [2.34
In this book attention will be directed more particularly to the deviation
of the actual pressure system from the hydrostatic system/>U. Intermolecular
forces at distances large compared with the molecular dimensions have little
effect on the pressure deviations, except for very dense gases; accordingly we
ignore them. The finite size of molecules, which results in a reduction of the
effective volume of the vessel containing the gas is also of importance mainly
for dense gases; this is taken into account in Chapter 16.
2.34. Molecular velocities: numerical values
The quantities/) and p appearing in the equation
P - \pC* (2.34. 1)
are directly measurable. From their experimental values the corresponding
values of C* can be found. For example, at standard temperature and pressure
(1-013 x 10* dynes/cm.*) the densities of hydrogen and nitrogen are respec-
tively 8-99 x io~*g./cm.* and 1-25 x lo^g./cm.'. The corresponding values
for V(£*) *re 1839 and 493 m./sec. For a gas in a uniform steady state,
C" ^{C*)l voS6
(cf. (4.11, 4))*, thus the corresponding values of C are 1694 and 454 m./sec.
These mean speeds are very large, and appear at first sight startling.
Evidence confirming their order of magnitude is, however, supplied by
other phenomena. If the basic assumptions of the kinetic theory of gases are
valid, sound must be transmitted by the motions of individual molecules,
and the velocities of sound in these gases are known to be similar in order of
magnitude to the above molecular mean speeds. Again, the speed of effusion
of a gas from a vessel into a vacuum through a small aperture should be of
the same order of magnitude as the mean molecular speed; experiment
shows that the speeds of effusion for different gases are, in fact, of this order.

2.4. Heat
The amount of translatory kinetic energy possessed by the molecules in the
element r, dr at time t is ndr\mc*. Writing c =• c 0 +C, we may express the
energy in the form , , ,., 7;, ;—.

or \pdr.(%+ndr.\mC*.
Since pdr is the mass of the gas contained in dr, the first term in the last
expression represents the kinetic energy of the visible or mass motion of the
gas. The second term is the kinetic energy of the invisible peculiar motion:
its ratio to the first is C*/cJ- In 2.34 it was shown that C is very large for
ordinary gases at S.T.P., the value of ^/( £*) being several hundred metres per
second. Hence unless c0 is much greater than is usual, C*/cJ is very large,
and there is much more hidden energy of peculiar motion than visible kinetic
2.41] DEFINITIONS AND THEOREMS 37
energy; for example, in the case of hydrogen at S.T.P., if r 0 = iocm./sec.,
the ratio is 3-4 x 108. In addition, there may be further hidden molecular
energy, kinetic and perhaps also potential, corresponding to molecular
rotations, vibrations, and so on.
In the kinetic theory this hidden molecular energy, or rather that part
which is communicable between molecules at encounters, is identified with
the heat energy of the gas. Thus the heat energy per unit volume, or the
heat-density, in a gas whose molecules are point-centres of force and there-
fore possess only translatory kinetic energy, is \nmC* or \pC*. This is true
also for a gas whose molecules are smooth rigid elastic spheres, for though
these may also possess rotatory energy, this is not communicable between
molecules at collision. In general, however, the molecules will possess other
kinds of communicable energy, whose amounts vary from one molecule to
another; the total heat energy £ of a molecule is the sum of this com-
municable energy and the peculiar kinetic energy \tnC*, and the heat-
density is nE.
Here JwC 2 and E are supposed expressed in mechanical units, and the
heat-density nE will be in the same units. If expressed in thermal units, the
heat-density is nE/j, where J is Joule's mechanical equivalent of heat
(4-185 x io T ergs/cal., or 4-185 joules/cal.)

2.41. Temperature
Two systems of temperature-reckoning are in common use among physicists.
In experimental work they generally use the empirical temperature measured
by expansion (mercury or gas) thermometers: in theoretical work they use
the absolute temperature of thermodynamics. In the kinetic theory, on the
other hand, the temperature T of a gas in a uniform steady state at rest or in
uniform translation is defined directly in terms of the peculiar speeds of the
molecules, by the relation ^ _ ^ ^ f)

where k is a constant, the same for all gases, whose value will be assigned
later (2.431); it is called the Boltzmann constant.
At a given density and temperature, the interchange of energy between
the translatory and internal motions at molecular encounters establishes a
balance between the mean translatory and internal energies. Thus, for a gas
in a uniform steady state, the total mean thermal energy J? of a molecule is
a function of the T defined by (2.41, 1). For a gas not in a uniform steady
state, the temperature T at any point is defined as that for which the same
gas, when in a uniform steady state at the same density, would have the same
mean thermal energy E per molecule at that point.
The kinetic-theory definition of temperature, being applicable whether
or not the gas is in a uniform or steady state, is more general than that of
thermodynamics and statistical mechanics, where only equilibrium states
38 PROPERTIES OF A GAS: (2.42
are considered. It is of importance to examine, however, whether the kinetic-
theory definition is in agreement with that of thermodynamics if the gas is in
equilibrium. Before so doing we proceed to deduce certain relations from
the definition (2.41,1).

2.42. The equation of state

An immediate consequence of the definition of temperature is that the
hydrostatic pressure/) of a gas in equilibrium is given by
p = \nmC* = knT. (2.42, 1)
This formula applies also to a gas not in equilibrium if the molecules possess
only translatory energy; for a gas with internal energy it need not be exact,
though still correct to a close approximation. The formula is in agreement
with the well-known hypothesis of Avogadro, according to which equal
volumes of different gases, at the same pressure and temperature, contain
equal numbers of molecules.
Consider now a mass M of gas contained in a volume V. The number of
molecules in the mass M is Mjm; the number-density n is therefore M/mV.
On substituting in (2.42, i), this takes the form
pV = M.M/m) T. (2.42, 2)
This important relation embodies the well-known experimental laws of
Boyle and Charles, which are closely followed by many gases, at moderate
or low densities, and at temperatures well above their critical temperatures.
These laws may be stated as follows:
Boyle's Lam: For a given mass of gas at constant temperature the product
of the pressure and the volume is constant.
Charles's Law: For a given mass of gas at constant pressure the volume
varies directly as the absolute temperature.
The deduction of these laws from kinetic-theory principles and definitions
affords some measure of justification for the latter. By itself, however, it
does not suffice to show that the kinetic-theory definition of T is in accord
with the thermodynamic definition for equilibrium states.
The relation betweenp, V and T for a given mass M of gas in equilibrium
is called the equation of state of the gas. The above simple form of this
equation is only an approximation to the equation of state as found for actual
gases; the error of the simple formula pV oc T becomes considerable at high
pressures and low temperatures. This is to be ascribed, not to any fault in
the above kinetic-theory definition of temperature, but to the neglect, in
deriving the expression JpC* forp, of such factors as the finite size of mole-
cules, their fields of force at large distances, and, when the gas is near the
point of liquefaction, their tendency to aggregate into clusters. The effect
of these neglected factors becomes considerable in precisely those con-
2.43] DEFINITIONS AND THEOREMS 39
ditions of high pressure and low temperature under which large deviations
from (2.42, 2) are observed.
From (2.42, 2) other important relations can be obtained. The (chemical)
molecular weight W of a gas is defined as izm/mc, where m is the mass of a
molecule of the gas, and mc that of an atom of carbon, = 1*993 x io""**g.
A mass W grams of the gas is called a gram-molecule or mole of the gas;
naturally it is different for different gases. It is a convenient mass to consider,
because for all gases it contains the same number of molecules, W/w or 1 i\mc.
This number is Loschmidt's number (6-022 x 1023). Some writers call it
Avogadro's number; but this term should strictly be applied only to the
number of molecules in one c.c. of gas at S.T.P., 2-687 x I0 '*-
Suppose that the mass M used in (2.41, 2) is a gram-molecule W. Then
/>V = RT, (2.42, 3)
where R = k\V/m = 12k/mc. (2.42,4)
Clearly R has the same value for all gases. It is called the gas-constant per
mole.
Again, from (2.42, ., 4) p=*pT. (2.42,5)

2.43. Specific heats

Let unit mass of a gas in equilibrium be enclosed in a constant volume. To
increase its temperature from Tto T+ST, a certain amount of heat must be
added, which, if ST is small, will be proportional to ST; we write it as
cvST.
The coefficient cr is called the specific heat of the gas at constant volume.
Since the gas does no mechanical work against external pressure during the
process, the added heat cvST must go entirely to increase the heat energy of
th(0gas. The number of molecules in unit mass is (i/m), and so the initial
heat energy is E/m; this is increased by cvST when T is increased by ST;
hence
8E/m = c.8T,

or, proceeding to the limit, c„ = - i ^ \ , (2.43, 1)

where (dEldT)y denotes the rate of increase of E with respect to T, when V

is kept constant.
If the thermal energy consists only of energy of translation of the mole-
cules, E = \mC*, and E = \kT, by (2.41, 1); hence

c
c= (243. 2 )
40 PROPERTIES OF A GAS: (2.43

in mechanical units, or c. = -*— (2.43,2')

v TJ
" 23m '
in thermal units.
If, instead of the volume, it is the pressure of the gas which is kept constant
while T is increased by ST, the volume V will increase by 8V, and on putting
M - 1 in (2.42, 2), we find
pSV »(k/m)ST.
In the expansion mechanical work of amount £<JV will be done, and the heat
supplied to raise T must provide this energy p8V as well as the increase
SEjm in the heat energy of the gas. Writing cp8T for the required amount of
heat, we have
cp8T-p8V + 8E/m

- {k8T+dE)lm.

Hence , , . ^ +1 ( ^ (2-43.3)
in mechanical units, the suffix/> denoting that the pressure is kept constant.
An increase in pressure, or density, of a gas of assigned temperature can
affect E only by increasing the number of pairs of molecules whose fields of
force overlap, and which in consequence possess mutual potential energy.
However, in the relatively rare gases for which Boyle's and Charles's laws
are valid, at any given moment all save a negligible fraction of the molecules
are independent systems, and E may be taken as depending only on T, and
not on p and V. Thus for such a gas
*
cP = - + c . (2.43.4)
jn

in mechanical units, or c„ = — +c„ (2.43,4')

in thermal units. 1.
If the specific heats c„ and cp are multiplied by the molecular weight W,
we obtain the specific heats C„ and Cp per mole of the gas. Thus in mechanical
unit8
„ WdE , ,

W/, dE\
cv
-«+C„, (2.43.6)
by (2.42,4). It is an experimental fact that Cp — C„ has nearly the same value
for all actual gases under moderate conditions of pressure and temperature,
a fact which further supports the principles and interpretations here used.
2.431] DEFINITIONS AND THEOREMS 41
The ratio cp/cv or CPICV of the specific heats is denoted by y. Thus

vr = C-S = ~J?
c C '
and (2.43, 4, 6) can be written in the forms

<^(y-i) = */'«. (243.7)

C„(y-i) = fl. (2.43,8)
For a gas possessing no communicable internal energy, it is clear from
(2.43, 2, 7) that y - f = r 6 6 .

2.431. The kinetic-theory temperature and thermodynamic

temperature
It is now possible to establish the consistency of the kinetic-theory definition
of temperature and the thermodynamic definition. Let a mass M of a gas
undergo a small change of state, such that the temperature increases by ST
and the volume by S\r. Then the energy it receives is

Me.*T+P*V-T[*£*T + %™\.

by (2.42, 2). Since cn like E, is a function of T alone, the expression in the

bracket denotes the increment of a function S of T and V. If the gas under-
goes an adiabatic change, i.e. a change in which no energy is supplied, the
change in S must be zero, that is, S is constant; in an isothermal change, at
temperature T, the energy received is T multiplied by the change in S.
Suppose now that the gas is taken round a Carnot cycle working between
lower and upper temperatures Tt and 2J; then S returns to its original value
when the cycle is completed. Since it is unaltered during the adiabatic
processes, its increase AS at temperature 7j must be equal and opposite to
its decrease at temperature Tt, and the heats gained and lost at temperatures
7i and Tt respectively are Ty AS, 7^ AS. Thus the efficiency of the cycle is
7i J AS-r i AS I _7\\
r»AS V Tj-
This proves that our T is proportional to the temperature on the thermo-
dynamic scale: the function S is the entropy.
It remains to consider the constant k introduced in (2.41, 1). This was
taken to be the same for all gases: this implies the assumption that the mean
peculiar kinetic energy of translation is the same for molecules of different
gases at the same temperature. We shall show in 4.3 that in a gas-mixture in
equilibrium the mean peculiar kinetic energies of molecules of the different
constituent gases are the same, which is a similar result: but to establish the
actual result it is necessary to consider the equilibrium of two gases separated
3 CUT
42 PROPERTIES OF A GAS: [2.44
by a diathermanous wall. This problem lies in the domain of statistical
mechanics rather than of kinetic theory, and the reader is referred to books
on that subject for a proof of the result in question. Alternatively we may
regard the result as established by experiment, since Avogadro's hypothesis
and the law Cj, — C, =• const, both follow directly from it.
In practice, k is chosen such as to make the kinetic-theory temperature
coincide with the thermodynamic temperature measured in degrees K. That
is, k is such that the difference between the temperatures of melting ice and
boiling water at standard atmospheric pressure is ioo°; the zero of tempera-
ture is found to be approximately equal to —273-15° C. The value of A can
be determined if we know either the number-density of molecules in a gas
at given/> and T, or (cf. (2.42,4)) the mass of a molecule. The determination
of either of these is a matter of some difficulty, but the values of each have
been found for several gases. The different determinations agree in giving*
k - 1*3806 x io- M ergs/degree.
The determination of the gas-constant R is much easier. It is found that
R •- 8-314 x ioT ergs/degree,

or, in thermal units, — = 1-9865 cal./degree.

J

2.44. Specific heats: numerical values

The values of y, Cp, Cv, cv, Cp - C„ are given for several gases in Table 1.
The values refer to a temperature of 150 C. and a pressure of 1 atmosphere;
departures from the Boyle-Charles law affect the values appreciably, even at
this pressure. The units of Cp and Cv are calories (at 150 C.) per degree per
mole; for c„ the units are calories per degree per gram.f
As Table 1 shows, y has very nearly the value 5/3 for the monatomic gases
helium, neon and argon, which are on many grounds believed to possess no
internal energy communicable at ordinary encounters.} This is a further
confirmation of the kinetic-theory interpretations.
For other gases, let us write
% = iNk, (2.44.1)
• B. N. Taylor, W. H. Parker and D. N. Langenberg, Rev. Mod. Phyt. 41. 375 (1969).
t The valuea given in this table are based on the following sources: J. R. Partington and
W. G. Shilling. The Specific Heats of Gates, p. aoi (Benn, 1924); J. Hilsenrath et at.,
U.S. National Bureau of Standards, Circular 564 0955); F. Din, Thermodynamic
Functions of Gases (Butterworth. 1956). The values given for D, and Xe are values
calculated by a method that gives good agreement with experiment for other gases
(A. Michels, W. de Graaff and G. J. Wolkers, Physica, as. 1097 (1959). for D,;
A. Michels, T. Wassenaar, G. J. Wolkers and J. Dawson, Physica, at, 17 (1956)» for
Xe). The value of y for Ne was found by W. H. Keesom and J. A. van Lammeren,
Physica, 1, 1161 (1934).
X The deviation from y — 5/3 for xenon is due to departures from the perfect-gas law*.
2.45] DEFINITIONS AND THEOREMS 43

Table 1. Specific heats

Gas r c. c. C c,-c.
Air 1402 6960 4965 0172 «•995
He 1668 497 298 0-745 109
A 1-669 4975 298 00747 1-995
Ne 1-668 ,
Xe [1-679] [5-03*l [2996) [0-0228] [2036]
H, 1 408 687 488 2-42 199
D. [1399] [698] [499] [•238] [••99]
N, 1403 696 496 0177 2-00
CO 1-402 697 497 0177 2-O0
NO 1-400 700 500 O167 2'O0
1-398 703 S03 200
o, 804
0157
CI, '356 5 93 O-084 2-11
H,S 1 340 815 608 0-178 2-07
CO, 1-303 885 6-79 OI54 2-06
N.O 1-300 885 6-81 0-155 204
SO, 1-284 962 749 0-II7 213
NH, i-3'8 8-75 664 O-390 2-II
CH, 1-310 849 648 0404 201
C,H4 1250 10-25 820 0293 205
C,H, I-20 12-42 1036 0345 2-06

so that for monatomic gases N = 3, while for other gases we expect that
W>3.Then kN k / N\
C = C =
° ™- » mV + V' (244 2)
'
V=1+N- (2-44.3)

If c„ is independent of T, as is found to be approximately the case for many

gases (monatomic and otherwise) over a considerable range of temperature,
then N is independent of T. Consequently, apart from a possible additive
constant, E = INkT.
For many diatomic gases, as the above table shows, 7 = 1-4 very approxi-
mately ; this corresponds to N = 5, indicating that the communicable internal
energy is two-thirds the peculiar kinetic energy of translation. For polyatomic
molecules the values of y are less than 1 -4, and the corresponding values of
N are greater than 5; this implies a still larger proportion of non-translatory
energy.

2.45. Conduction of heat

An important flux-vector (cf. 2.3, p. 31) is that giving the rate of flow of heat
energy, corresponding to <j>(C) = E. We denote this flux-vector by q,
so that —--
q = nEC. (2.45, 1)
3-a
 44 PROPERTIES OF A OAS: [2.5

Thus the rate offlowof heat across a surface through the point r, normal to
the unit vector n, is equal to —
H
f . n = nECn
per unit area. The vector q is termed the thermal flux-vector.
2.5. Gas-mixtures
All the above definitions and results may be generalized to apply to a mixture
of gases. The definitions of the number-density and velocity-distribution
function of each of the constituent gases are analogous to those employed for
a simple gas in 2.2 and 2.21. The velocity-distribution function/,(c„ r, f) and
the number-density n, of the rth constituent are connected by the relation
"$-ffJLc»r>t)<ICr (2.5.1)
The number-density n of the whole gas is given by the relation
» = 2»»,. (2-5.2)
a
The different masses and mean molecular speeds of the different constituents
render it pointless to consider the velocity-distribution function of the
whole gas.
If the mass of a molecule of the rth constituent is m„ the partial density of
this constituent is p„ where , .
P. = ».«.. (2.5. 3)
and the density p of the whole gas is given by
P = 2/». = 2 » . « r (2-5.4)
« *
We shall frequently refer to molecules of the rth constituent as molecules m,.
If $ is any function of the velocities of the molecules, its mean value <J, at
any point, for molecules m„ is given by
«.& - / / . & * „ . (2-5.5)
and the mean value ^ for all molecules of the mixture is given by
«? = 2 > . ? . = £f/.&<fc„. (2.5,6)
The mass-velocity c„ of the gas at any point is defined by the equation
pc0 = 2 jf.m.c.dc, - 2p.c/. (2.5. 7)
» »
it is not the mean velocity c of the molecules, but a weighted mean, giving
to each molecule a weight proportional to its mass. The momentum of the
gas per unit volume is the same as if every molecule moved with the mass-
velocity €¢.
The peculiar velocity C, of a molecule m, in a gas-mixture is defined by the
CqUati0n
C.-c.-c. (2.5.8)
Clearly 2 P . C , = 0. (2.5, 9)
2.5J DEFINITIONS AND THEOREMS 45
The temperature T of the gas at any point is defined, as for a simple gas,
by the equation
\mC* = I E / / . 1 « . C J A . = \kT, (2.5. to)

for a gas in a uniform steady state. For a gas in a more general state it is
defined by E = E(T), where E(T) is the value of E for a uniform gas in a
steady state, with the density and composition of the actual gas at the point
considered, at the temperature T.
The partial pressure of any constituent, on the surface of a containing
vessel or on an internal surface moving with the mass-velocity of the gas, is
defined as the mean rate at which momentum of that constituent is com-
municated to, or transferred across, unit area of the surface; the total pressure
of the gas on the surface is the sum of the partial pressures of the constituents.
It follows as in 2.31 that the pressure-tensor p is defined by the equation

P = S P, = 2 n,m,C,Ct = nmCC. (2. 5 , n )

a a

As before, the pressure on a surface normal to the unit vector n is p.n.

The mean hydrostatic pressure/) of the gas at any point is Jp:U, as in
(2.32, 3); hence for a gas in equilibrium, by (2.5, 10, 11),
p = JnmC* = knT, (2.5,12)
which is equivalent to Boyle's and Charles's laws. By (2.5, 2), this implies that
/> = E {kn,T).
a

Thus the hydrostatic pressure of the mixture at a given temperature is equal

to the sum of the hydrostatic pressures p, which would be exerted by the
constituents if each separately occupied the same volume and were at the
same temperature. This is Dalton's law.
Finally, the vector q of thermal flow is connected with the molecular
energy E by the equation —
y 4
*' q = nEC, (2.5,13)
as for a simple gas.
 3
THE E Q U A T I O N S OF B O L T Z M A N N
A N D MAXWELL

3.1. Boltzmann's equation derived

In Chapter 2 it was shown that the macroscopic properties of a gas can be
calculated from the velocity-distribution function/. This function can be
determined from a certain integral equation first given by Boltzmann. In
order to indicate Maxwell's association with the ideas thus formulated by
Boltzmann, Hilbert* termed it the Maxwell-Boltzmann equation.
In deriving the equation it is assumed that encounters with other molecules
occupy a very small part of the lifetime of a molecule. This implies that only
binary encounters are important.
Consider a gas in which each molecule is subject to an external force mF,
which may be a function of r and t but notf of c. Between the times t and
t + dt the velocity c of any molecule that does not collide with another will
change to c+Fdt, and its position-vector r will change to r+edt. There
are/(c, r, t) dcdr molecules which at time t lie in the volume-element r, dr,
and have velocities in the range c, dc. After the interval dt, if the effect of
encounters could be neglected, the same molecules, and no others, would
compose the set that occupy the volume r+cdt,dr, and have velocities in the
range c + Fdt, dc: the number in this set is
/(c+Fdt, r+cdt,t+dt) dcdr.
The number of molecules in the second set will, however, in general differ
from that in the first, since molecular encounters will have deflected some
molecules of the initial set from their course, and will have deflected other
molecules so that they become members of the final set. The net gain of
molecules to the second set must be proportional to dcdrdt, and will be
denoted by {djjdt) dcdrdt. Consequently
{f{c+Fdt,r+cdt,t+dt)-f(c,r,t))dcdr = ^dcdrdt.
On dividing by dcdrdt, and making dt tend to zero, Boltzmann's equation
for/is obtained, namely
df df df df „df „df 'df dj

* D. Hilbert. GrundtSgt enter allgemeinen Theorie der linearen Integralgleichungen, p. 269

(Teubner, 1911).
t A special case in which the force on a molecule depends on its velocity is considered in
Chapter 19.
[461
3.11] THE EQUATIONS OF BOLTZMANN AND MAXWELL 47

*/- % <3-i.a)
where ^/denotes the left-hand side of (3.1, 1); in vector notation

The quantity def/dt defined above is equal to the rate of change, owing to
encounters, in the velocity-distribution function / at a fixed point. It will
appear later that 8ef/St is expressible as an integral involving the unknown
function/. Thus Boltzmann's equation is an integral (or integro-differential)
equation.
The generalization for a mixture of gases is

J,
~ dt+c''dr+ a
dc, dt' U
'I,4J
where w.,F, denotes the force on a molecule m, at r, t, and S^fjdt denotes the
rate at which the velocity-distribution function f, is being altered by en-
counters. The equation can also be modified to apply to more general
molecular models (2.1, 2.21); in the case of rotating molecules possessing
spherical symmetry,/depends only on c, r, t and the angular velocity to, and
the equation for/ has the same form as(3.1, 1). For more general models/
will involve further variables, specifying the orientation and other properties
of a molecule; terms corresponding to these variables must in general appear
in Boltzmann's equation.

3.11. The equation of change of molecular properties

Another important equation may be derived from Boltzmann's equation as
follows. Consider first a simple gas. Let <j> be any molecular property as
defined in 2.22. Multiply Boltzmann's equation by 4>dc and integrate
throughout the velocity-space; it is supposed that all the integrals obtained
are convergent, and that products such as ^/tend to zero as c tends to infinity
in any direction. The result may be written as
f<f>®fdc = nA$, (3.11, 1)

where ntymjt?j[dc. (3.11,2)

The significance of A^ is readily seen; (def/dt)dc measures the rate of

change in the number of molecules with velocities in the range c, dc, per unit
volume at r, t, owing to encounters. Consequently $(dj\d()dc represents
the rate of change in the sum ~L<f> extended over all the molecules of this set,
due to the same cause. Similarly f$(dff/dt)dc is the rate of change by en-
counters in £ 0 , summed over all the molecules in unit volume. But 2 ^ = n$,
48 THE EQUATIONS OF BOLTZMANN AND MAXWELL [3.12

and, since encounters do not in themselves modify the number-density n,

the rate of change of $ by encounters is equal to

sj^*-*
Equation (3.11,1) can be generalized to apply to a mixture of gases; for
therthconstituent f<f>.9J,dc, » n,Aft, (3.,,, 3)
where A$, is equal to the rate of change of ^, by molecular encounters.
A modified form of (3.11, 1) is, moreover, satisfied by the generalized
velocity-distribution function of 2.21, 0 being then a function of further
variables besides c.
3.12. ^/expressed in terms of the peculiar velocity
If the peculiar velocity C ( s c-c 0 ) is used as an independent variable
instead of c, the meanings of 8/8t and d/dr are changed, since now C, not c,
is to be kept constant while performing the differentiation. Hence in (3.1, 3)
df/dt and dfjbr have to be replaced respectively by

8t BtdC Br \8r c V'9C

in order to take account of the implicit dependence of/on t and r through the
dependence of C on c0. Also df/dc becomes df/dC; hence the expression for
S>/becomes

YrTtdC+(Co + C)
{&r-{drC°) 3 C / + F - dC
D 8 8 ,
Let m = ^ + c0.^, (3.12, .)
so that D/Dt is the 'mobile operator', or time-derivative following the
motion, as in hydrodynamics. Then
Df 8/ / Dc0\ 8f 8f 8
^ ' m + Cdr+[F-^bl)-BC-&CCBrCa- (3>I2 2)
'
3.13. Transformation of fy&fdc
Suppose that ¢, like/, is expressed as a function of C, r, t. The various terms
obtained on substituting for £tf from (3.12, 2) into j(f>@fdc (or J$&fdC)
can be transformed by means of relations such as

J V | r f C = ~.\*CfdC-\d£.CfdC= -^.niC-nC.^, (3.:3, 2)

/ < f ^ - / / K : > - - / & * - - • & • <>••»>

3.2] THE EQUATIONS OF BOLTZMANN AND MAXWELL 49
In (3.13, 1, 2) the variable C is not included in the differentiations D/Dt,
d/dr, since C is now regarded as an independent variable. In (3.13, 3) the
term involving <f>f, obtained by integrating by parts, vanishes because, by
hypothesis, <fcf-*-o as U-* +00. Clearly (3.13, 3) is one component of the
equation _
J ^ c = -«g£. (3.13,4)
By a precisely similar argument

J^C«—.Jic—^U-."ga (3.13,5)
where (cf. i .3) U denotes the unit tensor.
Using these results in (3.11, 1), we obtain the equation

-n\Di + C
S?^[F-Dt)dC-BCC-B?Cn}- (3,3,6)

This is called the equation of change of 5$; it is a generalization by Enskog of

an equation of transfer due to Maxwell. Maxwell's equation refers to a
function <fi(c) of c alone, and does not introduce the peculiar velocity.

3.2. Molecular properties conserved after encounter; summational

invariants
Some important results can be deduced from the equation of change without
actually evaluating A^, because certain functions of the velocities of the
molecules are conserved during encounters; that is, their sum for the mole-
cules participating in an encounter is unaltered by the encounter, so that
A^ = 0. Such functions, which may be termed summational invariants for
encounters, are of fundamental importance in the theory of gases, both
because they can be measured throughout the fluctuations in the condition
of a gas, and also because, as will appear later, their mean values, if completely
specified at any instant as functions of position, in general determine the
whole condition and future course of the gas.
For every gas, of whatever kind, three such summational invariants are
^0) = I( d/» = mc, ^ = E, (3.2, 1)
where, as in 2.4, E denotes the total thermal energy of a molecule. For
special types of gas there may be one or more additional summational
invariants (cf. 11.24).
The statement A^ 1 ' = 0 merely implies that the number-density of mole-
cules is unaltered by encounters. Similarly A«[»(t> = 0 expresses the principle
SO THE EQUATIONS OF BOLTZMANN AND MAXWELL [3.21

of conservation of momentum (relative to axes moving with velocity c0),

while A^ 3 ' — o expresses the principle of conservation of energy. Although
d/*> is a vector it is convenient to refer to the three conserved functions by the
single symbol ^ ° .
It is sometimes more convenient to refer the momentum d/*> and the
energy xjf™ to axes fixed in space, instead of to axes moving with the gas.
Thus, for example, if the molecules possess only energy of translation the
summational invariants can be taken as
^i) = It yn = mc, ^ > = \mc*. (3.2, 2)
In the case of binary encounters, the conservation of (^° is expressible in
the form ptt+^w _ ^ 0 _ yt> „ 0> ( 3 2> ^
where ^ ° , ffl refer to the two molecules before encounter, andft**,$F to
these molecules after encounter.
Any linear combination of the three conserved functions ^ ° is also a
summational invariant, but no further summational invariant, linearly
independent of ^ \ 0 a ) and ^ ' as given by (3.2,2), can exist for molecules
whose energy is purely translatory. This is because an encounter between
two such molecules involves two disposable geometrical variables (for
example, if the molecules are rigid elastic spheres, these two variables may
be the polar angles 0, <p of the line of centres at collision; cf. 3.43); when these
two variables are eliminated from the six scalar relations which express the
six components of c', c[, the velocities after encounter, in terms of the six
components of the initial velocities c, elt only four general scalar relations
between the two sets of six components are obtainable. But we already have
four such relations, expressing the conservation of energy and of the three
components of momentum; hence no additional independent relation, valid
for all encounters, is possible.

3.21. Special forms of the equation of change of molecular properties

By substituting each of the functions 5^0 for <f> in (3.13, 6), important special
forms of the equation of change are obtained.
Case /. Let ^ •= i^> = 1; then ? = 1, $C = o, 00/3C - o, DQjDt - o,
dip/dr = 0, A^ =» 0. Thus (3.13,6) becomes

J P»9fdc • 5 7 + » | ; . « » - o; (3.21, 0
this is the equation of continuity, expressing the conservation of number of
molecules (or mass) in the gas. It has the alternative forms
D\nn d
Dt + ^ c « = °>
(321. 2)
Dp 8 D\np B
3.21] THE EQUATIONS OF BOLTZMANN AND MAXWELL 51
2)
Case II. Let <j> = ^ = mU; then 3 = o, H0C = pUC=px (cf. 2.31),
00/3r = o, £>0/£>< = o, 00/3C = (m, o, 0), (d<f>jdC)C = o, A3 = 0. Hence
the equation of change becomes

jW<*/dc = l.pt-p^-tgf) = o;

this (cf. (1.33,9)) is one component of the equation of momentum for the gas,
namely . „ . D.
fWdc-frP-pp-jg) =0. (3.21,3)

Equations (3.21, 2, 3) are identical with the equations of continuity and

momentum derived for a continuousfluidin hydrodynamics; they provide a
justification for the hydrodynamical treatment of a gas.
Case III. Let <j> = ^ = E; then nlf>C = q (cf. (2.45, 1)), 8$/dr = o,
D<j>/Dt = 0 and, since E depends on C only through the contribution of the
kinetic energy of translation, 8$/dC = \m 3C*/3C = mC\ thus 00/3C = o,
and n(d$i'dC)C = /»CC = p. Since A3 = o, the equation of change becomes

J^W^-D-gf) + ^ a ^ . c 0 + ^ . g + p : l c 0 = o. (3.21,4)

Since dEjdT = \Nk (cf. (2.44,1)) equation (3.21, 4) may be transformed to

Dt - - J V t e | P : 5 ^ + 8^*J' (3 2,,S)
-
using (3.21, 1). This, or more properly (3.21,4), is the equation of thermal
energy for the gas. Let equation (3.21, 4) be multiplied by drdt: then
D(nE)/Dt.drdt represents the increase of thermal energy during the time
dt in a volume dr moving with the gas. This may be interpreted from a
macroscopic standpoint as the sum of (i) the energy brought into dr by the
net inflow of molecules into the element, (ii) the gain due to the greater
energy, as distinct from the greater number, of the inflowing as compared
with the outflowing molecules, and (iii) the work done on the element by the
pressures on its surface as it varies in shape and volume during the time dt;
these three quantities are represented by the last three terms on the left of
the equation, with their signs reversed.
Equations (3.21, 2, 3, 5) represent the maximum information which can
be derived from the equation of change without determining the form of the
velocity-distribution function. To determine it we must first find explicit
forms for the expressions 8ef/dt and A3. This involves an investigation of
the statistical effect of encounters.
52 THE EQUATIONS OF BOLTZMANN AND MAXWELL [3.3

3.3. Molecular encounters

Exact expressions for S^/dt and A$ can be given only when the nature of the
interaction between molecules at encounter is known. Only in relatively
few cases can atomic theory describe the process of encounter exactly. It is
therefore necessary to assume some law of interaction: the appropriateness
of the assumed law can be tested by comparing the results deduced from it
with experimental results.
Physical data for gases, relating to the deviations of the equation of state
from Boyle's law, show that, at distances large compared with molecular
dimensions, molecules may exert a weak attractive force on each other,
whereas at distances of the order of molecular dimensions they repel each
other strongly. Moreover, at an encounter between complex molecules
possessing internal energy some interchange of this energy with energy of
translation may occur. Assumptions as to the nature of the forces between
molecules at encounter must take these facts into account.
Various special models, chosen for their physical simplicity, or for the
mathematical simplicity of their laws of interaction, have been studied.
One of the earliest and simplest molecular models is a rigid, smooth, and
perfectly elastic sphere. The impulse between two spheres at collision here
represents the repulsive force between molecules at a close encounter. The
representation can, however, only be approximate, since molecules, being
complicated electronic structures, cannot closely resemble rigid spheres;
the interaction between them varies continuously as they approach one
another. This fact is better represented by treating a molecule as a point-
centre of force, the force depending on the nature of the interacting mole-
cules and on their distance apart. One simple assumption is that the force
is always repulsive and varies inversely as some power of the distance.
A better representation of the facts is afforded, however, if the force is
supposed to change sign at a certain distance, beyond which it is an attraction.
The elastic sphere model may also be improved by supposing the spheres to
attract one another weakly, with a force depending on the distance.
If the molecule is represented either as a smooth sphere or as a point-
centre of force, no provision is made for a possible interchange between
internal energy and energy of translation. Such molecules may be termed
smooth: their internal energy can be neglected, as it does not vary with the
temperature. In most of this book, only smooth molecules are considered:
but models permitting interchange between internal and translational energy
are considered in Chapter 11.
All the smooth models we consider possess the property of spherical
symmetry. The molecules of a monatomic gas closely approximate to such
symmetry, but diatomic and polyatomic molecules diverge widely from it,
by reason of the concentration of mass in the atomic nuclei. Thus our
theoretical results do not apply strictly to diatomic and polyatomic gases;
3.41] THE EQUATIONS OF BOLTZMANN AND MAXWELL 53

since, however, in our calculations we average over all possible orientations

of pairs of molecules at encounter, many of our results may be expected to
apply approximately to such gases, if we ascribe to our spherically sym-
metrical molecules a field which is the average of the true field over all
possible orientations of the molecules.

3.4. The dynamics of a binary encounter

Consider the encounter of two molecules of masses mlt tnt. Since only
smooth and spherically symmetrical molecules are considered, the force
which either exerts on the other is directed along the line joining their
centres, A, B; it may arise only at contact, or may act when the molecules are
at any distance from each other, and be equal to some function of the
distance A B. It is supposed that any external forces (gravitational, electric,...)
which act on the molecules are so small compared with those brought into
play during the encounter that their effect can be neglected in a consideration
of the dynamical effect of an encounter.
The phrase ' before the encounter' will refer to the time before the mole-
cules have begun to influence one another appreciably, so that each is
moving in a straight line, or (more accurately) close to the asymptote of the
orbit which it describes under the influence of the other; the phrase 'after
the encounter' is to be interpreted in a similar way. With these conventions
the velocities before and after the encounter have definite values, which will
be denoted by clt c 2 (before) and c[, c 2 (after).* It is desired to express either
pair of velocities in terms of the other pair, and of any geometrical variables
required to complete the specification of the encounter. Since the motion is
reversible, the relation between the two pairs of velocities must be reci-
procal. The details of the encounter are of no importance for our purpose;
we wish only to know the relation between the initial and final velocities.

3.41. Equations of momentum and of energy for an encounter

Let m0 = m, + m2, M, = mjm0, Mssmt/m0, (3.41,1)
so that A/, + M 2 = i . (3.41,2)
The mass-centre of the two molecules will move uniformly throughout the
encounter; its constant velocity G is given by
m0G = mlCl + mzct = m ^ J + mjcJ. (3.41, 3)
Let gIlt g'2l and glt, g'it denote respectively the initial and final velocities
of the second molecule relative to the first, and of the first relative to the
second, so that
£21 = c * - c i = - t f i s . £21 = c 2 - c i = -tfi*- ( 3 4 1 . 4)
* In 3.52 c\, c'2 are used to denote the initial velocities, and c,, c, the final velocities, in an
inverse encounter. These uses of the symbol c' must be distinguished from that in 3.3
(p. 27), where c' denotes the velocity of moving axes of reference.
54 THE EQUATIONS OF BOLTZMANN AND MAXWELL [3.41
The magnitudes of gn and glt are equal, and can both be denoted by;;
likewise for the final relative velocities; thus
gn'gWg, g'n-i'n~g'> (3-4*. 5)
By means of (3-41, 3, 4) we can express cx, c„ c[, c't in terms of G, gtl,
and^thus ei-G-Mtgn, ct = G+Mlgtu (3.41,6)
c[-G-Mtg'n, c't = G+Mlg'tl. (3.41,7)
Hence a knowledge of G and gn or of G and g'n is equivalent to a knowledge
of c t and c, or of c[ and c't, that is, of the initial or final state of motion.

Fig. a

The mutual potential energy of the two molecules is zero both before and
after the sncounter; thus the equation of energy gives
1(1»! cf+m t c$ = i( w i c i*+ w t c i*)-
Using (3.41, 6,7), it is readily shown that
l(m,ci + m,el) = '+M1Af^»), J
im^+MiMtg*),
(3-4I. 8)
»+M1M1*'*)J
i K t f + «,<?) = Jmo(G»-
Hence g-g\
3.43] THE EQUATIONS OF BOLTZMANN AND MAXWELL 55
so that the relative velocity is changed only in direction, and not in magnitude,
by the encounter. The dynamical effect of the encounter is therefore known
when the change in direction of g2l is determined.
These facts are illustrated in Fig. 2. The initial velocities clt c2 are repre-
sented by Or,, Oct, and G by OG, where G divides c^ in the ratio m2:mv
The ends of the lines Oc[, Oc2 representing c[, c2 are likewise collinear with
G, which divides c[c'2 in the same ratio m2:mv The lines ctct and c[c't repre-
sent g21 and g21. Thus CjC2 = c[c2, c t G = c\G, c2G = c2G.

3.42. The geometry of an encounter

Considerations of momentum and energy alone do not suffice to determine
the direction of g'2V As will now appear, this direction depends not only on the
initial velocities C,, c2 (or on G, g2l) but also on two geometric variables which
complete the specification of the encounter.
Consider the motion of the centre B of the second molecule relative to the
centre A of the first (or to axes moving with A). Since the force between the
molecules is directed along AB, this motion will be confined to a plane
through A; let the curve described by B be LMN (Fig. 3 a). The asymptotes
PO, OQ of this curve are in the directions of the initial and final relative
velocities, gtl and g'2V and so the plane of LMN is parallel to the plane
cj Gc[ of Fig. 2. Let P'A be a line parallel to PO, so that it is in the direction
of g2V The direction of AP' is then fixed by the initial velocities c „ ct; the
orientation of the plane LMN about AP' is, however, independent of these
velocities, and is thus one of the additional variables of the encounter. We
specify it by the angle e between the plane LMN and a plane containing
AP' and a direction fixed in space, such as that of Oz.
The angle x through which g21 is deflected depends, in general, on the
magnitude £ of the initial relative velocity, and on the distance b of A from
either of the asymptotes. This distance b is the second of the additional
geometric variables of the encounter.
The functional relation between x< ° a n d g depends on the law of inter-
action between the molecules. This law is involved in the following discussion
solely through the dependence of x o n ° a l , d g- Hence, both for generality
and brevity, x will be retained as an unspecified function of b and g as long
as possible.

3.43. The apse-line and the change of relative velocity

The orbit LMN of the second molecule relative to the first is symmetrical
about the apse-line, or line joining the two molecules when at the points of
closest approach. This apse-line passes through O, the intersection of the
two asymptotes, and bisects the angle between them. In Fig. 3 a the direction
of the apse-line is represented by OAK, K being the point in which OA
produced cuts the unit sphere of centre A. The unit vector AK is denoted
56 THE EQUATIONS OF BOLTZMANN AND MAXWELL [3.43

,K

My

A
Fig- 3(<0- Direct encounter.

Ptn

-k ..*»

Fig- 3W- Invene encounter (*ee p. 6a).

3.44] THE EQUATIONS OF BOLTZMANN AND MAXWELL 57

by k. The components of gtl and g'n in the direction of k are equal in

magnitude, but opposite in sign, so that g'n.k = - f f « k ; the components
perpendicular to k are equal. Hence g2l and g'n differ by twice the com-
ponent of gn in the direction of k, so that
ffsi -8'n = 2(*2i • k) k = - 2(g'il. k) k. (3.43, 1)
Combining this with (3.41, 6, 7) it follows that
c[ - c, = 2M/gu. k) k = - 2M^g'n. k) k , |
U 43
c i - c , = -2M1(^l.k)k = 2Ml(^l.k)k.J ' '

Thus when k, c „ ct are given, the velocities after encounter are determinate,
and knowledge of k is equivalent to knowledge of the geometrical variables
b and e.
If the force between molecules is always re-
pulsive or always attractive then for a given
direction of gn the point K can range over one
or other of the unit hemispheres having AP'
as axis; if the molecules repel one another, the
pole of this hemisphere is in the same direction
as gn, while if they attract it is in the opposite
direction. The possible positions of K are
similarly related to —g'n: for a repulsive force
£ 2 , . k > o , and g'n.V.<o\ for an attractive
force these inequalities are reversed.

3.44. Special types of Interaction

Figure 4, drawn by Maxwell,* shows a number
of the orbits described by one molecule, relative
to another molecule represented by S, when Fig. 4
they exert a mutual repulsive force varying as
the inverse fifth power of their distance apart; they correspond to equal
values of g, but different values of b. With this law of force, or, more
generally, when the force varies as any inverse power of the distance, the
families of paths for different values of g differ only in scale; with more
general laws of force this is not true.
When the molecules are rigid elastic spheres the apse-line becomes
identical with the line of centres at collision. In this case the distance <r„
between the centres of the spheres at collision is connected with their
diameters cr„ <rt by the relation
0-,, = 1(0-, + 0-,), (3-44.')
and (cf. Fig. 5) b = o-„sin \fr = o-,2cos Jx, (3.44, 2)
• J. C. Maxwell, Collected Papers, vol. i, 42; Phil. Trans. R. Sec. 157, 49 (1867).
58 THE EQUATIONS OF BOLTZMANN AND MAXWELL |3.5
where ijr is the angle between gn and k; clearly i/r = $("-#)• This model
is unique in that X depends only on b, and not on g.

3.5. The statistics of molecular encounters

In evaluating def/dt and A$5, we suppose that encounters in which more than
two molecules take part are negligible in number and effect, compared with
binary encounters. This implies that the gas is of low density, so that
encounters occupy only a small fraction of the life of a molecule.
The probability is zero that at a given instant, in a finite volume of gas,
dr, there shall be any molecule whose velocity is exactly equal to any

F'g- 5

specified value e out of the whole continuous range; it is necessary to con-

sider a small but finite range of velocity, dc. Thus the probable number of
molecules of the first kind, in dr, having velocities within the small range
Cn dclt is fidc^dr, where /i stands for /ifo, r, t); likewise the probable
number of molecules of the second kind, in dr, having velocities within the
range c„ dct, hftdctdr, where/, stands for/2(c2, r, t).
The probable number of encounters in dr, during a small interval dt,
between molecules in the velocity-rangesrfc„dclt will in the same way be
zero if the geometric encounter-variables b, e are exactly assigned; it is
necessary to suppose that b, e also lie in small finite ranges db, de. The ranges
dcvdct,db,de are regarded as positive quantities; as they are small, the
average number of encounters of the type considered is proportional to
the product dc^dcxdbdtdrdt.
In considering such encounters between molecules having velocities
within assigned ranges, it is assumed that both sets of molecules are
distributed at random, and without any correlation between velocity and
position, in the neighbourhood of the point r* Also the interval dt which
* This' assumption of molecular chaos' was considered by J. H. Jeans, Dynamical Theory
of Gates (4th ed.), chapter 4 (1925). See also H. Grad, Handbuch dtr Physik, vol. 12,
aoS-94 (1938).
3.5] THE EQUATIONS OF BOLTZMANN AND MAXWELL 59

we consider is supposed short compared with the scale of time-variation

of macroscopic properties, but large compared with the duration of an
encounter.
In an encounter between two molecules as specified, the velocity of the
second relative to the first, before encounter, is c 2 — ex, or gtv
Consider the motion of the centre B of the second molecule relative to the
centred of the first, or relative to axes moving with A For such an encounter
to occur, the line PO of Fig. 3 must cut a plane through A, perpendicular to
AP', within an area, of magnitude b db de, bounded by circles of radii b,b + db
and centre A, and by radii from A including an angle de. Also, since the
relative velocity is giv and dt is large compared with the duration of an

Fig. 6

encounter, it follows by an argument similar to the one used in 2.3 that at

the beginning of dt the point B must lie within the cylinder indicated in
Fig. 6, having the area bdbde as base, and generators equal to -g3idt; that
is, it must lie in a vo\ume(gdt) {bdbde) orgbdbdedt.
We can imagine such a cylinder to be associated with each of the/,«/Cj A*
molecules of the first kind, within the specified velocity range, in dr. If db
and de are small, it can safely be assumed that the cylinders do not overlap
to any significant extent, so that the total volume dv of all the cylinders is
given by dv =flgbdbdedcldrdt.
In many of these tiny cylinders there will be no molecule of the second kind,
having a velocity within the range ct, dct; and if db, de and dc% are sufficiently
60 THE EQUATIONS OF BOLTZMANN AND MAXWELL [3.51
small, we can ignore the possibility that in any one cylinder there are two
such molecules. The total number of such molecules in the whole combined
volume dv is/. dct dv, which is therefore the number of' occupied' cylinders
in which such a molecule occurs. Each occupied cylinder corresponds to an
encounter of the specified type, occurring within dv during the time dt.
Inserting the above expression for dv in ftdctdv, the number of encounters
is found to be f^bdbdedc^dc^rdt. (3.5,1)
Let e, e' be unit vectors in the directions of gn, g'tv so that gtl =gc,
g'n = ge'. Then (following Waldmann*) (3.5, 1) may be expressed in terms
of the element dt', which (since e' is a unit vector) represents an element of
solid angle (cf. 1.21). The angles x< e m a v be regarded as polar angles
specifying the orientation of e' relative to an axis in the direction of e; thus
dc' = ainxdxde

Hence we may writef bdbde = aw</e' (3.5,3)

where the positive scalar <xlt is a function of £ and b, or of; and X< given by

On substituting altde' for bdbde in (3.5, 1), we obtain the alternative

expression /J^de'dc^drdt (3.5.5)
for this number of encounters.

3.S1. An expression for Af£

If there are several gases in a mixture, the rate of change, by molecular
encounters, of the mean value (¾) of <f> for molecules of the first gas can be
divided into the parts A,^,, A, ^ , . . . , due respectively to encounters with
molecules of thefirst,second, ..., gases. Thus
^ i " A , ? i + A»?, + .... (3-Si.O
In an encounter of a molecule of the first gas, the value of <px for the mole-
cule, which when written in full (as in 2.22) is </>i(cur, t), is changed to ¢[,
signifying $i(c|, r, t). Thus the <f> for this molecule is altered by the amount
<f>\ — <j>v The change in 2^, due to all encounters of the special type con-
sidered in 3.5, between a molecule of the first gas and one of the second gas,
is therefore ^, _W / i / | f t | A ^ ^ W i (3.5,, a )
• L. Waldnwnn, Handbuch dtt Phytik. vol. ta, 205~5>4 (1958).
t In tome caiei this transformation needs careful treatment, because ?b/?x is not always
of the same sign for molecular pairs of particular types, so that b is not a one-valued
function of X- Where such difficulties arise, the symbol a„<fe' may be regarded as merely
a convenient brief notation for bdbde.
3.52] THE EQUATIONS OF BOLTZMANN AND MAXWELL 61
Integration, first over all permissible values of e', and then over all values of
c, and c,, gives the total change during it in E^„ summed over all molecules
of the first gas in dr, due to their encounters with molecules of the second gas.
Since the number of molecules m, in dr is nl dr, this integral must equal
rtjrfr A, ¢5,(¾. Dividing by drdt, we get
»iA,& = fffWi- $i)fjtg*\td*'dcxdcv (3.51, 3)
The variable c'x in <f>[ is a function of c„ ct, and e', given (¢£.(3.41,6,7)) by
c[-cx = -M^g'ti-gti), (3-51. 4)
where g'tl=ge'.
The value of n,A,^| can be obtained from (3.51, 3) as a special case by
replacing m, by m, in the relation between c[ and clt c„ e', and using the
relation between a „ and e' (or between x and A,;) appropriate to the law of
interaction between two like molecules « , instead of that between the unlike
molecules m, and m t ; the symbol <xlf is then replaced by <xv To distin-
guish between the initial velocities of the two encountering molecules, one
velocity is denoted by cv as before, and the other is written without suffix,
as c. Similarly the two functions / , , / , are written as / l t / , being now
identical except that in the former the variables are c)tr,t and in the latter
they are c, r, t. Thus
»A?i = ///(#- faffigzide'dcdcx. (3.51, 5)
When the gas is simple, that is, when molecules of one kind only are
present, the suffix 1 in the symbol " I A , ^ , may be omitted; but it must be
retained in the integral, in order to distinguish between the initial velocities
of two molecules (now of equal mass) involved in an encounter, since the
two velocities are separate variables of integration.

3.52. The calculation of d.fjdt

Like A$v SJjdt may be divided into the parts (defJ8t)u (de/Jdt) , due to
the encounters of molecules w, with molecules respectively; thus

(3 S2,,)
I T " ^ / , + 1 ^ , + -- -
When an expression for (djildt)t has been obtained, the values of the other
parts of Sef-Jdt can be derived by changes of suffix.
Consider the set of molecules of the first kind, situated within dr, which
have velocities within the range c„ dcx; the expression

signifies the net increase, during dt, in the number of molecules of this set,
due to encounters with molecules of the second kind (without restriction as
62 THE EQUATIONS OF BOLTMANN AND MAXWELL [3.52

to the velocity of these latter molecules). This net increase is the difference
between the numbers of molecules of the first kind, within dr, which during
dt enter and leave the set, owing to encounters with molecules of the second
kind.
Every encounter of a molecule of the set results in a change of velocity,
and so involves the loss of the molecule to the set. Thus the number of
molecules lost to the set during time dt, owing to the particular group of
encounters with molecules m, such that c, and e' lie in ranges det and de\
» by (3.5. S) equal to fxUg(tx%dt! dCxdc%drdt.

The total loss for all values of c, and e' is found by integrating with respect
to these variables; this gives
dCidrdtjjfJtgc^tde'det. (3.52,2)
The number of molecules m, entering the set c„ dcx owing to encounters
with molecules m, during dt may be found in like manner. We must, for this
purpose, consider encounters such that the velocity of a molecule w, after
encounter lies in the range c„ dcv Such encounters will be termed inverse
encounters; those in which the initial velocity of the molecule m, lies in the
range, eu dex may be styled direct encounters. Corresponding to any direct
encounter with initial velocities et, c, and final velocities c\, cj, there is an
associated inverse encounter in which the initial velocities are c[, c't, the
final velocities are clt ct, and the apse-line has the direction — 1c. The
correspondence is illustrated in Figs. 3«, b (p. 56), which refer to the
encounter of molecules behaving like centres of repulsive force, and show
the motion of one molecule relative to the other.
The number of inverse encounters with initial velocities e'v e't in the
ranges dc'v de't, and such that the direction e of the final relative velocity lies
in the solid angle dt, is
/ifig»i^g>X)d«deide'tdrdt; (3.52,3)
s
the factor a„ = <*idg<X) > die same in (3.52, 2, 3) because; and x have the
same values in the direct encounter and the corresponding inverse encounter.
Now let -,,,G _ x
j = d( '*«i)
d(cltct)
denote a Jacobian similar to those of 1.411. Then (cf. (3.41,6,7) and
(I.4U,I,2)) d{cx + Mtgn,gn) ^djcygn)
1
" d(cltct) %cltct)
_8{ct,et-cl)_ 8ieltet)
afo.c,) «(<?„€,) = 1 (3S2. 4)

and similarly J' a S r r * ^ = »•

3.S2] THE EQUATIONS OF BOLTZMANN AND MAXWELL 63

Hence if c„ ct lie in ranges dcv dc2, by the theory of Jacobians G, gtl lie in
ranges dG, dgn* where the sixfold (positive) differential elements dcldc1
and dGdglx are connected by the relation
dGdg2l = |J|dc x dc t = dcxdc2\ (3.52, 5)
and similarly dGdg'2X = dc[dc't.
Again, since de, de' are elements of solid angle, and g2X = ge, g'n = ge',
dgti = g2dg de, dg2X = g*dgde'.
On combining these results,
dedc'xdc2 = dedGdg'ix = gidgdGdede'
= de'dcxdct (3-5 2 . 6 )
by symmetry. Hence (3.52, 3) can be expressed in the form
f'xf'2gccltde'dcxdc2drdt. (3.52,7)
This represents the number of encounters such that the final velocities lie
in the ranges rfc„ dc2, and the direction e' of the initial relative velocity lies
in the range de'. On integrating over all possible values of e' and ct, the total
gain by encounters during dt to the set of molecules m„ c,, dcx in the volume
dr is found to be . , ,rf,,., ,,, . „
dcx drdtjjf'J2 gxxtde' dct. (3.52, 8)
Combining (3.52, 2 and 8), we get
dcxdrdtfj(f'xf2 -fj2)gctxide'dct
for the net gain to this set.
This net gain is denoted by {Stf\l^t)2dcx drdt. Hence, dividing by dcxdrdt,
we find that

This is the required expression for (SefJ8t)t.

* This statement is not strictly accurate. To a volume-clement detdet in the stx-
dimensional space in which the coordinatcsof a point are the components of c t and c, there
corresponds a volume element 9 in the six-dimensional space in which the coordinates
are the components of G and g„: but i cannot in general be put into the form dGdg„,
any more than the element of area between the pairsof curves£ = 4>(x,y), { + ( / { » 4>(x,y),
and v = iK-v, y), 7 + 1/7 = $(x, y), can in general be expressed as equal to an elementary
rectangle dxdy. However, just as the small area between these pairs of curves can be
divided up into a large number of still smaller rectangles dxdy, so t can be divided up
into elementary volumes dGdgn. It is necessary in equations like (3.52. 5) to regard the
notation dGdgn as referring to the general volume-element S, formed from a sum of
products of three-dimensional elements in spaces in which the coordinates are the
components of G and gn respectively. With this understanding the expression (3.52, 7)
is readily derived, as in the text.
64 THE EQUATIONS OF BOLTZMANN AND MAXWELL [3.53
From this and (3.11, 2) we obtain a second expression for A,^„ namely

»|A,?i - ) ^ 1 ( ¾ 1 ) ^ .
- JSfWJ't -fJi)g»»^'dCldct. (3.52,10)
The corresponding formulae for (d«/i/9/)i and A , ^ are
(3S2 J)
( ^ ) i ' If ^ -JW^*'*- -'
"Aft - fffWf'i-fttgtidt'dcdcv (3.52,12)

3.53. Alternative expressions for nA$; proof of equality

The equality of the expressions for »t At^j given by (3.51. 3) and (3.52,10)
can readily be established. In the integral
JJM/i/;**ii<fe'«fc|<fct (353.0
let the variables of integration c„ c„ e' be changed to c[, e't, e, these being
functions of clt ct, e' such that dcxdcxdt! - dc\dc'tdt (cf. (3.52,6)). Then
the integral becomes
flfrj'J',g*i4g,X) d*dc[dc't.
Now cj, cj, e are variables specifying a certain encounter, namely, the
encounter inverse to that specified by c„ c„ e': and integration over all
possible values of c[, cj, e is equivalent to a summation over all possible
inverse encounters, or, since every encounter is inverse to another encounter,
over all possible encounters. Since c[, c't, e are variables specifying an
encounter, c lt c„ e' may be written in their stead; the variables cu c„ e'
that specify the inverse encounter must then be replaced by c'v c't, e.
Hence the integral becomes equal to
JJJViW.*'. t)fi(ci,r, t)Mct,r, ^ga^^de'dCidet
or, in brief, to jfjfJJtg^it^'^.^r (3-53>a)
From the equality of (3.53, 1) and (3.53, 2) that of the two expressions for
A,$, at once follows.

3.54. Transformations of some integrals

The proof in 3.53 is independent of the nature of the functions <f>,f, and
similar arguments serve to establish a number of analogous analytical results
which we quote here for later reference.
First, if F, G, and <j> are any functions of velocity, position, and time, such
arguments show that
^'t>iFiG,lg^ld*'dexdet - jjjUF^ge^Mdc^dc* (3.54,1)
3.6] THE EQUATIONS OF BOLTZMANN AND MAXWELL 65

In this equation replace <f>x by unity and Fx by <}>XFX; then it becomes

JJJtf **iG*t*«M*'<Mc! = jfj<i>xFxGtgaudt'dcxdct.
From this equation and (3-54, 1) it follows that
fffti(FiGt-F'xG't)gxxlde'dcxdct = - jfftl(FxGt- F'xG't)gciX2d^dcxdct
= hjJI(tl-t'i)(FiGa-FxGi)g*xadc'dcxdct. (3.54,2)
From (3-54, 2), omitting the suffix 2, the corresponding equation for
encounters between pairs of molecules mx is obtained; this is
fff<t>x(FxG-FxG')gccxde'dcxdc
= Ij/Jtfi- ftWtG-FWgandt'dCydc. (3-54, 3)
Since c, and c both refer to molecules m,, an interchange of c, and c does
not alter the value of either integral. Thus, making this interchange on the
right-hand side of (3.54, 3), we have
Hftx(FxG-F'xG')g*xde'dcxdc = \\l\^-^){FGx-F'G[)gcLxd^dcxdc.
Adding to (3.54, 3) this equation and two similar equations, in which
FXG — F'XG' on the left-hand side is replaced by FGX—F'G'V we get
ffftx(Fx G + FGX- F[ G' - F'G[)g»x de'dc, dc
= ljjf(<f> + * , - * ' - &)(FXG+FGX -F[ G'-FGx)gzxde'dcxdc.
(3-54. 4)
In this equation put F = G. Then
fjftm-F'FDgx^'dctdc
= lfjj(<t> + t1-<!>'-<f>lHFFx-F'Fi)gzxde,dcxdc. (3.54. 5)

3.6. The limiting range of molecular influence

The integration with respect to e' in 3-51-3.54 was tacitly supposed to be
taken ovet all permissible values of e'; that is, since aitdc' = bdbde, over all
values of b from 0 to 00, and all values of e from 0 to 2»r. But this statement
requires some qualification when applied to integrals such as (3-52, 2), since
such integrals become infinite if the range of integration with respect to b is
infinite. Some upper limit for b, representing the limiting range of molecular
influence, should really be taken; this distance might vary to some extent
with^, but with gases of moderate density it is usually considerably less than
the mean distance between neighbouring molecules. To make the distance
definite we may, for example, choose to ignore all 'grazing' encounters
during the whole course of which the deflection x of th e relative velocity is
less than a very small angle S.
The value of the integral (3.52, 2) will depend entirely on the upper limit
chosen for b, but in such integrals as (3.51, 3, 5), (3-52, 9-12), where the
66 THE EQUATIONS OF BOLTZMANN AND MAXWELL (3.6
integral contains a factor such as fa - fa or/J/J - / , / „ the case is different;
as b tends to infinity, e[, c't approach the initial values c„ c„ so that fa
becomes equal to fa, and/J/J to/t/ t . It is found that, when the relation
connecting x> * and j- corresponds to the laws of force that hold good in the
case of most actual gases, the larger values of b contribute very little to the
integral, and the result is not appreciably affected if, for analytical con-
venience, the integration is extended up to b = oo.
Where this procedure is illegitimate (as, for example—cf. 10.34—when
the molecules repel or attract according to the inverse square law) b must be
restricted, and a discussion of the effect of encounters in which more than
two molecules participate is required. In this case, moreover, it is not true
(as was assumed in the derivation of Boltzmann's equation) that molecules
are appreciably influenced by forces of interaction only along a small portion
of their paths; hence the results derived from Boltzmann's equation can-
not be expected to give more than the correct order of magnitude of the
quantities concerned.
 4
BOLTZMANN'S ff-THEOREM A N D THE
MAXWELLIAN VELOCITY-DISTRIBUTION

4.1. Boltzmann's //-theorem: the uniform steady state

Consider a simple gas whose molecules are spherical, possess only energy of
translation, and are subject to no external forces. If its state is uniform, so
that the velocity-distribution function / is independent of r, Boltzmann's
equation (3.1, 1) reduces to

I = jjU'f'i -//Jft*'*i. (4-1. 0

after substituting for £,//?< from (3.52,11).
Let H be the complete integral (that is, the integral over all values of the
velocities) defined by the equation
H = jf\nfdc. (4.1,2)
Then H is a number, independent of r, but a function of t, depending only
on the mode of distribution of the molecular velocities. Also

£-J|(/1n/)*-J(.+ln/)g*
= JJJfi + In/)(//I -//,)**,*'*:*, (4.1, 3)
by (4.1, 1). Hence, using (3.54, 5),
~ = \ J J J d + ln/+1 + In/, - r - In/' - 1 - In/J)

= ^ J J J l n (//»//'/;)(/'/; -ffl)g<zldt'dcdev (4.1,4)

Now In (ffjf'f'i) is positive or negative according as//', is greater or less than
/'/J, and is therefore always opposite in sign to/'/J —ff\- Thus the integral
on the right-hand side of (4.1, 4) is either negative or zero, and so H can
never increase. This is known as Boltzmann's //-theorem.
Since H is bounded below,* it cannot decrease indefinitely, but must tend
• This is so because H =• — 00 only if if In fdc diverges. The integral If. \mc'dc certainly
converges, since it represents the (finite) total energy of translation of the molecules;
hence if J/ln/<fc were to diverge, —In/would have to tend to infinity more rapidly
than c' as c -+00. But this would imply that/ tends to zero more rapidly than «'**, in
which case [fin fdc certainly converges.
167]
68 BOLTZMANN'S H-THEOREM AND THE [4.1
to a limit, corresponding to a state of the gas in which 8H/8t — o. By (4.1, 4),
this can occur only if, for all values of c, c„
/yw/i. (4-1. sr
or, what is equivalent, In/' + ln/J = ln/+ In/,. (4.1, 6)
Comparing (4.1,1) and (4.1, 5), we see that, if 8H/dt = o, then df/dt « 0
also, so that the state of the gas is steady as well as uniform. Conversely, if
the gas is in a uniform steady state, not only must df/dt — o, but also, since H
depends only on/, dH/dt - o, for which (4.1, 5) is necessary. That is to say,
the solution of //^.-//,)1¾^. - 0 (4.1.7)
is (4.1, 5) or (4.1, 6).
Equation (4.1, 5) implies that the number of encounters between mole-
cules with velocities in the ranges dc, dcv such that the direction e' of the
final relative velocity lies in the range dt', is equal to the number of inverse
encounters of a similar type that result in molecules entering these velocity-
ranges (cf. 3.52). This is expressed analytically by the equation
/'/i«i(* X) dtdc'dc[ =//,«,(*, xWdcdc,,
which is satisfied in virtue of (4.1, 5) and the relation dtdc'dc[ = dt'dcdcx
(cf. (3.52, 6)). Thus not only is the state of the gas steady, so that encounters
as a whole produce no effect, but also the effect of every type of encounter is
exactly balanced by the effect of the inverse process. This is an example of
detailed balancing, recognized as a general principle in statistical mechanics.!
Equation (4.1, 6) shows that In/is a summational invariant for encounters
(3.2). Thus it must be a linear combination of the three summational
invariants ^ ° of (3.2, 2), so that
InZ-So'O^o^an' + aW.Mc-a^.Jwc1, (4.1,8)
where, since In/ is a scalar, a™ and a&> are scalars and a** is a vector; all
three must be independent of r, t, since the state of the gas is uniform and
steady. This equation is equivalent to
In/ = a™+m[a.fu+aj*t>+a!*w) - \aflhn(u*+v* + w*)
- In a»> - oF>. \m{(u - a?/a™)* + {v- a * / ^ ) 1 + (a - a*/a»>)»},
where a™ is a new constant. Thus, if C - c - or*/***,
/-o«»e-««»'jmcrt (4.1,9)
This result was first obtained by Maxwell, and a gas in the state defined by
(4.1, 9) may be said to be in the Maxwellian state.J
* Equation (4.1, 5) first occur* in J. C. Maxwell, Phil. Tram. Roy. Soc. 157, 49 (1867);
Collected Pttptrx. a. 45.
t R. H. Fowler, Statistical Mechanics, p. 417 (1919). or p. 660 (1936).
t The variable C of (4.1, 9) should not be confused with the peculiar velocity of a
molecule after collision, for which the same symbol C will later be used.
4.11 MAXWELLIAN VELOCITY-DISTRIBUTION 69
m m 3
The constants <x , a and a* ' can be evaluated in terms of the number-
density n, the mean velocity c„, and the temperature T. First we have
n = ffdc = ^ - ' " ' K ' i C ' ,
whence, on expressing C in terms of polar coordinates C, 0, f,

n = a«» ^C'*e-'"'lmCltdC' {\indd0 P"' d?

Jo Jo Jo

using (1.4, 2). Again,

nc0 = Jc/rfc
=/(<*<«/««>+ C')/rfC'

The second term vanishes because the integrand is an odd function of the
components of C; hence c0 = or^ya*3' and C is identical with C, the peculiar
velocity (2.2), so that (4.1, 9) takes the form

Finally, from the definition of temperature in the uniform steady state,

\kT = JmC"2

= -2-

by (1.4, 2). Hence «<3' = i/kT.

Consequently (4.1, 9) is equivalent to

Mi5r) l *- 0MM '' (4,,,0)

which is the usual form of Maxwell's velocity-distribution function.
It therefore appears that when the density, mean velocity, and temperature
of a uniform gas are assigned, there is only one possible permanent mode of
distribution of the molecular velocities, and that the actual mode, if different,
will tend to approach this mode.
70 BOLTZMANN'S H-THEOREM AND THE [4.11

4.11. Properties of the Maxwellian state

The number of molecules per unit volume with velocities in the range c, dc
isfdc or fdudvdw. Hence the number of molecules per unit volume, in the
Maxwellian state, whose component velocities lie between the limits u and
u + du,v andv+dv, and n and tv + dto, may be written
n(m/2irkT)i rt'«u-*l'ikTdu.e-l'*v-,>»mTdv.e-l,«u>-w*>lkTdw.
This indicates that the distribution of u is independent of the values of v, to;
thus the probability that the x-component of the velocity of a molecule lies
between given limits is independent of the value of the component per-
pendicular to Ox; the ^-component is distributed about its mean value u„
proportionately to the Gaussian function

or *-•', where s* = lm(u-u0)*lkT.

Writing dc = iC = CttAnOdCdOdy, and integrating with respect to 0
and 9, the number of molecules per unit volume whose peculiar speeds lie
between C and C+dC'a found to be

which is proportional to s*e~^, where J* — mC/2kT. Graphs of the two

functions e~**, J* e-*' are given in Fig. 7. The first illustrates the distribution
of any component of the peculiar velocity, and the second shows that of the
peculiar speed C.
The mean value of any function of the molecular velocity for a gas in the
Maxwellian state can be found from the equation

„? - jtfdc = n (^)*JV»™*iVC.
If the function is of odd degree in any component U, V, or Wot the peculiar
velocity, its mean value vanishes.
The mean value of the peculiar speed C is given by

by (1-4. 3)-
 4.11| MAXWELLIAN VELOCITY-DISTRIBUTION 71
The mean value of C2 is given by
3kT
C2 = (4- »• 3)
(cf. the definition of temperature in (2.41,1)). The root-mean-square of the
peculiar speed is defined as equal to y/(C2). It is thus not equal to the mean
speed; in fact ^ ) = c V ( 3 W8) = .086C. (4.,,,4)

10

y=e

05

'y~s'e-

0-5 10 15 20 2-5
Fig-7

Another mean value needed later is that of the ar-component of the peculiar
velocity, averaged over those molecules at a given point for which this
component is positive; it will be denoted by W+. Since the number-density
of such molecules is \n, .
in1P+=)+fWdC, (4.11,5)

where the integration on the right extends over all values of C for which
72 BOLTZMANN'S H-THEOREM AND THE [4.12
W > o. Hence, using (1.4, 2, 3),

-JC. (4.11,6)
The components of the pressure tensor (2.31, 3) can readily be found.
Since the mean value of any function of the velocity odd in U, V, or W is
zero, the non-diagonal terms of the tensor vanish, while by symmetry
Pxx "Pn -Pu = K/>*z +Pv»+P*z)
~p = knT, and p = *»rU. (4.11,7)
Hence in this case the pressure system is hydrostatic (2.32).

4.12. Maxwell's original treatment of velocity-distribution

The above law of distribution of velocities, (4.1, 10), was first given by
Maxwell* for the case of a gas at rest. His original argument is of historical
interest, though not mathematically rigorous. He assumed that, as the com-
ponent velocities u, v, and to of a molecule are perpendicular to each other,
the distribution of one of these components among the molecules will be
independent of the values of the other components. Assume, then, that
F(u)du is the probability that a molecule should possess an x-component of
velocity between u and u+du, and that F(u) is independent of v, to. Then the
probabilities that its y- and x-components of velocity should have values
between v and v + dv, to and to+dto, are similarly F(v) dv, F[to) dto; hence if
f(u,v,w)du dvdto
denotes the number of molecules per unit volume whose component
velocities lie in the ranges du, dv, dto,
f(u, v, w)dudvdw = n F(u) du F\v) dv F\to) dto.
Now in a gas at rest there is nothing to distinguish one direction from
another; thus/(14,v,w) can depend on u, v, to only through the invariant
u* + v* + to*. Thus
nF\u)F{v)F{to) =/(«,», w) - #«» + »» +to*),
say. The solution of this functional equation is given by
F(u) = x«Y«',
/(«,», to) = ttit + tfi + to*) = »x3eY("'+t,+"^,
* J. C. Maxwell, Collected Papers t, 377; Phil. Mag- (4). 19. » (i860).
4.13] MAXWELLIAN VELOCITY-DISTRIBUTION 73

where x, y are arbitrary constants; this agrees with the form derived above
for/, taking nx3 = txm, Y = -|ma (3 >.
The unsatisfactory feature of this proof is the assumption that the distribu-
tion of each of the three velocity-components among the molecules is
independent of the values of the others. As these three components do not
enter independently into the equations governing a collision, it would be
natural to suppose that the distributions of the components are not
independent.
On account of this defect, Maxwell* attempted a second proof, which was
also imperfect; he showed only that i/in a gas the Maxwellian distribution of
velocities was once attained, it would not alter thereafter (since/'/J = / / i ,
so that df/dt = o). Boltzmann first showed, by his //-theorem,! that the gas
would tend to the Maxwellian state. The demonstration was improved later
by Lorentz.J substantially to the form given above in 4.1. This proof also is
open to some objection, because of the assumption in 3.5 that there is no
correlation between the velocity and the position of a molecule. In very
dense gases it is, in fact, probable that the velocity of one molecule is related
to the velocities of other neighbouring molecules, in whose close proximity
it remains for some time by reason of the close packing of the molecules; but
for gases under ordinary conditions the assumption appears to be valid.§

4.13. The steady state in a smooth vessel

Maxwell's form for / also applies to the steady state of a gas at rest in a
smooth-walled vessel under no forces.
Consider //„, defined by
H0 = JHdr = jjflnfdcdr (4.13, 1)
(cf. (4.1, 2)), the space-integration extending throughout the volume of the
vessel. In the time-derivative of this equation,

^ = JJ ( l n / + s ) | r f c ( / r > (4.I3>2)
we substitute for dfjSt from Boltzmann's equation (3.1, 1), omitting the term
containing F, since F = o. We thus have

2-JjW">(¥-8**
d
= jj(\n/+1) -ddcdr-jjc.^dcdr. (4.13, 3)
• J. C. Maxwell, Collected Papers, 2, 43; Phil. Trans. R. Soe., 157, 49 (1867).
f L. Boltzmann, Wien. Silz. 66, 275 (1872).
X H. A. Lorentz, Wien. Sit*. 95 (2), 127 (1887).
{ See J. H. Jeans, Dynamical Theory of Gases (4th ed.), pp. 59-64 (1925), and H. Grad,
Handbuch dtr Physik, Col. 13, 218-33 (1958).
4 CUT
74 BOI.TZMANN'S H-THEOREM AND THE [4.13

The second term becomes, on transformation by Green's theorem,

-\\cJ\nfdcdS

or -jncnlnfdS,
where cn is the component of c along the outward normal to the element dS
of the surface of the vessel. Consider the contribution to this integral from
any element dS, which we may without loss of generality take perpendicular
to Ox, since the directions of the axes of reference are arbitrary. Since the
vessel is smooth, the molecules that leave are the same as those that strike it;
their x-components of velocity are exactly reversed, and their y- and *-
components are unaltered. Hence near dS
f(-u,v,tv)-*f(u,v,w),

and so cn In/ = u In/ = o.

Thus the contribution to the integral from this element vanishes. Since the
element dS is arbitrary, the integral as a whole vanishes.
It follows that
^=Jj{lnf+1)Mdcdr. (4..3,4)
After substituting for d,f/dt from (3.52,11), we may show, as in 4.1, that

and hence that, when the state of the gas is steady (so that dfjdt = 0 and
BHJdt «= o), In/must be a summational invariant, as in (4. 1,8). Consequently
0,//5/ = 0, and
1 n
\2tTkT) '

the variable c taking the place of C because the gas is supposed to be at rest.
The quantities n, T in this equation might conceivably be functions of r,
as the state has not been assumed uniform. But Boltzmann's equation for/,
which now reduces to „,

must be satisfied for all values of c; this implies that n and Tare independent
ofr.
When the walls of the vessel are not smooth, no such simple proof that
Maxwell's formula remains valid seems available.*
* On thii point MC R. H. Fowler, Statistical Mtchatrict, pp. 697-99 (1936).
4.14] MAXWELLIAN VELOCITY-DISTRIBUTION 75

4.14. The steady state in the presence of external forces

We next consider the case when an external force mF acts on each molecule
(F being, as before, independent of the velocity c of the molecule). As in 4. i

_=J( I + ln/)^c
-J(-+W)(¥-..£-F.B*
.J(«&Q.,.figfl_ F .«gfl) fc (4,4>I)
In the last term we have three components of the integral
f g(/ln/) j . .
JnST-* (4-14.»)
to consider. In the component

/// OU

an integration with respect to u gives

which vanishes since / I n / must vanish when c or any of its components

tends to + 00. Hence the integral (4.14, 2) vanishes.
If the gas is in a smooth-walled vessel at rest, or if its density tends to zero
in all directions, it may be shown, as in 4.13, that the second term on the right
of (4.14, 1) contributes nothing to //„, defined as before by
H0 = JHdr.
Thus again it follows that dHJ8t < o, and in the steady state SJIdt •> 0 and
V,BCW (4 I43)
'="fe) '
where C = c-cn and n, c0, T are independent of c and t, but may now
-'
depend on r.
To examine this dependence, we substitute from (4.14,3) into Boltzmann's
equation (3.1, 2), which then reduces to @f = 0. We use the form for Sf
given in (3.12, 2), since in (4.14, 3 ) / is expressed in terms of the peculiar
velocity C. Since the state is steady, DjDt in (3.13, 2) may be replaced by
c0.8/dr. Hence, on dividing by/, Boltzmann's equation becomes

c ain/ +c ain/ r 1 e\ l a i n / ain/ a

>--w •-sr+r-h *rn^c -*c c*c*=°-
(4-14.4)
4-«
76 BOLTZMANN'S H-THEOREM AND THE [4.14
Since I n / - In(n/ty-m&lzkT + constant,
. a in/ ain(»/r») mC* er a in/ mc
we have - ^ - ^ - + ^ ^ - , -^,-=--.
Using these values, we can express the left-hand side of (4.14, 4) as the sum
of a part independent of C, and parts involving C to the first, second, and
third powers of its components; these must vanish separately, since the
equation is an identity in C.
The part of the third degree in C is
mC* _ BT

whence it follows that dT/dr = o, that is, the temperature must be uniform
throughout the gas. Allowing for this, the part which is of the second degree
in C becomes
-CC-
t?rCC:
LC 5-co = 0>
kT 9r

whence it follows that -.- c0 - o,

or
or e = 0 in the notation of 1.33. This implies that

Bx** By Bz ' Bz By ** Bx Bz = By 8x
The solution of these equations* is
c0~c'+wAr, (4.14, 5)
where c' and to are arbitrary constants. Thus the mean velocity of the gas at
any point is the same as the velocity of a rigid body moving with a screw
motion.
First consider the special case when c„ = 0. This ensures the vanishing of
the part of (4.14,4) which is independent of C\ using the conditions c0 = 0
and BTjdr » o, the remaining part, of the first degree in C, becomes
, /aInn mF\
•\W~kT)-
Since this vanishes for all values of C,
a Inn m
F. (4.14,6)

The steady state is therefore possible only if F is the gradient of the scalar
function (kTlm)lnn, so that die field of force must possess a potential V,
satisfying the equation ,~
V- In n +const.
m
* Cf. A. E. H. Love, The Mathematical Thtory cf Elasticity, } 18 (1917).
4.14J MAXWELLIAN VELOCITY-DISTRIBUTION 77
The density distribution is given in terms of T by
n = n0e-mrikT, (4.14, 7)
where n0 is a constant, being, in fact, the number-density at points at which
Y = 0. Hence the complete expression for/is

^"»(^r),^ , Y
^ a t r
- (4.14,8)
This result wasfirstgiven by Maxwell* as a deduction from his equation
(4.1, 5). Boltzmannf later gave the same result (apparently unaware that
Maxwell had already published it); his proof was based on the //-theorem,
and supplied a needed foundation for Maxwell's deduction.
Consider next the case when c0 does not vanish. Let Oz be taken as the
axis of the screw motion; then the components of c„ are ( — toy, tox, c'), and
WCfind
(c±)c-»b
where T 0 = - *to\x* +y*). (4.14, 9)
The equations obtained on equating the terms of first and zero degrees in C
to zero are now
., /dlnn mF m 5Y0\ 8\nn . .
c
(-i)F-kT+kTi») = 0- c "-3r = °- (4, ,0
* >
The first of these implies that F is again derivable from a potential T, and
that the number-density n is given in terms of T by
n = II0«—^•"•V*1'. (4.14, 11)
Comparing this with (4.14, 7), we see that the effect of the motion on the
density-distribution is the same as if a field of centrifugal force, of potential
Tp, acted on the gas.
Using this expression for n, and remembering that Cg.dYg/dr = o, we
obtain from the second of equations (4.14,10) the condition

expressing that the motion of the gas must at every point be along an
equipotential surface Y = constant. Thus if w = o and c' + o, T does not
depend on z, and if c' = 0 and w + o, *F must be symmetric about Oz: if
both c' and <o do not vanish, Y is constant along spiral curves with Oz as axis.
If the gas is enclosed in a smooth stationary vessel, the motions must
be consistent with the shape of the vessel: that is, the gas must in general
• J. C. Maxwell, Nature, Land. 8, 537 (1873); Collected Papers, a, 351.
t L- Boltzmann, Wien. Ber. 7a, 427 (1875).
78 BOLTZMANN'S H-THEOREM AND THE (4.2
be at rest, but if the vessel possesses symmetry about an axis a rotation
about this axis is possible.
When c' «= o and F « o the form of/ is

/ - «o ( ^ 7 7 « P t - (m/2kT) {u*+«*+w* + 2w(«y - «)}].

This form of the velocity-distribution function for a rotating gas was first
indicated by Maxwell.*

4.2. The Jf-theorem and entropy

For a gas in the uniform steady state the quantity H, defined by the equation
//«f/ln/«/c-»in7,
can be expressed in terms of n and T. For in this case
I n / - ln» + |ln(m/2JT*r)-mC*/2Ar,
whence H - »{ln n + } In {m/2irk T) ~ J}
(cf. (2.41,1)). If the total mass of gas present is M, the volume occupied by
the gas is Af/p, or Af/mn. On integrating H through this volume, it follows
*"** H, m JHdr - (Af/«){ln» + fln(m/2>rAr)-i}.
The entropy S of the gas (cf. 2.431) is such that
„ „/ lTk*V\

Since ct « 36/2*1 in this case (there being no communicable internal energy),

and nV is the total number of molecules, which is constant,

«-?e?-?).
whence, on integration, we find
Mh
S = — I n (r*/f«) +const.;
tn

•xM
thus S + kH9 •* - ^ - {In (znk/m) +1} + const.

The right-hand side of this equation is independent of the state of the gas;
hence, except for an additive constant,
S = -«/.. (4.2,1)
* J. C. Maxwell, Nature, Loud. 16,144 (1877).
4.21] MAXWELLIAN VELOCITY-DISTRIBUTION 79

This relation connects //„ with the entropy* when the gas is in a uniform
steady state. For non-uniform or non-steady states there is no thermo-
dynamic definition of entropy, but the concept of entropy may be generalized
to such states, taking equation (4.2, 1) as the definition. Boltzmann's
//-theorem, which shows that for a gas not in a steady state //„ must
decrease, is a generalization of the thermodynamic law that entropy
cannot diminish. This association of H with S was indicated by Boltzmann
in 1872.

4.21. The //-theorem and reversibility

Suppose that at a certain instant the velocity of every molecule in a mass of
gas in a uniform state under no forces is reversed; the value of H or n In/
is unaltered by this process. The molecules will now retrace their previous
paths. Since, in general, BH/dt < 0 before the change, one infers that
dH/8t > 0 after the change, which contradicts the //-theorem. Thus a
paradox arises.
The //-theorem is, of course, a probability theorem. Its proof depends on
probability concepts, e.g. in the definition of the velocity-distribution func-
tion and the calculation of the number of encounters of a given type. Thus
the theorem is to be interpreted as implying, not that H for a given mass of
gas must necessarily decrease during a given short interval, but that it is
far more likely to decrease than to increase. Even so, the theorem appears
inconsistent with reversibility, which implies that to every state of the gas
for which H is decreasing there is one for which H is increasing equally
rapidly.
The following considerations help towards resolving the paradox. The
argument leading to the expression for 9ff/8t assumes molecular chaos before
encounters, i.e. that the velocities of two molecules just before they collide
are uncorrected. This is reasonable, because in gases at moderate densities
the two molecules will have come from regions far enough apart to rule out
the possibility of their having recently influenced each other. The velocities
of the molecules immediately after encounter are not, however, in the same
sense uncorrected. If the velocities of all the molecules are suddenly
reversed, encounters are described in the reverse sense, and one is no longer
considering a state with molecular chaos before encounters. Thus the value
found earlier for 8H/8t no longer applies. This in part resolves the difficulty;
on physical grounds, states characterized by molecular chaos before col-
lisions are far more probable than states deviating widely from such chaos.
* In Boltzmann's papers of 1872 and 1875, in which the //-theorem was introduced, he
used the symbol E (presumably because it is the initial letter of entropy) for what is
now denoted by H; S. H. Burbury (Phil. Mag. 30. 301 (1800)) seems to have intro-
duced the symbol H, though later he used B to denote an almost identical function.
Boltzmann used the symbol E as late as 1893, but in 1895 adopted the letter H; cf.
Nature, Land. 139, 931 (1937).
80 BOLTZMANN'S H-THEOREM AND THE (4.3
But the mathematical difficulty remains that to each state characterized by
molecular chaos there corresponds one (that with velocities reversed) not so
characterized.
The reversibility paradox can be related to another occurring in the kinetic
theory. Consider, for example, the following paradox relating to an atmo-
sphere in a steady state under gravity. Any one molecule has a constant
downward acceleration due to gravity, and since with the velocity-distribu-
tion of (4.14, 8) velocities in all directions are equally likely, collisions with
other molecules may impede, but will not completely destroy, the downward
motion of the molecule.* Hence the gas as a whole must descend; that is, it
cannot be in a steady state.
This second paradox is easily resolved. If the atmosphere is held up against
gravity, it must be held up by a surface, and collisions with this surface
interrupt the steady descent. To see how the steady state is maintained at
a level well above this surface, consider two neighbouring horizontal planes
A and B, of which A is above B. Because of the action of gravity it is more
probable that a molecule which at the beginning of a short interval is at A will
sink to B during that interval, than that a molecule initially at B willriseto A.
As, however, the density of molecules is greater at B than at A, the smaller
proportion of the molecules initially at B which rise to A can exactly balance
the larger proportion of molecules initially at A which sink to B. Hence each
molecule can tend to diffuse downward with a certain velocity, and yet the
mean velocity of molecules at a given point can vanish.
A comparison with the second paradox helps further to clarify the
reversibility paradox. The value of dH/dt found earlier is, as it were, a velocity
of diffusion, analogous to that of a molecule at A in the preceding paragraph,
with which H tends to approach its minimum value as the actual state of the
gas varies among the different possible states. It may be negative, even
though the mean value of the 'velocity' dH/dt for all possible states with a
given H is zero, provided that the possible states with a smaller value of H
are more numerous than those with a larger value, i.e. provided that smaller
values of H are intrinsically more probable than larger. This agrees with the
result of statistical mechanics that the Maxwellian function gives the most
probable distribution of the molecular velocities.!

4.3. The /f-theorem for gas-mixtures; equipardtion of kinetic energy

of peculiar motion
The velocity-distribution functions for a mixture of gases in a uniform
steady state under no forces may be obtained by a generalization of the
argument of 4.1. For simplicity, we consider only a binary mixture. By
analogy with (4.1, 1), the equations satisfied by the velocity-distribution
• Cf. 5.5. on the persistence of velocity after collision.
f J. H. Jeans, Dynamical Theory o/ Gates (4th ed.), chapter 3 (1925).
4.3) MAXWELLIAN VELOCITY-DISTRIBUTION 81

functions/,,/, are

= =
° 1» / / ( ^ ^ - ^ ^ ^ ^ ^ + //0/^1-/1^)^11^^1.
(4-3. 0
where, for the present,/, has been replaced by F„ in order to emphasize the
distinction between the functions /,(0,, r, t) and/,(c,r, t) for molecules of
the first kind, and the functions/,(c2, •", 0 and/,(c, r, r) for molecules of the
second kind: these are represented in the above integrals by/,,/andF t , F.
Multiply the first of equations (4.3, 1) by ln/,</c„ and the second by
InFtdct, and integrate over all values of c,, c, respectively; then, on trans-
formation by (3.54,2) and (3.54, 5) they become
j///(In/, + I n / - In/; - In/')(/'/J -J/,)ga.de'dcdc,
+ hflfQn/l-Wi)(/iF'i-/lFt)gxltd*'dcldcs = o,
J//J(ln Ft + In F - In F'% - In F') (F'F't - FF^go^ de' dc dct
+ HSJOnF,-]nFQ{flFi-/lF^galldt'deldet = o,
whence, on addition, we obtain
k!!JH//iir/i)U'/i-//i)g^de'dedel
+ \\\\\n(/lFil/\F't)(flF-t-/lFt)gxxtdti'dcldct
+ ifff\n{FFt/F'F-t)(F,F-t-FFt)gaLtde'dcdct - 0.
In none of these three integrals can the integrand be positive, so that their
sum can be zero only if, for all values of the variables, the integrands vanish.
Hence for all types of encounter, between like or unlike molecules,
/ . / - / J / ' . /iF,-/iF» F,F=FiF, (4.3,2)
and therefore ln/„ InF, are solutions of the three equations
^+^-^1+^, fc+y.-ft+Y,. v,+Y = y;+v. (4.3,3)
x
The first and third of these equations show that \jflt Vt are of the forms
^, = «iI> + «?».m,c1 + a(,!,>. Jm,«!. V* = «?>+ ^,31. «,<:, +«?>.*«,<:$;
the middle equation then requires that a?'= «£*> = ^(say);*'," - af> « a°>
(say), to satisfy the equations of conservation of momentum and energy at
encounters of unlike molecules. Thus
In/, = a'," + a™. w, c,+a 91 . Jm, cj = In Ax + a"'. Jm, S(«, - u9)\
In/, -ai l >+ a«>.m,c,+a<».tm,c$ = In^,+«<,>.Jm,E(M,-«0),,
82 BOLTZMANN'S H-THEOREM AND THE [4.4
where Av A2, u0, v0, w0 arc new constants. It may be shown, as in 4.1, that
M0, fg, tfg are the components of the mean velocity of either constituent, and
so of the mixture: and that the mean kinetic energies of peculiar motion of
molecules of the two constituents are the same, and equal to -3/20101.
Hence, if T is the temperature of the gas, kT = — i/oc"*, and the velocity-
distribution functions/,,/, may be expressed in the forms

/l exp miC JzkT)

" "» (ZSTT)* (- ' '
(4-3. 4)
/1 = »* ( ^ ) exp(-MjCI/aAT),

The result that the mean kinetic energies of peculiar motion of molecules
of the different constituents are equal is a special case of the statistical-
mechanical theorem of equipartition of energy, referred to in 2.431.
The results of 4.13, 4.14 may also readily be generalized to apply to a
mixture of gases. Thus, for example, if the molecules of the two gases are
subject tofieldsof force of potentials *Fj, *F2, the velocity-distributions will
be given by equations of the same form as (4.3, 4) where «,, nt are given by
«! - Ai«—T«*r „, = ^ - ^ ^ , (4.3,5)
Nv Nt being constants.

4.4. Integral theorems; 1(F), [F, G], {F, G)

To conclude this chapter certain integral theorems, similar to those of 3.54,
will be proved. Only a binary gas-mixture will be considered; the corre-
sponding results for a simple gas, also required later, are merely special
cases of these.
The velocity-spaces of the different sets of molecules will be regarded as
distinct domains; thus functions of the velocity may be differently defined
in the two domains. Let/W) denote Maxwell's velocity-distribution function

/W, = n
( 2 l^) , '" m C , / t t T ' (44>,)

where the suffix 1 or 2 must be appended to/W), n, m and C, while an accent,

as in fm', will indicate that the variable C" replaces C. Then, by (4.3, 2),
/f/w-/r/», fpjF-fpjp, /r/w =/r/m. (44.2)
Let F be a function of the velocity defined in the first velocity-domain;
we write „ ; / l ( F ) s ///JW/^f, + F-F'1-F)gct1Je'dc. (4.4.3)
The quantity It(F) is similarly defined when F is defined in the second
velocity-domain. Again, if K is any function of c t and c„ and K' is the same
4.4] MAXWELLIAN VELOCITY-DISTRIBUTION 83

function of c[ and cj, we write

»i"i h*K) = tifTtfXK-K^gandtdc* (4.4,4)
«i»»/»(K) = \lA»ff\K-K-)g«ltd*-dcv (4.4, 5)
Since F and K appear linearly in the above expressions,
/tf + ^) = W) + /W, /(«fl««/(0, (4.4.6)
where a is any constant, and I may have any of the above suffixes. The
functions I possess a certain similarity to 8efj8t in being only partly inte-
grated, so that /,(F), Ii^K) are functions of c„ while In(K), /»(F) are
functions of c,.
Complete integrals related to these functions are defined as follows. First,
if F and G are functions defined in the first velocity-domain, we write
[F.Gk-jG,/,(*•)&». (4.4.7)
Then, by (4.4, 2) and (3.54,4),

[F.G], = ± jjjm^F+F.-F' -F3(G + Gi-C-GDj«i*'*<fcx,

(4-4. 8)
whence, by symmetry, [F, G], = [G, F],.
For functions F, G defined in the second domain, [F, G], may be defined in
like manner.
Again, when F, H are defined in the first domain, and G, K in the second,
let
[F1 + Gt, H1 + KJlt s frlM + KJdc^iCUHi + KJJc* (4-4.9)
Then, by (4.4, 2) and (3.54, 3),
[Ft + C H1 + KJlt = -L-JJJ/Joy(0)(Fi + G t _ F ; _ G i )

- [ H 1 + K,„F1 + C1lli. (4.4.10)

If «„ nt both refer to the same gas, so that <x„ = a„ it follows from (4.4, 9)
** [ F ^ + GJu-LF.GL (4.4,11)
These complete integrals bear a certain resemblance to the expressions
for A,^,, £,¾. The following compound of them is likewise analogous
to *(& + &):
n*{F,G} = "ffF.Gl^n^J^+F,, G ^ G J , , + ^ ^ , ^ , , (4.4,12)
where F and G are each defined in both velocity-domains, and n *= «! + »4.
84 BOLTZMANN'S H-THEOREM AND THE [4.41
On account of the linearity of these complete integrals in the functions
F, G, etc., relations typified by the following hold good for each of the
functions defined in (4.4, 7, 9,12).
{F,G) = {G,F}, {F,G+H} = {F,G} + {F,H}, {F.aG} = a{F,G),
(4-4. 13)
where a is any constant.
In {F, G) and [F, G] the functions F, G may be either scalars or vectors
or tensors, so long as they are of like kind. The incomplete integrals /(F)
are of the same nature as F; the integrands in [F, G] and {F,G} are under-
stood to contain the scalar products of G and /(F).
If the functions F, G do not involve the number-densities nl or n,
explicitly, the functions /j and /,„ and the square-bracket expressions
[F, G]j, [Fj + Gt, Hi + Kt]n,..., are absolutely independent of the number-
densities; the curly bracket expressions {F, G}, on the other hand, in general
depend on the concentrations njn, njn, though they also are independent
of the total number-density n.

4.41. Inequalities concerning the bracket expressions [F, G], {F, G)

It follows from (4.4, 8) that

[F,F]l - ^fjjr/riF+F^F'-FWga^'dcdct > o, (4.41,1)

since the integrand is essentially positive. Similarly, by (4.4, 10)
[Fi + G+Fi + GJnZo,
and so {F,F} > 0. (4.41, 2)
The sign of equality in (4.41, 1) is valid only if
F+F^F + Fl, (4.41,3)
i.e. if F is one of the summational invariants ^ ° of 3.2, or a linear combina-
tion of them. Thus, if F is a scalar quantity, the complete solution of the
equation It(F) = o, which, by our definitions, has [F,F]1 = 0 as an im-
mediate consequence, is
Fl = a\I>+«i».«lCl+«?>F1, (4.41.4)
1
where aj ', «P\ a? are arbitrary magnitudes independent of c„ which may,
however, be functions of r, t. Similarly
[Fj + G.^. + G J ^ - o
implies that Fl + Gt = F[ + G'it
and {F, F} = 0 (4.41, 5)
requires that
[ F . F h - o , [F I +F t ,F l +F t ] l i -o, [ F , F ] , - o ,
4.41] MAXWELLIAN VELOCITY-DISTRIBUTION 85

and hence that

F+K-F + K, F,+F,-*", + *"„ Ft + F = F't+F'. (4.41,6)
Thus, as in 4.3, if F is a scalar, the solution of (4.41, 5) is
F,-ep + al».MlCl + at»E„ Ft~a!ti + aP>.mtCt + afl>Er (4.41,7)
The results (4.41, i, 2, 3, 6) apply equally to scalar, vector and tensor func-
tions of the velocity.
 5
THE FREE PATH,
THE COLLISION-FREQUENCY A N D
PERSISTENCE OF VELOCITIES
5.1. Smooth rigid elastic spherical molecules
The work of this and the following chapter refers to molecules that are
smooth rigid elastic spheres not surrounded by fields of force. In this case
the molecules affect each other's motion only at collisions. The path of
a rigid molecule between two successive collisions is called a free path. For
non-rigid molecules an encounter has no definite beginning and end; the
concept of a free path then involves difficulties, and will therefore not be
employed.
Consider the collision of two molecules of diameters <rlt crt\ let

The angle ty made by the relative velocity gn with the direction k of the
line of centres of the molecules at collision can take any value between o and
nJ2. The deflection X °f the relative velocity in the collision (see Fig. 5, p. 58)
is given by ,

and so x takes all values from 0 to it. Also, as in (3.44, 2), the encounter
variable b satisfies the equation
b = <rltcoaix,
whence, by (3.5,4) a u = Jo-J,. (5.1,1)
e
Again, the angles x> arc polar angles giving the orientation of e' about an
axis parallel t o * n ; thus ^ _^ ^ ( j h 2)

5.2. The frequency of collisions

Consider collisions occurring between pairs of molecules m„ mt in a gas-
mixture at rest in a uniform steady state. The number of collisions per unit
volume and time such that e' lies in rfe' and the velocities of the colliding
molecules lie in ranges clt dcv and c„ dct, is, by (3.5, 5),
/i/t£<Xii<fe' de^ct
or, using the value of altdt' found in 5.1,

186]
5.21] THE FREE PATH 87
The total number of collisions occurring per unit volume and time between
pairs of molecules w„ m2 is obtained by integrating over all values of e', cv ct;
thus it is Nlt, where
Nn s IjffffJtgvlisinxdxdedcydct.
The integrations with respect to x and e offer no difficulty. The limits of
integration are o and n for x, and o and zn for e; hence, integrating and
substituting the forms for/j,/ a appropriate to the uniform steady state,

Nu = ^ * J J ^ JJexp{-(» 1 cf + « 1 i)/2*7-} f rfc.rfc,.

(5-2, 0
To evaluate this expression, the variables of integration are changed from
c,, c3 to the variables G, g2i introduced in 3.41. Then, by (3.41, 8),

Also, by (3.52, 5) the element dcxdct may be replaced by dGdgti, and so

Nu = ^^lL^JJexp{-m0(G« + M1Ml^)/2Ar^rfG^1.
(5-2. 2)
On integrating over all directions of gtl and ¢7, we get
2
*.t = -^^^—^\l\\M-^^ + MtM^)l2kT}g^dGdg.

Thus, using (1.4, 2) and (1.4,3) t 0 effect tne

integrations with respect to
G and g,
2 1
N„ . Z ^ K ( ¾) j\xp{-moMlMtg>/2kT}g>dg (s.2,3)

= 2»,n,o-? 2 ^-^-^-?j . (5.2,4)

5.21. The mean free path

Changing the suffix 2 to 1 in (5.2, 4), we get
Nu-inloJLnkTImtf. (5.21, 1)
The number of collisions between pairs of molecules mlt per unit volume and
time, is lNllt because Nn counts each collision between a pair of molecules
A, B (say) twice over, once regarding A as the molecule with the velocity c1(
and once as that with the velocity ct. On the other hand, the average number
of collisions of any one molecule of the first constituent, per unit time, with
similar molecules, is Nn/nlt not Nlj2ti1, since each collision affects two
molecules at once.
88 THE FREE PATH, THE COLLISION FREQUENCY (5.22
The average number of collisions undergone by each molecule per unit
time is called the collision frequency. Thus the frequency for a molecule m,,
for collisions with like molecules, is Nu/nl; for collisions with molecules m,
it is Nltjnv etc. The frequency for collisions of all kinds is

the number of terms in the bracket being equal to the number of con-
stituents in the mixture. The collision interval, or mean time between
successive collisions, is therefore rt, where
T, = nJ(Nn + NU+...). (5.21, 2)
The mean distance /, travelled by a molecule m, between successive collisions
in a given time t is called its mean free path. This is found by dividing the
total distance nlclt travelled by molecules mx in this time by the total
number nlt/rl of their collisions; thus
/, _ f,T, = nlcJ(Nu + Nu+...).
From (4.ii| 2) f, = C, = 2{2kT/mnl)i. Using this and the values found
above for Nu, Nlt, etc.,
/,«- i/7r{n,<7;V2 + ;!,o1,V(i + «,/*») + - } • (5.21, 3)
In particular, if only one gas is present,
/ = ijitna*^2 = o-joj/nna*. (5-2i. 4)
Another kind of mean free path was used by Tait,* who defined it as the
mean distance moved by a molecule between a given instant and its next
collision. The calculation of Tait's mean free path involves the evaluation
of an integral by quadrature; for a simple gas its value is
0-677/»« <r*.

5.22. Numerical values

For a gas at S.T.P. the number of molecules in a cubic centimetre is approxi-
mately 2-687 x I0'*- F° r hydrogen the molecular radius, found by com-
parison of the experimental values of the coefficient of viscosity with the
formula deduced on the assumption that the molecules are rigid elastic
spheres, is about 1-372 x io~'cm.; the radii found similarly for other mole-
cules are of the same order, though in general somewhat larger. As the mass
of a hydrogen molecule is 3-347 x io -M g., the number of collisions (|/V„)
occurring per second between hydrogen molecules in 1 c.c. at S.T.P. is found
to be 2-05 x 10**, and the collision frequency is 1-5 x io u sec.~ 1 ; also the
mean free path of a hydrogen molecule is 1 -112 x i o - t cm., and the mean time
between two successive collisions is 6-6 x io~nsec.
* P. G. Tut, Tram. R. Soc. Edmb. 33, 74 (1886).
5.3] AND PERSISTENCE OF VELOCITIES 89

The length of the mean free path docs not depend on the mass of the
molecule, nor on the temperature (unless the diameters of the molecules are
supposed to vary with the temperature; see 12.3). Thus, as the molecular
diameters for different gases are of the same order, the mean free path in any
gas at S.T.P. is of order io~ 5 cm., and so is several hundred times the diameter
of the molecule. This is the justification for the assumption of molecular
chaos made in 3.5: at the beginning of a free path which terminates in the
collision of two molecules, these are at a relatively large distance apart, so
that an appreciable correlation between their velocities is improbable.
The mean free path is inversely proportional to the density; thus in a
highly rarefied gas at, say, 0-01 mm. pressure, the free path is of order 1 cm.,
and so may be comparable with the dimensions of the containing vessel.
For a gas at a pressure of, say, icoatm., on the other hand, the free path is
comparable with the dimensions of a molecule; in this case the assumption
of molecular chaos may be invalid.

5.3. The distribution of relative velocity, and of energy, in collisions

The total number of collisions occurring in a gas-mixture, per unit volume
and time, between molecules ml and mt is, by (5.2, 3),

Nn = n . n . o l ^ i g i Q , ) 1 j ^ exp(-mtMtMtg*/2kT)g*dg.
The element
ninio-U2n)i i ^ f tSp(-m0MlM,^l2kT)g'4g (5.3, 1)

of this expression represents the number of such collisions in which the

relative velocity £ lies in the range dg. Hence the number in which g exceeds
an assigned value £ 0 is

»1 »• aU*")* (~£f) * J " exp( - m0 Mx Mtg*/2kT)g»dg.

I f we write x = g N/(w0 Ml Mj2k T), x0 = g0 x/(m0 A/, MJzk T), this integral
reduces to (2„mukT\i f- „ . .

i.e.to mint^^P^r-i(4+^ (S-3.2)

In the theory of activation of gas reactions it is sometimes assumed that a

reaction between two unlike molecules occurs at collision in a certain pro-
portion of those cases in which the kinetic energy relative to their mass-
centre exceeds a critical value E0. As the kinetic energy in question is
\m0MxMtg* (cf. (3.41, 8)), the number of encounters per unit volume and
time for which it exceeds E0 is given by (5.3, 2), provided that
E0 = \m0MxMtg\ = kTx%. (5.3, 3)
90 THE FREE PATH, THE COLLISION FREQUENCY [5.4

We can also use (5.3, 1) to find the mean value of any function <Xg) of the
relative velocity; averaged over all collisions. This will be ji, where

tfuftf) = nlntaU2n)i(^ji^tKP(-mi)M1Mtg*l2kT)^dg.
(5-3.4)
In particular, if $(g) = \m9Mx Mtg* s E\ the initial or final kinetic energy
of a pair of colliding molecules relative to axes moving with their mass-centre,

NltE' - n1nta^2n)i(^jij\xp{-m0MlMig*/2kTUm0MlMtg^g

-4w!,(^)V)».
whence, substituting for Nlt from (5.2,4),
E' = Jw0MlMtg* = zkT. (5.3, 5)
This may be compared with the average kinetic energy ($kT) for such pairs of
colliding molecules, relative to axes with respect to which the gas is at rest.

5.4. Dependence of collision frequency and mean free path on speed

The number of collisions occurring during dt between pairs of molecules
i»„ mt, such that clt c„ x> * lie in ranges dcv dct, dx, de is
dt
• i/i/*£fff*sin x</x<fe</ci</ct.
The total number of such collisions during dt such that c t lies in dcx is found
by integrating over all values of ct, x<e< it will be proportional to dt, and to
the number/,(ci)rfc, of molecules m, within the given velocity-range. As is
shown below, it is independent of the direction of c,. Hence this number may
be denoted by P^c^f^dc^dt; the function Pn(Ci) thus signifies the average
number of collisions per unit time per molecule of speed c,, with molecules of
the second kind. It is called the collision frequency for a molecule ml having
the speed c„ with molecules mt.
Dividing byf^dc^dt, we obtain
fwfo) = \\\\ftgo\t*mxdxdedev
The integration with respect to x a "d e ' s simple, the limits being 0 and
it for x< and 0 and 2it for e; thus
Plt(c,) = n<r\t\ftgdct. (5.4,1)
Let c, be expressed in terms of polar coordinates ct, 0, 9 about c, as axis;
then dct = c\sinddctd&d<y, and
g* m c\ + ^ - 2 ^ , ^ 0 0 3 0 . (5-4. 2)
5.4] AND PERSISTENCE OF VELOCITIES 91

Hence, on integrating with respect to cp between the limits o and zn, we get
Plt(Cl) = 27rV?8J//8£c?smddc2d6. (54, 3)
In this we change the variable of integration from 6 log; then, by (5.4, 2)
gdg = c,f,sin0</0,
and the limits of g are c, ~ ct and ct + ct. Hence
jgsmedd = -±-Jg*dg

3clc2

3M

or =-^-(3<1 + <1) 'f f*>M-

3cs
Using this result in (5.4, 3), and substituting the value of/,, we obtain
1
PM -1^.^.(¾)
x Ic- I exp ( - m2 cJ/2* T) c%c\ + 3cf) </c4
\ iJo

+J exp ( - mlf^2kT)c^y\ + c*) dct).

The second integral is equal to

the first, after two integrations by parts, reduces to

x I exp( —mtc\J2kT)dcv
Thus
P12(Cl) = » , a f t ( ^ ] * { e x p ( - i i i t £ ! / a * r ) + I ( ^ - ! + i )

x I exp( —m2 ¢1/2^71) «fc2}-

In terms of the error function Erf (x)* defined by the equation
Er((x)= J%'y'dy, (5.4,4)
• The notation is that of E. T. Whittaker and G. N. Watson, Modern Analysis, 4th ed.,
footnote to p. 341.
92 THE FREE PATH, THE COLLISION FREQUENCY [5.4

we have Plt(Cl) - nt<(^~f) V * ' + ( * * + i/*)Erf(*)>, (5.4. 5)

where x = Ci^mJzkT). (5.4,6)

The frequency Pti(C|) of collisions of the molecule with molecules of the
same kind is given by a similar expression; and the total collision frequency
of the molecule is D, . „ , . , „ . . ,
^1)=^11(^1)+^1)+-
Hence the collision-interval rx(c^) for molecules of speed cx is given by

i/'i(q) = Pi(ct) = P n ( c i ) + A ^ i ) + . » , (S-4. 7)

and the mean length /,(<:,) of their free paths is equal to CjT,. In particular,
when only a single gas is present

- x»lninl<r\E(x), (5.4,8)
where E(x) denotes the function
E(x) = xe-** + {2x*+ i)Erf (*), (S.4> 9)
and * is now given by * = ctJ(mj2kT). (5.4, 10)
Values of E(x) have been tabulated by Tait.* Using his table, a table
giving the ratio l(c)/l is obtained which is given below.f

Table 2

cn *• mn
0 0 0
025 0-080 0-3445
o-5 0-318 0-6411
0-627 o-5 07647
0-886 t 0-9611
1-0 '»73 10257
1-253 2 I-I340
t'535 3 1-2127
1-77» 4 1-2572
2 SO93 1-2878
3 11-459 1-3551
4 — 1-3803
5 — l-3°a3
6 — I-3989
00 00 1-414»

• P. G. Tait, Tram. R. Soc. Edinb. 33, 74 (1886).

t This table it taken from O. E. Meyer'i Kintiic Thnry of Gout (Longman, 1899),
P- 4»9-
5.5] AND PERSISTENCE OF VELOCITIES 93

5.41. Probability of a free path of a given length

Let />(/', ri) denote the probability that a molecule moving with a speed ct
shall describe a free path at least equal to an assigned value /'. The probability
that it should undergo a collision while travelling a further distance dl' is
(dl'/cJPAcJ-dl'/lfa).
Hence the probability/^/' + dl', c t ) that the molecule describes a free path at
least equal to I'+dl' is ^ ^ { i _dVjli{Cx%

and so p(i;Cl){i -dl'Mc,)} =p(l>,Cl)+dl'?£^,

whence Slnpij^) i_
whence
di' i^y
Integrating, and using the fact that/>(o, c,) = i, we find that
McJ-rHM. (5.4,,,)
The probability p{l') that a molecule of any speed should describe a free
path at least equal to /' is not e-*1**, because of the variation of the collision
frequency with speed. Jeans* has found by quadrature that for a simple
gas, over the range of/' for which p(l') is appreciable, it never differs by more
than about i % from e~l<M", which is the value for molecules moving with
speed (<Jn)£/2.
It is clear from (5.41, ,) only a very small proportion of molecules have
free paths many times as long as the mean free path.

5.5. The persistence of velocities after collision

After a collision with another molecule the velocity of a given molecule will,
on the average, still retain a component in the direction of its original motion.
This phenomenon is known as the persistence of velocities after collision.
As a consequence, the average distance traversed by a molecule in the
direction of its velocity at a given instant, before (on the average) it loses its
component motion in that direction, is somewhat greater than its (Tait) mean
free path.
The collision frequency for a molecule mv having the speed c„ in collisions
with molecules mt, is Pitfo); for the particular set of such collisions in which
the velocities of the latter molecules lie in the range c „ dct, and Xi £ lie in
ranges dx, de, the collision frequency is
iftg<rl*siaxdxdedcv
Let cj(cj) denote the mean velocity, after collision, of molecules m, which
before collision have the velocity c,, the average being taken over collisions
* J. H. Jeans, Dynamical Theory of Gases (4th ed.), p. 158 (19x5).
94 THE FREE PATH, THE COLLISION FREQUENCY (5.5
e
with molecules mt for all possible values of c„ x, '< then

ci(Ci) - 7^- jjj/tc'igvltsmxdxde^

or, using (3.51, 4),

ci(«?i) = •jjrJj)fi{ci + MAgtl-g'n)}g<r*lt*inxdxdedct.

The components of g'n parallel and perpendicular to gtl are j> cos ^, g sin x-
As e varies, the second component rotates about gn; hence its integral with
respect to e is zero, so that

cl(Ci) - jprJ)ft{Ci + Mtgn{i -costfteo-fssinxrfxrfc,.

The integration with respect to x between the limits 0 and n is elementary,

and yields the result

cl(«o) - jrJMci+Mtgn)g<r\tdct = £-j/^A/^+MjC,)^,^,,

or, using (5.4,1),

€[(€,) ~MlCl+^0!Jftctgdct.
The integral may be evaluated by the method used in 5.4, expressing c,
in terms of polar coordinates et, 0,9 about c, as axis. On integration over all
values off, the component of c, perpendicular to c, vanishes; as the com-
ponent in the direction of C, is c, cos 0, the mean value of e[ is w^c,) c„ where

«,,(£,) - ^, + -^,--^11/,^08^^8^^^,^.
The quantity 101,(0,), which is the ratio of the mean value of the velocity of
a molecule after collision to the velocity before collision when the latter
velocity is of magnitude c„ may be termed the persistence-ratio for molecules
of speed cv
On changing the variable of integration from 0 tog, as in 5.4, it follows that
1
kco.*.in^ = ^ / 1 + ^ 1 ^

2Ct
•=^( C «-S«i) i f c
i>c«

2CX
j£i(c\-s<f) if *,>*,.
5.511 AND PERSISTENCE OF VELOCITIES 95
The expression for m,2(c,) accordingly reduces to

».£,) -Ml + l A ^ ? 2 jjjt | (4 - St) dc2 +fj,cM - 54) &,).

(5-5. i)
The second of the integrals in the bracket can be evaluated in finite terms,
and the first integral can be expressed in terms of the error function Erf (x)
of (5.4, 4); on so doing, the value of tolt(cj) is found to be
. . ,, n~M,<r\a (zitkT\iI 1 ., 1 - 2 * * , . . .1

(5-5- 2)
where x is given by (5.4, 6).
The value of TO,2(C,) given by (5.5, 2) varies between M , —Mj/3 and M I
as x varies between 0 and 00, i.e. as c, ranges from 0 to 00. For collisions
between like particles, for which Ml = Mt = J, n , ^ , ) lies between J and J.

5.51. The mean persistence-ratio

An expression for tn12, the mean value of Wii(fi) averaged over ail values of
cv may be derived as follows. The number of molecules m, per unit volume
whose speeds lie between cx and c, + dcx is 4/1/, c\dcx; the number of collisions
per unit time with molecules of the second gas in which these molecules
participate is ^nf^ld^.P^. Hence, by (5.5, 1), as the total number of
collisions between molecules ml and m2 per unit time is Nlt, the mean value
of BT12(CI) averaged over all possible collisions is given by

AT12c7I2 = A/, 4^/1cJPl2</cI + j-—»

Jo IS

* (JT7, f! l!*ft i(c| •sc!)dci+JT/,c,M ~sc|) **] **) •

or, since Nn = I 4 ^ / 1 ^ ¾ ^
Jo
, ,, i6M.»rV?,
by .,,-^, + - ^ - ^

*! J/1 c! [ £/s I(c| •sc?) * • + I / J C J ( C ? •sc|) dct]dci) •

On inserting the known expressions for / t , / 2 , and Nlt (cf. (5.2, 4)), this
becomes

M +
"it = i ^ p ^ ( { 0 " J o r , e x P { - ( « l C ! + m 8 C i)/2Ar}^( C l-sci)*,</c 1

+ I exp{ —(mlcJ + OT2c|)/2A7'}c{c2(c5-5c|)A:j</c1>.

96 PERSISTENCE OF VELOCITIES [S.S1
In this write ct — 6cx; then

where /(5) - f expf-cftmt + m^/zArjcfefc!

_sisV»r/_2*r_\*
™ 16 Vi^+m,^*/ '

- -^^:,¾^¾¾.
whence, on evaluating the integrals,
• u - J3f1+iMJMi-Mn[(A/^ + i)/Aff]. ( S . S i, i)
The value of wlt given by this equation increases from zero to unity as
mjnti increases from zero to infinity; its value for the collisions of like
molecules, when Mt — Mx — J, is - + —;-In (1+,^2), or 0-406. For the
4 4^/2
encounter of heavy molecules with light molecules, the persistence ratio
for the heavy molecules is nearly unity, and that for the light molecules
nearly zero: this implies that, as might be expected, the heavy molecule
continues its path nearly undisturbed, while the light molecule bounces off
in a direction unrelated to that of its previous motion.
 6
ELEMENTARY THEORIES OF THE
TRANSPORT PHENOMENA

6.1. The transport phenomena

The phenomena of viscosity, thermal conduction, and diffusion in non-
uniform gases represent tendencies towards uniformity of mass-velocity,
temperature and composition. The kinetic theory attributes these tendencies
to the motion of molecules from point to point. This tends to equalize
conditions at the two ends of each free path, by transporting to the further
end an average amount of momentum and energy that is characteristic of
the starting-point. Hence we may speak of the phenomena in question as the
transport phenomena, or the free-path phenomena.
It was shown in Chapter 4 that a gas in a uniform steady state has a
Maxwellian velocity-distribution function. When the gas departs slightly
from a uniform steady state, Maxwell's function gives a first approxima-
tion to the actual distribution of velocities. Thus the results of Chapter 5
are still approximately true; they may therefore be used in obtaining
approximate expressions for the coefficients of viscosity, thermal conduction
and diffusion in a gas composed of rigid elastic spheres of diameter <r.

6.2. Viscosity
Consider a simple gas, uniform in temperature and density, moving parallel
to Ox with a mass-velocity « 0 which is a function of z alone. Thus the gas is
in laminar motion parallel to z = o, and te0 = o, to = W.
Consider the rate of transport of ^-momentum across unit area of the
plane z = 0. The number of molecules crossing this area in unit time from
the negative side (that on which * is negative) to the positive side is

f wfdc = f WfdC,
the integration extending over all values of C for which W is positive.
Similarly, the number crossing from the positive side to the negative is

fj-w)fdc-fj-W)fdC,
the integration extending over values of C for which W is negative. Since
the gas has no mass-velocity parallel to Oz, these two numbers are equal.
Their value, to a first approximation, is {nC, the same as for a gas in a
uniform steady state (cf. (4.11, 5, 6)).
197 1
98 ELEMENTARY THEORIES OF THE [6.21
The mean x-velocity of molecules crossing the area from the negative
side to the positive is not the value of u„ appropriate to the plane z = o. It
corresponds to the mean layer in which their free paths began, or, if allowance
is made for the persistence of velocities after collision, to a somewhat more
distant layer. Thus their mean x-velocity is that of a layer z = — u/, where
I is the mean free path of the molecules, and u a numerical factor of order
unity. Their total x-momentum is
i»Cm(«o)«~ uh
or, as / is usually very small compared with the scale of variation of the mass-
properties of the gas, , - .

the values of u, and dujdz here refer to the plane z =» o.

Similarly, the mean x-velocity of molecules crossing the area per unit time
from the positive side to the negative is that corresponding to z = + u/, and
their total x-momentum is
ipCK).-„, = i/>c(«.+u/^°),
u0 and dujdz in the last expression again referring to * = o. Thus the net
rate of transport of x-momentum across unit area of the plane z * o from
the negative side to the positive is
JpC(„ 0 -u/§)_ bC ( Ko+ u / | . ) . _ i u p C / § .
This momentum transport is equivalent to a force of this amount per unit
area, exerted parallel to Ox, by the gas on the negative side of z = o, upon
the gas on the positive side. According to the usual definition of the coefficient
of viscosity /*, this force is — ft dujdz. Hence

ft = jupc/ = —t ^^-j--. (6-2« 0

by (4.11,2) and (5.21,4). The value of u is found (cf. (12.1,6)), by the
exact methods described later in the book, to be 0-1792»' or 0-998.

6.21. Viscosity at low pressures

In a gas at a sufficiently low pressure the mean free path is comparable with
the dimensions of the containing vessel; this produces an apparent diminu-
tion in the viscosity of the gas. Consider as a typical example of this pheno-
menon the motion of gas between two infinite parallel walls ar = 0 and *«=</,
the first of which is at rest, while the other moves parallel to Ox with speed q,
the motion of the gas being supposed due solely to that of the walls, and not
to imposed pressure-gradients. Let qv qt denote the mean x-vclocities of
6.21] TRANSPORT PHENOMENA 99
molecules just before striking and just after leaving the second wall; then
the mean *-velocity of the gas at this wall is $(¢,+q t ).
Experiment indicates that some of the molecules striking the moving wall
enter the material, and later leave it with the temperature of the wall, and
a mean ^-velocity equal to its velocity q. The remainder are reflected
elastically. The proportion 0 of those that enter depends on the nature of
the wall and the gas, and on the condition of the surface of the wall. The
mean x-velocity of molecules leaving the wall is a weighted mean of qx (for
the elastically reflected molecules) and q (for those that enter the wall) in
the proportion i — 6 to 0; that is,
qt = (l-0)q1 + 0q. (6.21, l)
Now, as in 6.2, the value of qx is equal to the mass-velocity of the gas at a
distance u/ from the wall, which differs from that at the wall by ul(dttjdz),
where dujdz denotes the (constant) gradient of the mass-velocity of the gas.
Thus -

On combining (6.21,1) and (6.21, 2) we find

so that the gas near the wall is slipping along the wall with a speed
2-0 ,Bu0 .,
-g- u/ -£. (6.2.,3)
There is a similar speed of slip at the other wall; hence the difference between
the mean velocities of the gas near the two walls is
2-0 ,du0

and since this is equal to (duJSz) d,

0«o. 1(..
qlld+zul* .2-0\
dz if\-'"- e r
The viscous stress transmitted by the gas is /1 BuJSz per unit area. In the
absence of slipping at the walls this would be equal to fiq/d; if it be denoted
by/i'q/d, . 2-<A
p' =fid \d+2ul-g-). (6.21,4)

The effect of slipping at the walls is that the apparent viscosity, /1', of the
gas, is smaller than the true viscosity /«. The reduction is very small for
ordinary pressures, but for pressures so small that / becomes comparable
with d it is very pronounced.
100 ELEMENTARY THEORIES OF THE [6.3
The above discussion fails when I exceeds d, since molecules pass direct
from one wall to the other, and the transport of momentum ceases to be a
free-path phenomenon. In this case experiment shows that /i' decreases to
zero with the pressure.*

6.3. Thermal conduction

The simple theory of thermal conduction is similar to that of viscosity. We
consider now a simple gas at rest, whose temperature T is a function of z;
it is required to find the rate offlowof heat across unit area of the plane a = o.
Let E denote the total thermal energy of a molecule; then E, the mean
thermal energy of molecules at a given point, is a function of T, and so of a:;
also, as in 2.43, the specific heat cv of the gas is given by the relation
e
*~M*)'
The number of molecules crossing unit area of the plane * » 0 from the
(6.3.1)

negative side to the positive in unit time is \nC, as in 6.2. Each of these
carries with it thermal energy equal, on the average, to the value of E, not
at the plane * = o, but at a = - u7, where I is the length of the free path, and
u' is another numerical constant which, like u (cf. 6.2), is of order unity. Thus
the total thermal energy which they carry across * = 0 is
\nC(E)^^\nC(E-Wldly
where, in the last expression, E and dEjdz refer to z = 0. Similarly, the
total thermal energy carried by molecules that cross unit area of the plane
* — 0 from the positive side to the negative is
J«C(£U 7 -JnC(£+u'/^),
where again E and dEjdz refer to * = 0. Thus the net rate offlowof thermal
energy across unit area of the plane z •= 0 from the negative side to the

'"'"*" ^(E-u7|f)-(.c(j+u'/g)

«-$«»CVfc,,7j-,
by (6.3, 1). In the theory of the conduction of heat this rate offlowis written
— A dT/dz, where A is the coefficient of thermal conduction of the material.
Accordingly A = W'C<., (6.3,2)
• W. Crookei, Phil. Tront. 17a, 387 (188a). See alto M. H. C. Knudwn, Kinetic Theory
of Gostt (Methuen, 1946).
6.4] TRANSPORT PHENOMENA 101

whence, by (6.2, 1), A = f/tcv, (6.3,3)

where f is a new numerical constant, equal to u'/u.
The constant u' of this section is not equal to the constant u of 6.2, because
of the correlation between the energies of molecules and their velocities;
molecules possessing most energy are in general the most rapid, and there-
fore possess the longest free paths. In consequence, u' is in general greater
than u, and f is greater than unity. For the transport of internal energy,
however, which is but slightly correlated with the molecular velocity, an
equation similar to (6.3, 3), with f roughly equal to unity, may be expected
to hold; the symbols A and <"„ now refer to the coefficient of conduction, and
the specific heat, of the internal energy. In general, the larger the ratio of
internal to translatory energy, the smaller is f.
It is immaterial, in the formulae of this section, whether thermal energy is
measured in thermal or mechanical units; the effect of a change from one
set of units to the other is that E, c0, and A are all multiplied by the same
factor.

6.31. Temperature-drop at a wall

Just as the mean velocity of gas near a moving wall differs from that of the
wall, so there is a difference between the temperature of a hot body and the
gas that conducts heat away from it. This temperature-difference is, by
analogy with (6.21, 3), equal to

-ru/i*. (6 3I>I)
-
where dT/dz denotes the temperature gradient near the body.
One method of measuring the thermal conductivity of a gas is to determine
the loss of heat through the gas from a hot wire of small diameter. In such
experiments the temperature-drop at the surface is usually very important.

6.4. Diffusion
Consider a binary gas-mixture, uniform in temperature and pressure, whose
composition varies with z. If «,, nt are the number-densities of the two gases,
then, as the pressure is uniform,
dp ,Te{nx + nt)
° dz ~ dz '

whence ^ = - ¾ . (6.4,1)

We assume that the pressure remains uniform as diffusion proceeds in the

mixture; this is very nearly true.
The number of molecules m, crossing unit area of the plane z = 0 from the
negative side to the positive in unit time is J/tj Clt as before: in this expression
102 ELEMENTARY THEORIES OF THE [6.4
ftj no longer refers to z = o, as the gas is non-uniform, but to z = — u x / lt say.
Since collisions between pairs of molecules mx do not directly affect the mean
diffusion velocity Wv lt is taken to be the mean free path of molecules tnl
between successive collisions with molecules «,; ux is a number of order
unity. The number of molecules in question is thus
J^O.-u./.'J^^-U,/^).
The number of molecules m, crossing unit area of z = o in the opposite
direction per unit time is similarly
K\(*iU*-K\(«i+Vi||f).
Hence the net number-flow of molecules ml per unit area and time across
* — o from the negative side to the positive is

^.(-1--.45)-^(^+^)-10,4¾¾.
Similarly the net number-flow of molecules m, per unit area and time
across z = o in the opposite direction is, by (6.4, 1),

where /, is the mean free path of molecules mt between successive collisions

with molecules mlt and u, is another number of order unity. This does not
balance the number-flow of molecules mt; however, to ensure that the
pressure remains uniform there must be no resultantflowof molecules. It is
therefore necessary to assume that the first trace of a pressure difference in
the gas immediately produces a massflowof the gas, with a velocity

just sufficient to counter the resultant number-flow. When this mass-flow is

taken into account, the number-flow of molecules m, becomes

»,+n, Bz'
the number-flow of molecules mt is equal in magnitude, but opposite in
direction. We write the two as -Dltdnjdz and -Dudnjdz, where D„ is
called the coefficient of mutual diffusion of the two constituents in the
mixture. Then ,, ,n . 1 r>\
n ^ , ^ / , ^ , + ¾¾¾¾ (6 }

In (6.4, 2), /1 and /, are respectively inversely proportional to n, and n^,

thus (6.4, 2) implies that Dlt is inversely proportional to the total pressure,
6.5] TRANSPORT PHENOMENA 103
whereas the coefficients of viscosity and thermal conduction are independent
of the pressure. This result agrees with the exact theory, which shows,
however, that D12 varies somewhat with the proportions of the two gases in
the mixture (cf. 14.3). Thus Uj, ut must be supposed also to depend somewhat
on the proportions.
When the two gases in the mixture are identical, the process under con-
sideration is the diffusion of certain selected molecules of the gas relative to
the rest. In this case Dxt is replaced by Z) n , which may be termed the
coefficient of self-diffusion of the gas; since noww 2 /j = «,/, = (z^ + n,)/,
where I is the free path when all the collisions are taken into account, we may
write
Du = i u u / C . (6.4,3)
Hence, by (6.2, 1), Du = u'n/i/p, (6.4,4)
where u'n denotes a new numerical constant of order unity.
In a gas-mixture, diffusion can also arise as a consequence of a temperature
gradient. No really satisfactory simple theory of this 'thermal diffusion' can
be given, employing only the simple free-path ideas of this chapter. The
reason is that thermal diffusion is an interaction phenomenon, arising from
the processing by collisions of the heat-flow which is the immediate conse-
quence of the temperature gradient. Similar remarks apply to the inverse
'diffusion thermo-effect', in which diffusion due to inhomogeneity of
composition is found to be accompanied by a heat flux in a gas-mixture
initially at uniform temperature.

6.5. Defects of free-path theories

The free-path formulae derived above depend on the undetermined numbers
u, u', u1( u2, f, u n , uj,. Elementary arguments are often used to derive rough
values for these numbers. Implicit reliance on the formulae so obtained may,
however, lead to serious difficulties, especially in the theory of diffusion.*
Better estimates of the numbers can be obtained by taking successively
into account such factors as the difference between the lengths of the free
paths of different molecules, the persistence of velocities after collision,
and so on. This is the plan adopted by Meyer and others, f Even in its most
refined form, however, this mode of attack leads to somewhat inaccurate
results, and the precise magnitude of the error has to be found by other
methods. Also the free-path methods apply only to rigid spherical molecules.
Similar objections apply to other approximate theories, such as those
described later in this chapter. To obtain reliable results, it is better to
proceed by evaluating the velocity-distribution function, as in the later
chapters of this book.
* See S. Chapman, 'On approximate theories of diffusion phenomena', Phil. Mag. 5, 630
(1928).
t Cf. O. E. Meyer, Kinetic Theory of Gases, or J. H. Jean*, Dynamical Theory 0/ Gases,
chapters 11-13.
104 ELEMENTARY THEORIES OF THE (6.6

6.6. Collision Interval theories

A rather different kind of approximate theory is based on the collision
interval r instead of the free path.* Consider a simple gas under no forces.
As was shown in Chapter 4, collisions cause the velocity-distribution func-
tion/to approach the Maxwellian value/10' corresponding to the local values
of n, ca and T. The collision interval theory assumes, as its basic approxima-
tion, that during a time dt a fraction dtjr of the molecules in a given small
volume undergo collisions, which alter their velocity-distribution function
from/to/0". This is equivalent to assuming that the rate of change dtf/dt of
/due to collisions is — {J~fm)lT' *° th** *he Boltzmann equation becomes

Because of the smallness of r this equation implies that, in a gas whose state
is not varying rapidly with time,/—/™ must be small. Replacing/ by/i* on
the left of (6.6,1) we findf
f.fm-rlg-erg. ( 6 .6,,)

For a gas of uniform density and temperature, streaming parallel to Ox with

a mass-velocity ue which is a function of x alone, equation (6.6,2) becomes

The viscous stress in the x-direction across a plane z = const, is

A» - /«(« - «o) «>/<&• (6.6,4)
0
Substitute for/from (6.6, 3); then, since/' ' is an even function of «-«„,
(6.6,4) reduces to p„ = - ft dujdz, where
f df®
'T
rjm(ti -«,)»» -^- dc
= 0 - { T Jm(H-«0)KtyWrfc}+Tfmw*/<<»</c.

The integral in the brackets vanishes, the integrand being an odd function of
* Such theories, uted in the theory of ionized gate* for a number of years, have attained
wide vogue through the work of P. L. Bhatragar, E. P. Grots and M. Krook, Phys. Rtv.
»4. $«« 0954).
t The exact solution of (6.6, 1) ia
/ - j " e-,,"f>(c, r-ef, t-f) T->«ft';
this can also be derived directly by classifying the molecules in the rsnge c, dc that
reach the volume r, dr at time t according to the times f — (' at which they began the
free paths that bring them into the volume. Equation (6.6, a) follows from this integral
on expanding/'" in powers of t' and neglecting ('* and higher powers.
6.61] TRANSPORT PHENOMENA 105
u-u0; the second integral is equal to p, by (142, 1) and 2.32. Hence the
viscosity misgiven by ^= ^ (66>s)

Similarly for a gas at rest, uniform in pressure but with a temperature

which is a function of z, equation (6.6, 2) becomes

f-r-ir^- (6-6,6)

If the molecules have no internal energy, the corresponding heat flow

- A dTjdz in the z-direction is {^mchofdc. On evaluating the integral by
(1.42, 1) and (1.4, 2) the thermal conductivity A is found to be
8
A= T
ar J*"**"* = 4(1^)
= i/"Cv, (6.6,7)
since p is constant in the gas, and c„ = 36/2»».

6.61. Relaxation times

The collision interval r as used in 6.6 takes no account of persistence of
velocities after collisions, or of the variation of collision frequency with
molecular speed. Thus it must be regarded as an ideal collision interval,
which may be expected to differ from the collision interval of Chapter 5 by
a numerical factor analogous to the numbers u, u' of 6.2 and 6.3. For example,
to make (6.6, 5) and (6.2, 1) agree we must take
T = 3u/Ci/2Cr» = 4u//ffC,
whereas the T of Chapter 5 was l/C,
To elucidate further the significance of the present T, consider a uniform
gas with a non-Maxwellian velocity distribution. For such a gas (6.6, 1)
becomes
8i---f-. (6-61.0
m _ T
so that/—f decays like the exponential « '' . Thus T is the time in which
a disturbance to the Maxwellian state decreases to i/e of its original value;
it is called the time of relaxation of the disturbance.
Exact theory shows that an equation of form (6.61, 1) holds good only
when/—/"" is small, and then only for a specific set of 'eigenfunctions'
/ - / < 0 ) , the relaxation time differing from one eigenfunction to another
(cf. 10.321). If /(1),/(8),..., are the successive eigenfunctions, and Tvrt,...,
the corresponding relaxation times, any small deviation of/from/"' can be
expressed by a series , m _ v . - ,
m

corresponding to this |f = _ £ ^"i". (6.61,2)

106 ELEMENTARY THEORIES OF THE [6.62
Equation (6.61, 1) is to be regarded as an approximation to (6.61, 2), the
T in (6.61, 1) representing an average value of rlt T„ ..., which varies from
one function f—f® to another. In particular, T in equations (6.6, 5) and
(6.6, 7) may be expected to stand for two distinct average relaxation times,
unequal but similar in order of magnitude, characteristic respectively of
viscosity and heat conduction.*

6.62. Relaxation and diffusion

An approximate theory of diffusion can be given in terms of a relaxation time
without explicitly determining the velocity-distribution function. Suppose
that diffusion in a binary mixture leads to a force between the two gases
proportional to their relative velocity Cj — ct, tending to destroy this velocity.
Let T U be a time such that the forces per unit volume on the two gases are
± (PiPtlfTit) (?i—^). Then, if the factors leading to diffusion were suddenly
to cease to operate, the two gases would experience accelerations
-PI(CI-CI)/^„, Pi(ci-ct)lprlt
and their relative acceleration would be - (ct -ct)/rlt. Thus the velocity of
diffusion 2 i - c t would die away exponentially like e"*/r'»; that is, T„ is the
relaxation time for diffusion.
In steady diffusion, the two diffusing gases are assumed to move inde-
pendently, each under the action of its own partial pressure p, an,kT,
together with the drag ± {PiPtlprlt) (ct - c») exerted by the other gas. Since
the gases are unaccelerated,
- Vpi = P1P&1" ct)fprlt = V/>„ (6.62, 1)
where V m d/dr (cf. 1.2). In the case of diffusion at constant temperature,
this give. _ _ ^ _ _pkTTitpnjiPlPr (6.62, 2)
According to 6.4, in this case
»1*1 - -0itV»i - -ntct.
These equations are consistent if
#ii = pr^kTItm^mt. (6.62, 3)
Provided that this equation is satisfied, the results of the relaxation theory
duplicate those of free-path theory. In terms of Dlt, the drag forces which
the two gases exert on each other per unit volume are
±PiPt(ci-ct)lpDlt. (6.62,4)
We may relate r lt to the collision processes by assuming that there are Nlt
collisions between unlike molecules per unit volume and time, and that these
* For • more detailed discussion of the relation of relaxation times to the transport
phenomena, see E. P. Gross and E. A. Jackson, Phys. Fluids, a, 43a (1959).
6.62] TRANSPORT PHENOMENA 107
reduce the mean velocities cx and c 2 of the colliding molecules to a common
mean velocity c'. By conservation of momentum
mlc1 + mtci = (ml + mt)c'; (6.62,5)
hence the mean momentum change of a molecule mx at a collision is
w
i ( « ' - « i ) . o r -m1mt{c1-Ct)l(ml + m2). The drag -PiP a (ci-c,)/pTi, on
the first gas is Nn times this change of momentum, and so
T I2 = > W » i + mt)/pNlt = ( p , ^ +p1nt)/pNlt. (6.62, 6)
Thus T, 2 is a weighted mean of the collision intervals nlfNla, nt/Nu for
molecules mlt m2 in mutual collisions, with weighting factors p,, p,. If wij/m,
and P2/P1 are small, T12 approximates to nJNlt, the collision interval for
molecules m2.
By (6.62, 3, 6), = (w1+w2)/>1/>2
12
m^tpN^ • V • • /;
Since A^oc n,n2, this formula indicates that Dlt is, at a given temperature,
inversely proportional to the total pressure, and independent of the pro-
portions of the mixture. As noted in 6.4, exact theory indicates a small
dependence of Da on the proportions of the mixture, so that to this extent
the relaxation approach is inadequate. Also the relaxation theory, like the
simple free-path theory, is not able to explain thermal diffusion. However,
because of its simplicity a relaxation argument has often been employed in
complex problems, particularly in relation to ionized gases.
Consider, for example, a moving binary mixture in which, in addition to
the drags (6.62, 4) and their partial pressures, the gases are subject to applied
forces m,F, per molecule. Since the velocities of the two gases differ only
by the small velocity of diffusion, the acceleration of each gas is, to a first
approximation, equal to the acceleration DcJDt of the mass as a whole.
Thus the equations of motion of the two gases are

Pi ^ -P1F1- v>i - ^ (*i - ct), (6.62, 8)

Pt 5 f = P8 F* ~ ty* + & £ (c, - ct). (6.62, 9)

Adding these equations, we get the approximate equation of motion of the

gas as a whole, n

On the other hand, eliminating DcJDt,

ci - ct = T^Ft. - PT lV/>t - (F 2 -pi 'Vp,)}. (6.62,10)
This is a general equation of diffusion. It implies that relative diffusion
occurs whenever the accelerations that would be given to the two gases by
5-a
108 ELEMENTARY THEORIES OF THE [6.63

the forces p, F,—Vp, acting on them are unequal; the diffusion velocity is r lt
times the difference of those accelerations.
The diffusion equation can also be derived from the Boltzmann equations
of the two gases, written in the approximate forms

|tCl.|+fl.|._4^, (6,,„„

where T, = Hj/Af,,, rt « nJN^, and /J", _/}" denote Maxwellian velocity-

distributions about the mean velocity c' given by (6.62, 5). Correct to
terms of first order in r lt T, this gives results agreeing with (6.62, 8, 9).
However, modified forms of (6.62,11, 12) may be used in certain more
general diffusion problems for which the simple relaxation method is
inadequate.*

6.63. Gas mixtures

An approximate theory of the viscosity of a binary or multiple mixture of
gases can be given as follows. Each gas is assumed to have its own relaxation
time; that of therth.gas is rf, where r~x is the collision-frequency 2 A^/n,.
t
Write Nu - /ij\ntkT; then
r,-1 - kTXntfij1. (6.63,1)
1
It is further assumed that, since the viscosity is a measure of the ease with
which momentum is transmitted by the molecular motions, the viscosity of
the gas as a whole is the sum of contributions from the separate gases,
calculated as in (6.6, 5), i.e.
/* = VPS. = S : r r r ^ V r - • < 6 6 3* 2>
By taking all save one of the quantities n, equal to zero, /tu can be seen to be
the viscosity of therthgas when pure; /iH can similarly be interpreted as a
' mutual viscosity' of therthandrthgases. More correctly /<^ '«„ or NJn, k T,
is to be interpreted as a measure of the extent to which collisions with mole-
cules mt impede the transfer of momentum by a molecule m,. Thus (6.63, 2)
implies that the contribution of therthgas to the viscosity is proportional
directly to the number-density n„ and inversely to the sum of terms /t.l1"!,
rti1 «ti • • •. measuring the extent to which collisions with molecules m1( w„...
impede the flux of momentum.
Equation (6.63, 2) is similar to one derived by Sutherland! by free-path
methods. His derivation, like that above, gave Mit " Pu- However, he then
* P. L. Bhatnigar, E. P. Grow and M. Krook. Phyt. Rtv. 94, 511 (1954).
f W. Sutherland, Phil. Mag. 40, 4*1 (189s)-
6.63] TRANSPORT PHENOMENA 109
introduced a semi-empirical correction to allow for differences of velocity-
persistence of the two colliding molecules, equivalent to taking
ft* = ft.K/wt)1-
Even if the possibility fi# 4= fit* >s admitted, a formula of the form (6.63, 2)
can only be approximate; collisions do not simply obstruct the flux of
momentum, but transfer it from one gas to another.
An approximate formula similar to (6.63, 2) was given by Wassiljewa*
for the thermal conductivity; it is

A = S r^nrr--,---i---. (6.63,3)
where AM is the thermal conductivity of the rth gas when pure, and Art
a 'mutual conductivity'. A formula of this type, with Art 4= Ate, is found to
give a fair degree of agreement with experiment for the variation of A with
composition, even for gases whose molecules possess internal energy.
Finally, the methods of 6.62 can be extended to apply to diffusion in a
multiple gas-mixture.f In such a mixture, collisions with molecules of the
rth gas lead to a drae force t t

per unit volume acting on therthgas (cf. (6.62, 4)). Here D# is the coefficient
of mutual diffusion of the rth and rth gases, at the pressure of the actual
mixture (it should be noted that pD^ is independent of the pressure). Hence
equations (6.62, 8 and 9) can be generalized to read

These equations can be solved to give DcJDt and the velocities of diffusion,
relative to an arbitrarily chosen gas of the mixture, of all the remaining gases.
In certain problems, notably those referring to ionized gases in a rapidly
alternating electric field, it is inadequate to equate the accelerations of the
different gases to DcJDt. In such problems, DcJDt is to be replaced on the
left of (6.63, 4) by D,cJDt, where DJDt is the derivative following the
motion c, of the rth gas.
• A. Wassiljewa, Phys. Z. 5, 737 (1904).
t See M . H . J o h n s o n and E. O. Hulburt, Phys. Rev. 79, 80» (1950); M . H . J o h n s o n ,
Phys. Rev. 82, 298 (195:); A. Schltlter, Z. Naturf. 50, 7a (1950), and 6 a , 73 (1951).
 7
THE N O N - U N I F O R M STATE
FOR A SIMPLE GAS

7.1. The method of solution of Boltzmann's equation

The present chapter deals with the method by which Enskog solved
Boltzmann's equation in the general case. We consider here a simple gas,
and in Chapters 8 and 18 binary and multiple gas-mixtures, whose mole-
cules possess translatory energy only. The corresponding theory for a gas
whose molecules possess internal energy is given in Chapter 11.
Enskog's method is one of successive approximation. Boltzmann's
equation can be expressed in the general form £(/) = o, where £(/) denotes
the result of certain operations performed on the unknown function / .
Suppose that the solution is expressible in the form of an infinite series*
/-/»+/w+/»+„.. (7.1,1)
Suppose also that when £ operates on this series, the result can be expressed
as a series in which the rth term involves only the first r terms of the series
in (7.1,1); that is,

(7.1.2)
The functions /"*, which as yet are subject only to the condition that their
sum is a solution of £(/) — o, are now assumed to satisfy the separate
equation, W " ) —. (7-.3)
W V ^ - o , (7.1,4)
W W * ) - © , (7.1,5)

which together ensure that £(/) = 0. The function f® is determined from

(7.1,3), and /"V/ 18 ,... are then found successively from (7.1,4,5,...);
each equation contains only one unknown when all the preceding equations
have been solved.
The division of £(/) into the parts £^£^,..., if possible at all, is not
unique, because any term £(r) can be divided into any number of parts, any
of which can be transferred to any subsequent term without altering the
general form of the expression. The division must be made in such a way
* All the series introduced are supposed to converge uniformly.
( UO]
7.11] THE NON-UNIFORM STATE FOR A SIMPLE OAS 111
that the equations (7.1, 3,4,...) are all soluble, and regard must also be had
to the convenience of solution of these equations.
Arbitrary elements may enter into the solution of the various equations:
they must be so chosen or grouped that the number of arbitrary elements in
the final result does not exceed that appropriate to £(/) = 0. The expressions
f®)j<s»+fO.\f<m+p-)+fii\... wiH be successive approximations t o / .

7.11. The subdivision of £(/); the first approximation/ 10 '

In the case of Boltzmann's equation for a simple gas,

&J)-4ffd + 9fi <7-".»)

where J(FGt) a jjiFG,, - F'G'^go^ d*'dcx (7.11,2)
so that 7(//,) = - BJIBt, by (3.52,11); also, as in (3.1, 3),

•t-t+'l+'-t "•"•"
Substituting from (7.1, 1) into (7.11, 1), we have

Enskog wrote
j«sj«(f,f,..,f)
= Af(%r))+Afw/r ^)+...+^^), (7.11,4)
corresponding to the mode of grouping of the terms in the expression for the
product of two infinite series S.rr and 2/y„ as a third series
Z(x0yr + *,.yr_i + . . . + *r.yo).
Enskog also divided 2^/< r) into a series of parts 3^, but not by the obvious
method of writing S*r> = &>flr). His method of division will be explained later
(7.14); it suffices here to say that he took

S"-o, (7.11,5)
r) (0> (, 1)
and for r > 0 he made & depend only on/ ,...,/ - . Thus, writing
gw = yw+^<» (7.11,6)
(0) m
we have £ = J = o, (7.11,7)
£« = / 0 1 + ^ 0 = 0 (r>o). (7.11,8)
Now(7.n,7),or J{f>)fim) = o,
is identical in form with the equation (4.1,7), which determines the velocity-
distribution function/in the uniform steady state. The general solution is
112 THE NON-UNIFORM STATE FOR A SIMPLE OAS [7.12
therefore of the form found in 4.1; that is, In/"" is a linear combination of
the summational invariants ^ ° of (3.2, 2), say
In/"" - a®+or*. me+a!*>. \m<?t
where ofl\ a™ and a** are arbitrary quantities, which are independent of c,
but may depend on r and t. By a simple transformation as in 4.1, we get

/ro = w
(j5r) f '" mC " , * r ' (7 >9)
"
where C - c - c 0 ; », c^ and 7 here denote arbitrary quantities related to
aP, or3*, and ofi> by simple relations. So far as (7.11, 7) is concerned they are
not necessarily identical with the number-density, mass-velocity and tem-
perature of the gas. Their values are, however, entirely at our disposal, and
it is convenient to identify them with these quantities. This amounts to the
choice, as a valid first approximation to/at r, t, of the Maxwellian function
corresponding to the number-density, mass-velocity and temperature at r, t.
Later, in 7.15, it is shown that this choice is the only one that leads to a
properly ordered form for the whole solution/.
In consequence of our identifications of n, c0, and T,
fff»dc = n - jfdc,
and similarly
ff>»mCdc = jfmCdc, lf™\mC*dc m fflmC*dc.
These relations may be written
/(/-/W<fc = °.
where now j ^ 0 = 1, mC, JwC*. By (7.1, 1), this is identical with

2/o-y)</c = o. (7.11,10)
/,
It follows from (7.11, 9) that
WP-FJF- ( 7 » . >»)
7.12. The complete formal solution
Equation (7.11, 8) may be written in the form

there being r terms on the right (when r = 1 the single term on the right is
-9®). The right-hand side involves only/<0»,/(1>, ...,/"-1,( and these are
known from the solutions of the previous equations {*» = o, ...,ii,~1) = o;
the unknown function/"'' appears only on the left, and occurs there linearly.
Hence if .Pr> is any one solution, any other solution will differ from F<r) by a
7.13] THE NON-UNIFORM STATE FOR A SIMPLE OAS 113
r)
quantity xf which is a solution of

F<r) and p^*-' correspond to the particular integral and complementary func-
tion of a linear differential equation. The most general solution of (7.12,1)
is the sum of F<r> and the most general solution of (7.12, 2).
To obtain the general solution of (7.12, 2), write tfr) = d^f®, so that 0*r)
becomes the function to be determined. Now, if 6ir) is any function of c, by
(7.11,2) and (4.4, 3),
• W W ) + J(fw¥r)fi0)) = / / / " y W " + W - 0(r)' - #"')*«! rfeVc,
= »'W»), ( 7 . " . 3)
where the suffix has been omitted from /(0), since the gas contains molecules
of one kind only. Thus (7.12, 2) takes the form /(0<r)) = o, of which the
solution was shown in 4.41 to be

where a"'", a" ,r) , a' 3,r) are arbitrary functions of r, t. Hence
^r) = (aa.r) + a a r ) - m C + a G ! . r ) . JwC*)/"".

Moreover, as any solution/'' 1 ' of (7.12, 1) is of the form F"' 1 +^ r ) , by a suitable

choice of a(1'r), c^2,r), <x(S,r) the quantities
ff<)f»dc (1 = 1,2,3)
can be made to take any arbitrary values. It is convenient to choose <x(,,r),
a**-", a (3,r) such that, for all values of r greater than zero,
J/w^0rfc = o (»=1,2,3). (7-12.4)
This choice ensures the satisfaction of (7.11, 10), which is the only restriction
as yet placed on the functions/'", beyond the fact that they satisfy (7.12,1):«
also implies that at every stage of the approximation t o / t h e new constants in-
troduced depend on no parameters other than n, c0, and T, and their space-
and time-derivatives of all orders, at r, t. This choice of a a , r ) , a*", and a"' r)
may seem unduly restrictive, since it substitutes, for the one condition
(7.11, 10), the infinity of relations (7.12, 4); but actually the same value is
found for the sum E/ (r) , whether this is made to satisfy the single relation
(7.11,1 o), or its parts are made to satisfy the set of relations(7.12,4) (see 7.2).

7.13. The conditions of solubility

Equation (7.11, 8) is as yet indefinite because iS/has not so far been divided
into its component parts S*r). The division cannot be performed arbitrarily,
for, according to the theory of integral equations, the equation is soluble
only if 3^ satisfies certain conditions. For, if the equation is satisfied,
/(y"->+^ r >)^"Vc = o,
 114 THE NON-UNIFORM STATE FOR A SIMPLE GAS (7.14

where ^ ° denotes any one of the summational invariants of (3.2, 2). Now
yw is divisible into pairs of terms such as Jif^)+Jiffi*), together with
a term J(Jip)f^'>) if r is even. When these are multiplied by i/i^dc and inte-
grated over the whole range of c, integrals similar to those appearing in
(3.54,4, 5) are obtained; these vanish since
^ + ^1-^-^1-0-
Hence a necessary condition for (7.11, 8) to be soluble is that
\3^f>de - o. (7.13,1)
This is, moreover, a sufficient condition; for if we write
/*»-/"•», (7.I3.2)
then, using (7.12, 3), equation (7.11, 8) becomes
n*i(W) = - ^ - y c r y r 1 ' ) - . . . -/(/^yi 1 '). (7.13,3)
It can be shown* that /(4^), which is a function of c, can be expressed
in the form ,
KJc)W«(c)+JK(e, c j •TX'XcO dcv
where K(c,c^} is a symmetric function of e, c,. Consequently (7.13, 3) is a
linear orthogonal non-homogeneous integral equation of the second kind;
the associated linear orthogonal homogeneous integral equation of the second
kind is n 1 / ^ " ) «»o, whose independent solutions arefl*'' «• \^° («'= 1,2,3),
as was shown in 4.41. From the theory of integral equations! '* follows that
(7.13, 3) possesses a solution if, and only if, the associated 'conditions of
orthogonality* are satisfied, namely
jy<>{ _ &* - yr/tiyj- »>) - . . . - J(f-»fi»)} dc = o.
On omitting vanishing terms from this equation, it reduces to (7.13,1).
Hence this is also a sufficient condition for the solubility of (7.11, 8).

7.14. The subdivision of 3>/

The equations (3.21, 3, 5), giving the time-derivatives of c0 and T,
involve mean-value functions, p and q (cf. 2.31 and 2.45), which can be
evaluated only when/is known. This causes difficulty in the determination
of/by successive approximation, because the time-derivatives of n, c0 and
* This was fint proved by D. Hilbert (Inttgralgltichungm, p. »67; Math. Atmaltn, 7a,
367 (191a)) for rigid elastic spherical molecule*; the extension to general spherical
molecules was made by Enskog, whose proof was included as 7.6 in earlier editions of
this book.
t See, e.g., R. Courant and D. Hilbert, Methoden dtr Math. Phyiih, vol. 1 (and ed.),
99. « 9 .
7.14] THE NON-UNIFORM STATE FOR A SIMPLE GAS 115
<0)
T determine that of/ , which affects the equation from which/°> is deter-
mined. At this stage, however, a s / is only incompletely known, p and q also
cannot be completely evaluated. This difficulty was overcome by Enskog by
an appropriate method of division of the part 3j of 1(/).
Let 4> denote any function of c, r, t; then

-HiZfjdc
o

-£?», (7.,4,,)

where ?<" = i [fflT)dc. (7.14, 2)

In particular, taking in turn <f> = mCC, <f> = EC, we have

p = 2p"», *-£«". (7-14.3)

0 0

where p"> = fmCCf>dc, q*> = JEC/^dc; (7., 4, 4)

1
in the present chapter, of course, E = \1t1C and (cf. 2.44) N = 3.
Owing to the form of/<0), _m. . ..
6
-7 g«» = o; (7-I4.S)
also the non-diagonal elements of p<0) vanish, and each diagonal element is
equal to knT. Consequently
p<°> = knTU = Up, (7.14, 6)
where U is the unit tensor (1.3), and p is the hydrostatic pressure.
The expressions (3.21, ,, 3, 5) giving 8n/dt, dcjdt and dTjdt may now be
written ^ .
8t - - 3 ^ -
fc0 I 8\_ . „ 1 8
8t
8T 8T

These time-derivatives are divided into parts, as follows,

K7
8t~dt' 8t 8t ' dt 8t • *' ''
116 THE NON-UNIFORM STATE FOR A SIMPLE OAS [7.14
where the quantities on the right are not themselves time-derivatives, but
are defined by

"¥ " ~^ne<^' (7-14. 8 )

-;T-o (r>o), (7-14.9)

--(c>ih+F-iZ- (7-14. io)

%-ll-r fr>* (7-14. " )

dBT dT 2 ( „, d 8 _)

dT 2T d , ,
= -Co^---g^Co. (714.12)

¥—3K(^!^+W (r>0)
- (7 H ,3)
' '
It is convenient also to write
D0 d0 d ,
DtBdt + e'^' (714.14)
so that DJDt denotes afirstapproximation to D/Dt, and
D0n d D0c0 _ 1 dp , .
l»--tt}>rc« -DT-F-P%> <7-i4.is)

The notation dr/dt can also be applied to functions involving n, c0 and T

and their space-derivatives, dr/St being assumed to obey the same laws of
differentiation as d/dt. Thus, for example, from (7.14, 15, 16) it follows that

g ( » / r i ) = o, (7.14.17)

so that, to thisfirstapproximation, the variation of temperature during the

motion of the gas follows the adiabatic law
pec Ti.
7.14] THE NON-UNIFORM STATE FOR A SIMPLE OAS 117
Again, when dT/dt is followed by a space-derivative of n, c 0 , or T, it is
understood that . - - o
=
didr o>&"
Also, if F is a function which depends on t only through parameters X,
(these being n, c0, T and their space-derivatives of any order),
BrF 8F 8r\, /,,.,«
I T -?M.~dT' (7-H.I8)
We now assume that 63011/^ in (7.11, 1) depends on t only through n, c 0 ,
7" and their space-derivatives. Then

Enskog's division of ^ / c a n now be indicated. He tooled 0 ' = o, and, for

r > o,
Cfrr) _ <2°->(/«»,/<l>, ...,/f-l))

= - 8 7 - + - 3 ^ + - + - 3 ^ - +0 - ¾ ^ 4 -
^-¾ - 1
(7-14.19)
(0)
each term in which is determinate when/ , ...,/^-^ are known. This pro-
cedure overcomes the difficulty mentioned at the beginning of this section.
It can be shown that with this mode of division the conditions of solubility
f&ry<0dc = o (1=1,2,3) (7-»4.20)
m
are satisfied. This is so because S = o, while for r > 0 the conditions
(7.14, 20) are equivalent to (7.14, 8-13), the equations for the subdivision
of 8n/8t, 8cJ8t, dT/Bt.
To see the truth of the last assertion, we note that the values of i^Sfdc
given in(3.21, 1, 3, 4) can be analysed at sight to give the parts corresponding
to the separate expressions 3^. In these parts, p, q are replaced by their
parts p <r_,) , qir~lh, also, whenever the time-derivative 8$/ St of the mean value
of a function <f> occurs, it is replaced by

8t 8t 8t •
corresponding to the subdivision of 8f/8t among the expressions 3^. But in
the equations of 3.21 the only time-derivatives present are those of mean
values of the functions t^'\ which, by (7.12, 4), are such that
r )
^ = o (r>o).
118 THE NON-UNIFORM STATE FOR A SIMPLE OAS (7.15
Hence the only time-derivatives appearing in JtJWjjrfOrfc are d^n/Bt,
d^xcjdt, dr_1 T/dt. The equivalence of (7.14, 20) with (7.14, 8-15) follows
directly. The rules for the subdivision of Bn/8t, dcjdt, dT/8t could, in fact,
have been derived from the integrability condition (7.14, 20).
Since the conditions of solubility are satisfied,/can be determined to any
desired degree of approximation. The time derivatives of n, c0, and T are
then known to the same degree: they do not appear as arbitrary adjustable
parameters, but are uniquely determinable in terms of n, c„ and T and their
space-derivatives at r, t.

7.15. The parametric expression of Enskog's method of solution

The preceding description of Enskog's method will now be summarized.
Boltzmann's equation is expressed in the form £(/) -= o, and a solution
/ - £/"•> is assumed; ft/) is expressed in the form £#", where £"» = Jm+&'\
and J"*, 3P> are given by (7.11,4) and (7.14,19). Thefirstterm (/,0>) of/ is the
same function of the number-density n, the mass-velocity c„, and the
temperature T at the point in question as would represent the velocity-
distribution function in a uniform gas in the steady state in which every-
where the number-density, mass-velocity and temperature are n, CQ, and T.
The later terms, /^, are the solutions of the equations f" — 0 such that
ff«W>dc - 0 (1-1,2,3);
the solutions thus defined are unique.
The division of & into Z£(r) was expressed concisely by Enskog in the
following way. He introduced a parameter 6 into the series for/, writing

/ - i/«»+/»>+0/«>+0*/w+... (7.15, 1)

(if we put 6 — 1 this agrees with (7.1, 1)). Then

where J(r) is the same as in (7.11, 4). Similarly the equations of 7.14 may be
modified as follows:

J? - I]* ( ^ + ^ + ^ + -) de

or ^=20^, (7-15.2)
7.2] THE NON-UNIFORM STATE FOR A SIMPLE GAS 119

*MH;+e*+FK)(j/'"+/™+-)

where i^ r) is defined as in 7.14. It follows that

£(/) = ^ + ^ + ^ ) + -

where fr) = J*"+££<", as before. I f ? " = 0 for every r, then £(/) vanishes for
all values of 6. The preceding description of Enskog's method corresponds
formally to the case 0 = 1.
On evaluating/(0),/(,>)..., with the above choice of arbitrary parameters,
it is found that/10' is proportional to n (or p),fw is independent of H,/*8' is
proportional to i/n, and so on. Hence the power of ljd appearing in any
term of the equations is the same as that of n appearing in the same term.
If a different choice of arbitrary parameters were made, the solution would
not be thus simply ordered according to powers of the density, though the
value of the complete solution would be unaltered (7.2).
When the density of a gas is comparable with that of the atmosphere near
the ground, and the non-uniformities in n, c0, T are such as ordinarily occur
in laboratory experiments, the 16™™/'''' in/decrease rapidly as r increases,*
and/ (W +/ ll) is a sufficiently good approximation to/for most purposes. In
rarer gases, naturally, the later terms are relatively more important, so that
for some purposes it is worth while to determine / (,) . Unfortunately the
complexity of the successive terms f-r) increases rapidly with r.
Hilbert expressed/by an equation of the form(7.i5, 1), but his discussion
of Boltzmann's integral equation did not afford a convenient determination
of/beyond the first term/' 0 ', because he did not introduce (7.15, 2) before
proceeding to the division of @f.
7.2. The arbitrary parameters in /
The solution / of Boltzmann's equation which is obtained by Enskog's
method depends on no parameters other than the values of n, c0) and T
throughout the gas. The conclusion that the physical state of the gas depends
• The ratio off1" t o / " " " is in general comparable with //£ for rigid spherical molecules,
where / is the mean free path, and L a scale-length characteristic of the spatial variations
in density, mass-velocity and temperature; for more general molecules the ratio is
comparable with C'r/Z., where r is the relaxation time (cf. 6.61).
120 THE NON-UNIFORM STATE FOR A SIMPLE OAS (7.2

on these parameters, and on no others, appears to be in accordance with

experiment. However, Enskog's solution of Boltzmann's equation is not the
most general one: it is possible to assign an arbitrary value t o / a t an initial
instant, and Boltzmann's equation merely determines the way in which /
subsequently varies.
It might be supposed that Enskog's solution lacks the arbitrariness of the
general solution because of the special values which have been adopted for
n, c0, and T in the expression (7.11,9) for/"", and the special values adopted
for the arbitrary constants a (,,r) , a*" and a a , ) , appearing in the expression
for/1" in 7.12. This, however, is not so: it can be proved (though the proof is
too long to be given here) that, subject to certain conditions of convergence,
every solution given by Enskog's method of subdivision, corresponding to
a given distribution of n, c„ and T, is identical with the one obtained with
these special values of the arbitrary elements. The lack of arbitrariness is
due to the fact that df^jdt does not contribute to the equation determining
/"»: if it did,/ 1 " could be arbitrary, and the equation would merely determine

To see the relation of Enskog's solution to the general one, suppose that
the initial velocity-distribution function/corresponds to values of n, c& and
T, whose scale-lengths of spatial variation are large compared with a free
path, but that/is otherwise arbitrary. Then in general/at each point in the
gas begins by varying rapidly on account of molecular encounters, with a
time-scale of variation comparable with the mean collision-interval. How-
ever, by analogy with the results for a gas in a uniform state,/can be expected
quickly to approach a limiting form in which molecular encounters no
longer produce such rapid time-variations. The limiting value of/ is called
a ' normal solution' of Boltzmann's equation, and the corresponding state
of the gas is a 'normal state*. The time of relaxation in which the gas
effectively attains a normal state is of the same order as the collision-interval
(cf.6.61).
If the gas is uniform, the normal solution is the Maxwellian distribution
function. In a slightly non-uniform gas the normal solution is a function of
r and (, though at any time it approximates locally to the Maxwellian form.
The normal state is characterized by a balance between the effects of non-
uniformity, which tends to produce departures from a locally Maxwellian
state, and those of molecular encounters, which tend to destroy such
departures.
Normal solutions of Boltzmann's equation are distinguished from non-
normal solutions by having no' irregularities' which can be rapidly smoothed
out by molecular encounters. In consequence, they depend on fewer para-
meters than the non-normal solutions. Since the mass, momentum and
energy per unit volume are unaffected by encounters, the values of n, c„
and T throughout the gas are among the parameters on which a normal
solution must depend; conversely, since there are only three summational
7.3] THE NON-UNIFORM STATE FOR A SIMPLE OAS 121

invariants, we can expect that n, c 0 and T (determined by these invariants)

are the only local properties on which a normal solution can depend.
Actually, since the state at the point P is appreciably affected only by the
gas immediately adjacent (i.e. within a distance of a few mean free paths)
a normal solution at P can be taken to depend on the values at P of n, c 0 , T
and their space-derivatives; it may also depend on the force mF acting on the
molecules. Enskog's method of approximation provides a solution depending
on just these parameters, and on no others. Since a normal solution depends
on a minimum number of parameters, Enskog's solution is such a solution,
and so the sequence 7.1, if it converges, defines a normal state of the gas.*
It must be stressed that Enskog's method is justified only because the
collision-interval is normally small compared with the time-scale of variation
of the macroscopic properties. This justifies our regarding df^/St as smaller
in order of magnitude than the contribution of/"' to 0e/<r>/0f (cf. (7.11, 4)
and (7.14,19)). It also has the consequence that deviations from uniformity
affect / o n l y slightly; the first approximation f" depends on «, c0, and T,
but neglects their space-derivatives; the second approximation /w>+/w
involves their first derivatives linearly; the third,/ < 0 , +/ < 1 , +/ < , , , includes
terms involving second derivatives linearly, or quadratic in first derivatives;
and so on. When conditions are such that the time-scale of variation of any
macroscopic variables is not large compared with the collision-interval, the
whole method of approximation breaks down.

7.3. The second approximation to /

The remainder of this chapter is devoted to the detailed evaluation of the
second approximation to / (involving the determination of Wl>); this leads
to the determination of qf1' and p(I>, and hence to expressions for the
coefficients of thermal conductivity (A) and viscosity {ji).
The equation from which J™ orfm^1) is to be determined is
£ » • <2*»+/0) = c, (7.3,1)
(1 0
The differential part &* of | > depends only on/™. As/* ' is a function of C,
it is convenient to express £&"' in a form analogous to that of 2f given in
(3.12, 2), namely

a
Dt + ° 3 r + r Dt)dC 8C°8r0'
or, on substitution from (7.14, 15),

(7-3. 2)
* A very full discussion of normal and non-normal solutions of Boltzmann's equation
has been given by H. Grid, Handbuch der Physik, vol. la, 241-51, 266-93 (Springer,
1958); Phyt. Fluidt, 6, 147 (1963).
122 THE NON-UNIFORM STATE FOR A SIMPLE OAS 17.3
Now, by (7.11,9)
ln/ro = const. + In («/r») - m&lzkT,

•ndso £ = kT,
while, by (7.14,16,17),
D.ln/ro mC* D0T mC* d
Dt ™ 2kT* Dt " 3*rar C °*
Thus the sum of the first and last terms in the bracket on the right-hand
side of (7.3, 2) is m » e

0
where CC is given by (1.32, 2). Also the sum of the two middle terms is
/31nT m dp\_r 8ln(f»>lnkT)
*"\ 8r pkTdr) ~ ° dr
_/ainr-i mC* dT\

(mC* 5\^dlnT

Accordingly (7.3, 2) may be written in the form

-/•((^-DC.^+aflftf^e.}, (7.3.3)
where V is a dimensionless variable defined by the equation

• .(£)'« (7.3.4)
with components #, f^ if" and magnitude V. In terms of this variable

^ " - f e ) ' ' " ' >"*=y*'™- (7.3.5)

m
The integral part J of g"' is given by the equation
/»=-wil>)+JU°W).
Substitute/™ - / ^ 4 ^ : then, using (7.12, 3), we get
jm = „*/(<i>u>). (7.3( 6)
7.31] THE NON-UNIFORM STATE FOR A SIMPLE OAS 123
n
The equation £ > = o is therefore equivalent to
„»/(«jXi>) = _/«»{(«*_!)C.Vln T+2VV: Vc0}, (7.3,7)
introducing the notation V in place of 0/3r (cf. 1.2).

7.31. The function *<"

Since <t>(1), like /"' itself, is a scalar, it suffices to consider only the scalar
solutions of (7.3, 7). Now /(O*") is linear in <l>(,>, and the right-hand side of
this equation is linear in the space-derivatives of T and of u0, v0, tv0. Hence
the most general scalar solution 4>(I> is the sum of three parts: (i) a linear
combination of the components of VT: for this to be a scalar, it must be
given by the scalar product of VTand another vector; (ii) a linear combina-
tion of the components of Vc 0 : this must similarly be the scalar product of
Vc, and another tensor; (iii) the most general scalar solution of the equation
/(4)0)) = 0. Of these (i) and (ii) correspond to the particular integral of a
differential equation, and (iii) to the complementary function. Thus we
can write
¢0) = _ I l?tl\k A.Tj\nT--b: Vc 0 + a(l-» + a'*.».»iC+a<»-w.imC1,
n\ m } n
(7-3». 0
where A and at*" are vectors, and B is a tensor: A, B are functions of C,
while a (,,,) , a*1', a 0,1 ' are constants.
Substituting from (7.31, 1) into (7.3, 7), and equating the coefficients of
the different components oiVT, Vc0, we find that A, B are special solutions
of the equations n / ( i 4 ) = ^0)(¾¾ _ j ) v> (?.3, y 2)

«/(B) -/Wflfr. (7.31. 3)

It is easily verified that the conditions of solubility of these integral equations
are satisfied, i.e. (cf. 7.13) that
jf»^ _ | ) fgpfidc = 0> ffWjfopiiJc = O («=I,2, 3).
The only variables involved in (7.31, 2, 3) are C, or 9g, and the values of
n, T at the point in question; c0 does not appear explicitly, but only as
involved in <€. Hence A and B must be functions of », T and V. The only
vector which can be formed from these elements is "8? itself, multiplied by
some function of n, T and V, the latter being the one independent scalar
connected with S?. Hence we can write
A-A(ir)9. (7-3L4)
A{$) being a function of 'jf (and of n, T).
Equation (7.31, 3) can be separated into nine component equations;
typical among these equations are
124 THE NON-UNIFORM STATE FOR A SIMPLE OAS [7.31
Addition of the three equations of the first type shows that
/(B^ + B^ + BJ-o;
also, from those of the second type,
/(B^-B^-o.
Thus the nine equations possess solutions such that
Bxx + Bn + Bu - o, flx„ - B¥X, etc.,
i.e. such that B is a symmetrical and non-divergent tensor. Now B depends
only on n, T and V, and the only symmetrical non-divergent tensors that
e
can be formed from these elements are multiples of W by factors that are
functions of n, T and V. Hence for the solution of (7.31, 3) we can write

B-«%fl(<f), (7.31.5)
where B(¥) is a function of "if, n and T.
The constants a0-", a*1', a*M in (7.31,1) are to be chosen such that the
corresponding form for/11' satisfies equations (7.12,4), i.e. that

_ J>yw> j - I p £ Z y ^ W V . vin 7-^^^: VC,

Neglecting vanishing integrals, and simplifying by (1.42,4), these equations,

for 1 — 1,2,3, become
jf^cfi-»+a* ». imC*) dc - o,

j/^(-jtA{V)VlnT+mefl-AmCtdco,

//*»(a(»-w + a*w. J«C») \mC*de - 0.

Thefirstand third of these relations show that a0-" «• o,a^-u « o; the second
shows that a* u is proportional to VT, so that in (7.31, 1) the term cr*u.mC
may be absorbed into thefirstterm on the right: hence we can also write
er*w •» 0, when the second of these relations becomes
ff™A(V)V*dc - 0. (7.31, 6)
Also (7.31, 1) reduces to

<tm I
'- n(^)iAVlnT-2nB:Vc^ (73,>7)
7.4] THE NON-UNIFORM STATE FOR A SIMPLE OAS 125

7.4. Thermal conductivity

Before considering the details of the method by which A, B are determined,
we shall derive general expressions for the second approximation to the
thermalfluxq (this is g*11, since g*0' = o; cf. (7.14,5)) and to the deviation of
the pressure system from the hydrostatic pressure given by p*0* or/>U; this
deviation is, to the approximation considered, given by p<1). The equation
giving g«> is ^ , _ jmjya>cic</c - imff<eW»C*C(lc.
On substituting for O*1' from (7.31, 7), and neglecting integrals of odd
functions of the components of %? or C, this becomes

*"' " - ^ / / ^ W ^ ) W ) « ' - V , n T i c

"~s(i?y ffBygvtAcirfvvinT'i<:
= - - =mn^ - Vln T IfPWAmdc
3 J
by (1.42, 4). Using (7.31, 6), we get

= -^£vTJvA<P)-f*(V-V9dc,
whence, by (7.31, 2) and (4.4, 7),

^v--^^JA.I(A)de

= - — [A,A\VT.
3«
zk*T
Thus, if A - — [A, A], (7.4,1)

so that, by (4.41,1), A is essentially positive, the value of g^ is given by

g<» = -AVr. (7.4,2)
The thermal flow is therefore, to the present order of approximation,
opposite in direction and proportional in magnitude to the temperature
gradient; this agrees with the approximate theory of Chapter 6. The
quantity A is the coefficient of thermal conduction of the gas.
126 THE NON-UNIFORM STATE FOR A SIMPLE OAS [7.41

7.41. Viscosity
The equation for p™ is
p<» - mfp»CCdc = mffWtPKlCdc.
In this we substitute for W from (7.31, 7): then, on omitting integrals of
odd functions of the components of C,

pa> =, _ 25 jfw>B(,<g)(<ifc: Vc0) CCdc

4kT{
r
- [f»B{<g)(<ifo: Vc0)Wdc
n
whence, using the theorem of 1.421 and (7.31, 3), and writing e for Vc0
(cfi-33). tkT^f 0 0
po> = -*2± gl/«»B(«Jf)(««: <&ig)dc

5n J
=.-t*rl/B:/(B)<£c
--tA7-[B,B]e\
Thus, if /t-|*7[B,B], (7.41,1)
so that (i, like A, is an essentially positive quantity, the above expression for
p<» becomes _o_
p™ = - 2/tt s - 2/iVco. (7.41, 2)
By (7.14, 3, 6), the second approximation to the complete pressure tensor
pis/>U + p(,), or 0
pV-2/iVco,
by (7.41, 2). Thus the values of the second approximations to the typical
elements ptxl p„ of the pressure tensor p are

,l
\dz+ dyj-
These are identical with the expressions for the corresponding stress-com-
ponents in a medium of viscosity /1. They are generalizations of the result
of 6.2, which corresponds to the case when t>0 = tc„ = o, and tig is a function
of z only; the above approximations to the six components of stress then
reduce to
Pxx - Pn " P- ™ P> Px»—Pvx — -l*-fa, Py, = Pt» " ° " Pa - Pxf
7.5] THE NON-UNIFORM STATE FOR A SIMPLE GAS 127

7.5. Sonlne polynomials

The quantities A and B are determined by expressing them in terms of
certain polynomials S^*),* defined as follows.
Let s be a positive number less than unity, and x, m be any real numbers;
also let
5 = */(> -s). (7.5, 1)
Then the polynomial S%Xx) is defined by the expansion

( 5 / 0 - + ^ - ^ = (1-*) _ m - 1 «- j r r t , - ) = 2 snS%\x). (7.5,2)

n-0
Now (I _j)-m-l e -x«U-»> = 2 ( - * 0 " ( l -S)-»-m-llp\
P
= S 2 ( - xs)"f(m +p + q\lp! q\,
p 1
where rq denotes the product of the q factors r, r— 1, ...,r — q+i; selection
of the terms such that p + q = n gives

fiiT'W - £ ( - * ) * ( « + »)„-plP !(»-/>) !• (7-5. 3)

p-D
In particular S^(x) = 1, 5J»» = m+ 1 - * . (7.5, 4)
The polynomials 5^°(x) are numerical multiples of Sonine's polynomials,!
which arise in the study of Bessel functions.
Since, writing r = r/( 1 -1),
(1 - o ^ - ' c - 0 : m _ 1 1 "e- a "+ s+r) « m «&

= (1 - 0 - m - , ( i - 0 - m - 1 f°'V*l-«<»<l-»><1-*>*m</.x
- ( i - i O - ^ ' H r n + i),
on equating coefficients of spf> on the two sides of this equation, wefindthat
' r - Stf>(*) 5g>(*) *"• <*r = 0 (/»*?)
/: (7-5. 5)
= T(m+p+i)//>! (p=?).J
* These polynomials were first used in the kinetic theory of gases by D. Burnett, Free.
Land. Math. Soc. 39, 385 (1935). The polynomial called 5iT'(x) by Burnett differs by *
constant factor i/r(m + n + i ) from that defined above. An alternative discussion in
terms of generalized Hermite polynomials was given by H. Grad (Comm. Pure Appl.
Maths, 2, 331 (1949)). These polynomials are associated with an expansion in Cartesian
coordinates in the velocity space, whereas the Sonine polynomials are associated with in
expansion in polar coordinates.
t N. J. Sonine, Math. Ann. 16, 41 (1880). The typical Sonine polynomial is equal to
r(m + «+i)(-i)"S!:'(x).
It is related to the generalized Laguerre polynomials L used by E. Schrftdinger {Ann.
Phys. 80, 483 (1926)), which are such that
128 THE NON-UNIFORM STATE FOR A SIMPLE OAS [7.S1

7.S1. The formal evaluation of A and A

Suppose that the function A(1f) can be expanded in a convergent series of
the form# „,
^W-z«„sfW. (7.51,1)
p-0
where the coefficients ap are independent of V. The relation (7.31,6) is now
(cf-(7-3.5)). . r

nrri 2 flpl e-ixmSf\V)dV - o,

or, on integrating over all directions of V, and noting that SyC***)«- 1,

0 - znirt £ a, j * e-^VS^fWSYXV^dV*

- znjr-Jrd)*,,
b
y (7-5. 5)- Hence a0 - o, and
AmA(V)V- S-p**'. (7-51. *)
p-i
where <r*> - SfX^V. (7.51,3)
The values of a,,a*... are to be determined from (7.31, 2), which is
equivalent to nJ{A) _ . ^ i w v . ( w 1§ 4 )

We multiply this by «t*> - Sjf>(«*)«" and integrate with respect to c. Then,

using (4.4, 7) and (7.3, 5), we obtain
[<&,A)-ar (7.51,5)
where « , - - 2 r l j " e^V^IF) Sff^ d{<g*)

by (7.5, 5). Let ^ , - [<•»,*#]. (7.51, 7)

Then, in virtue of (7.51, 2), equation (7.51, 5) becomes

2«p«p,«a, («-«1« »)• (7-5». 8)

p-i
As the functions a*' are known, a^ may be regarded as known. Hence
(7.51, 8) is an infinite set of equations to determine the infinite set of
coefficients ap. We need not go into the theory of infinite sets of linear
equations: it is sufficient to state that when they have a unique solution (and
* Problem! of convergence were considered for certain molecular model* by D. Burnett.
IK. at. For the general theory of the expresubility of a function by a seriea of poly-
nomial*, tee R. Courant and D. Hilbert, Mtth. dtr Math. Phyx. vol. 1, chapter a.
7.51] THE NON-UNIFORM STATE FOR A SIMPLE GAS 129
physical considerations make it clear that this is so in the present problem)
the solution can be found by a method of successive approximation. Enskog
assumed that A(m\ given by
if«- i«<v) (7.51,9)
m
where £ «<,"%, = <x, (q= 1,2,...,«) (7.51, 10)
p-i
can be taken to be an mth approximation to A. This approximation amounts
to neglecting the terms in (7.31, 2) and (7.51, 8) depending on a">, where
r > m. It is assumed that, asm tends to infinity, aj,"1' tends to«p,andil<m)tOi4.
Using the values of a, given in (7.51, 6), we obtain
ao»> = _i^m V j / < m > ( g = I 2 m) (7.51,11)
as the solution of equations (7.51, io). # Here sf^" is the symmetric deter-
minant with elements apq(p,q = 1,2,...,m) and j / j j ' is the cofactor of
a,s in s^m).
The thermal conductivity A depends only on the coefficient a,; for, from
(7-4> ')> , U T
A = — [A, A]
J
3«

3W P-I
Sk*T
zm
using (7.51,5 and 6). Since the molecules of the gas are supposed to possess
only translatory energy, the specific heat ce of the gas is 3k/2m. Thus, using
(7.51, 11), we may write
A = *£cvkT lim jfftW*. (7.51, 12)
m-*oo
The limit on the right can be replaced by an infinite series. By Jacobi's
theorem on the minors of a determinant and its reciprocal (adjugate)
determinant,! j/^j/s-i)-(jrffcfli = ,«/fi-»WW,

Also s/ft/s/111 = i/a n ; hence, by repeated application of (7.51, 13),

Mr _ 1 , Kg? , , w)'
j/(»» «..
»11 \ Z » / » j/i»-»rf"i'
* In practice, when numerical valuta of the a„ are known, the moat convenient way of
finding a',"" ia by one of the standard numerical techniques (e.g. that of Gauas) for the
solution of a system of linear equations.
f See, e.g., L. Miraky, Introduction to Linear Algebra (Oxford, 1961), p. 25.
130 THE NON-UNIFORM STATE FOR A SIMPLE OAS [7.52
Substituting from this into (7.51,12), we get

By (7.51, 7), the quantities «p, do not depend on the number-density n;

hence the same is true of the series in (7.51,14), and so of A.
Every term of the series in (7.51,14) is essentially positive; for if
*„ x, xm are arbitrary non-zero parameters,

[p-1
*» m

p-1
"I

J
m m

p-1 «-1
The expression on the left is essentially positive; it cannot be zero, since
m
Yi x„ap is not a summational invariant. Thus the quadratic form on the
K-l
right is positive definite, which can be true only if the determinant J/""' of
the coefficients is positive. This holds for all m; since an = j / " ' , the series
in (7.51, 14) is one of positive terms.
When the expressions aM are known, the mth approximation to A is found
by taking m terms of the series in (7.51,14). The calculation of aM is
effected in Chapters 9 and 10.
7.52. The formal evaluation of B and /t
The tensor B and the coefficient of viscosity /1 are evaluated by methods
0
similar to those used in 7.51. Since B is of the form tUf?BCf), we express it
as a series of the form °°
B= S*Pb'P>, (7.52,1)
p-i
0
where b*> = «-tf.Sf-"(«") (7.52, 2)
and the coefficients bp are constants to be determined. Multiplying (7.31, 3)
by b(»\ and integrating with respect to c, we obtain
[b">,B] = /?, ( 7 . S 2,3)

or, by (7.52, 1). S *p*pj " fiqt (7-5*. 4)

p-l
where b„ = [bf>\ b»], (752. s)
( ,
fi. = Uf^h-.^\ e*) ivdc.
On using (1. 32. 9). (7-3. 5) and (7.5,4,5) this gives

A- . - ^ f'V*,«*Sj['-,>(,if*)d(«*)

•1 to-1).
•o (y>i). (752.6)
7.52J THE NON-UNIFORM STATE FOR A SIMPLE OAS 131
The viscosity ft depends only on 6,, for (cf. (7.41, 1))
/»-|A7TB.B]-1*7*2*,[*». B]
- kTbt
bv and 2
(7-52. 3) (7-5 . °)- Thus, solving (7.52, 4) by the same method of
approximation as that used in 7.51, we find
/1 = f k T lim «ff>/«w (7.52, 7)

~ l*u ^ " ^ + J ^ S « ) + - " J * t7<S2

' *'
tm)
where # is the determinant with elements bpq(p,q=i,2,...,m), and 0$
is the cofactor of 6 „ in ,#(m>. The expression for /*, like that for A, is inde-
pendent of the number-density; also every term in the series in (7.52, 8) is
positive, so that the successive approximations to fi form a monotonic
increasing sequence.
Series similar to (7.52, 8) and (7.51, 14), were given by Enskog.*

Historical note
Enskog's results appeared in his Uppsala Dissertation in 1917. f A little earlier,
Chapmanf had independently given formulae for A and ft, and for the
coefficient of diffusion in a gas mixture, essentially equivalent to those of
Enskog. Chapman, following Maxwell, considered equations of change of
molecular properties rather than attempting explicitly to solve the Boltzmann
equation. Equations (7.51, 8) and (7.52, 4) are readily identifiable as equa-
tions of change for the molecular properties «VP>, b (p) , and similarly for the
corresponding equations in Chapter 8. Essentially, Chapman used such
equations in approximate forms like (7.51, 10), whence the agreement of
his results with those of Enskog. §
* Inaugural Dissertation, Uppsala (1917). Enskog also showed that of ill functions Ain"
given by an equation of the form (7.5. 9). the one in which the coefficients a"" ire given
by (7.5. 10), is the one that makes the essentially positive quantity [A—A'?, A—A""]
a minimum. This implies that the coefficients ajT" are chosen such as to make A"*\ in
a certain sense, the best possible spproximation to A of the form (7.5, 9). A similar
argument appliea to B. The minimum principle can conversely be used to establish
(7.5,10) and the corresponding equations for 4J,"°: see E. J. Hellund and E. A. Uehling,
Phys. Rev. 56. 818 (1939). and M. Kohler, Z. Phyt. X24, 77a (1948).
t See also D. Enskog, Arkiv. Mai. Aslr. Fyt. 16, 1 (1921).
X S.Chapman,Phil. Trans. R. Soe. A,216,179(1916), and217, lis (1917);Proe.R.Soe.A,
93, 1 (1916).
i See also the Historical Summary (p. 407).
 8
THE N O N - U N I F O R M STATE FOR A
BINARY GAS-MIXTURE

8.1. Boltzmann's equation, and the equation of transfer, for a binary

mixture
The general problem of evaluating the velocity-distribution function in a
gas-mixture, when the composition, mass-velocity, and temperature of the
gas vary from point to point, is solved in a manner analogous to that em-
ployed for a simple gas. In this chapter we consider a mixture of two
constituents; the corresponding solution for a multiple gas-mixture is
given in Chapter 18.
The definitions of mass-velocity, partial pressures, densities, and so forth,
for a gas-mixture, are given in 2.5, p. 44.
The functions/,,/, that specify the distribution of velocities of molecules
of the two gases satisfy equations of the form
3
^>J-/. ^»4.17 #» - «/» (Ri ,\
+e +F (8l,l)
«F »-5F »-E;-li-'

the values of 0«/i/«fr, B^Jdt being given by (3.52, 1, 9, n ) and similar

equations. If/, is expressed as a function of C„ r and t instead of c„ r and t,
the left-hand side of (8.1,1) becomes
Dfi.r *fi.(F DcA dfx dft d
o>

there is a similar expression for the left-hand side of (8.1, 2).

The equation of change of a function ¢,( C„ r, () of the velocity of mole-
cules of the first gas is (cf. (3.13, 2))

W*+*H-*+x*tt
-ni\^r+Ci-8F+lF>-DrJ8c;-6c,Ci^Cor"»**•(8,,3)
a similar equation holds for a function ^«(Ct, r, () of the velocities of the
molecules of the second gas.
As in 3.21, AfS vanishes for certain values of ¢-, the corresponding
equations of change are of special importance.
I 132]
8.1] BINARY GAS-MIXTURE 133
Case i. Let 0, = i; thenfl,A<^tis the rate at which the number-density
of the first gas is being altered by encounters; this rate is, of course, zero.
Thus the equation becomes

£j»+"igf-*s + g ^ ( « i C i ) - * (8.1,4)

which is the equation of conservation of molecules of the first gas. On

adding this to the corresponding equation for the second gas, the equation
of conservation of molecules of the mixture is obtained; this is

^ + tifrC0 + ^.(nlCl + ntCt) = o. (8.1,5)

Again, multiply the_equations for the two gases by m, and mt and add:
then, since /»,C, +/>,C, = 0 (cf. (2.5, 9)), the resulting equation is
D 8
P /O ZA

which is the equation of conservation of mass for the mixture.

Case 2. Let <j>x = «1,(7,; then the equation of change is

jtac.)+ftC, (!.«.) + i. f t c,c,-ft (*,-§*)

+p,C, ^ c 0 - n^mfit. (8.1, 7)

On adding this to the corresponding equation for the second gas, and using
n^nifC^n^mfix = o,
expressing the fact that the total momentum of the molecules is unaltered
by encounters, the equation of motion of the mixture is obtained, namely

frP-PiFx+PtFt-Pjg. (8.1.8)

Case 3. Let ¢, = Ev Then, since the translational motion of a molecule

contributes a term {mlC\ to £„ 20,/3C, — mtCv The corresponding
equation of change is thus

«W,(A,)4.*-,,e..(*-£)
8 _
+PxClC1:^ct = »,££,. (8.1,9)
On adding this to the corresponding equation for the second gas, and using
the relation . ,ff . . . »
134 THE NON-UNIFORM STATE FOR A (8.2
expressing the conservation of energy at encounters, the equation of energy
of the whole gas is found to be
D(nE) P / a \ a 7,„ T,„ a
~Dt+ "E \dr • e°) +
to • 9 = PlCl
• F l +ptCt •Ft ~ P : &> C"'
which, like(3.2i,4), may beinterpreted term by term. Writing dE/dT= ±Nk,
as in (2.44,1), and using (8.1, 5), this equation reduces to the form

+PiC1.F1+ptCt.Ft-p:^c0. (8.1,10)

8.2. The method of solution

The equations (8.1,1,2) for/,,/, are solved, as in Chapter 7, by a method of
successive approximation. They may be written in the forms
^/1+-^(/^)+^.(/1/,)-0, 0 , / , + 7 , ( / , / ) + 7,,(/^,) = 0, (8.2,0
where # , / „ QJX denote the expressions on the left-hand sides of (8.1,1, 2)

JMI - JJ(/i/,-/;/;)*«i,<fe'«fc,. (8-2.3)

with similar definitions for the expressions 7,(/,/), 7,,(/,/,).
The expansions for/,,/, corresponding to (7.1,1) are
/ , - / ^ + / ^ + / 1 ^ + - , f,-JP+JP+JP+~: (8.2,4)
and the equations (8.2, 1) are similarly subdivided into the sets of equations
# i " + 7i'> = o, 0 i " + # > - o , (8.2,5)
e 0>
where 7i > « 7,(/1^9+7,,(/r/, ). (8.2,6)
JfT m 7,(/1^)+7,(/1^)+...+/.(/IT")
+7,^/1^) + -+-/1^/.^). (8.2,7)
l r l)
and &p is a function of/f>,/i \ -,A ' . but not of/, \ such that @[m = o,
(r

and „
2 sir=^1(/1^+/1^+/1^+-) - ®Jv (8.2,8)
r-1
Similar remarks apply to 3^ and ftr)-
The first approximations t o / , , / , are given by the equations
71« - JJJPf") + 'idC/P/V') = 0. A°> = 7^/i°>/«') + 7,,(/, 9 ^) - 0.
8.2] BINARY GAS-MIXTURE 135

These are identical in form with (4.3, 1); hence, as in 4.3, they possess the
solutions
/l<0> = l
" (^r) lexp {-5T [( "'-" o)2+(t ' 1 - t ' o) ' +(Wl - t0o)2] |'
(8.2,9)

A
*= "•(i5f) ,ex p{-^K«i-«o) t +(f«-^ , +(^-«g , l).
(8.2,10)
1
where »,, nt, c0, 7 are arbitrary functions of r, t. These are chosen, as in 7.11,
such that »,, n% are the number-densities of the two gases in the mixture,
and c„ and T are the mass-velocity and temperature of the mixture: conse-
quently for a given gas at a given time t they are known functions of r. This
choice implies that
I fide, = //!•»&„ \f%dc% = ffipdc*
fflmlCldcl +JftmtCtdet = JAm'»iCidel+jft,nmtCtdet,
JMdct+jfiBidc, = jfl^dc. + JA^dc,,

i.e. that f£yi"*i-o, fi/rVc.-o, (8.2,11)

J r-l J r-1

JJ;/f r , «iCiAi-+JJ^WiC,*:, = o, (8.2, 12)

f £ / P S i * . + f £ /T*, &, = 0. (8.2, 13)

J r-l J r-l
The equations from which f\r),JlT) (r > 0) are determined are of the form
(8.2,5). By an argument similar to that used in 7.12 it may be shown that, if
F\r\ F£} are any pair of solutions of these equations, any other pair will be of
the form F{r) + Xi< F*r)+Xt< where Xv X* a r e solutions of the equations

A(A0)x)+A(x.r')+AMFXM)+A,(x./30)) = o,
Wx)+JJIXMD+AiW 0, xi)+j n <xjn=°.
which, on writing Xi = /i o, 0i> Xi = /i o> 0a. reduce to

»! A(0i)+»1 »s A*(&+&) = °.
"I A(0l) + »1 »* ' « ( & + 0l) = 0,
in the notation of 4.4. We multiply these equations by 4>idcx, $sdct,
integrate over all values of c,, c2, and add. The resulting equation is
it, ¢) = °.
136 THE NON-UNIFORM STATE FOR A [8.21
the solution of which was found in 4.41 to be
A - ^ + tP».mlCl+c^El, ¢, - «i n + aP>.mtCt+ a>'>Et,
where ai u , a,n, a™, a® are arbitrary functions of r, /.
Any pair of solutions of (8.2,5) can be expressed in the forms F\r) +/[ 0 Vi.
*»"+/!"&• By a suitable choice of a[l), <#>, «»> and <*<», solutions/!",/&"
may be constructed such that
\f[*dcx-o, fflr)dct = o, (8.2,14)
J/l'^iCjrfc,+J/rXCrfc, = o, (8.2, i S )
r
JA %dCl+j^Etdet - 0. (8.2,16)
rt
We adopt these as the values of/}" and/J in the series (8.2,4) when r > 0.
Equations (8.2,: 1,12,13) are then satisfied.

8.21. The subdivision of Sf

The subdivision adopted for ^ , / , and 2JX is similar to that for ^ / i n 7.14.
We define pw and 9"» by
p" - \^^CxCxdc^\^mxCxCxdCt,
<P - \f[r'ElClde^l^ElCtdct.
In the same way
Q* - ijfi'^dc,, up - ~/yj"ct</c„
so that C, - 2 Q", Ct - £ <%\
f-0 r-0
r, ,,
and (8.2,15) becomes /^ 1 ¢} +/¾¾ = 0.
The time-derivatives in equations (8.1,4,8,10) are expanded in accord-
ance with the formal scheme

the subdivision being made as follows, for each of the variables n„ c„, and
T on which djdt operates in those equations; in each the operations djdl,
djdt,... are directly specified, while the operation djdt is specified in terms
of DJDt, signifying djdt + c0. d/dr:

^ * - | > , C h (r>o). (8.21,3)

8.21] BINARY GAS-MIXTURE 137

-PiFi+PtFt-jfc, (8.21,4)
p
^?s_s--p<r (r>0)> (8 2, s)
--
, LD„T , ,(d0T BT\ d

-&•**-?"••&*• ('><>)• ( 8 - 2 I -7)

In (8.2i, 6,7) it has been assumed that the molecules have only translational
energy, so that N = 3. Where dr/ct operates on any function of n„ nt, c0, and
rand their space-derivatives, it obeys the ordinary rules of differentiation,
as in 7.14.
With this subdivision of the time-derivatives, &ir) may be expressed in
the form
+ +
** * to " + to + Cl-
\ 8r+'|,acJ/l '
(8.21, 8)
analogous to (7.14, 19).
The equations ^ + J[^ = 0i gp + jp-o
are soluble if and only if certain relations analogous to (7.13, t) are satisfied
by 2f, QV- Using (3.54,2,4,5) and the definitions of AT\ J$\ it can be
shown that J ^ ^ . o . J^rfc. - o,

fAr)'"1CldCl + fW^tCtdct « o,
r)
JA^ifn^dc^ jA imtC\dct = 0.
Hence for the equations to be soluble it is necessary and sufficient that
f2[r)dCl = o, jSjpdc, = o,
j^m^d^ + j^m^Ctdc, = o,
V\miC\dcx + \3lV\mtC\dct = 0.
It may readily be verified that with the above choice of i£ir), Sjp these
conditions are, in fact, satisfied.
6 CUT
138 THE NON-UNIFORM STATB FOR A (8.3

8.3. The second approximation t o /

Attention will be confined to a study of the second approximations to/,,/,.
If/»a),/Sl> are written in the form/1«0>Oi,,,/i°><Di1»> then

in the notation of 4.4. Thus the equations satisfied by 4>',u, <J>i" are
9? = -nVAm-n^tliiiW + W), (8.3. 0
«,« - - ^ / ^ - ) - ^ , 1 , / , , ( ^ + ^ ) . (8.3, 2)
If/, is regarded as a function of C„ r, (,

^1 -/1 ( D / +C». ^ .

+
1Fl DOC0\ ein/r ain/f") a 1
l -orJ--3cr"acr Cl ^ c «')' (83,3)

as in 7.3; also, as in that section,

Dpln/j0' w,C|^ f W T _»»£,
Dt ~" 3*7- 8r ,c °' 8C, =
kT'

Thus the first and last terms in the bracket in (8.3, 3) reduce to

Since (8.21, 4) is equivalent to

*-%S(2+««-4 (8 3>4)
-
the middle two terms in the bracket in (8.3, 3) may be written in the form
Let/>„/>, be the partial pressures nlkT, ntkTo(the two gases; also let
ViMlMjikTpC,, Vt = {mj2kT)iCt, (8.3,6)
x, 3 »,/» = />,//>, x, = njn = />,//>, (8.3, 7)
so that x„ xf denote the proportions by volume of the two gases in the
mixture. Then (8.3, 5) can be written in the form
c,{(^!-l)^in7,+xr,rf,J,
where dlt s x,V In/., -&ji(F, - F . ) - & Vp. (8.3, 8)
Hence (8.3, 3) becomes
S i " - / H W - D C i - V l n r + x r 1 rfu.Ci+ aVfoiVcJ. (8.3, 9)
8.31] BINARY GAS-MIXTURE 139

If </,, is defined by an equation similar to (8.3, 8) we find in the same way

<?<•> = y J ° ' { ( « 1 - | ) C , . V l n T+xt-ldtl.Ct + 2V&:Vc0}. (8.3, 10)

Using the equations x, = />,//>, p = pl +pt, two alternative expressions
can be derived for </,,, i.e.

d
»~ pp
(F,-!^,-^,-^,)}, (8.3,,,)
dlt = v , l + ?.^za)vin,-^(F,-F,). (8.3, ,2)

From either of these (since Vx, = — Vx,) it follows that dn = — dlt.

The expression for dxt given by (8.3, 8) is much simplified if the gas is at
rest. Taking D0cJDt = 0 in (8.2,, 4) we find
Vp-PiFi+PuFt
and (8.3, 8) becomes
rf„ = x, V ln/>, -p.FJp = (V/>, -PiFJIp. (8.3,13)

8.31. The functions 0'", A, D, B

Since, by (8.3, 9, 10), the independent expressions VlnT, dlt (or — rf„)
and Vc 0 occur linearly in the left-hand sides of equations (8.3, ,, 2), we can
prove, as in 7.3,, that <t>{1}, Oi" are expressible in the forms

4>i» = - i l , . ^ - 0 , . r f l f -2B i : g^ (8.3,, ,)

4>£l> = -A2.-^r-Dt.dlt-2^:^,,, (8.3,, 2)

the functions A, D being vectors and the functions B being non-divergent

tensors, such that
A = CA{C), D = CD(C), B = CCfl(C), (8.31,3)
with the suffix , or 2 throughout. The functions A, D, B must satisfy the
equations ff^gt^)Cx = n\ I^AJ + », ntI^Ax + At),\ j
( 3 4)
/ A « 1 - ! ) C , = nlIt(At) + nintItl{A^At),f ''

l
-xi-Vf^C, = H|/,(i>,) + »,»,/„(A+Z>,),J *31' S '
/,°Wf,V, = H?/,(B,) + «,*,/,,(6, + B,),l
o (8.31,6)
f2»VtVt = »! /,(6,) + »,», / 8l (B, + B,).
These equations may be shown to satisfy the conditions of solubility, which
6-3
140 THE NON-UNIFORM STATE FOR A (8.4
are similar to those of the original equations (8.3, 1, 2). For (8.2,14-16) to
be satisfied, A, D must also be chosen such that
jf?»mlCl.Aldel + fjetmtCt.A,dea - o, (8.31,7)
lffnmlCl.Dldel + )pf>mtCt.l>tJ€t = 0. (8.31, 8)
Thus, to a second approximation,/, and/, are given by
/ , - / H i -^(C,)C,.Vln T-D^C,)0,.0-,,-2^,(0,)0,0,^1¾}.
(8.31.9)
/ , = / r { i -AACt)Ct.V\n T-DJiCt)Ct.dtt-2Bt(Ct)C°Ct:Vcg).
(8.31,10)
It may readily be verified with the aid of these expressions that, correct to
the second approximation, the mean kinetic energy of the peculiar motion
of the molecules is the same for each of the two constituent gases.
In virtue of (8.31,4-6), if a is any vector-function, and b any tensor-
function, each defined in both velocity-domains, then
n*{A,a) = //m-f)C,.o,rfc,+//i 0 >(^-!)C,.o,</c,, (8.31,11)
1 1 0
n*{D,a) = xr //,°>C,.a,rfc,-x,- //J >C,.a,</c,> (8.31,12)
»»{B, b} - //[•>*>,: b,rfc,+//5°'«>,: b,<fc,. (8.31,13)
8.4. Diffusion and thermal diffusion
The two constituents of the gas-mixture are said to be diffusing relative to
one another if the mean velocities of the two sets of molecules at r, (are not
the same, that is, if c, — c, or (what is equivalent) Cx — Ct is not equal to
zero. Now _ _ , r , r
Ci-C, - ^jAC^c-ij/.C.rfc,.
On substituting from (8.31, 9,10), the first and last terms in the expressions
for/, and/, give integrals of odd functions of the components of O; hence
these integrals vanish. The remaining terms readily yield the expression
Ci-c, - -1- [(^ J/rqw,)^,-! JVrq^orfc,} dlt
+ {~ f/PClAACJ dcx - 1 J>>CJ.4,(C,)rfc,)V In r ]

+ (^ J > C , .AtdCl - 1 J > C,.A,dcM} V In r]

- - i«[{D, D} o*„+{D, ^} V In T] (8.4, 1)
by (8.31,12).
8.4J BINARY GAS-MIXTURE 141
The velocity of diffusion Cy — C2 accordingly has one component in the
direction of —rf„,since the factor {D, D) is positive, by (4.41, 2). Since — dlt
itself (cf. (8.3, 12)) has components proportional, with different (positive)
factors, to — Vx„ to F, — Ft, and to - (mt — m,) V \np, the diffusive motion
of molecules of the first gas relative to the second has components in the
directions of these vectors. The first of these three components corresponds
to diffusion tending to reduce the inhomogeneity of a gas whose composition
is not uniform; the second component indicates that diffusion also occurs,
as one would expect, when the accelerative effects of the forces acting on the
molecules of the two gases are unequal; the third component shows that,
when the pressure is non-uniform, the heavier molecules tend to diffuse
towards the regions of greater pressure.
The velocity of diffusion also possesses a component in the direction of the
temperature gradient; at this stage no general statement can be made about
the sign of the coefficient (D, A). This thermal diffusion tends to produce
a non-uniform steady state in a gas enclosed in a vessel, different parts of
which are maintained at different steady temperatures.
The definition of the coefficient of diffusion Dlt given in Chapter 6 refers
to a state of the gas in which no forces act on the molecules, and the pressure
and temperature of the gas are uniform. In this case n does not depend on r,
and so (8.3, 12) reduces to

Thus c, - c, = Cx - C% = - |{A D) Vnv (8.4, 2)

The vector nlcl giving the flow of molecules of the first gas is equal, as in
6.4, to — DttVnv and similarly for molecules of the second gas. Hence

c, - Cj = - D12 ( i V», - i Vn.,j

-MiK)*-
Comparing (8.4, 2 and 3), we see that
J>ii-(»W3») {A*}- (8-4.4)
We also write DT a (n, njjn) {D, A) (8.4, 5)
and k r = DTIDlt = {/>, A}/{D, D}. (8.4, 6)
The coefficient DT is called the coefficient of thermal diffusion, and kT will
be termed the thermal-diffusion ratio. In terms of Z)„, DT and k r , (8.4, 1)
142 THE NON-UNIFORM STATE FOR A [8.41

may be written as

which is the general equation of diffusion (to the present order of approxi-
mation). It is also convenient* to introduce a thermal-diffusion factor alt
defined by , ,0 0.
' k r = x^oc,,. (8.4,8)
Then the term in ^l — Ct due to the temperature gradient becomes
-auDltVlnT.
As can most simply be seen by interchanging the suffices i and 2 in (8.4, 7)
and using the relation dtl = — du, the quantities Dlt, au satisfy the relations
Dn = Dw au - - a „ .

8.41. Thermal conduction

Since we are at present considering a gas-mixture in which all the molecules
possess only translatory communicable energy, the thermal flux is given
(cf. (2.5,13)) by
q - J/tJmtCJdrfcj+J/.Jm.qC.rfc,. (8.41, 1)
Thus
(ff/MVK«& + «A> = /A(«!-l)Ci*i+//t(«|-t)^iAr (8-41. *)
After substituting for/,,/, from (8.31,9,10), and omitting terms that do not
contribute to the final result, the right-hand side of this equation becomes
(to the present order of approximation)

-ilJP>m-m<0,.D^dn + (C,.A^ViaT}dev
or, using (8.31,11), - J«» [{A, D} du+{A, A} V In T\.
Hence our approximation to q is
q m 1421(¾¾ + ntCt)-lkn*T[{A,D}du + {A,A)VIn T),
or, eliminating dlt between this equation and (8.4, 1),
q - Ikfyfa + njC,)+^^-0,) ({A, D)I{D,D})-XVT
. -AVT+ikT\nlCl + ntCt) + knTkT(Cl-'5t), (8.41,3)
where A a j W [{A, A) - {A, 0}»/{0. D}}. (8.41,4)
* W. H. Furry, R. Clark Jonc* »nd L. Onnger, Phyt. Rtv. 55, 1083 (1939).
8.41] BINARY GAS-MIXTURE 143
We write
Al S
^1-¾¾^ = A V
I- T*>I> &t = At-kj-Dt, (8.41, s)
so that {A, A) = {A, A} - zkT{A, D)+kfT{D, D]
= {A,A}-{A,D}*I{D,D}.
Then (8.41, 4) becomes A = \knl{A, A}, (841. 6)
showing that A is essentially positive. If there is no mutual diffusion of the
gases in the mixture, so that Cv C% both vanish, the thermal flux becomes
equal to - AVF; this is the case when the composition has attained the
steady state corresponding to the given temperature distribution. Clearly A
is identical with the coefficient of thermal conduction, as usually defined.
Equation (8.41, 3) indicates that the thermal flux is in general made up of
three parts. First, there is the ordinaryflowof heat resulting from inequalities
of temperature in the gas. Next, when diffusion is proceeding, there is a heat
flow resulting from thefluxof WJCJ + n2Ct molecules per unit time relative
to the mass-velocity: each molecule carries, on an average, a quantity
§kT* of heat energy. Thisflowappears because the thermalfluxis measured
relative to the mass-velocity c0 of the gas, not relative to the mean velocity
c of the molecules; if it were measured relative to c, this term in the flow
would disappear. Lastly, even if theflowof heat were measured relative to c,
diffusion would contribute a term knTk^Cx — Cj) to the heat flow. Only
the first of these three parts of the thermal flux is usually measured in
laboratory experiments.
TheflowknTkT{Ct —Ct) is, however, important in certain circumstances.
It indicates a heatflowdue to diffusion, which may be regarded as an effect
inverse to thermal diffusion; it is called the diffusion thermo-effect. The
direction of the flow is such that, if it generates a temperature gradient,
* The factor J occurs instead of the usual } because the kinetic energy of the molecules,
as well as the thermal flow, is that measured relative to the mass-velocity c„ not the
mean velocity c. The following analysis is due to Enskog. By analogy with (8.41, 1),
the flow of energy relative to the mean velocity c of the molecules is
»(/>i(Ci - c ) . ( c , - c ) ( c , - c ) + p t ( c , - e ) . ( c , - c ) (c,-c))
= K/^C.-Q.CC-CMC.-O+p.fC-CMC-CKC.-C)).
Expanding this, and using the fact that P,C,+p,C, = o, we obtain the expression
i(plC'lCl+p,Cf1Cs)-ipC.CC-i(p,Cil+p,'cl)C-piC^C'1.C-p,C,C,.C.
Now C is a small quantity: thus we may neglect its square, and its products with other
small quantities. Hence we can neglect \pC.CC, and in the product p,C,C,.C we can
replace C, C, by its first approximation C, Cf" =• IkT/m,) U, where U is the unit tensor.
Then the expression becomes
q-\hnTC = fl-|*r(n,C, + n,C^.
Thus the term |*T(u,C, +n,C,) in (8.41, 3) arises because the thermal energy and
thermal flow are both relative to the mass-velocity c„, not to c.
144 THE NON-UNIFORM STATE FOR A [8.42
thermal diffusion due to this gradient decreases the primary diffusion. In
other words, the heatflowis in the direction of diffusion of molecules which,
in thermal diffusion, would seek the cold side. The equality of the coefficient
k r in the expression for the heat-flow with the thermal diffusion ratio is a
special case of a general reciprocal theorem in the thermodynamics of
irreversible processes.*
In a gas in which there is no diffusion, dlt = — k r V In T (cf. (8.4, 7)).
Hence for such a gas (8.31,1, 2) take the forms

( 4 ,7)
W--At.VlnT-2St:VcJ '
These relations give the vectors Ax, At an independent interest. The
conditions that (8.41, 7) should correspond to zero diffusion are
J/rH.C^-o; Jfi»At.Ctdct-o. (8.41,8)

8.42. Viscosity
As in the case of a simple gas, the first approximation to the pressure system
reduces to the hydrostatic pressure. The second approximation adds to this
the pressure system given by p<u, where
p<» = »l«,(ClC1)'» + »tw1(C2CI)<»
which represents, to this degree of approximation, the deviation of the
pressure system from the hydrostatic pressure p.
Only the terms in (8.31, 9, 10) that contain Vc0 contribute to the value
of p111; thus, as in 7.41,
p<» - -2«1J/i0>ClCI(Bl:Vc0)rfcl-2«,/y50>C,Ct(Bt:Vc0)«/c1

- - H * J / T c f o * ! * , + m1//S«,>C1°C,:Btdct] %
= -f*n«r{B,B}8
by (8.31, 3), the theorem of 1.421, and (8.31, 13). Hence, if we write
p & |An*r{B,B}, (8.42,1)
the last equation becomes
0

p«> = _ 2/t8 s - 2/1 ^ Co, (8.42, 2)

which is identical with the expression for the viscous stress system in any
medium (cf. 7.41). Hence /1 can be identified with the coefficient of viscosity.
As in the case of a simple gas, /t is essentially positive.
• !». OnMger, Fhyt. Rev. 37, 403 093'); 3*. « 6 5 ('930.
8.51] BINARY GAS-MIXTURE 145

8.5. The four first gas-coefficients

Thus, when the velocity-distribution functions for a gas-mixture are deter-
mined to a second approximation, it appears that there are four coefficients
concerned in the phenomena of the non-uniform state, namely Dn, DT, A
and /i. These we call the 'first' gas-coefficients for a non-uniform gas. They
and the associated ratios f and k r (cf. (6.3, 3) and (8.4, 6)) depend on the
four integral expressions
{A, A}, {A,D}, {D,D}, {B,B},
which may be evaluated by methods similar to those of 7.51, 7.52.

8.51. The coefficients of conduction, diffusion, and thermal diffusion

It is assumed that At, At, 2),, D, may be expanded in series, which, for
convenience in later work, we write in the forms

A, = s"«,«iw, At = r-p-ST'. (8-5'. 0

_ 00 —00

D,^£dpaf\ D,-*£d„^\ (8.51,2)

— 00 — 00

Here ap> • M/p,*,//), ai°> m -Mlpjtjp (8.51, 3)

(where A/, = mJmB, Mt = Wj/m0, m0 = w, + mt as before), and for values of
/greater than zero a[P) s # ^ ^ a<-p>so,|

The expansions (8.51, 1),(8.51, 2) resemble those of 7.51. It will be noted

that in the two equations (8.51, 1) the same coefficients ap are written, and
likewise the coefficients in (8.51, 2) are the same: but except when p = o,
one or other of oV" and a*/1 is always zero, so that the equality of their
coefficients has only formal significance. The notation Z' implies that the
summations in (8.51, 1) do not include a term corresponding top = o; this
is shown to be the case by an argument similar to that leading to (7.51,2), the
equations (8.41, 8) taking the place of (7.31, 6). The equality of the coeffi-
cients of «i0> and aj?' in (8.51, 2) follows from equation (8.31, 8). On sub-
stituting from (8.51, 2) into the latter equation, the parts of the integrals
involving the functions efp) vanish (as in 7.51) when p * 0. Hence, since the
same coefficient has been chosen for oi0) and ai0), the equation reduces to
fA»mlCl.4»dcl + jjr>mtCt.alPdct = 0.
With the values of ai" and a»°' given by (8.51, 3), this is equivalent to
/exp ( - <**) <g\dSgl - /exp ( - «?) <*\dVt = o,
146 THE NON-UNIFORM STATE FOR A [8.51
which is obviously satisfied. Thus (8.31, 8) is satisfied by expressions for
0,, Dt of the form (8.51, 2) if the coefficients of 4", <4" are equal, but not
otherwise.
To determine the coefficients ap, dp we note that, by (8.31, n , 12),
[A^-a* {D,a">}-^ (8.51,5)
where n\* jffm~flC,.**"*,+//^-1)0,.^4¾. (8.51,6)
1 1 0
n*8Q s xr //I°'C1.at>&1-xt- //i 'C1.ai«&1. (8.51,7)
Evaluating these integrals by the methods of 7.51, using (7.5,4, 5), we find

°.-^'(f)'. «--^(f)' *-±(£)'.

(8.51.8)
whereas for all other values of q
av = 0, iq - o. (8.51,9)
On substituting from (8.51, 2) into (8.51, 5) we find
2 dpa^-S^ (8.51,10)
J>— — 00

where a„ m {a*"',^"} m am. (8.51,11)

These equations determine the coefficients dp. Again
{A.cfi*} - {i4,o«"}-kr{D,ate>} - a , - M , . (8.51,12)
Now (J, — oifq + o; hence, on substituting from (8.51, i)andusing(8.5i, 11)
we have m
Z' Op**-«fl (¢+0). (8.51,13)
This provides one equation less than (8.51,10), corresponding to the fact
that there is no coefficient a,, to be determined.
Equations (8.51,10) and (8.51,13) can be solved by successive approxi-
mation in the same way as (7.51, 8). The solutions are
</p = * 0 lim J ^5>/j^-> (8.51,14)

and ^ - lim («,j/i' p m> +a_ 1 j/'it,)/j/'"- ) (P+o). (8.51,15)

Here j/^m) is the determinant of 2m +1 rows and columns with elements

apq(-m^p,q < m^andj/j^Msthecofactorofa^intheexpansionof^"0;
also s/,<m) denotes J/JJ *, and J / ^ ° is the cofactor of « w in the expansion of
 8.52| BINARY GAS-MIXTURE 147
By (8.51, 1,2, 5, 9 and 12),
{D,D} = -Ldv{<tP\D}=d080,
{D,A} = Xdp{cf»\A} = rflBl+ <#_,<*_„
{A, A] = T,'ap{<fl>\A} = a ^ + a . j a . , .
Hence, using (8.51, 8, 14, 15), the formulae for the transport coefficients
given in 8.4 and 8.41 become

D
» =; ^Sr-lim ^ ' " " V ^ " " . (8-s'. 16)
2 7lm0 m-M)
k r = - I lim { x 1 M l - » ^ r + x,Af i -*^f2 1 }/ J /'<'-), (8.51, 17)

A = ^k*T lim {xjmf1 j/JV"' + 2x,x a (m l w J )-i j/?"/

+ xImi-W^!,}/j3r'<"". (8.51, 18)
As in the latter part of 7.51, it is possible to prove various theorems about
the determinants introduced above, and in particular that the successive
approximations to Dlt and A form monotonic increasing sequences. On
account of the complexity of the results for gas-mixtures, however, these
results are not given here.
From (8.51, 3, 4 and n ) the quantities a M depend on the composition,
but are independent of the total density for a mixture of given composition.
Thus the same is true of A and k r ; however, Dlt is inversely proportional
to the total density.

8.52. The coefficient of viscosity

As in 8.51, it is assumed that B t , B, can be expressed in series, of the form

B»= r bpb["\ B , -25 W , (8.52,1)

with the same coefficients in the two series: the functions b 0 " are denned in
the two velocity-domains by the equations

S2 2
Vf> m Sf-i'dittiv'Vx, b^sSy-W^M?, (/>>o).j '
As in 8.51, the notation S' signifies that the value p = 0 does not appear in
the summation.
From (8.31, 13) {B, b<«>} = /?,, (8.52,3)
where «*/?, = / / 1 ^ 1 ^ , : VF^+JjpVjr,: b^dcr (8.52,4)
Integrating as in 7.52, we find

A-I5- £-1-¾. A-o (?*±i). (8.52,5)

I
148 BINARY GAS MIXTURE [8.S2
On combining (8.52,1,3) we obtain

r *pft„ = /?„ (8.52,6)

p « — «0

where b„ « {b">, b*»}. (8.52, 7)

These equations are similar to (8.51,11,13). Their solution is
6„ - lim {ftrfp+fi^lW,)!*". (8.52,8)
Here &*> is the symmetric determinant, having 2m rows and columns, in
which the elements areftp,,p and q taking all values other than zero between
—m and m; £f$ is the cofactor of bJp in the expansion of if"*.
From (8.52,1,3, 5)
{B. B} - SAP{B. *»»} »/?,*, + /?_,*_,. (8.52,9)
p

Hence (8.42,1) becomes, using (8.52, 5,8),

/i - | * r lim {x?^17)+2x1x1^i"L>1+xJ^,."i_,}/^'-'. (8.52,10)
m-»<D

As before, it can be shown that the successive approximations to ft form a

monotonic increasing sequence.
Like A, the viscosity ft depends on the composition, but is independent of
the total density for a mixture of given composition.
 9
VISCOSITY, THERMAL CONDUCTION, AND
DIFFUSION: GENERAL EXPRESSIONS

9.1. The evaluation of [a<«", a'"] and [b"»,b<«>]

In order to determine the coefficients of viscosity, thermal conduction, and
diffusion of a gas by the methods of the last two chapters, it is necessary first
to evaluate {a*',**91} and {b^, b'"*}, and therefore, by (4.4, 12),
M»an. K\«n.. [afun,, WPM'% iw.vn* ivr.wu
This involves integration over all the variables specifying an encounter
between pairs of like or unlike molecules; such integration can be completely
effected only when the nature of the interaction between the molecules is
specified. As noted in 3.42, the special law of interaction affects only the
relation between the deflection (x) of the relative velocity at encounter, and
the variables; and b. The integration over the variables other than; and b
can be performed without a knowledge of the law of interaction, and is
effected in this chapter. In the next chapter various special laws of inter-
action are considered, and the corresponding values of the above ex-
pressions are determined.

9.2. Velocity transformations

Certain general relations are grouped here for convenience of reference.
The velocities c l t c, of two molecules of masses mv m, before encounter are
given in terms of the variables G, gn by the equations (3.41, 6), i.e.
c, = G - Mtgn, ct = G + Mrfn,
where, as in (3.41, 1),
A/, 3 «i/»»o, A/, s mj/wj, m0sm1 + tnt,
so that A/j + A/, = 1.
Similar relations connect the velocities after encounter, c'v c't, with the
variables G, g'lv The magnitudes of gtl, g'tl are equal, and are denoted by;.
Let C 0 denote the velocity of the mass-centre of the pair of molecules,
relative to axes moving with the mass-velocity of the gas, so that
G0 = G-c0. (9-2.0
Then the peculiar velocities C„ Ct of the molecules are given by the
equations C, = G,-Mtg„, Ct = G 0 + Mlgtl; (9.2,2)
[ 149 1
 ISO VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [9.3
similar equations give C[, C't. From (9.2, 2) it follows that
i m ^ + l w . C S - imJGl + MiMJ) (9.2,3)

and (cf. (3.52,4)) * | £ f g > - 1. (9.2,4)

By analogy with the equations (8.3,6), namely

Vtm(mJakT^Cu Vam(mJikTyC* (9.2,5)
we define new variables S?0, f, f' by the equations
90a(mj2kT)lG„ rmfaMiMJakTpga, f'm^M^MJikT^g^,
(9-2,6)
so that f » f'.
From these definitions and (9.2, 2-4) it follows that
Vj-Mto-MJr, Vt-MlV0+Mif, Vi-Mfo-MiV, (92,7)
*!+*!-•!+*», (9.2,8)
*«»>) _ ( « » r ) a(Qe.g«) „ K ^ t ) * , Q , a.
wz,9;
«(€„*) «(©»#«)• «(c„<y (a*rp*
Also, since the angle between y and f' is the same as that between gtl and g'u,
f.f' - f*CMX- (9-2, 10)
Any function of the velocities of two molecules after encounter may be
transformed into the corresponding function of the velocities before
encounter by taking x — 0.

9.3. The expressions

[SPQVuSVWVJu and [S(*I)9?fa, 5(«D»iVtlii
1
By (4-4.4.9) "w (35. 3).
[SjfTO * i , 5 f W ) * J u (9-3. 0
is equal to

which, by the definition (7.5) of Sfg\x), is the coefficient of s*P in the

expansion of

x exp ( - r*f|) gbdbdedcidct, (9.3, 2)

s = T
where r b ' 'Fri* (9-3.3)
9.31] GENERAL EXPRESSIONS 151
0, 0)
Inserting the values of/[ ,/j in (9.3, 2), we get

x exp ( - 7¾¾ gbdbdtdc^dc%,

or, using (9.2, 8, 9),
(«•/*)• n-3ll\l exp (-91- f) (exp ( - stfj)«, - exp ( - riffl *Q. «Pt
x exp ( - T%\) gbdbdedV0dy.
Let //It(,Y)s/exp(-Srg-^-s^;*-7^i)^.«?trf»0. (9.3,4)
Since any function of #J, "<PS is transformed into the corresponding function
of «\, «", by putting * = o,
tf18(o) = /exp ( - 0 | - f t _ ,s*| _ rtfj) V ^ d V , .
Thus the expression (9.3, 2) is equal to
(«•/*)• r*JSS{H^o) - Hlt(x)}gbdbded9. (9.3, 5)
Similarly, it may be shown that
[Sf>(*l) * & , 5J"(«1) * t °»J M (9.3. 6)
is the coefficient of j"f» in the expansion of
(sT/st)i rr-'JJJ{Llg(o) - Ln<oc)}gbdbded9, (9.3, 7)

where £,,(*) = J e x p ( - ^ - 7 « - s t f ^ - 7 ^ ) ^ : ¾ ^ <f*o. (9.3.8)

9.31. The integrals //„(*) and Llt(^)
By (9-2. 7) ,q _ (MJ ^ + A/f y). (A^ ®0 + M\9)
= Mt9l + Mx9* + 2(MxM^V0.f, (9.31, 1)
and similarly V* = M^l + Mt? - 2(M1 M$ » 0 . y'. (9.31, 2)
Hence
n+f+s&'S+rVl = «'i,^+*«? 8 +2(M 1 M 2 )»®o(T?-5F'),
where i l t s 1 + A/, s + MtT, itl a 1 + Af, 5+Af^. (9.31, 3)

Let v m ® 0 + i - (Aft Af,)* (TV - sf'), (9.31. 4)

'11

so that a change of variable from ®0 to v is equivalent to a change of origin

in the ®0-space. Then
a S + ^ + ri^ + rfri - «„f 8 +f„? 2 -(M 1 M 2 /« 1 2 )(7y-sy').(7y-sy')
- 'u»*+;W. (9-3'. S)
152 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: (9.31
where ; „ a i H - ( ^ / , ^ 1 , , ) ( ^ + T»-2srCOS^). (9-3 1 ' 6 )
Again, writing t>, a (A/,/i„) (iy - sy') + f',}
(93,,7)
Mi»-*)-f.\
Vt a {MJi :

we have, by (9.31, 4) and (9.2, 7),

tf, = Afjc-Afji
Hence
V[.Vt - (M,M 8 )»^-MAf,t>, + Af,t>,) + (Af, Af,)»c,.0,, (9.31,9)
« 7 = Af, 1* - 2(M, Af,)t». t», + Af,t^,
«1 = Af,»« - 2(M, A/,)* v.», + A/,«i,
and so
*I»*| - A/,A/,»« + »*(A/J^ + A/irf) + 4A/,A/ l (».»,)(».»,)
+ Af, Af,wj vj + odd powers of v,
1
(^.^2) - Af.Af.r^Af.Af.t**,.», + {*.(Af,t>, + Af,»,)}*
+ A/, Af,(©,. 1¾)1 + odd powers of v.
Thus, using (1.32, 9), (1-42, 2), and the theorem of 1.421, we may write
(9.3, 8) in the form

= 47rAfiA/ t re-'»"-'»''{K+^» I r 1 .»,+(»,. tF,)*-J*}t$»*<fc

- fftAf,Mt^»',ir,i{f + -^/,^,.^, + 1 1 ^ , . ^ - JflitlfJ}.

(9-3 L ">)
Again, by (9.31, 9),
#i«(X) - (Af 1 Af i )l/«-*"« , -*i» , (««+B I .«»)*

- 4ff(M,A/^» f V ' i t f - ' i t ' V + t ^ . i g v * *

- »r»(M,Af,)»«->••»*if,*(| + !„©,.»,). (9.31, 11)

It remains to evaluate o,. t>, and v\«j. From (9.31,7), using (9.31,3,6) and
(9.2, 10), we find
t>,.t>,

- (Af,Af,/ii,)(Tf -Sf').(7*-sy') + (1/1,,)(Af,/ - A f , y ) . ( T * - s y ' ) - 9 . ?

- (AfjAf^if,)^^T*-2srcosx) + WhJ{~hi+' + ( ' » - ')«»X)
-f*co$x
- V/I'IIXI -7.2-cosx). (9-3«. " )
9.32] GENERAL EXPRESSIONS 153
Again, | c , A ©,1* = v\ » | - (©,. »,)*,
andsincey'A f = -f A ?',? A y = o = y'A ?', wefindfinally,using(9.3i, 3),
» , A » t = (f A/)/',*•
Hence, since the magnitude of f A f' is j 2 sin ,y,
v\v\-(»,.©,)* = (y/«?,)sin*x. (9.31. '3>
By using (9.31,12,13) in (9.31, 10, 11) we find
H»(X) = J r l M A f J I e - A . ' i a l f l + O -;„-cosx) ? «}, (9-3'. M)
At(x) = 1 ^ ^ , ^ ^ 1 , ^ + 5(1 - y „ - c o e X ) f «
+ {(1 - A * - cos *)« - i sin* *}?«]. (9.31, 1 s)

9.32. #„(*) and £„(x) as functions of s and t

Write now >>„ = A/,$ + A/,f, s „ = 2A/,A/,rt(i - c o s # ) . (9.32,1)
-1 -1 x
Then, since 5 = 5 - 1, i~ = r ' - i , w e find from (9.31, 3, 6),
hJsT = (1 -ylt)/st, (9.32, 2)
>1* = {(I + 5) (I + T) - 2Afj MtST{l - COS*)}//,,

-(I-*!*)/(»-yu)- (9-32.3)
If jyla and :7,2 are regarded as independent,

Ir = rV- (9-32.4)
Hence, from (9.31, 14),
(S7-A0»(Af,M,)-»,H//„(x)

= '-''"'(i - J i t H H + O -ht-cosx)?}

V'I*
But, by (9.32, 3) and the expansion defining Sonine polynomials,

«-/..«•(, -,,,)-1 = e-»'/u-v„>S (fi*p-'(, - , , , ) - 1

r **'
»V

r n •'
and from a similar relation

04» r n ™ *
154 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: (9.32
Combining these results and using (9.32, 1), we find

- # - ^ 2 Z{2MiMttt(i-cosx)Y—, (Mts + M1l)n

r n "
x {(»+ i)&r\y>(f')+9\i -cos X )S>rW)}. (9-32. S)
Using the binomial theorem we can expand the sum on the right of
(9.32, 5) in a series of powers of s and t. Since Af,, Mt appear only in the
combinations Mxt, Mts, the coefficient of s"f in this series has M$M\ as
a factor, but is otherwise independent of M„ A/,. It is in general a polynomial
in f* and cos ^, of degree p + q+i in ?*, and of degree in cos x equal to the
lesser of/>+1 and q+1. Thus, on expanding, (9.32, 5) becomes

(932. 6 )
where Ap<t,l is a pure number, independent of Mx and Mt.
Similarly
UsTls&MMJ-iiriLJjd

+y4{(i - cos*)* - J sin"x) *-'••»*( 1 -^,,)-1

- «-»'SL{2Wi^i*'(« - c o s x W M A V + A/,/)"

X [(«+ 1)(11 + 2 ) ^ - / ^ + 2(- + I)f«(l -CWX)5ft*,W(f»)

+y«{(i - c o s X ) » - J s i n ^ } ^ , ^ ) ] . (9.32, 7)
On expanding this in powers of s, t, the coefficient of s"f is found to be
*-»*Aff Aff multiplied by a polynomial in «* and cos x, of degree £ + ? + 2
in y1, and of degree in cos x equal to the lesser of p + 2 and 9 + 2. Thus
Ksr/xOWAf,)-! r-lLufcO-r*' S iWA/.^A/.O^cos'x,
p.«.'.*
(932.8)
where B^ is a pure number, independent of M, and A/,.
Explicit expressions for -dp,*, Bp^can be obtained from (9.32,5,7), using
the expression (7.5, 3) for .SJUX*)- 1° v ' e w °f t n e complication of these
expressions it is, however, better in practice to calculate any desired values
of A
pvi> Bi*t directly from (9.32, 5,7).
9.33] GENERAL EXPRESSIONS 155

9.33. The evaluation of

[5(^^,5(.1)^,, and [5(^) tffo, 5(¾¾ tf.Vj,,
By 9-3. [^(^)^,^(^1)^,,
p
is the coefficient o(s f in the expansion of (9.3, 5), that is, of
(ST/st)i n-^\\{Hu{p)-Hu(x)}gbdbdedf.
Hence, using (9.32, 6)

- jHAff+»Jlff+ J / / k » * S^ P ,r,? 2r (i -cos'X)gbdbdedf.

r.l
The integrand here is independent of e and of the direction of f; thus, on
integrating over all values of e and all directions of f, we find that

= 8ff»A/r-*A/r*ff«-*'£^ w rf ,,+, (i -cosfx)gbdbdf

r, I

= 8A//+* AJ?+» 2 4 * * a B M . (9-33.1)

r. '
where Qi'^r) is defined by the equation
QBKr) = ».//«-«* **+«( i - cos'x)gbdbdf, (9.33, 2)
so that QiS'(r) = o, Q$(r) > 0 if / > 0.
Similarly, by (9.32, 8) and (9.3, 6-8),

= S7T-lMf+lMl+1jffe-<'ZBlKirl9*'(i -cosfX)gbdbdedf

r.l
+1 +1
= v«r Aff 2 fl^niftr). (9.33.3)
r. J

In addition to £2i'2'(r), it is convenient to introduce $}, defined by*

$i = 2nf( 1 -cos' x)bdb. (9-33.4)
In terms of this, by (9.2,6),
OSW = (kTlznmaA/, Ms)5 J " f - ' V + W * - (9.33, 5)
• In earlier editions of this book. (S{'J was defined as J(i — cos'x)gbdb; however, it it
advantageous to give ¢ ^ the dimensions of an area, so that it can be related to a mole-
cular cross-section.
156 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [9.4
Expressions for
[SfKVV «?„ S f W VJ„ and [Sf >(*!) *fo. -Sj^D «t*tlii
in terms of the functions Qii(r) are given in 9.6 for some special values of
pandq.

9.4. The evaluation of

[S<PQ9V S p r » » J u and [5(<fJ) * > , , 5(¾¾ » > , ] , ,
The methods here required are similar to those of 93-933, so that only the
main steps need be indicated.
The expression ^ , ^ ^ ^ , , ^ ^ fe^ j}

is equal to the coefficient of s"f in the expansion of

{sT/st)in-^jjjiH^-H^gbdbdedf (9.4, 2)
in powers of t and f, where
#i(*) - J e x p ( - a r » - , « - ^ i « - 7 ^ ) * i « i ^ (9-4.3)
and 5 and T are defined by (9.3, 3).
The expression
[sy>(»f) *?,*?„ s j w v , *j l t (9.4,4)
p
is likewise equal to the coefficient of s P in the expansion of
(sT/st)i n-*\\\{Lx{o) - L^gbdbdedf,

where Lfo) • Jexp(-9\-f-dr x *-rSf»)«>i:1?> l irW r (9.4,5)

Using relations analogous to (9.31,1, 2) we find that
W+9*+sV? + rffj = . , ^ + 1,^-2(^/,^/,)^,.(5^ + 77),
where i, • 1 + A/,(s+r), », • 1 + M,(s+r). (9.4,6)
Hence (9.4, 3, 5) can be evaluated by making the substitution
V - « , - (A/, MtfW + Tf)liy
Integrating as in 9.31, we find
"lOO - nUiie-^UMi + iMAi - / , ) + M, cos*} 7*], (9.4.7)
Li(x) " l**H*rMVtM\+ 5 W , ( i -/O+M.cos*}?*
+ {Mi(i -;i)+M t oo»Xr , f*-»JI#|f l iin , xl, (9-4.8)
where j \ • 1, - (A/, MJit) (s* + T* + asr cos x)- (9-4i 9)
9.5] GENERAL EXPRESSIONS 157

By (9.4, 6, 9) and (9.3, 3),

«, - 1+A#, + Af,— ^-. VA "!\ —t (9-4t*°)

l-s 'i-t (i-f)(i-r) ' \y-n '
jx = 1 + MJs + T)-M1M^S%+T*+2ST cos xVU+M^s+f)}
= {(i + s)(i + r)-sr(A/|+Aff+ 2^^,0)8^))/(1 +Afj(s+r)}
= {1 - st(M\ + A/J + zM, Af, cos *)}/{i - Af,(* +1) + (M, - A/t) rf),
(9-4. " )
so that Ht(x), Li(x) can be expressed as functions of s, t. The expansions
corresponding to (9.32, 5, 7) are
(sT/st^n-iH^) = e-»'SSMAff + AfJ + 2A/1Af1cosX)}'(?«'/r!)
r n

x {M& + 0 - (A/, - Af,) 51}» {!/,(» +1) •%}» V )

+ (AfI + ilftcofx)y«5'r?.»|(y«)}, (9.4.")
l(sT/st)l n-\Lx{x) - «-»' S S MM? + A^ + 2Af,Af,cosx)]'{y7r')
r n
x [A/1(5+0-(M1-Afl)rf]"[Affri + i)(« + a ) 5 a ^ V )
+ 2(»+ O^Af^Mj + M . c o s ^ ^ V V )
+ ^{(Af1 + A/,cos X )*-lMisin« X }5<y | ( ? »)]. (9.4,13)
If the expansion of (9.4, 12) in powers of s and (is expressed in the form
r+ S A'^ff^eo^x, (9-4, «4)
P.ff.r.l
then [SfWVv^^VAn-ifA'priOWr). (9-4. > 5)
which is analogous to (9.33, 1).
Likewise the value of (9.4, 4) may be derived from the coefficient o(spf>
in the expansion of (9.4, 13). Expressions for (9.4, 1 and 4) in terms of the
functions £l\l(r) are given in 9.6 for special values of p and q.

9.5. The evaluation of

[S(fD*i. 5W)»Ji and [S(»J) < » „ « ( » ! ) * > J ,
It is clear from (4.4, 11) that [Sf X<V\) 9U S^\Vf) VJi can be derived from

by taking m, equal to m, and adopting the law of interaction between pairs

of molecules of the first gas instead of the law for the interaction of pairs of
unlike molecules. The effect of these changes is that in every result Af„ Af,
158 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION! [9.6

must each be replaced by 1/2, and O'u(r) must be replaced by the similar
integral Qi"(r), relating to the encounter of pairs of molecules of the first gas.
On putting Mt = Mt =» \, the expression (9.31,14) for H^x) becomes
identical with the expression (9.4,7) for //,(#), except that cos^ has the
opposite sign (cf. (9.31, 6) and (9.4,9)). This implies that the coefficient of
&ii(r) >n the expansion (9.33,1) is ( - i / times that in the expansion (9.4,13).
Thus
[Sf>(«i) vlt s$x<e\) «-j u +[spofl) *„ sfyrQ <€ j u
is equal to the expression derived from [S^\1f^Vu SJ"(*S)*JM by sup-
pressing those terms involving Q$(r) for odd values of/ and doubling the
remaining terms. Hence [5^(¾¾¾^. ^"CSPS^J, involves only expres-
sions Q'i"(r) f° r e v e n values of /.

The value of [Sy'fSffl tffo, 5 ^ ( ¾ ¾ ^ ¾ » similarly derived from

Expressions for
tsjpcifD*,, spc*D«M, and [sp(*l)*fo. sfrtvnvfo]!
in terms of the expressions fl'/V)are given in 9.6 for certain special values
of p,q.
9.6. Table of formulae
The following special cases of the results of 9.33, 9.4,9.5 are tabulated for
convenience of reference:
[»„ » J „ - -8(M,Mt)\OlV(i), (9.6,1)
pPfSyWDVali. = 8(MtaMt)»{iii»(2)-»QiV(0}. (9-°. 2)

* { W ( 0 -S«JV(a) + QiV(3)"»Qff(*)}. (9-6. 3)

[«i,«,]i. = 8M,QiV(i), (9-6,4)
t»i.SfXWDVJu = -8Afi{Q{i'(2)-fQii'(')}. (9-6, S)
[Sy»C»D«i. S f W ) » J u = 8Mf{i(6Af! + sM!)QiV(i)-sMifia'(2)
+ «|Qffl(3) + ai«fiM,08Xa)}. (9-6,6)
[^,^(^)^,-0. (9.6,7)
The last result also follows directly from (4.4, 8), by the principle of con-
servation of momentum. Again,
P I W »1. S J W »,]» = 4QPX2), (9.6, 8)
S)
P i W * » S^W) * J i = 7"i (2) - 2«i*'(3). (9-6.9)
[Sf»(»» «1. SfC*!) »Ji = W ( * ) ~ 7"ia>(3) + ^'(4). (9-6. ">)
9.7] GENERAL EXPRESSIONS 159
3,
W » * i . ^ (*i)*.]i =¥«i"(2)-1¾%) + IQPX4). (9-6. »)

(9-6,12)
[ S f ^ W . Sp<*D * J , = ^HOPW-HWfoHW"i*'(4)-1^(5)
+iW(6)+iQi«>(4), (9.6, 13)
[ « > , , « > , ] „ = -VMiJIf.feaiKO-IOffW}. (9-6, 14)
[»>», * > J „ = - ^ ( 5 ^ / ^ ( 1 ) +IM.Qift*)}. (9-6,15)
[»>,, *?>,], = 4Qi»>(a), (9.6, 16)

l ^ W « f a . «">.]i = 7«i'*(2) - »QPX3). (9-6,17)

[5f W ) « f a , Sf X*D tf*Ji = W ^ ' ( 2 ) ~ 7^(3) + ^»(4). (9-6, 18)
Pf W ) * > „ * > , ] , = VOPXa) ~ f^ 0 ) + W U ) . (9-6,19)
[spc*D*X 5J>Y»DV>J = ^ W W - W ' ( 3 )
+¥«i* , (4)-ini* ) (s). (96.20)
[ - ^ ( ¾ ¾ ^ . sf T*»«fa]i = H i F G P W - H W O )
+ W"i2,(4) - IQPts) + A W ) + ia«(4). (9-6. 21)
9.7. Viscosity and thermal conduction in a simple gas
In the case of a simple gas the elements a^ of the determinants J / , J / " " ' of
7.51 are defined as equal to [t/p\a"']; that is, by (7.51, 2) to

Hence (9.6, 8-13) give the values of am for values of p, q from 1 to 3.

Similarly equations (9.6, 16-21) give the values of the elements b^ of
(7.52, 5) for the same range of values of p, q (cf. (7.52, 2)).
In (7.51,14) and (7.52,8) the coefficients A and /t are expressed as infinite
series. By taking one, two,... terms of this series we obtain first, second,...
approximations to A or p. These will be denoted by [A],, [A], or [ft]v
[/*]„ ..., so that, for example,
r , 5*7" . . . 2SC,kT . ,
20,, 40,4

1* - ?*»£•££) - ?*"*£* <"• •>

Similar notation will be used for successive approximations to Dlt, DT, A,
H and k r for a mixture. The first non-vanishing approximations are obtained
160 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: (9.71
by deleting the limit sign in such expressions as (8.51,16-18) and (8.52, 8)
and giving to m the value 0 in (8.51,16), and 1 in the other cases. Similarly
for the ratio f of (6.3, 3), so that we write [f]t 9 [X\Jcv{ji]v
This manner of indicating successive approximations differs from that
used for the velocity•distribution function/, the successive approximations
to which a r e / m , / m + / n \ . . . ; but there is no reason why a similar notation
should be employed. All the quantities A, /1, £>,„ DT, f, k r depend on the
second approximation to /, and approximations to them are really sub-
approximations related to phenomena depending on/ m +/ t l ) .
From (9.6, 8,16) it follows that
^ - 2 5 ^ 7 7 1 6 ^ ( 2 ) , ^,-5^7-/80^2). (9.7,3)
Thus, whatever the nature of the interaction between molecules,
Mi - W>]i. (9-7.4)
This shows that the first approximation [f]l to the ratio A//tcv is f for all
spherically symmetrical non-rotating molecules.
In considering the second approximations [A], and [/*], we make use of the
results (9.6, 8,16 and 9,17, 10,18), giving

we thus find [A]t/[A], - (1 -a\Ja lx a t ^r l

the sign of equality corresponds to the case when au — blt — 0. Thus

[A], > KM*
No similar result valid for all degrees of approximation has been obtained:
but numerical calculations for special molecular models suggest that the
ratio A/c„/( increases as successive degrees of approximation are taken into
account, and that the limit is only slightly greater than f.

9.71. Kihara's approximations

Simplified second approximations to A and /t were derived by Kihara,*
using the fact that for most molecular models the non-diagonal elements
apv AM of the determinants J/""', &m) are considerably smaller than the
diagonal elements app, bpp.
From (9.33, 2), since f* - m, M, M^*l2kT,
Qfl(r) - ni ^^hyHje-m.ulM,^»kTgtr+*^ -CM'X)bdbdg.

• T. Kihare, Imperfect Gam (Aaakuta Bookitore, Tokyo, 1949; Engliih translation by

U.S. Office of Air Research, Wright-Patterson Air Base): tee alto Rev. Med. Phyt.
15, 831 (1953). Kihara's approximations were suggested by results for Maxwellian
molecules (cf. 10.3a, below); they amount to giving certain ratios of collision-integrals
flfi(r) the values which they take for such molecules.
9.71J GENERAL EXPRESSIONS 161
Here X is a function of; and b (cf. 3.42) but not of T. Hence

T^jjQ = -(r+S)Ot2(r) + ni?(r+i) (9-7«. 0

with a similar relation between ili"(r) and fli"(r+ ')• Using this relation and
(9.6,8-10,16-18) we find
a
«n = V u *= *w - - lT-Jr<

""'^Y dT*+21 dT +
2an)' "n 4\rdT*+Zl dT+ i"11)-
(9.71, 2)
For all physically reasonable models for electrically neutral molecules,
a n (= 4Qi8)(2)) varies over small temperature-ranges roughly as 7^, where
n lies between about 0 and \ (cf. 12.31). Thus in the expressions (9.71, 2) for
aM and b„ the terms involving dan/dT and dtalljdTt are fairly small com-
pared with the an terms; likewise alt and blt, depending on dan/dT, are
fairly small compared with an, a„ and bn, bn respectively.
Kihara obtained his simplified second approximations by assuming that,
in the Enskog approximations
tA], = tA] 1 /(i-aV« u a„), [ji]t - [/t]J(i -^/MM). (9-7'. 3)
terms of higher order than second in the temperature-derivatives can be
neglected. This amounts to writing a,, = Vau> *«» = ff^n m the second-
order expressions a\t/anatl, b]Jbnbn, so that approximately

The formulae (9.71, 4, 5) are simpler than the exact Enskog formulae
for [A], and \ji]t. They also usually give values closer to the exact values of
A and ft. The neglect of the temperature-derivatives usually decreases a„,
btt, and so increases [A], and [/<],; thus it is similar in effect to taking a later
Enskog approximation. Mason* has indicated an extension of Kihara's
method to higher approximations, based on the retention of terms of higher
order in the temperature-derivatives.
• E. A. Mason. J. Chem. Phyt. rj, 75. 782 (1957). Aa regards the accuracy of the Kihara
approximations, see also M. J. OfTerhaus, Physica, a8, 76 (1061).
162 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: (9.8

9.8. The determinant elements a„, b„ for a gas-mixture

The expressions for the elements a^, b„ of the determinants j / t m ) , ^"°
discussed in 8.51 and 8.52 are simpler than those for the general complete
integral {F, G) of (4.4,12), by reason of the conditions (8.51,4) and (8.52,2).
111,18
a„ = {«r»>,o<*} - xJ[oi">,ai"]1+x1xt[ai">,ai<»]1,
• x j a ^ + x ^ a ^ (p,q>o), (9.8,1)
p, 1
*„ = x 1 x 1 [4 .ai'l,»+xiW ,^'].
•XIXI«P«+X?«M (p,q<o), (98,2)

• XjXja^ (p>o>a), (9.8,3)

and similarly for b^. As the quantities cf"\ 6°" do not involve the number-
densities if p + o, these appear in the above equations only as shown
explicitly.
The values adopted for of", a?1 in (8.51,3) lead to specially simple forms of
«0,. By the principle of momentum it follows from (4.4,8) that [V, J7], - o,
[V,F]t — o, whatever the function F. Thus, if q > o, a^ =» xtx,agp where
«*, = [<+4% 4'V
Again, from (4.4, 10), [mlCl + mtCitF]lz = o, so that
Ml[VvF\xt--M\[Vt,F)xr
Hence we find that d^~ M|[tf„oi"],,. (9.8,4)
Similarly, if? < o, «o,--AfJ^ai"],,; (9.8,5)
likewise if q » o, applying the same method a second time, aM ** x^a^,
where
^-M^oi-'+af)],,
--(M.M,)^,*,],,. (9.8,6)
Now, by (9.6,1, 6,8), a,, - tMMQHK*),
*'n - 8 ^ , ^ 6 ^ + 5 ^ 1 ) 0 ^ ( 1 ) - ^ ^ 1 ( 5 ^ ( 2 ) - 0 ^ ( 3 ) ) + 2 ^ / ^ , 0 ^ ( 2 ) ] .
^-40^)(2),
whence, by (9.8, 1), we know a„; similarly for a'_i_i, ali_i, anda_,_„
using (9.8, 2). Again by (9.6, 3, 5),
< - -8A/J M!{0iy(2)-i0iV(i)},
-Li - 8MjMJ{Oiy(2)-fOii>(i)},
«I-i - -SfAf.M,)![^-0^(0-(5^(2)-0^(3))-20^(2)].
9.81] GENERAL EXPRESSIONS 163
Let four constants A, B, c, E be defined by the equations

A s Qg(a)/sQftXi), B = (sQftX*)-iiiV(3)}/5«iV(0.1
''} (9-8,7)
c + i.aQtJ^VsfliVti).
E = Ar/8M,Af 2 QiV(i), (9-8,8)
so that A, B, c, depending on ratios of the functions (I (cf. (9.33, 2)), are
pure numbers. Then
«io = WE, (9.8, 9)
a^^-scMtkT/zMh, (9-8,10)
<*i_i = s c M ^ r ^ A / l E , (9.8,11)

4 - 1 = - 5(A*1 **,)» * 7 W - H - 2A)/E, (9.8, 12)

,
ai. = S * 7 { J ( 6 A ^ + 5 M ? ) - M 1 B + 2 M I M , A } / M 1 E , (9.8, 13)

and similarly for a'_i_t. Again, if [/<J]J denotes the first approximation to the
coefficient of viscosity of the first component gas of the mixture at the
temperature T, by (97,3) ,;, _ ^ r ^ ^ ,,,

with a similar expression for a l i _ i in terms of [//a]i-

The values of bu, *,_, and 6_,_1 may likewise be deduced from (9.6,14, 15,
16), using the definition of bpq given in 8.52. It is found that
*l-i = -VMiMifeflitO-IOBW}
--S*7T(f-A)/E, (9.8,15)
b'n = V^rfS^iQfttO + IM.QBXa)}
-5*7X1+ ^/^/^,)/8, (9-8,16)
% = 4^(2) = 5^/2^,],, (9.8,17)
with similar results for 6_i_i, 6_i_|.
These expressions for the quantities a w ,fcMgive first approximations to
the coefficients of viscosity, thermal conduction, and diffusion in a gas-
mixture.

9.81. The coeflSdent of diffusion Dn; first and second approximations

[D11L. [O1J1
The first and second approximations to Dlt, written [Dlt]v [Dlt]t, are
obtained (cf. 9.7) by deleting the limit sign in (8.51, 16), and putting m = 0
or m = 1. Thus
m 1_ 3 ^ 3*7" _ 3E
1
" J l " 2w» 0 a;, ~ r o n f f ^ p f j O i V O ) ~ 2»«o' ''
164 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [9.82
The second approximation, [D lt ]», is obtained by multiplying [D^ by the
factora^j/JiV-i*™. 'e- ty
«n «1-1i I /I «ii «io «1-1
«1-1 -1 / «01 «00 «0-1 (9-8i. 2)
|/ I «_ii «-io «-i-i
so that t^ia]i-Pii]i/(i-A), (9.81,3)
where «OO(«II«-I-I-«I-I)A • «!i«-i-i+«!-i«ii-a«-ii«w«o-i-
Using the values of the quantities aM given in 9.8, A is found to be given by

A SC (981 4)
QxxJ + Q.xS+C.x.x, ' '
where p, • M^/fr*^, p, m MtBl{jt^lt (9.81, 5)
pM = 3(M1-A/1)*+4Af1M1A, (9.81,6)
On • 3(^i-^i)*(S - 4 B ) + 4 ^ i ^ i A ( " - 4 B ) + 2P 1 P„ (9.81, 7)
Q, • Pi(6J»fJ+$M\ - 4AfjB+8M, Af, A), (9.81,8)
with a similar relation for Qt.
The general expression for the third approximation to D l t is extremely
complicated, and is not considered here.

9.82. The thermal conductivity for a gas-mixture; first approximation

The mth approximation to the coefficient of thermal conduction for a gas-

mixture was found in (8.51, 18) to be
[A]m - V**7Xx?mris/'ir+2x1x/,m1mt)-is/'^i +i$mi W*i?L,)/.*"<">.
The first approximation is thus
[A]i - ¥*t7TxJm,-la_1_1-2x1xt(m1»»,)-»«,_,
+xJ«s- 1 « u )/(«u«-i-i-«!-i). (982,1)
The values of an, a.^i, at_t have already been found in (9.8,12-14) m
terms of E, [/*,], and [/tjj,..., but it is convenient here to express an, a l 1 _x
in terms of [A,], and [A,l„ the first approximations to the coefficients of
thermal conduction of the constituent gases in the mixture at the tem-
perature T. By (9.7,4),
[AJi - ft[*Lfe)i - ijjj
4«! [ftL [AJ, - 404
i g Wi.
and the equations (9.81, 5) for p l t P , can be written
15AE iS*K . a .
9.84] GENERAL EXPRESSIONS 16S
Thus (9.8,14) becomes

with a similar equation for «!.i_i. Substituting the values of a u , «,_j, a_i_i
into (9.82, 1), we obtain
X
m = I Q I [ A I ] I + X I Q I [ A » 1 I + X I X «QM /Q8, .)
1 Jl w-
xfQ, + x|Q f +x 1 x l Q 11 ' *•'
where Qj, Qt, Q „ are given by (9.81, 7, 8) and

Q
» = ^ T {p>+ p * + ( ' l ~ 4B ~ 8 A ) M I M*>- (9-82> S)
If n, or nt is equal to zero, this first approximation to A reduces, as it
should, to the first approximation to the conductivity of a simple gas.
9.83. Thermal diffusion
By (8.51, 17), the first approximation [k 7 ], to the thermal diffusion ratio is
given by
nj i = 5 *iA*r*(gQi<»-i-i-flo-itfi-i) + x,Mf*(a 0 _ 1 a u -api'h-i)
2 «n«-t-i-«i-j
The first approximation to the thermal diffusion factor is [a lt ]i a [kfJj/XjX,.
Hence, using the values of <JM given in 9.8,
x
[«iJi = ^ = 5c , A-^*» , (9.83,1)
1 D VV
"" X,X, xfQ1 + xjQt + X,X 1 (} li ' "" '
where Ql( Q„ Q„ are as defined in (9.81, 7, 8) and
8, s M I P 1 - M 1 { 3 ( M 1 - M 1 ) + 4M 1 A}, (9.83, 2)
s, . M , P I - M 1 { 3 ( M 1 - M I ) + 4 A/ 1 A}. (9.83, 3)

9.84. The coefficient of viscosity for a gas-mixture;firstapproximation

Mi
By (8.52, 10), this is given by
M i = f*7Xx?*-i-i + xl&n - 2xlx,6l_1)/(6,,6-,_1 - *?_!). (9.84,1)
With the values of i w found in 9.8, this becomes
r«1 — x1R1 + xtR1 + x1x1R1> . _ .
Wl W
~ x!R1/t//l], + x!R 4 /^ a ] 1 + x,x t R l t ' -° 4 > 2)
where R, = f + M, A/A/„ R, m } + Mt kjM„ (9.84, 3)
"it = E/2L"Ji + E/2[/<d1 + 2(| - A), R„ • E/2^,], \jl J , + 4A/ 3 EM, Af,.
(984. 4)
166 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION (9.8S

If x, or x, is equal to zero, this expression reduces to the first approxi-

mation to ft for the corresponding simple gas.

9.85. Kihara's approximations for a gas-mixture

By (9.71, i) and (9.8, 7)
c 2_T_dM!) B _2 . d /rdQffl(i)\ 8 .
B (98Sl,)
SQSKi) dT ' 4" sm^dTV^ dT ) •
Hence, if (cf. 9.71) terms of third order in the temperature-derivatives are
neglected, we can approximate in the expression (9.81,4) for A by taking
B « }. In this case (9.81,7, 8) take the simpler forms
Q,t - 6(M 1 -M I )* + 32M,MtA + 2P1pt, (9.85, 2)
Q, - 2P,(3M\ + M\ + 4M, Mt A), (9.85, 3)
which is Kihara's approximation. Kihara took B — } also in the formulae for
[A]j and [kT]„ so that he used the values (9.85, 2, 3) for Qlt, Q„ together with

Qi. - ^ {P, + r,+8(i -A)MXMJ. (9.85, 4)

For most molecular models B < f. It can be shown that the effect of
increasing B to } is to increase the magnitudes of [Dlt]lt [A], and [k r ] ( . The
effect of taking later approximations is likewise to increase [Z>u]i and [A]„
and usually [k r ], also; hence Kihara's approximations are often closer to the
true values than Enskog's, and they are easier to calculate. However, whereas
in 9.81 the approximation B = | is made in a correcting term already known
to be small, in 9.82 and 9.83 it affects the whole of [A], and [k r ],; hence in
these it is a correspondingly cruder approximation. In particular, the values
of [A], calculated from it are sometimes found appreciably to exceed the
exact values of A.#
• E. A. Mann and S. C. Sixena, J. Chtm. Phyt. 31.511 (1959)-
 10
VISCOSITY, T H E R M A L C O N D U C T I O N ,
A N D DIFFUSION: THEORETICAL FORMULAE
FOR SPECIAL MOLECULAR MODELS

10.1. The functions 0(rj

The expressions derived in Chapter 9 apply to any type of spherically
symmetrical molecule possessing only energy of translation, but they involve
the functions ii^r) of (9.33, 2), which can be evaluated only when the law
of interaction between molecules is known. These functions will now be
determined for certain special molecular models.
It is convenient first to introduce a non-dimensional variant if^f) of
Qt&ir). Let <rlt be a conveniently chosen length, lying within the range of b
that contributes most to the integral infli'i(r).Then <Hftf/) ' s defined by*

so that (cf. (9.33, 2))

*Tft(r) = p-«y + */(i -coz>X)(bl<rxi)d(bl<Tlt)d{f*). (10.1, 2)
The formula corresponding to (10.1, 1) for like molecules is

In terms of the quantities T^^\r), the expressions (9.7, 1), (9.81,1) for
Mn Wi a n d lAdi become

• The quantity *18(r) is twice the Ji"fi(r) used by J. O. Hirschfelder, R. B. Bird and
E. L. Spotz, J. Chem. Pkys. 16, 968 (1948). The notation a'*-" is used by J. O. Hirsch-
felder, C. F. Curtiss and R. B. Bird (The Molecular Theory of Gases and Liquids, chap. 8)
to denote the ratio of fl"j(r) to its value for rigid spherical molecules of diameter c,,.
The use of their notation introduces a number of additional numerical factors into the
general formulae.

I 167 1
168 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [10.2

RIGID ELASTIC SPHERICAL MOLECULES WITHOUT

FIELDS OF FORCB

10.2. An encounter between two rigid elastic spherical molecules of

diameters o-,, er, occurs only if b < <rlf> where <r„ - J(<rt + <r-t). At an
encounter, by (3.44, a), b - 0-Itcos}x> and bdb - - J(rJ,sin^</^. Hence if
<rlt is identified with the <rlt of 10. i, by (IO.I, 2),

imtr) - ! j \ i -cos'x)ainXrfxJ,o%-,*l^+•4y,)

•;[ a ~/^ ( , + ( " I ) 0 ] ( '" , ' , ) ! ' (I02,I)

In particular, HT^{i) - 1, *rft(2) - 2. Clearly ^"(r) - lT8(r).

Again, by (9.33,4)

-I»«*[»-J^7(I+(-IW]. (IO-2.»)
In particular, $\J is equal to ncr\t, the ordinary collision cross-section.

10.21. Viscosity and conduction for a simple gas

By (10.1,4), thefirstapproximations \jt]u [A], to ft and A for a simple gas are
given in terms of the molecular diameter <r by

IA-A'SBP. W . - I M . ' . - ^ Q V <•"•••>

Later approximations to A and ft are obtained by multiplying the above
values by successive partial sums of the respective series
1+^ ( - ^ ) ^ - ^ + ^ 1 ( - ^ ^ - ^ - ^ + . » , (10.21, 2)
| , B ,
1 +MW/* w*+M*8D /» W" +». (>o.2i. 3)
(cf. (7.51,14), (7.52> 8)). In the present case each term of these series is a
pure number, since the factor ijir*^(kTjm) occurs in each element of the
determinants si,91, and so cancels throughout.
From (9.6,8-13,16-21), using the above expression for &\r), it is easy
to deduce that
«it - - J«ti. <*u - tt«n. «13 - -1V11. «» " -fH«n. <hx - \Hl<»iv
*»--i*n. *«-W*n. *i» = -A*u. bn--Wbu, 6M-W6„.
10.3] FORMULAE FOR SPECIAL MOLECULAR MODELS 169
Hence the first three terms in the above series are
i (in)2
H + 6 5 " -i-J~— = I+002273+0-OO209
/J y
44 88x66951
and 1+ - - - + 0 - 0 - - - = 1+001485+000103.
T J J
202 808 x145043
The fourth terms of these series have also been calculated,* though the
results necessary for the calculation are not given in 9.6; these terms are
0-00031, 0-00012,

so that, to the fourth degree of approximation, the expressions for A, /t are

[A]4 = i-025i3fA]„ [/<]4 = IOI6OO[//]„ (10.21, 4)

and [A]4 = 2 - 5 2 2 ^ ] ^ . (10.21,5)

The convergence of the above approximations is so rapid that these values
may be taken as correct to within one-tenth of 1 per cent.f

10.22. Gas-mixtures; [D„]„ [£>„]„ [A]„ [k,.],, [ft],

The results for gas-mixtures, found in 9.81-9.84, can be completed in the
present case by means of (10.2, 1), from which the following values of A, B,
c and E are obtained:
A = 2/5, B = 3/5, c = 1/5, (10.22, 1)

E
I 2kTm0 \i 1
= -»*-»* I 5~r- (10.22, 2)

MOLECULES THAT ARE CENTRES OF FORCE

10.3. Another important simple molecular model (first suggested by
BoscovichJ in 1758) is a point-mass surrounded by a field of force. During
interaction, let two molecules w,, mt exert forces P, — P on each other.
Let the position-vectors of the two molecules at the time t be ru r f ; then
the position-vector rn of the second relative to the first is Vi — r^ The
equations of motion of the two molecules are
tnlr1 = — P, ms¥t = P.
Hence tnxtnt¥tl = m^mj^x - f,) = (ml + m,) P = tn0P,
* A numerical error in this part of Chapman's calculations was corrected by D. Burnett,
Proc. Ijond. Math. Soc. 39. 385 (1935).
+ The values of A and /1 were recalculated by C. L. Pekeris and Z. Altermann (Proc. nam.
Acad. Set. U.S.A. 43. 998 (1957)) by a different method, applicable only to rigid elastic
spheres; they found A = 102513^],, /i = 1016034(/1)1.
J R. G. Boscovich, Philosophiac naluralis thtoria (1758).
 170 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION! [10.31
so that the motion of m, relative to w, is the same as the motion of a particle
of unit mass about a fixed centre of force m„ Plmt mt. If polar coordinates r, 0
be taken in the plane of the orbit, the equations of angular momentum and of
energy for such a particle have the forms
«= const. = gb, (10.3, 1)
J(r* + f*d*) + m0vlJir)lmlmt = const. = lg*. (10.3, 2)
H e r e ; and b have the same meanings as in 3.41, 3.42, and r It (r) is the
potential energy (taken to be zero at r = 00) of the force P.
By eliminating the time between (10.3, 1, 2), we get the differential
equation of the orbit,

« H - "-=£?•
2

This has the integral

(since y* = mlm2gtlzmllkT), if 0 is measured from an axis parallel to the

initial asymptote of the orbit. The apse r = R of the orbit corresponds to
*/tf-o.i.e. m „
=1 (, 3,4)
W ~*' °-
if this equation has more than one root, R is the largest root.* The angle
between the asymptotes of the orbit is twice the value of 0 corresponding to
r «• R. Since the angle x ' s the supplement of this angle,

"-'-JTW-aPM"'* <»**
This shows that x depends only onft,on kTg* and on the force-law. Thus
the integral 1?~\!i(r) of (10.1, 2) depends only on kT and the force-law; in
particular, it does not depend on the molecular masses.

10.31. Inverse power-law force

Suppose now that P is a repulsive force given by the power laws
P^Kjr, Vttfi-KuHv-t)**-1. (10.31,1)
If pure numbers v, u0, vM are defined by
vab/r, v0sb{2kTf*lKu)^-«, v^sb/R, (10.31,2)
* When (10.3, 4) hat more than one root, bound states of the same energy can occur, in
which the molecules remain indefinitely at a finite distance from one another. Such
bound states are ignored here.
10.31] FORMULAE FOR SPECIAL MOLECULAR MODELS 171
(10.3, 5) becomes, on substitution for vl2 from (10.31, 1),
X=7 2
'- j'o''[l-Vi-^^J 1 'dV- ( I °- 3 I , 3 )
The quantity i'w is the (unique) positive root of the equation
2 /v\"-1
1-v2 (-} =0. (10.31,4)
v-i K /
Hence x depends only on v and t»0. The corresponding formulae for an
attractive force (Kl2 negative) are found by replacing K12 by |/c12| in (10.31, 2),
and reversing the signs of the terms involving zj{v— 1) in (10.31, 3, 4);
t'oo is the least positive root of the modified equation (10.31, 4).
The integral 5¾ of (9.33, 4) can now be transformed as follows:

$[J = 2TT (°°(i-cos'X)bdb

\ vw-n
( K
4$p) M»). (IO-3L 5)

where 4 ( 0 = 1 ( 1 -cos'x)o 0 </i/ 0 ; (10.31,6)

/j,(e) is a pure number depending only on / and v. The integrals il'-lt(r) can
now be expressed in terms of gamma functions, for (cf. (9.33, 5))

°8M = ( ^ ¾ ¾ i (2f-) i(V " ,, ^^o % -^ 2r+3 ""- ,,rf?

%-UlJ U) ^ ) ^ ( ^ 2 - - ) . (.0.31,7)
The length <rw introduced in 10.1 may here conveniently be defined by
cr12 = (* l 2 /2*7y<-» (.0.31, 8)
so that it depends on the temperature, as well as on the law of force. Then
(10.1, 1) becomes 1 , \
^,(0 =4 ( 0 ^ + 2 - ^ ) . (10.31,9)
Clearly if'xi\r) has the same value as 1(r{l\,(r) if v has the same value in both.
From (10.31, 7) and (9.8, 7, 8) the following are found to be the special
forms of A, B, c, E appropriate for molecules which interact in accordance
with (10.31, 1):
. (3"-5)4t(0 B= (3^-5)(^+0 c "-5
&-*)*&)' s(v-i)» ' 5(1,-1)-

The numbers A, B and c thus depend only on v.

172 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [10.32
The constants A^y) have to be evaluated by quadrature.* The following
table contains values of A^v), At{v) for several values of v, and also some
values of A:
Table 3

V ^,(") *M A

5 0421 0436 0517

7 0385 0-3S7 0493
9 0381 o-33* 0478
11 0383 0319 0-465
"5 0-393 0309 0450
at — 0307 —
*S — 0-306 —
00 o-J 0-333 04

Rigid elastic spherical molecules may be regarded as a limiting case of the

present model, corresponding to v = 00; the force vanishes if r > <rlt and
becomes infinite at r = trlt. These conditions are satisfied if
P oc lim (<rlt/rY.
p —*«o

10.32. Viscosity and conduction for a simple gas

The first approximations to A and /1 for this model are given by

Mi - 5(kmTln)i (2kTlK)^-^jSr ( 4 - ^ ) .4,(,), [A], = §[/<],cv.

(10.32,1)
Both [A], and [/i]t are proportional to T', where

The limiting value of the index s as v tends to infinity is \, the value already
found for rigid elastic spherical molecules. For smaller values of v the
index is larger, becoming equal to unity when v = 5.
The exact values of A and ft are obtained as in 10.21 by multiplying the
above first approximations by the series (10.21, 2, 3). The sums of these
series are pure numbers; by an argument similar to that used in 10.21, it may
be shown that they depend neither on the temperature of the gas nor on the
• -Ms). A^S) were calculated by J. C. Maxwell (Collected Paptrt, vol. a, a6), and by
K. Akhi and A. Tanakadate (aee D. Enskog, Arkiv. Mat. Attr. Fyt. 16,36 (1921)); A,(v)
and A^v) were given by S. Chapman (Monehetttr Lit. and Phil. Soc. Mimoin. 66, 1
(1921)), for v — 5,7,9, n , 15; -1,(21) and ^,(25) were found by J. E. (Lennard-)Jonea
(Proc. R. Sot. A, 106, 421 (1934)), and -1,(9) and ^ 9 ) by H. R. Haste and W. R. Cook
(Proc. R. Soc. A, 135, 196 (1939)). The value* given in the text are thoae of Chapman
and (Lennard-)Jonea.
10.33] FORMULAE FOR SPECIAL MOLECULAR MODELS 173

force-constant of the molecular fields, but only on the force-index v. The

second terms of these series can be determined, using the result

OH*) V I'-IL,'
which is an immediate consequence of (10.31,7); substituting in the
equations of 9.6, we get

*» - " ( J " l £ i ) *»• "» = (l6-ITTi + (TTIji) "»'

6
^ = -(^-^,)^ "=(^-^7+(7^)6»-
This leads to the second approximations

M.-M.(' + « ^ ^ £ ^ ) . (I0.32,2)

^-^( ,+ ^i?SW))' (,032,3)

which are identical with the first when v = 5; for values of v between 5 and
infinity, the numerical multipliers of [/1^ and [A]t lie between unity and the
values found above for rigid elastic spherical molecules. The ratio A//tc„
also lies between 2-5 and the value 2-522 appropriate to elastic spheres.

10.33. Maxwellian molecules

When v = 5 the formulae of 10.31, 10.32 become specially simple. The
simplicity of this case was first perceived by Maxwell,* who found that it
was possible to develop the theory for a gas whose molecules are centres of
force repelling proportionately to r~5, without determining the velocity-
distribution function/. For these ' Maxwellian' molecules, by (10.31, 2),
gbdb = (m^Jm^^i v0dv0, (»°-33. ')
so that the element gbdb does not depend on g, but only on c0. The integrals
of 9.3-9.5 accordingly take simpler forms: thus, for example, from (9.31,14),

fJHMfgbdbd?

= I ** (£-*) JOWis)"* (i - cos x) v9dv0,

• J. C. Maxwell, Collected Papers, vol. 2, 4»; Phil. Trans. R. Soc., 157, 49 (1867).
174 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [10.33
and so, by (9.32,1-3),

{sTlst)* n-»jjj{H^o)-H^X)}gbdbded9

= - 3" ( ^ ) J { » - 2M1 Mtst( 1 - cos x)H (1 - cos x) «>o<*V

Now the coefficient of s"t? in the expression on the left is equal to
[SH*»*i.SfC*»*i],.:
this accordingly vanishes unless p = q. Similarly from 9.5
[5^(^)^,5^(^)^ = 0
unless p ™q. An analogous argument shows that the same is true for
[Sf\V*) «&, 5J»(<if») 3 ¾
Using these results, and the known values of aq, /}„, the sets of equations
(7.51, 8), (7.52, 4) reduce to
a a a a
i u"*i» » n = °> « » « » " 0 . •••!
=
KKi Pv K^» — °t Kb» = 0, •••,
so that all the coefficients ap, bp vanish, save «,, 6,. The formulae for A, p
therefore reduce to the first approximations given in 10.32, i.e.

>-m) l kT

It may similarly be proved that all the quantities aM, bM of 8.51, 8.52
vanish unless/> = qorp — —q, and hence in the series (8.51,1,2), (8.52,1) all
the coefficients ap, dp, bp vanish save a„ «_,, d0, 6, and &_,. Consequently
the expressions for the coefficients of diffusion, viscosity, and thermal con-
duction for a gas-mixture become identical with the first approximations
given in 9.81, 9.82 and 9.84, while the coefficient of thermal diffusion
vanishes.
These results imply that for a simple gas the expansions in series of the
functions A, B given in (7.51, 2), (7.52,1) reduce to the single terms
,4 = ^ = «,(*-«*) 9?, (10.33,2)
B = 6,b<" = 6 , ^ . (10.33, 3)
Similar results hold for a gas-mixture.
Since A and B satisfy (7.31, 2, 3) it follows that with suitable values of
•»•*»• -•^/{(•M)*}-/"^-!)*. (10.33.4)
*,»/(*&) =f»&. (10.33,5)
 10.331] FORMULAE FOR SPECIAL MOLECULAR MODELS 175

A direct proof of these results was given in earlier editions of this book.
Similar results hold for a gas-mixture.
For a simple gas, by the results of 7.51, 7.52 and 9.7,

<*! = - ¥ - . A-t. «11 - *H - 5 " " / » *

Hence «, = - ^ , h = -£f (10.33, 6)

and (10.33, 2, 3) become

A=
i^~^' * = {T™- <IO-33.7)
Thus (cf. (7.31, 1)) for Maxwellian molecules/'" is given by

yu> = _r> |3£ ^ * ( V . _ j ) V . v i „ r+?£«fr:Vc.).

(10-33. 8)

10.331. Eigenvalue theory

Equations (10.33, 4' 5) a r e particular results of a general eigenvalue theory
which has been extensively discussed, in particular by Kihara, by Wang-
Chang and Uhlenbcck, and by Waldmann.* For a uniform gas under no
forces, let , , mi/ , . . . .
/=/(0)(,+0), (10.331,1)
where <I> is small, be a non-normal solution of the Boltzmann equation. Then
approximately _»
-„*/(4>)=/«»^, (10.331.2)

as can be seen on substituting (10.331,1) into the Boltzmann equation, and

neglecting products of functions ¢ . If 4> is suitably chosen, it decays
exponentially like e~tT, say, so that (cf. 6.61) r is the relaxation time for ¢ .
Then (10.131,2) becomes ,T/*>\ lirmA , *
v 3i
' ' n*I(Q) = r-ywQ. («o-33i>3)
In (10.3 31,3) T _ 1 may be regarded as an eigenvalue of the collision operator
/ ; <I> is the corresponding eigenfunction. Thus, for example, (10.33, 4, 5)
0
imply that for Maxwellian molecules ("Jf* — f) V and '&€ are eigenfunctions
of / ; the corresponding relaxation times T are — ajn and bjn, or (cf.
(10.33, *>)) 3fll2P a n d flip- These are the relaxation times effective in heat
conduction and viscosity respectively (cf. (6.6, 5, 7)); the results derived by
the relaxation approach of 6.6 become identical with those of the exact
theory for Maxwellian molecules.
* T. Kihara. Imperfect Gases, section 21; C. S. Wang-Chang and G. E. Uhlenbeck,
University of Michigan Project M999. 1952; L. Waldmann, Handbuch dtr Phytik,
vol. 12 (Springer, 1958), 367-76.

I
176 VISCOSITY, THERMAL C O N D U C T I O N , AND D I F F U S I O N : [10.34

The general eigenfunction for Maxwellian molecules is found to be

• t r - S K i C * " ) * «(*.?). (10.331.4)
where Yj(6,<?) is a surface harmonic function, of order/, of the polar angles 0,
9 that specify the direction of V. The corresponding relaxation times are
small compared with fijp when either r or / is large.
For more general molecular models, the determination of the eigen-
functionsisamatter of considerable difficulty.* However, just as (10.33,2,3),
which are exact only for Maxwellian molecules, none the less represent first
approximations yielding values of A and /1 close to the true ones for other
molecular models, one may expect that the first one or two, at least, of the
eigenfunctions for Maxwellian molecules give reasonable approximations
to the eigenfunctions for more general models. This provides some measure
of justification for the relaxation approach used in 6.6 and 6.62.

10.34. The inverse-square law of Interaction

If a large proportion of the molecules of a gas are ionized, electrostatic forces
play a dominant part in the molecular encounters. It is therefore of interest
to consider the case when the force P between pairs of molecules satisfies
the equation j»_Wf (,0.34,,)
(where c lt et denote the electric charges of the molecules), ignoring any
small part of the force which varies more rapidly with r.
If «„ et have the same sign, the variable e0 of (10.31, 2) is given by
v0 = 2kTbf*lelet, (10.34, 2 )
n
and the angle ^ by x " ~2 | (1-01 I dv,

where vM is the positive root of the equation

The expression for x •* integrable in the present case, the result being

X-a-in-'^-^. (10.34.3)
v*— 1
whence it follows that cosy =-¾— .
vjl+i
* In addition to discrete eigenvalues corresponding to regular eigenfunctions. there may
also in special cases be a continuous spectrum of eigenvalues, possibly associated with
singular eigenfunctions; see H. Grad, 3rd International Rarefied Gat Dynamics Sym-
posium Report, p. 26 (Academic Press. 1063). Grad also points out that for molecules
repelling like r~", where v < 5. the eigenvalues may be indefinitely small, corre-
sponding to indefinitely great relaxation times.
10.34] FORMULAE FOR SPECIAL MOLECULAR MODELS 177
If ¢,, et have opposite signs, e,e 2 in (10.34, 2) is replaced by |«i«,|, but the
final expression for cos x ' s unaltered.
We accordingly find that, in either case,

In particular, # V = zn V^fjA In (1+1^), (10.34,4)

where vm denotes the upper limit of i>0. In the previous discussion the upper
limit of i>0 was taken as infinite: but this is not a valid approximation here,
as it would give infinite values for ¢$, 0 $ , and the corresponding values of
the coefficients of viscosity, diffusion, and conduction of heat would all
vanish.
The difficulty arises because electrostatic forces, being proportional only
to the inverse square of the distance, decrease with distance much more
slowly than the ordinary forces of interaction. Hence important contri-
butions to the integrals (10.34, 4, 5) are made by encounters in which the
mutual distance of the molecules is always large. However, a molecule at a
large distance from a given molecule will also be under electrostatic attrac-
tion or repulsion of many other molecules, so that these' distant' encounters
are not really binary; but throughout our analysis account is taken only of
binary encounters. When ' inverse-square' forces are dominant, the mole-
cular fields interpenetrate one another to such an extent that all encounters
might be regarded as multiple. It is, however, possible, in certain circum-
stances, to derive approximate values for the coefficients of viscosity, con-
duction, and diffusion, considering only binary encounters.
The average value of fl, or m0 Mx Mtg2J2kT, for pairs of colliding mole-
cules is 2 (cf. (5.3, 5)), so that i>0 in (10.34, 2 ) *s ' n general comparable with
4A7,i/|e1«2|; thus, by (10.34,3), x '« sm^l if * is large compared with
le^l/^kT. The mean distance between pairs of neighbouring molecules is
comparable with n~i. Suppose now that
|*i*,|/4*r « n-i. (10.34, 6)
Then x is small unless b <S n~i, i.e. large deflections are possible only if two
molecules approach much closer than n~i. The distant encounters pro-
ducing small deflections are thus more numerous than the encounters
producing large deflections; they are also more important, as the divergence
of (10.34, 4, 5) for large y01 implies. As noted earlier, the distant en-
counters cannot strictly be regarded as binary. However, because the
deflections due to distant encounters are small, their effects are roughly
additive; that is, the change in velocity produced by one such encounter can
 178 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [10.34

be calculated without taking into account the forces simultaneously being

exerted by other relatively distant molecules, just as if there were a succession
of binary encounters.
The argument breaks down when b becomes too large; molecules at a
sufficiently great distance from a given charge are shielded from its field by
a polarization of the intervening gas. Debye* showed, on the basis of certain
approximations, that, as a consequence of a similar polarization, the potential
of a point-charge in an electrolyte varies as (i/r)exp( — rjd), where d is a
distance known as the Debye shielding length. If the Debye potential is
applicable in an ionized gas (it strictly applies only to the field of a charge at
rest) the inverse-square force can be regarded as ceasing to be effective at
about the cut-off distance d. That is, the binary-encounter formulae
(10.34,4, 5) can be used, assuming u0 to have an effective upper limit vn,
whose correct order of magnitude is given by taking b "din (10.34, 2 )-t
In a binary mixture the Debye length is given by
4mP(»t«i+»««!) - *r. (10.34.7)

Since, by (10.34, 6), ^kTjn^\eyet\ is large, d> n-*; thus molecules not its
immediate neighbours contribute appreciably to the deflection of a given
molecule.
In actual calculations it is convenient to set
vn - 4^*7Vl«i'*|. ( IO -34,9)
equivalent to taking b •» d in (10.34, 2) and replacing f* by its mean value 2.
Uncertainties in vol are in any case relatively unimportant, because the
functions appearing in (10.34, 4> 5 ) v a r v o n ' v slowly with vn. If the cut-off
distance in (10.34, 9) were taken as n~i instead of d, the effect would, by
(10.34, &)> m n o caae reduce 0fi| and 0¾ by as much as one-third. J
With vol given by (10.34,9)> the constants ^1(2), A^z) of (10.31, 6) are
found to be
AX(2) = In (1+ «fc). Afr) « 2 (in (1 + <&) - - ^ } .
(10.34,10)
• P. Debye end E. HUckel, Phyt. Z. 24, 185, 305 (1933).
t R. L. Liboff (Phyt. Fluids, a, 40, (1959)) calculated the transport integrals for the
Debye cut-off potential, obtaining results sgreeing with those given here so far as the
leading (logarithmic) term is concerned.
X The first proposal of a cut-off distance to overcome the divergence difficulty in electro-
static interaction was made by S. Chapman, Mm. Not. R. Attr. Soe. 8a, 394 (19*1):
he took the cut-off distance as n~*. The first discussion of electrostatic shielding in
ionized gases was given by E. Persico, Mm. Not. R. Attr. Soe. 86, 294 (1936). The
general adoption of the Debye length as the cut-off distance is largely due to Spitzer;
cf. R. S. Cohen, L. Spitzer and P. McR. Routly, Phyt. Rev. 80, 330 (1950), L. Spitzer
and R. Hlrm, Phyt. Rev. 89, 977 (1953), and papers referred to in these. At high
temperatures ( > 10* deg. K.) the cut-off distance is modified by quantum effects.
10.4] FORMULAE FOR SPECIAL MOLECULAR MODELS 179

In terms of these, the first approximations to the coefficients of viscosity,

conduction, and diffusion are expressible in the forms
Ml =
8 (*?)' ffPfh**' [A]l = *M>^ (,0
-34. ")

A simple gas composed of molecules all with charges of the same sign is not
met in nature. However, since cos y does not depend on the signs of ev «„ the
formulae (10.34, 11) can be taken to apply to a binary gas composed of equal
numbers of ions of equal mass and charge, half positive, half negative.
From (10.32, 2, 3) [fi]t = I - I S O J I , [A], = 1-25^],; also if mx = m„
e, = -et, «! = nt one finds that [Dls]t = i'i5[/>u]i- Thus the errors of first
approximations to the gas coefficients are considerably greater for the
inverse-square law than for large values of the force-index v. Further
approximations have been considered by Landshoff • in the important case
when mj/m, is negligibly small (corresponding to ion-electron mixtures).
Taking e, => —Ze2, n2 = Znv corresponding to Z-fold ionized atoms in an
electrically neutral mixture, he found the values shown in Table 4 for the
ratio of the mth approximation to Dlt to the first approximation. These
figures show that, whereas the second approximation differs considerably
from the first, the corrections added by later approximations are relatively
small.
Table 4. Values of [D 12 ] m /[Z)„] I for a gas
of electrons and Z-fold ions

m Z = 1 Z = a Z = l Z » 00

2 i-93*o 23180 25290 3*5

3 1-9498 »•3210 25290 33906
4 1-9616 23256 *-53<>5 3 3945
S 1-9657 23268 25308 33950
QO
— — — 33953

MOLECULES POSSESSING BOTH A T T R A C T I V E

AND REPULSIVE FIELDS

10.4. The Lennard-Jones model

As noted in 3.3, in actual gases the force P between uncharged molecules is
repulsive at small distances, and more weakly attractive at larger distances.
• R. Landshoff. Phys. Rev. 76, 904 (1949) and 8a, 442 (1951). Landshoff's cut-off
procedure differed somewhat from that described above, but the change does not
greatly affect his numerical values.
180 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: (10.4
This behaviour is most simply represented mathematically by the Lennard-
Jones model,* for which p _ - / „ „ » /,• /,„ . ,\
f = KuJr-Knjr, (10.4, 1)
where v > v'\ thus the interaction approximates at small distances to a
repulsion KtJf, and at large distances to an attraction K'lt/r"'. The potential
energy corresponding to (10.4, 1) is

Let <Tj,, eu be defined by

Then cr„ denotes the'low-velocity diameter' such that ^(c„) = o; e„ is the

depth of the potential well, i.e. — e„ is the numerically greatest negative
value attained by v{r). In terms of <rlt, e11( (10.4, 2) can be written
Hr) = fatiivitlry-i - (crlt/r)"'-»}, (10.4, 4)
where £. _ j£,--^ . (10.4,5)

The model is usually specified in terms of the powers appearing in the

expression for v(r); thus (10.4,1) corresponds to the Lennard-Jones v—i,
v' — 1 model.f For a simple gas, *,„ K'1%, <rxl, e„ in (10.4,1-4) are replaced by
*, K ' , <r, e.
For fixed values of v and v' (10.4, 4) can be written in the form
I'M - • u ^ W ' i i ) . («o-4.6)
where "f is a non-dimensional function. Whenever v(r) is expressible in
this form, the transport coefficients conform to a law of corresponding states.
The non-dimensional quantities #"„ of (10.1, 1) must be functions of kT/eu
alone, this being the one non-dimensional combination which can be formed
from the parameters <ru, e„ and kT, on which alone 1P\t depends (cf. 10.3).
There is a similar result for iTx\ thus for a simple gas, by (10.1, 4),
. . (kmT)i ... (AT)* , ,
IA1I (10,4,7)
M'TtiVmr "»JMA7W
where F,, Fx are two non-dimensional functions. On dimensional grounds,
later approximations can modify (10.4, 7) only by altering the functions
Ff, FA, so that equations similar in form to (10.4, 7) also apply to the exact
• Thit model, first introduced by J. E. (Lcnnard-)Jones (Proe. R. Soe. A, 106, 441, 463
(1924)), ha* been widely used in discussing the equation of state of gases, and crystal
properties, as well as in the theory of the transport phenomena. It was a generalization
of the model (see p. 18a) earlier introduced by W. Sutherland, Phil. Mag.. 36, 507
(«893);i7. 380(1909).
t The notation differs from that of Note A (p. 39a) of the 1951 edition of this book,
which referred to (10.4, 1) as giving the v, v' model.
10.41] FORMULAE FOR SPECIAL MOLECULAR MODELS 181

values of fi and A. These equations embody the law of corresponding states

for a simple gas. The quantities e and a in (10.4, 6, 7) are, of course,
constants, not (like the <r,t of (10.31, 8)) functions of the temperature.
Similarly in a binary mixture, by (10.1, 5),
[D l ! ] l = (kT)l/{(m0M,Mt)i n<r\tFD{kTleu)), (10.4, 8)
where FD'isz further non-dimensional function. This gives a law of corre-
sponding states for Dlt, to a first approximation. However, later approxima-
tions make the function FD depend somewhat on the parameters njnt, mjm^
(rj<rlt, <rt/cri2, ej/e,,, e^e^ as well as on kT/elt, even if the function "f in
(10.4, 6) is the same for like and unlike molecules. Thus there is no simple
exact law of corresponding states for Dlt; the same applies, but with greater
force, to ft, A and k r for a gas-mixture.

10.41. Weak attractive fields

When in the Lennard-Jones force-law (10.4, 1) the attractive part of the
force is weak, it is more convenient to replace (10.4, 7, 8) by
M i = D»Ji/^(«/*n Wi = [AJi/^(«/*r), («o-4>. 0
[Oid, = Wu\y^D(eltjkT), (10.41, 2)
where t/«0]i. [A0]i> [(0n)o]i denote the first approximations to /t, A and Dxl
obtained when the attractive force (K'/f' or K[JT"') is neglected. As the
notation implies, the (non-dimensional) quantities y^ and \jrD are functions
only of e/kT and elt/kT; the same function appears in the expressions for
M i and [A]t because both depend on the same collision-integral £^(2).
For small values of K', K'it, the attractive fields can be regarded as small
perturbations, whose effects on ^,,, ifrD depend linearly on the strengths of
those fields, i.e. on K', K'1%. In virtue of the form of the expression (10.4, 3)
for elt we can therefore write
j, _ . . 'i(".">i« / I 0 „ ,x
Vl>
~ (/-1)(/c„)<-- 1 W'- , >(Arf-''«'- 1 ' v 4
'3>

and similarly ^ = 1 + ( / _ t ) J&Zi*kTp-*»-i>. ( I 0 - 4 ' . 4)

where t',(i>, v') and it(v, v') are pure numbers. Hence (10.41, 1,2) reduce to
the forms
M i = [/<oli/(' + ST^-^-% [A], = [AJ,/(i + ST-+**-% (10.41, S)
[£»]« = [(0i.)o]./(«+s n r-^-'M'-"), (io. 4 i, 6)
where
0 »i('-,Of' «, ^."'Xt
l
{V - I)/cO'-lW'-"^'-''')*-"* * (v' — 1)(K.J/y-Wr-*>lfi>-*»-V•
(»0-41, 7)
182 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [10.41
The evaluation of the constants t'„ i, is fairly simple in two special cases.
The first is that of the Sutherland model, a smooth rigid elastic sphere sur-
rounded by a field of attractive force. This may be regarded as a limiting case
of the Lennard-Jones model, v being taken as infinite, and *c„ SO adjusted
that when r > <r„ ( a , , now being the sum of the molecular radii) the force
Klt r~* is zero, and when r < <r„ it is infinite. In the limit, (10.41,5, 6 and 7)
becomc
Mi-WiAi+s/r), [A]»-[Aj,/(i+s/r), (10.41,8)
[*>idi - [ ( J > » U / ( > + s i t i n (10.41,9)

Note that 5 is proportional to K'I(V' — 1)0* - 1 , which is the potential

energy of the mutual attraction of two molecules when in contact; this
gives S an independent physical interest. A similar remark applies to S „ .
For the Sutherland model, molecules which would not have collided in
the absence of the attractive field are deflected by it through an angle x
proportional to K'U ; for these molecules the factor 1 — cos'x appearing in the
integrand of 0 $ is of order K[\, and so is neglected. This means that in
evaluating 0 $ we need only consider encounters such that b < ¢7-,,. For such
encounters, the apsidal distance of the relative orbit (i.e. the least distance
between the centres of the colliding molecules) is equal to <rlt; also in
(10.4, 2) the repulsive term in v(r) is to be omitted when r > <r„. Hence,
from (10.3, s)

where u =• b/r, v w «= b/crlt. Thus to the first order in K'U

where Xo >s ™ e deflection 2 cos - 1 (b/<rlt) in the absence of the attractive field.
Using this equation in the expression for fi$(i), and retaining only terms of
first order in n'lt, we may obtain an expression for 1,(00, v')\ and similarly for
4(00,0.
Values of 1,(00, v'), 1,(00, v') have been determined* for certain integral
values of v', and are given in Table 5. They are all positive; that is, the
attractive force always increases the effective molecular cross-section.
The second case in which I,(P, v'), ij<y, v') can be easily calculated is when
v' — 3. Lennard-Jonesf showed that in this case x c a n be found in terms of
the deflection in the absence of the attractive field. This facilitates the
calculations; however, v' = 3 is not appropriate to actual gases, and so his
* D. Enikog, Uppsala Dissertation; C. G. F. James, fVot. Cam*. PAtV. 5 « . 20,447(1911);
and R. C. Jones, Phyt. Rev. 58, 111 (1940).
f J. E. (Lennard-)Jones, Proc. R. Sot. A, 106, 441 (1924)
10.42] FORMULAE FOR SPECIAL MOLECULAR MODELS 183

Table 5. Values 0//,(00, v') and *'2(oo, v')

3 1(|2 —n*) = 0266} 1(17«-8) = 02337

4 3-4 In 2 = 02274 1(3 In 2 — 2) — 0-2118
5 1(3"'-28) = 0-2011 i(io — IT*) = 01956
7 i = 01667 A(9"*-88) = 01722
9 So = 01444 A = 0-1556

results are not given here in detail. He found that t\(i>, 3) and (2(^,3) are
positive for large values of v (greater than about 12 for f,, and about 16 for »'»),
but negative for small values; thus the effect of the attractive force is to
decrease /1 and Dlt when v is large, but to increase them when v is small.

10.42. Attractive forces not weak; the 12, 6 model

When the attractive force is too strong to be treated as afirst-orderperturba-
tion, the calculation of the collision integrals for the Lennard-Jones model
involves a succession of numerical integrations. A first integration is required
to determine x f°r each of a set of values of kTf*lelt and b/a12; further inte-
grations have then to be performed, with respect to b/crlt and $*, in order to
determine each of the non-dimensional quantities itrlii{r) of (IO.I, 2) as a
function ofkT/e12. The early work was limited to values of p, v' which led to
some reduction in the labour of such integrations.
Hasse and Cook* considered the Sutherland model v = oc, v' = 5 and the
Lennard-Jones 8,4 model (v = 9, v' = 5), each when the attractive force is
not restricted to be small. For the first model, the deflection x caXi D e
expressed in terms of elliptic integrals; a slight simplification of the numerical
work is also possible for the 8,4 model. Clark J ones t pointed out that a
similar slight simplification is possible for the 12,6 model (v = 13, v' = 7).
In view of the power of modern electronic computers, such a simplification
would now be regarded as unimportant; however, on physical grounds also
the 12,6 model has some advantages.
For non-polar gases (i.e. gases whose molecules possess no permanent
electric dipole moment) quantum theory + suggests that the principal part
of the intermolecular attraction is a force K^Jr1, due to dipole moments
mutually induced in each of the two molecules; thus it is reasonable to take
v' = 7. Quantum theory gives no such clear indication as to the value of v;
it actually suggests that the principal part of the repulsive force is pro-
portional to an exponential e-"r rather than to a power r~". However, the
• H. R. H a s * and W. R. Cook, Phil. Mag. 3, 978 (1928), and Proe. R. Soc. A, xas, 196
(1929)-
t R. C. Jones, Phyt. Rev. 59, 1019 (1941).
t See. e.g., N. F. Mott and I. N. Sneddon, Wave Mechanics and its Applications (Oxford,
«948). PP. 164-75.
184 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [10.42
repulsive force is important only over a small range of r, within which it can
be represented almost equally well by a variety of expressions, in particular by
a power law Kur~", where v > v'. The choice v •» 13 is suggested as reason-
able, both because viscosity measurements for gases with weak attractive
fields suggest a repulsive force varying at least as fast as r~" (cf. 12.31, below),
and because studies (by the virial method) of the deviations from Boyle's law
indicate that the 12,6 model gives a good approximation to the force-law.
For the 12,6 model*
v(r) = 4e„{(<r1,/r)» - (<rlt/r)«). (10.42,1)
The model has been investigated independently by Kihara.f by de Boer
and van Kranendonk.J and (more completely) by Hirschfelder, Bird and
Spotz.§ Machine integrations have also been made for this model by
Monchick and Mason, and by Itean, Glueck and Svehla. || Different methods
of numerical integration were used by the different workers, but in general
their results agree well. Values of the quantities *TiV( 1), l^SX 2 ). A. B and c
based on the results of Monchick and Mason and of Itean, Glueck and
Svehla*] are listed for various values of kT/eti in Table 6.
Curves illustrating the variation of log 1^¾ 1), log iff&i), A, B and c with
log (kTjett) are given in Figs. 8,9. In no case does the slope of the curve vary
monotonically. When kT/eit is large, the repulsive part of the field is
dominant, and the slope of the curves for log* AV(i), logTfV»(2) is 2/(^-1),
or I; when kT/elt is small, the attractive part is dominant, and the slope is
2 / ( / - 1 ) , or J. However, the slope does not decrease steadily as kTjelt
increases, but has a maximum greater than § when kT/elt is about unity.
Similarly c is 0-1333 when kT/elt is large, and 0-0667 when it is small, but
takes negative values when kT/elt is between about 0-45 and 0-95.
Hirschfelder, Bird and Spotz proceeded as far as a second approximation
to the viscosity coefficient. This differs very little from the first approximation
when kT ~ e,„ but (as (10.32, 2) implies) it is about 0-168 per cent greater
when kT/elt is small, and about 0-667 P e r c e n t greater when kT/elt is large.
• E. A. Mason {J. Chtm. Phyt, aa, 169 (1954)) u m the alternative expression
•W - «.,{(r./r)"-i(r./r)«).
where rj, = 2<r}a; r„ denotes the value of r for which ->'(>•) attains its maximum value <„.
t T. Kihara, Imperfect Gaitt, section 28.
X J. de Boer and J. van Kranendonk, Physica, 14, 44a (1948).
I J. O. Hirschfelder, R. B. Bird and E. L. Spotz, J. Chem. Phyt. 16, 968 (1948); Chtm.
Rev. 44, 205 (1949).
II L. Monchick and E. A. Mason, J. Chem. Phyt. 35, 1676 (1061); E. C. Itean, A. R.
Glueck and R. A. Svehla, NASA Report TN D-481 (1961).
% A detailed tabulation of the first two expressions, based on these results, was given by
P. E. Liley, Thtrmophytical Propertiet Research Centre, Report 1$ (Purdue University,
1963).
 (*)5U*>| ft-a-S J
r* 1 / 811=-
j
ilffiHif!!!!!!!

ertica
diagra
odel(
"jf / sO /

and
As
/ • Wl fi. a •

t>. /
< iEEHfffllfmtl
" / /o /
- / / /
/ / /
«/ / *"*
/
/
1¾ 11| mi § if !|f || i
V// / i «
ins
flgSwi *; -
«1V15
J\l
^¾
lV.\\l
^, — x £ u JO
•U
.«
Hum mmm o o o o o o o
•*
•—•
X
C
•* sO •*>
- * - « ^
/ A/ *

83*!
mis??i Riftffi

or a = »7) or do
ig. 8. Plots of lo
odel s. To avoid
splaiced by ione u
ould othervrise c
' »*» m !>• 0 f n ^ O «n 0 O O 0 0 0 O 0
6 6 6 - - -

CW.^Soi
186 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: (10.43
0-8 0-8

^*\ -
s«-12
07 4 07
^^ ^^^^^
-

0-5 0-5
«-12 .
12,6
0-4 0-4

- a - 17
^£¢5=53 12.6
01 0-1
-
\ '^
-0-05 . "—1 i
1
log (*r/t„)
Fig. 9. Plot* of A, B »nd c against log (*r/e„) for the 12,6 and exp;6 models.
The rap; 6 valuesof c ire shown for a • 12 and a «« 17; for A and B, curves are given only
for a — 12, aince the valuea for « — 17 differ only alightly from those for the 12,6 model
for A, and are not available for B.

10.43. The exp; 6 and other models

Advances in computing techniques have now made it possible to discuss
other and more complicated laws of force between molecules. One of the
most important of these is the exp;6, or modified Buckingham law, first
studied in connection with the transport phenomena by Mason.* For this
law, the potential energy is

where a is a number usually in the range 12-15. As with the Lennard-Jones

model, c u is the maximum value of - v(r): rm is the value of r for which this
maximum is attained. The quantum-theory arguments in favour of (10.43, 0
have already been mentioned in 10.42. As a consequence of thefinitenessof
the exponential at r = o, the potential energy (10.43, ' ) n a s a positive maxi-
mum at a distance r**r'(< rm), and is negative for very small r; to avoid this
non-physical behaviour, Mason proposed to take v(r) = 00 when r < r'. The
effects of such an assumption are not great, since few molecular encounters
are sufficiently energetic to reach r <° r'.
• E. A. Maaon, J. Chtm. Phyt. aa, 169 (1954)- The full Buckingham law includes an
attractive r* term in ?(r). See R. A. Buckingham, Pnt. ft. Soc. A, 168,264 (1938); also
H. Margenau, Phyt. Rtv. 38, 747 (1931).
10.43) FORMULAE FOR SPECIAL MOLECULAR MODELS 187
As with the Lennard-Jones model, the viscosity and thermal conductivity
conform to a law of corresponding states (for given a). If arn again is defined
as such that K(<T„) = o, tr12 130-8761 rm for a. = 12, and 08942 rm for a — 15;
for comparison, <r18 = 0-8909^ for the 12,6 model.
Detailed numerical results (for each of the values a = 12, 13, 14, 15) were
given by Mason in his original paper, and those for a. = 16 and a = 17 were
given later by Mason and Rice.* Those for a = 12 and a = 17 are illustrated
in Figs. 8 and 9; those for a = 14 and a. = 15 resemble fairly closely those
for the 12,6 model.
The r~l variation of the attractive potential energy applies only to
electrically neutral non-polar molecules. The interaction of an ion with a
neutral molecule involves an r~* part in v(r); to discuss this type of inter-
action, Mason and Schampf have calculated collision-integrals for the
potential energy

*tr)-Jeit[(i+y)(^)U-4y ( ^ - 3 ( 1 - 7 ) ( ¾ 4 ] .
(10.43. 2)
which reduces to the 12,6 model if y = 1; y is a parameter measuring the
relative importance of the r~* and r~* parts of the attraction. For polar
molecules, StockmayerJ proposed the form
v{r) = 4 ^ ( ^ 2 / ^ ~ K 2 / ' ) 6 ] -~*x<*t C/**; (10.43, 3)
here -rf',, ~#t are the dipole moments of the interacting molecules, and £ is
a function of their relative orientation. The energy (10.43,3) ' s n o t spheric-
ally symmetric, and the theory of smooth molecules is strictly not applicable
to it. However, Monchick and Mason § have discussed polar molecules as if
they are symmetric, assuming that molecular trajectories are negligibly dis-
torted by transfer between rotational and translational energies, and that
each encounter is characterized by a constant value of £ in (10.43, 3)I t 0
determine the collision-integrals they then average over all possible values
of?.
Other models studied in connection with the transport properties
include the pure exponential model| (v{r) = e 12 exp( — r/<ru)), the
attractive power-law models (V(r) = —K'/(I>' — i)r" , - , ) 1 the exp;exp
model** (V(r) = Ae-rl''-Be-rl'>') and the Lennard-Jones 9,6 and 28,7
models.ft For details of the results, the reader is referred to the original
papers.
• E. A. Mason and \V. E. Rice. J. Chem. Phys. 22. 843 (1954).
t E. A. Mason and H. W. Schamp, Ann. Phys. 4, 233 (:958).
t W. H. Stockmayer. J. Chem. Phys. 9, 398 (1941).
J L. Monchick and E. A. Mason, J. Chem. Phys. 35. 1676 (1961).
I!L. Monchick, Phys. Fluids, 2, 695 (1959). repulsive force; R. J. Munn, E. A. Mason,
and F. J. Smith, Phys. Fluids, 8, 1:03 (1965), attractive force.
1 T. Kihara. M. H. Taylor and J. O. Hirschfelder, Phys. Fluids. 3, 715 (i960).
** F. J. Smith and R. J. Munn, J. Chem. Phys. 41. 3560 (1964).
ft F. J. Smith, E. A. Mason and R. J. Munn, J. Chem. Phys. 42, 1334 (1965).
188 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: (10.5

THE LORENTZ APPROXIMATION

10.5. A specially simple expression for the velocity-distribution function in

a gas-mixture may be obtained if (i) the mass m, of the molecules of one
constituent is very great compared with that of the molecules of the second
constituent, and (ii) the influence of mutual encounters among the latter in
altering their motions is negligible compared with that of their encounters
with the heavy molecules. The latter condition is fulfilled if either (a) the
number or (6) the extension of the molecular field of force is much less for the
light than for the heavy molecules. The kinetic theory for a gas satisfying
these special conditions (i) and (ii) was first studied by Lorentz;* hence we
shall refer to such a gas as a Lorentzian gas.
On account of the condition (ii), the integral / , can be omitted from the
right-hand sides of (8.31, 4-6). Also, since the mean kinetic energies of the
peculiar motions of the two sets of molecules are approximately equal, the
peculiar velocities of the heavy molecules will be small compared with those
of the lighter molecules; hence the relative velocity; in the encounter of
molecules of opposite types is roughly equal to the peculiar velocity C, of
the lighter molecule. The problem, in fact, approximates to the determina-
tion of the distribution of motion among a set of light molecules subject to
deflection by stationary obstacles.
The velocity of one of the heavier molecules is not appreciably altered by
a collision with one of the lighter molecules: thus, in evaluating the integrals
/,„ / t l in equations (8.31, 4-6), we can put C\ = C„ A[ = Av D[ = D t ,
B[ - Bt. Since, moreover, /, is to be neglected, and g = C„ the second of
equations (8.31, 4) becomes
/PM-flC, = fflfiy°>(At-A't)Ctbdbdedct,
whence, on integration with respect to cv we obtain
(«1-1)0, = nJfcAi-A'JCtbdbde.
Since the relative velocity after encounter is equal in magnitude to that before
encounter, C't = C,. Hence this equation is equivalent to
(*l-i)Q = MsfCjJJ^-OC.MMs. (10.5, 1)
Similarly the second members of the pairs of equations (8.31,5, 6) reduce to

- ± C, = niD^Ct)jj(Ct-C't) Ctbdbde, ( I0 . 5 , 2)

%frC°Ct - n . ^ O / J c e J c . - c f c a C A r f * * . (10.5, 3)

* H. A. Lorentz. Proe. Amst. Acad. 7, 438, 385, 684 (1905).

10.5] FORMULAE FOR SPECIAL MOLECULAR MODELS 189

Since x and e now are polar angles specifying the orientation of C't about
Ct (c • 35). c - = c , cosx+ C t sin x(hcosc+t sine),

where, h, i are suitably chosen unit vectors perpendicular to C, and to each

other. Hence ^^ _ c^bdbde = 2W c,J(i - cosX)bdb

by (9-33- 4)- Also. sin ce C't = c a.

jf(cZct-C?C't)bdbde = 2»r/(i - cos**){C,C t - JCJ(Iih + H)} bdb.
Since CJCt, h and 1 are mutually orthogonal unit vectors, by (1.3, 9),
Ccyq+hh+u-u,
where U is the unit tensor. Hence

\\(c\ct-clc^bdbde = 3TrCtCtj(i-coa*x)l>db

= icfctfg.
Here 0$, 0¾ are functions of C2. Using these results in (10.5, 1-3) we get
A {C ) = (l s 4)
'> ^cJs' °'
D„(C2) = -n/n^Citfi, (10.3, 5)
Bt(C,) = mJ^kTCt0ft. (10.S, 6)
In the present case equations (8.4, 1, 7) reduce to
C, = - ^ {rf„//J0,C1Ot(C2) <fct + V In TfA«C\AACt) dcj

= £-{DltJlt + DT\nnT},

since here C t = 0. Thus, using (10.5, 4, 5),

(, 5,7)
^-^/^¾^ °-
In the conduction of heat the lighter molecules, because of their larger
velocities, are more effective, in proportion to their number, than the heavy
molecules. If the conduction is assumed to be due predominantly to the
light molecules, then in (8.41, 4), namely
A = JA«*[M. A) - {A, DUD, />}], (10.5, 9)
190 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [10.51
we must insert the following simplified forms of (8.31, 11, 12)
n*{A,A) - n^lJPPl-mcMdc (10.5, 10)
n{A,D) - - ( ^ - • J / M - I K C ^ & V * . . ('O-S. »)
{D, D] = totQ-yffiPMdCr (10.5, 12)
The contributions of the two sets of molecules to the hydrostatic pressure
are of the same order of magnitude, if their number-densities are comparable;
the viscous stresses due to the heavy molecules are, however, large compared
with those due to the light. If, nevertheless, the pressure system due to
the light molecules alone is considered, then (cf. 8.42) the corresponding
coefficient of viscosity /1 is given by

/< = \"tf/Pcfa:CtaBJLCJdc

(,os>,3)
-d&J^S**
using (10.5, 6) and the theorem of 1.421.
The meaning of equations (10.5, 7-12) can perhaps be better seen if we
make the substitution ,,,, „ , - , .
n,#y-i//(c,).
If the molecules are rigid elastic spheres, $»' = ncr\t, by 10.2, and so this
substitution implies that for such molecules
/(0=1/(7^,).
Since in the Lorentz approximation collisions between pairs of molecules
w, can be neglected, it follows that for rigid elastic spheres /(C,) is the mean
free path of molecules m, of peculiar speed C,: for more general molecules
we may interpret l(Ct) as an equivalent mean free path for molecules m, of
peculiar speed Ct.
For example, making the substitution, we get

3ntnJ 3"
a relation which may be compared with (6.4, 2).

10.51. Interaction proportional to r~"

These results take a specially simple form when the molecules are centres of
force varying inversely as the vth power of the distance, so that, using
(10.31, s), and putting? = #,,
10.S2] FORMULAE FOR SPECIAL MOLECULAR MODELS 191
In this case

(i°-5».3)
k
r = — —,—~\, ( , °-5 I >4)
v J T/
n z{v—i)

^ - ^ ( ^ ( ^ ^ ( 3 + ^ ) / 3 - ^ ( - 5 . . 1 )

A= 4».*.(^* ( ^ ^ ( 3 + ^ ) / 4 5 ^ 1 ^ ) .
(10.51, 6)
A/> = 1 5 ^ 4 ^ ) / 4 ^ 1 ( ^ ) = ( 0 , 5 ^ . ( ^ ) / 2 ^ ( ^ ) . (10.51, 7)
where (c„)t denotes the specific heat of the lighter gas.
These equations can be made to apply also to the case of rigid elastic
spherical molecules, by making v tend to infinity, and K\{ approach cr„.
In particular, since /l,(oo) = J, A^cc) = J, we then have

kr = A=
~tn' ^'(^ (10S*> 8)
For the above models k T does not depend on the temperature. This is not
true for more complicated models. As can be shown by applying to (10.5, 7)
the argument used in proving (9.71, 1), in general

kT
~n ITnT^ ( I O -5'.9)
the differentiation being performed keeping n constant. Thus, for example,
for both the Lennard-Jones and the exp; 6 models

k r = ^F(kT/elt), (:0.51, 10)

where F is a function whose form depends on the particular model.

10.52. Deduction of the Lorentz results from the general formulae

The results of 10.5 can also be derived from the general solution, by making
mjm1 tend to zero, and neglecting the terms arising from the mutual en-
counters of the light molecules. The results thus derived are expressed in
terms of infinite determinants, as in the general case: it may be shown*
(though the proof is not given here) that these determinantal expressions are
• See S. Chapman, J. Land. Math. Soc. 8, 266 (1933).
192 VISCOSITY, THBRMAL CONDUCTION, AND DIFFUSION: [10.52

equal to those found above, and the comparison of the two forms of the
results throws interesting light on the nature of the convergence of the
determinants in which the general solution is expressed.
The convergence may be illustrated by comparing the different approxi-
mations to the transport coefficients for inverse-power interaction fu/r*
with the exact values given by the formulae of 10.51. In Table 7 the values of
[^ltWfitli f° r m"" 2 *»d 3 a r e compared with the limiting value
£>u/[£>ii]i. The first line, for v - 00, corresponds to the case of rigid elastic
spheres; thus the first approximation to D„ for this model is in defect by
about 12 per cent; the second reduces the error to less than 5 per cent, and
the error of the third approximation is about 2 per cent. The errors of the
various approximations for values of v between 5 and 00 are less than the
errors for i> = 00; when v - 5 (corresponding to Maxwellian molecules) the
first and all later approximations are equal to the exact value. When v is less
than 5, the first approximation falls off greatly in accuracy, though the
second and third approximations are good.

Table 7. Values of [D,JJJD,J t and DJ[DXJ,

V [DMD,,], (O.J./tD,,], Ai/dU.

00 1-083 1-107 113a
17 1049 1-060 1-07a
13 1039 1-048 1-036
9 1033 i-oa7 1-031
5 1 1 1
3 lias 1-130 113»
a 3aso 339' 3306

Table 8 gives the ratios which the approximations [k^],, [k r ], bear to the
limiting value k r ; it shows that the first approximation to k7 is more in error
than the first approximation to Dn. When v •» 5, k r and every approximation
to it is zero, but the ratios in question tend to unity as v tends to 5.

Table 8. Ratios of the first and second approximations

to kr, to the exact value of k r

v — 00 «7 13 9 5 3 a

fkrli/kr - 077 0-83 085 0-89 1 t-11 077

[kr),/k|. - 0-88 o-9a o-93 0-95 1 1-0! 1-01

Next follows a table giving the ratios of [A]t and [A], to the exact value A;
these are calculated on the assumption (cf. 10.5) that the conductivity due to
the heavy molecules is negligible. It appears that the accuracy of the approxi-
mations to A is about equal to that of the approximations to Dn.
10.53] FORMULAE FOR SPECIAL MOLECULAR MODELS 193

Table 9. Ratios of the first and second approximations

to A, to the exact value ofk

V = 00 '7 •3 9 5 3 2

(1) = 0 8 5 0-91 o-93 096 1 O-82 014

(2) = 0-93 096 o-97 099 1 009 0-92

Finally, Table 10 gives the ratios of [/<], and [/«], to /«, where fi now repre-
sents the (small) contribution to the viscosity made by the light molecules
alone (cf. 10.5). These show that the approximations to /1 are father more
accurate than those to Dl2.

Table 10. Ratios of the first and second approximations

to ft, to the exact value of ft

V = 00 •7 «3 9 5 3 2

(1) = 092 095 096 098 1 0-92 0-46

(2) = 098 0-99 o-99 099 1 1-00 0-98

The inaccuracy of the first approximation when v = 2 is very marked in

all the Tables 7-10. However, the Lorentz approximation is not strictly
applicable to an ion-electron gas, in spite of the large mass-ratio, because
the effect of electron-electron interactions is normally not small in such a
mixture. These interactions reduce somewhat the inaccuracy of the first
approximation in such a gas (compare Table 7 with Table 4, section 10.34).

10.53. Quasl-Lorentzian gas

Another interesting special case is that of a mixture in which mljmt is large,
but njnt is so small that encounters between pairs of molecules mx do not
appreciably affect the velocity-distribution function/!. This case has been
studied by Kihara* and by Mason,f the latter of whom called such a gas
quasi-Lorentzian.
For this case Kihara verified, by direct substitution in (8.31, 5), that the
first approximation to the diffusion coefficient gives the exact value. Mason
confirmed this result by examining the orders of magnitude of the deter-
minantal elements. Unfortunately there is no corresponding simplification
for the other transport coefficients.
• T. Kihara, Imperfect Gases, section 20.
t E. A. Mason, J. Chem. Phys. 27, 782 (1957).
194 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION: [10.6

10.6. A mixture of mechanically similar molecules

Another specially simple case of the solution for a gas-mixture is that in
which the different sets of molecules are of the same mass and obey the same
law of interaction at encounter, so that they are mechanically similar. In this
case the thermal conductivity and viscosity are the same as if the molecules
were identical in alt respects, and the coefficient of thermal diffusion
vanishes; also the coefficient of diffusion is the coefficient of self-diffusion
Dn of a simple gas.
Since all the molecules are similar, we introduce the velocity-distribution
function of all the molecules. The first approximation to this is/"", where
fmln'f!ltlnv}%>lntare identical functions of the respective variables C, C„ Ct.
The second approximation iaf'+f11, where
/»(C)=/l 1, (C)+/ 1 , »(C). (10.6,1)
Since/does not depend on the relative proportions of the two gases in the
mixture, but only on the total number-density, /"> is unaffected by the
relative diffusion of the two gases, and so has no part depending on the vector
dH of (8.3, 8). Thus, equating the terms in (10.6, 1) depending on rf„, we
get (cf. (8.3., 9. 10)) niDi(C) m .^2,^). (l0.6i 2)

We denote the common value of each of these expressions by DJ[C). Now,

on account of the mechanical identity of the molecules of the two gases,
A W Q = /ittWi)+JVC.)}, /,{iW<y> - /,,(1),(0,)+0.(0,)}.
Using this and (10.6, 2), the equations (8.31, 5) become
x r V r C , - -/..{IWC,)}, x f ' / T C , = «/„{D,(C,)},
which are identical save for the different variable involved. Again, by (8.4,1),

3^1¾ J

so that, by (8.4, 7), D„ = ±-tjf»C.D0(C)dc.

Thus Dn does not depend on the proportions of the mixture, but only on its
density.
Approximations to the value of Dn may be derived from (9.81, 1, 3) by
putting ml = mt = m. The first approximation is [O,,],, where
[£„], = 3E/4P.
By (9.8, 8) and (9.7, 3),
E/M, = 4^(2)/5¾¾ 0.
10.61] FORMULAE FOR SPECIAL MOLECULAR MODELS 195

or, using (9.8, 7), and remembering that in the present case 0^(2) = ft?ii(2)>
£/[/<], = 4A. (10.6, 3)
Hence [/>„], = 3*[/i]Jp. (10.6,4)
The second approximation is obtained by multiplying [Z? u ] l by 1/(1 —A),
where A is given by (9.81, 4). Putting A/, = A/, in this equation, and using
(10.6,3), we find that ,,. , 0 . . , .
v
'•"' A = 5 C 2 / ( I I - 4 B + 8A). (»0-6,5)
In the Kihara approximation (B = J)
A = 5C2/8(I+A). (10.6,6)

On substituting for A, B, c, the values of 1/(1 — A) given by (10.6, 5 and 6) for

force-centres of index v are found to be:

i-«5 9 15 00

(10.6.5) >/(•—A) = I 1004 10083 1-017

(10.6.6) 1/(1— A) = I 1004 1009 1018

The true value of Dn will be slightly greater than the second approximation:
the exact factors by which the first approximations for these models are to
be multiplied may be estimated as 1, 1-005, I-OIO and 1-019 approximately.
Thus, in particular, for rigid elastic spheres {y = 00) the exact value is*
D =ioi (,o 6,7)
^ ^A^y -
_ 1-019 6 /1
~ 1 0 1 6 sp
= 1-204/1/p. (10.6, 8)

10.61. Mixtures of isotopic molecules

In a mixture of isotopes of the same element, the law of force between all
pairs of molecules is normally taken to be the same. This implies that
Tsr(f(r) = ir%(r) = yriV) (io.6i, i)
(cf. 10.1) and m\&?{r) = (2m0MlMi)\^r) = mjfi^r). (10.61, 2)
These results can be used to simplify the general formulae for the transport
coefficients in a gas mixture.
• This value of the coefficient of self-diffusion was first obtained by F. B. Pidduck, by
a quite different argument (Proc. Lond. Math. Soc. 15, 89 (1915)). C L. Pekeris (Proc.
nam. Acad. Set. U.S.A. 41, 661 (1955)), found the numerical factor in (10.6, 7) to be
1 01896.
196 VISCOSITY, THERMAL CONDUCTION, AND DIFFUSION [10.61

Save for the lightest gases, it is sufficient to regard (M, — «»)/(wl + m,) as
a small quantity whose square can be neglected. Then, for example, the
first approximation to the thermal diffusion factor a„ = k r /x,x, is inde-
pendent of the proportions xv x, of the isotopes in the mixture, being

[aiJl
"A(i.-4B+8A)iM^- < 10 - 6, '3)
In the Kihara approximation (B = }) this takes the form

^-¾¾¾ <"•'••*»
* R. C. Jones, Phyi. Rev. 58, t n (1940).
 11
MOLECULES WITH INTERNAL ENERGY

11.1. Communicable internal energy

Throughout the previous discussion the molecules have been regarded as
possessing no internal energy, or at least none that is interchangeable with
translatory energy at collisions. If interconversion of internal and trans-
lational energy is possible, extra variables specifying the internal motion
have to be introduced. A general theory covering this case can be given, but
the detailed discussion is very complicated for all save the simplest models.
However, certain general results for all molecules with internal energy can
be established. The internal energy is found to contribute, proportionately,
rather less to the thermal conductivity than does the translational energy.
Also the finite time required for interchange between translational and
rotational energy introduces a ' volume viscosity' opposing motions of con-
traction or expansion of the gas. This volume viscosity can considerably
increase the rate of damping of sound waves above that due to the 'shear'
viscosity /«.
Detailed formulae have been derived only for one or two decidedly
artificial molecular models. One is a perfectly rough, perfectly elastic and
rigid spherical molecule. This model was first suggested by Bryan;* the
methods developed by Chapman and Enskog for smooth spherical molecules
were extended to Bryan's model by Pidduck.f The model is mathematically
simpler than all other variably rotating models, in that no variables are
required to specify its orientation in space.
Jeans}: proposed as a model a smooth elastic sphere whose mass-centre
does not coincide with its geometrical centre; this model has been studied by
Dahler and his associates. § Curtiss and Muckenfuss || worked out the
detailed theory of sphero-cylindrical molecules; that is, smooth cylinders
with hemispherical ends. Both these models require two space variables to
specify the molecular orientation, as well as variables specifying the angular
velocity. The sphero-cylindrical model also has the awkward property that,
especially when the molecules are long and slender, 'chattering' collisions
• G. H. Bryan. Rep. Br. Ass. Advmt. Set. p. 83 (1894).
t F. B. Pidduck, Proe. R. Soe. A. 101, 101 (1922).
( J. H. Jeans. Phil. Trans. R. Soc. 196, 399 (1901); Q. J. Math. 35, 224 (1904).
5 J. S. Dahler and N. F. Sather, J. Chem. Phys. 35. 2029 (1961). and 38, 2363 (1962);
S. I. Sandler and J. S. Dahler, J. Chem. Phys. 43, 1750 (1965); 46, 3520 (1967), and 47,
2621 (1967).
II C. F. Curtiss. J. Chem. Phys. 24, 225 (1956); C. F. Curtiss and C. Muckenfuss, ibid.
a6, 1619 (1957) and 29, 1257 (1958).
[ 197 1
198 MOLECULES WITH INTERNAL ENERGY [11.2
can occur; that is, the reversal of angular velocities at one collision can lead
to one or more further collisions between the same molecules immediately
afterwards.
The artificial nature of such models, and the complexity of the formulae
derived from them, led Mason and Monchick to attempt a simpler approach.*
They based their discussion on the general theory, but determined the
collision integrals approximately in terms of relaxation times.

11.2. Liouville's theorem

The general theory is based on a dynamical theorem due to Liouville, which
is of fundamental importance in statistical mechanics. Let the state of a
system be specified by R generalized coordinates Q* and their conjugate
momenta P* (a -= i, 2,..., R). If H is the Hamiltonian function of the system,
the Hamiltonian equations of motion are

0j> dP* , .
(II2,2)
so that #--a?=-
The quantities Q*, P* may be regarded either as the components of R-
dimensional vectors Q, P, or as the coordinates of a point in a space of
2R dimensions (the phase-space).
Liouville's theorem refers to the points in phase-space representative of
an assembly of dynamical systems with the same H. It asserts that, if at time t
these occupy an infinitesimal volume V of the phase space, at any other time
they occupy an equal volume.
The proof is as follows. After a short time dt, Q" and P* become Q' + Q'dt
and P* + P*dt, and V becomes V+rfV; by the rule for transforming small
elements of volume
V+dV d(Q + $dt,P+Pdt)
V = d(Q,P)
The Jacobian on the right is a determinant whose non-diagonal elements are
all proportional to dt; the diagonal elements are of form 1 + (d()"j'c>Q*)dt or
i+(dP"/8Pa)dt. Thus, neglecting squares and higher powers of dt,

*(*•£*)-.•«$+$
- 1,

by (11.2, 2). Hence dV/dt = o; that is, V does not change with time.
* E. A. Mann and L. Monchick, J. Chtm. Phys. 36, t6ai (196a).
11.21] MOLECULES WITH INTERNAL ENERGY 199

11.21. The generalized Boltzmann equation

Consider a gas-mixture; a simple gas is regarded as a special case of this.
The position and motion of a typical molecule ms are specified by a set of
generalized coordinates and momenta Q*, Pf (a =1,2,..., R), or by the
/^-dimensional vectors Q„, P„ of which these are the components. The first
three of the coordinates are taken to be the Cartesian coordinates *,, y„ z,
of the mass-centre (the components of r„): the corresponding momenta are
the components mstt„ msv„ tnews of the linear momentum m,c,. The
remaining coordinates and momenta, specifying the internal state, are
represented collectively by (/? —3)-dimensional vectors Qi", PJ" in the
phase-space specifying the internal motion; the volume-element rfQ^'i/fJ''
in this internal phase-space is denoted by dSl,. The generalized velocity-
distribution function f,(Qa, P„ t) is then defined as such that the probable
number of molecules m, whose coordinates and momenta Q,, P, lie in the
volume dQ,dP, of the total phase-space at a time t is

^3UQ.,P.,t)dQ,dP, =f,(Qs,P„t)drsdc,dSi,. (H.2I, i)

As usual, any molecule is supposed to be interacting with others during a
negligible part of its total lifetime, and binary encounters alone are con-
sidered; also processes like chemical action and ionization are supposed
absent. The motion of a molecule ms between encounters is assumed to
satisfy equations of the Hamiltonian form ( n . 2 , i). The corresponding
Hamiltonian function H, is assumed to be an even function of the momenta,
not explicitly depending on the time, so that H, is equal to the total energy of
the molecule. If external forces act on the molecule, their potential energy
is assumed to be a function vj(r,) of the position r, only. Thus
H,= v,(r.) + H</\ (u.21,2)
1
where HJ* denotes the energy of free motion in the absence of the external
forces.
Consider now the rate of change of fs. Were it not for encounters, the set
(11.21, i) of molecules m„ would, after a time dt, constitute the set

*." V«(Q, + Q.dt, P, + Psdt, t + dt) dQ.dP,

(the volume of phase-space which they occupy is still dQ,dP„ by Liouville's
theorem). Let the net effect of encounters during dt be to increase the
number in the set by .,
m;*dt-'l*dQ,dP..
Then we find, as in 3.1, that
200 MOLECULES WITH INTERNAL ENERGY [11.22
where d/dQ(, 8/dP, denote A-dimensional vector operators with com-
ponents a/agj, a/apy, and Q..8/J9Q, = -ZQidfjdQZ, etc.
a
Equation (11.21, 3) is the generalized Boltzmann equation. Using the
Hamiltonian equations, we may express &J, in the form
a t 8/, 8H, 8f, 8H, 8/, . .
*•*>= 8t+bpliQ,-*Q.&: (,, 2, 4)
• '
11.22. The evaluation of 3,/./3/
In evaluating 8efJ8t, the encounter of a pair of molecules m„ mt is regarded
arbitrarily as commencing at one instant t'0, and ending at another instant f0.
To avoid difficulties arising from the extremely rapid variation of the
internal coordinates and momenta, the encounter is replaced by a dis-
continuous 'collision' as follows. The motions of the molecules prior to t'0
are supposed to be continued up to }('o + '0) a s if n o interaction were taking
place, the coordinates and momenta at the latter instant being Q'„ P', and
Q't, Pi- Likewise the motions subsequent to t0 are supposed to be produced
backward to J('o + 'o) as if no interaction were taking place, the coordinates
and momenta at }(<„ + (,) now being Q„ P, and Q„ Pt. Then the effect of the
encounter, at all times outside the (negligibly small) interval between t'0
and <„, can be represented by replacing the actual encounter by a ' collision'
occurring at the instant i(t0 + t'„), at which Q'„ P'„ Q',, Pi change dis-
continuously to Q„ P„ Qt, Pt.
Suppose that encounters arc replaced by collisions in this fashion. The
line of motion, before a collision, of the mass-centre of the molecule mt
relative to that of the molecule m, may be specified by impact parameters b, e
defined as in 3.42. Consider the number of collisions occurring during dt
between the dn,(sm^if,(Q„P„t)dQ,dP,) molecules m, in the phase-
element dQ,dP, and molecules m, with momenta and internal coordinates
in the ranges dPt, dQ\'\ the impact parameters b, e lying in the ranges db, de.
At the beginning of dt the molecule m, must lie in a volume gdtbdbde,g being
the magnitude of the relative velocity ct-c, (cf. 3.5). Thus the number of
such collisions is dn,dnt, where
dn, = mf'ftiQt.Pt.^gdtbdbdedPtdQ^.
Summing over all values of b, e, P„ Q}'\ the total number of collisions of the
dn, molecules m, with all molecules m, in time dt is found to be
dt{m.mt)-*UQ., P.) dQ,dP,jjjjft(Qt, Pt)gbdbdedQ[«dPt.
This is the number of collisions in dt which carry molecules m, out of the
phase-element dQ,dP,.
The number of collisions carrying molecules m, into the same phase-
element is similarly
dtim.m.^jjjjf.iQ'., P.)MQl, PDg'b'db'de'dQ'.dP'.dQ^dPi
11.23| MOLECULES WITH INTERNAL ENERGY 201

Here the integration is over all collisions with initial coordinates and
momenta Q's, P'„ Qi'y, P't a n d w 'th collision parameters b', e' such that, of
the final coordinates and velocities Q„ P„ Q„ Pt, the pair Q„ P, lie in dQ„
dP,. By Liouville's theorem applied to the pair of molecules regarded as a
single dynamical system,* if b, e are impact parameters specifying the motion
of the molecule mt relative to the molecule m, after a collision,
g'dfb'db'de'dQ'.dP'.dQpdP't = gdtbdbdedQ,dP,dQ^dPt. (i 1.22, 1)
Combining these results, we find the net gain of molecules m, to the phase-
element dQsdP, through collisions with molecules /¾ during dt to be
dt{m,mt)-*dQ,dP.H\\U',f't-fJt)gbdbdedQpdPt,
where /,', f't denote f,(Q'„ P',), ft(Qi, Pi), etc. Hence, from the definition
°(W, 8f

-%l=s«r3///j(/;/; -fjt)gbdbdedQpdpt („.22,2)

= 2/JJJW -U)gbdbdedctdSlt. (11.22, 3)

11.23. Smoothed distributions

The above derivation rests on the assumption that at any instant the pro-
portion of the molecules m, undergoing encounters is negligible, so that the
velocity-distribution function of all molecules m, is indistinguishable from
that of molecules not undergoing encounters. The justification of the
assumption is in the smallness of the duration t0 — t'0 of an encounter, which
was also invoked to justify the replacement of encounters by 'collisions'.
In the Boltzmann equation (11.21,3) t n e terms involving (J,, P, give the
change in / , due to the motions of molecules between encounters. Now
(cf. 11.34) the period of rotation of a molecule is in general comparable with
the time taken for the molecule to travel a distance equal to the effective
molecular radius, and so with the duration of an encounter. The periods of
molecular vibration are in general even shorter. Thus the terms involving
Q„ P, in the Boltzmann equation introduce fluctuations i n / , on a time-scale
comparable with the duration of an encounter. Since this duration is being
regarded as negligibly small, it is reasonable to remove the fluctuations by
averaging fB over a time large compared with the periods of molecular
rotation and vibration. So averaged,/, ceases to depend on quantities which
vary rapidly in the internal motions; it may depend on the energy of internal
vibrations, and on the angular momentum and energy of rotation, but not
* Liouville's theorem has to be applied three times to the pair of molecules; first to their
actual motion during an encounter, then to the (supposed undisturbed) motion from
l'„ to 4(/»+(J), and finally to the similar undisturbed motion from M's+'o) to f«. In the
last two cases H is the sum of the Hamiltonians of the two molecules, ignoring inter-
actions.
8 CMT
202 MOLECULES WITH INTERNAL ENERGY [11.24

on the phase of a vibration or the changes of orientation during rotation.

Corresponding to this averaging of/,, it is necessary also to average djjdt
over all states corresponding to particular values of the angular momentum
and energies of rotation and vibration.
This type of averaging replaces/, by a smoothed function which, however,
suffices to give all measurable properties of the gas.#

11.24. The equations of transfer

It is here convenient to employ summational invariants measured relative to
fixed axes, i.e. ^ , , , ^ + , » _ ^ ^»_ nj" (u.24, I )
where, as in 11.21, HJF) is the energy of free motion in the absence of
external forces. Each ijr, is a constant of the free motion; thus
G.-WQ.+p.-afc/ap.-o
in the free motion, so that

The equation of change of any summational invariant \}rt is got by multi-

plying the Boltzmann equation (11.21, 3) by jfr.dP.rfQJ", integrating, and
summing with respect to s. The summational-invariant property ensures
that the S,fjdt terms, representing the effect of collisions, then vanish. Thus

Again, H, = K^r,) + HJ n (cf. (11.21,2))and the force w.F.onamoleculem,

is given by dy
m F
> > = -^r- (11-24.4)
Using (11.24, 2> 4) m (n-24> 3). w e n n d

?//{l^4(«-A(«
+P
'-£WM-F':g/'}dP'dQ' " °- (,,24' 5)
In (11.24, 5), the third term of the integrand vanishes on partial integration
with respect to P„ since/, -+ 0 as P , -»• 00; the fourth term similarly vanishes
after a partial integration with respect to c,. In the second term, the co-
ordinates Ql can be divided into three groups. First there are the Cartesian
coordinates of the mass-centre; next there are Eulerian angles specifying the
orientation of the molecule -, finally there are coordinates giving the displace-
* The smoothing introduced here it similar in essence to that introduced in a.ai.
11.3] MOLECULES WITH INTERNAL ENERGY 203
ments in internal vibrations. The contributions from the Eulerian co-
ordinates to the term in question vanish on partial integration, because of
periodicity; those from the vibrational coordinates similarly vanish, because
f,-*-o when one of these coordinates becomes large. Since ft, does not
explicitly depend on t, the remaining terms of (i 1.24, 5) give

2 \ft{»^.) + ^,ctft,)-n$F,.^ft^ = 0. (11.24, 6)

Equation (11.24, 6) is identical in form with the equation obtained in the

absence of internal motions. Thus, applied to the summational invariants
fti*\ 4 ^ 1 ftf\lt must give equations of continuity, momentum and energy
identical with those of 3.21 or 8.1. In particular, the energy equation is

p
Wt ytn>E->'p) +&;•« =
SftC.-F.-p.-^co. (n.24,7)

Here p = ~Zp,C,C„ as before, and E, differs from HJ F) only by being

9
measured relative to axes moving with velocity c 0 .
In addition to the summational invariants (11.24, 1), there is also a sum-
mational invariant of angular momentum, given by
<J><4> = r, A mac, + h„ (11.24, 8)
where h, is the angular momentum about the mass-centre. Direct sub-
stitution from (11.24, 8) m t 0 ("- 2 4i &) yields, after simplifying by using
the momentum equation,

^ ( Z » A ) + ;£.(I>,C,A.) = O. (II.24, 9)

However, this equation is open to the objection that 4»J'' is a summational

invariant only in virtue of the difference between the values of r for the
centres of colliding molecules. Thus for consistency, in considering the
equation of change of «J*J4), it is necessary equally to take into account the
finite distance through which the internal angular momentum h, is displaced
in being transferred from one molecule to another at collision. This col-
lisional transfer of h„ like the corresponding transfer of heat or momentum,
is very small at ordinary densities, but it modifies (11.24, 9) appreciably,
because h, is normally very small (see 11.32).

11.3. The uniform steady state at rest

The determination of the velocity-distribution function in the uniform
steady state is complicated by the non-existence, for general molecular
models, of an 'inverse' collision exactly reversing the effect of a given
'direct* collision. However, a 'reverse' collision can be found, in which the
course of the direct collision is exactly retraced: the final velocities of the
8-3
204 MOLECULES WITH INTERNAL ENERGY [11.3
molecules in this are minus the initial velocities in the direct collision. Such
reverse collisions can be used (following an argument due to Lorentz*) to
determine/, in the uniform steady state, but only by invoking an additional
probability assumption.
Consider the uniform steady state in a simple gas at rest under no external
forces. In order to use reverse collisions, it is assumed that in such a state
f. =UQ.,P.) -MQ„ -P.) - / _ . (11.3.»)
i.e. that a given set of velocities is as probable as the same set reversed. This
assumption is reasonable, but not self-evident.
Let the Boltzmann equation (11.21, 3) be multiplied by In/, and inte-
grated over the whole range of (¾0 and P,. In the uniform steady state,
d/Jdt = o; since also/, and H, are both even functions of P„ the left-hand
side of the resulting equation reduces to the integral of an odd function of
P, (cf. (11.21,4)), which vanishes. Thus, on substituting for djjdt from
1122,2
' 0 = \l\l\\U'Jl-fJt)\nf,gbdbdedQpdPtdQydP. (11.3,2)
(the suffixes s, t are retained purely to distinguish between the two colliding
molecules). In (11.3, 2), Q'„ P'„ Q't, P't are the coordinates and momenta
before collision of molecules whose coordinates and momenta after collision
«« Q.. P* Qt. Pt
From (11.3, 1,2),
0
= \MW-rt--fMW^gbdbdedQpdPtdQydPr
In this equation, Q„ —P„ Qt, —Pt can be regarded as the coordinates and
momenta before the reverse collision which leads to thefinalcoordinates and
momenta Q'„ — P'„ Q't, — P{. Rename these quantities Q'„ P'„ Q't, P't (before
collision) and Q„ P,, Q ( , Pt (after). Then, using (11.22, 1), one finds
0
- IffljfWt-r.nw.gbdbdedQ<t«dptdQ™dP,. (.i.3,3)
Add this to (11.3, 2), and to the equations got from (11.3, 2 and 3) by inter-
changing the role of the colliding molecules (so that Q„ P, become Q ( , Pt,
etc.). Then (because the space-coordinates are to be regarded as identical
in Q, and Q,)
0
" \MU:fi-fJt)^UMJ't)gbdbdedQ^dPtd^dPt. O1.3,4)
As in 4.1, equation (11.3, 4) implies that for all possible collisions
ln/.+ln/ t = ln/; + ln/;. (11.3,5)
When this condition is satisfied, the integrand in de/J8t is identically zero,
and there is detailed balancing. The interpretation to be given to detailed
balancing here is, however, that collisions of a given type are balanced in
detail by their reverse collisions.
• H. A. Lorentz, Win. Sit*. 95 (2), 115 (1887).
11.31] MOLECULES WITH INTERNAL ENERGY 205
Equation (i 1.3, 5) implies that In/, is a linear combination of summational
invariants. As it is an even function of P„ it cannot depend on the linear or
angular momentum; thus , _ A g_M, (11 T 6)

where A, and a are constants. Since H„ the total energy, includes a part
£m,ej, the condition \m,c\ = §kT (cf. 2.41) leads as usual to a = 1/kT; A,
can be related to the number-density. The formula (11.3, 6) specifies what
is known as the Boltzmann distribution. It can readily be shown to apply also
to gas-mixtures.

11.31. Boltzmann's closed chains

Lorentz's argument given in the last section determines/, on the assumption
that a steady state has been attained in which/, is an even function of P,\ it
does not prove that the gas must tend to such a state. For this reason
Boltzmann* gave an alternative discussion on the following lines.
He considered chains of collisions such that the coordinates and momenta
of the colliding molecules, and the impact parameters, after the/th collision
of a chain, lie in the same elementary ranges as do the corresponding variables
(for other colliding molecules) before the ( / + i)th collision of the chain. He
argued that, if such a chain is extended indefinitely, ultimately a collision
will be reached such that the coordinates and momenta after it are indistin-
guishable from those before the first collision, so that the chain is effectively
closed. By using closed collision-chains as a generalization of inverse col-
lisions, he was able to prove that in a uniform gas the integral
jjf.lnf.dQlnJP.
tends to a minimum value, and that, in the state corresponding to this
minimum value, In/, is a summational invariant. For a gas at rest, the formula
(11.3, 6) for/, follows as before.
The weak point of Boltzmann's argument was its assumption of closed
chains. Ergodic theory has gone far towards meeting objections to this
assumption.! However, though there is little doubt as to its essential sound-
ness, its detailed mathematical justification is not easy.

11.32. More general steady states

A more general steady state of the gas under no forces is obtained by
assuming In/, to be a general summational invariant, i.e.
In/, = «i« + « « . « l c ( - ^ H i " + ^ . ( r A V ,+*,), (11.32, 1)
* L. Boltzmann, Witn. Sit*. 95 (a), 153 (1887). Boltzmann's argument is expounded in
somewhat greater detail in R. C. Tolman's Statistical Mechanic!, chapters 5 and 6
(Oxford, 1938). See also D. ter Haar, Elements of Statistical Mechanics, appendix 1
(Rinehart. New York, 1954).
t See D. ter Haar, op. cit.
206 MOLECULES WITH INTERNAL ENERGY [11.33
where aj", a™, af*>, a*® are independent of Q„ P„ t. Such a form makes
/,'/t =/,/(, so that dgfjdt = o; # also, since the summational invariants are
constants of the free motion specified by the Hamiltonian H(/"\ it makes
^ , / , = o. As the notation implies, in a gas-mixture aj,1' takes different values
for the separate constituent gases, but or*, <x(3), et(<) are the same for all.
As in 11.3, since H'P includes the part Jw.cJ, the value of a(3) is i/kT. The
part of (11.32, 1) depending on c, is
- (mj2kT) {cj - 2k 7c,. (a«> + a<«' A r)},
so that the mass-velocity c0 is given by
c0 = *r(a<»> + «<« A r). (n.32, 2)
Thus the whole mass of gas moves like a rigid body possessing both rota-
tional and translatory motion, its angular velocity being kTat**; the state of
the gas is similar to that considered in 4.14.
To see the significance of the term of^.h, in (11.32,1), consider the
special case of rigid rotating spherical molecules. If I, is the moment of
inertia about a diameter, and to, the angular velocity,
H ^ = J«.f*+ il.to*, h, - /,uv
Thus the terms of (11.32,1) which involve «0, are
- (IJikT) (wj - 2Ar<x<«. w,).
Since to, is a linear function of P„ it follows that the mean value of to, is
equal to kTat*\ the angular velocity of the mass as a whole. In terms of c0, by
(II 32,2)
' * 2O, = VAC 0 . (».32.3)
This result also holds exactly for non-spherical rigid molecules, and
approximately for non-rigid molecules. The velocities given to the particles
composing a molecule as a result of the mean angular velocity to, are
extremely small; they are comparable in order of magnitude with the change
in c0 in a distance equal to the molecular radius.

11.33. Properties of the uniform steady state

For a non-rotating gas (ct(4) = o), (11.32, 1) takes the form
/,-«,Z.-VW, (".33. x)
where n, is the number-density, E, the thermal energy (differing from HJ,"
only by being measured relative to axes moving with the mass-velocity ct;
cf. 11.24). Since f (x J j n / . . - . -\
*' "."JSf.JcJto* (u-33.2)
* To justify the inclusion of the a"1 term in (11.3a, t), it it necewary to modify the
Boltznwnn collision-integral (it.aa, a), to take account of the small difference between
the position* of the centres of the colliding molecules (cf. 11.34).
11.33] MOLECULES WITH INTERNAL ENERGY 207
s
the quantity Z, in (11.33, 0 ' the partition function given by
Z, = jje-B'Tdc,JSl,. (11.33, 3)
An equation similar to (11.33, 1) also applies to a gas at rest in a conservative
field of force; now n, cc e~v>kr, where v, is the potential energy of a molecule
(cf. (4.14, 7))-
The mean value of E, corresponding to (11.33, 1) is given by

= Z; lffE,e-E<«Tdc,dSl, (11.33,4)
kT*dZ, . ,
( 33,5)
-ZTrfT' "-
by (11.33, 3)- The specific heat (per unit mass) of molecules m, is given by
, . d /E,\ kN, . .,

(cf. 2.44). This specific heat can be expressed as the sum of parts (cv)'„ (et)",
corresponding to the translatory and internal motions, where
(cr)', = 3A/2OT,, (c£ = (TV, - 3) k/2tn,. (11.33, 7)
An alternative expression for (<:,.), is obtained as follows. From (11.33,1
and 4-6), Yl = Z-'jJE^e-^^dc.dSl,

_ Lft (dE. . E. dZ,'

\dT Z.dTJ
= kT*m,(cr), + (E,)*. (".33.8)
It is convenient to introduce the non-dimensional quantity «f, defined by
*, = E,/kT. (11.33,9)
Then, combining (11.33, *>, 8),
\N, = I\-{g,)\ (,1.33, io)
Similarly* if £," is the internal energy of a molecule, and &}° = Ep/kT,

(cX = (kThn.)-^(E^-(Efn W.-3)-&*-(&?• (»-33. »)

* The proof of (11.33, 11) follows the same lines as that of (• 1.33, 5) and (11.33, 8). It is
merely necessary in the expression (11.33, 3) for Z, to replace EJkT by
Im.CjkT' + E'flkT.
and to differentiate with respect to T only; 7" is set equal to T after the differentiations.
 Copyrighted Material

208 MOLECULES WITH INTERNAL ENERGY [11.34

11.34. Equipartition of energy

A simple interpretation of the number N„ defined by (i i .33, 6) can often be
given. Suppose that in a gas at rest Es is expressible as a non-degenerate
homogeneous quadratic function of N, of the coordinates and momenta.
(This is the case if Et is composed of independent parts, consisting of kinetic
energies of translation and rotation, and kinetic and potential energies of
small internal vibrations.) By writing Pf = (kT)ip*, Q% = (kT)ttf for each
of these N', quantities, we can express the integral (11.33, 3) f° r Z, as the
product of {kTf^ and an integral independent of T. Hence E„ = \N'„kT,
by ("-33» 5); and so (11.33,6) gives N, = N't.
The equation E, = \N,kT in this case expresses the general principle of
equipartition of energy, which asserts that each of the coordinates and
momenta appearing in the quadratic expression for E, makes a contribution
\kT to the heat energy per molecule. For molecules fully free to rotate, as
well as possessing translatory energy, Ng = 6; for diatomic molecules this is
replaced by Nt = 5, because quantum restrictions exclude rotation about
the line joining the atomic nuclei. Each full degree of vibrational freedom
would increase JV, by two, corresponding to the additional P? and Q"
involved; however, quantum restrictions normally prevent molecules from
taking the full equipartition energy of their vibrational degrees of freedom,
so that each such degree increases Nt by less than two. Thus the principle of
equipartition of energy, as derived above for classical systems, is to be
regarded as an ideal principle rather than as one applicable to real gases.
However, apart from the restriction on diatomic molecules noted above,
there is nearly exact equipartition of energy between the rotational and
translatory degrees of freedom, save for light molecules at low temperatures
(cf. 17.63).
Assuming equipartition in a gas of rigid rotating spherical molecules, if
I, is the moment of inertia about a diameter, and o>( is the angular velocity,

Since It is comparable with m, x (radius)2, this shows that the period 2TT/O)S
of rotation of a molecule is in general comparable in order of magnitude with
the time taken by the molecule to travel a distance equal to the effective
molecular radius (cf. 11.23).
11.4. Non-uniform gases*
The general theory of a gas not in a steady state is here discussed for a binary
gas-mixture; the results for a simple gas follow as a special case. The first
* The discussion of non-uniform gases follows (with modifications) one given by N.
Taxman, Phys. Rev. n o , 1235 (1956); this in turn is based on a quantum discussion by
C. S. Wang-Chang and G. E. Uhlenbeck, University of Michigan Report CM-681,
1951 (cf. 17.6). The general results derived in the classical and quantum discussions
are identical; differences appear only in the detailed calculation of gas coefficients.

Copyrighted Material
11.4] MOLECULES WITH INTERNAL ENERGY 209
approximation t o / , is taken to be the Boltzmann expression
y r - »,z.-v«**' (11.4,0
(
(cf. (11.33, '• 3))- I rnight appear more correct to include a term corre-
sponding to the rotational summational invariant tpl*\ as in (11.32, 1).
However, this would affect only the value of A,; a variable value of c 0 is
already permitted in (11.33, ') through the term involving «]*!,*'. Also the
mean angular velocity w, corresponding to A, can be expected, as in
(11.32, 3), to be comparable with the space-derivatives of c0. Since these
space-derivatives are normally taken into consideration only in the second
approximation to/„ it is reasonable to neglect Jt, in the first approximation,
as in (11.33, ')• Actually Tt, cannot properly be taken into account in the
second approximation unless account is also taken of the difference between
the positions of the centres of colliding molecules, here neglected.
As in 2.41, the temperature is defined as that which would give a gas of
identical composition, in a uniform steady state, the same mean energy E
per molecule as the actual gas. The hydrostatic pressure is equal to knT to
a first approximation only (its deviation from this value corresponds to the
appearance of volume viscosity).
When A, is neglected, the second approximation to/, depends only on the
parameters nvnt, c0, and T, and on their space-derivatives. The first approxi-
mations to DnJDt, DcJDt, and DT/Dt depend only on the first approxima-
tion /> 0 (s nkT) top; thus (cf. (8.21, 2, 4, 6))
8 D c
A»»._ „ „ n oo _ v „ p dPo / „ A ,\
-ft=-».frco> P-fif -ZP.*.-^. ( » 4 . 2)

here N is the mean value given by

nN=-£n,N,. (11.4,4)

To the second approximation,/, is expressible in the form

/ . =/»'(1+<D?'). (11.4.S)
The equation satisfied by •DJ," is

- ^ ) = ir+M^-jjQ-;^. (»-4.6)
where
J#V*>) = £ jj\jfimAm(<^) + *P~4>'.lr-W)gbdbdeJc,dat.
("•4.7)
Since the parts of E, and H, involving the internal coordinates and
momenta are identical, by (11.4, 1) the terms involving differentiations with
210 MOLECULES WITH INTERNAL ENERGY [11.4
respect to these coordinates and momenta can be omitted on the right of
(i i .4,6). Thus, if / j " is regarded as a function of C„ r, t instead of c„ r, t,
equation (11.4, 6) becomes

(cf. (8.3, 3)), in which ff is regarded as depending on r and t only through

n, and T. Since E, depends on C, only through its part Jm.Cj, we get on
substitution from (11.4,1-3) and (11.33,5)>

I D9e0\ m,C, m, r r . d \

-/r[^C,.d. + ^,-i.-i)C..VlnT

+ 2V°V.:Vct+ ( § ( * * - 1 ) - J ( / . - / , ) ) V.«,j. (11.4.8)

Here <fj, <^ are the vectorsrfu,rf„( s -rf„)of (8.3, n), with/> replaced by
/>„; and S?„ / „ as earlier, denote the non-dimensional variables
V.-(mJzkT)lC„ *.-EJkT. (n.4,9)
The functions 4>^\ ¢ ^ must satisfy certain extra conditions in addition
to (11.4, 8), in order that, correct to the second approximation, n„ c0 and T
may represent the number-density of molecules m„ the mass-velocity and
the temperature, and that the total internal angular momentum shall be
negligible, as assumed. These conditions are

jjfl»Wdc.da. - o, SIJf2»Wm.C.Jc.Ja. - o, (11.4,10)

2jjfm^E.dc.da. - o, xjjf?Wh.dc.ja, = 0.
(11.4, n )
The conditions of solubility of (11.4, 6) can readily be found. For each
of the summational invariants i^H, tj/", $™,
2 fjuw*) V?dc,da, - o, (11.4,12)
since the expression on the left represents the contribution of the second
approximation to the total collisional change of a summational invariant.
(An explicit proof of this statement can be given along the lines of the
argument of 3.53.) The conditions of solubility are obtained on substituting
for JJiW0) from (11.4, 8) into (11.4,12). It is easy to show that these are in
fact satisfied.
11.5] MOLECULES WITH INTERNAL ENERGY 211

11.41. The second approximation to f.

The solutions of (i 1.4, 8) must be of the form (cf. (8.31, 1, 2))
¢ ^ = -.*!.VIn ^ - 1 ) , . r f , - 2 B i : Vc0-2B1V.c0, (11.41, 1)
,
¢ ^ =-.4.,.Vln7 -Z>2.rfl-2B2:Vc0-2B2V.c0. (11.41, 2)
In these, As, D, are vectors, B, is a non-divergent symmetric tensor, and B,
is a scalar. The quantities A,, D, and B, must derive their vector and tensor
characteristics from vectors associated with the motion of molecules m,. This
may at first appear to allow an excessive latitude to these quantities; but if/,
is regarded as having undergone the smoothing of 11.23, it can depend only
on vectors which remain constant during rotation and internal vibration.
Thus A„ Z>, and B, can depend on only two vectors, C, and the internal
angular momentum h„.
Since ft, is a rotation-vector (cf. 1.11), not an ordinary vector, the only
ordinary vectors that can be formed from C, and h, are linear combinations
of C„ h, A C„ h, A (A, A C,),.... But
h, A {h, A (*, A C„)} = - A? h, A C„
and so A, can be expressed in the form
A, = AlsC, + A]%AC, + Alllh,*(h,*C,), (11.41.3)
with a similar expression for D„. In (11.41, 3) A\, All, A\u are scalars,
which must be functions of quantities remaining constant during the
internal motions, like C\, h\, £,, etc. Similarly B, is a linear function of the
six symmetric non-divergent tensors that are quadratic functions of
C„h, A C , , h„ A (A, A C , ) , together with possibly the tensor h°hs; the
coefficients again are quantities not altered by the internal motions.
In virtue of the forms of A„D„ B„B„ equations (11.4,10,11) are found to
imply that jjjrBJc.M.-o. (,,.4..4)

xjjf»A,.m,C,dc,dSl, = o, (n.41, s)

^jjfS»D,.m,C,dc,da, = o, („.41,6)

%jjf«»B,E,dc,dSl, = o. (11.41.7)

11.5. Thermal conduction in a simple gas

The coefficients of viscosity and thermal conduction are here evaluated only
for a simple gas. From (11.4, 8) and (11.41, 1) the equation satisfied by A for
such a gas is J{A) mjmp_j_ l)C ( l}
212 MOLECULES WITH INTERNAL ENERGY (11.5

Also equations (i 1.41, 3, 5) become

A"A1C+AllhAC+AmhA(hAC), (".5, a)
fjf»A.mCdcda-Q. (11.5,3)
As a first approximation to A we adopt the form
nA - a£€*-\)C+aW-i®)C, (11.5,4)
where ax and ax are suitable constants; it may be verified that this form for
A satisfies (11.5, 3). However, (11.5, 4) is not the most general expression of
the form (11.5,2) and of third degree in the velocities which satisfies (11.5,3);
it is a special case of A1C only, and contains no terms corresponding to A11
and Am. It is closely analogous to the first approximation to A for a gas
without internal energy, but is liable to give rather less accurate results,
because of the extra omitted terms. If different parts of the internal energy
interchange with translational energy at altogether different rates, it may be
desirable to replace the second term of (11.5,4) by a number of separate
terms depending on these different parts of the internal energy.
The expression (11.5, 4) will, of course, not satisfy (11.5,1), since it gives
only an approximation to A; it is taken to satisfy the two equations obtained
by multiplying (11.5,1) by (V*-%)CdcdSl and (6®-£<•*>) Cdcdtl and
integrating.* This gives
«,«„ + «,«„ = V. (11-5. 5)
a, a,!+ «,<!„ = j(iV-3), (11-5,6)
where
«n - [<v-i)v,(v*-im alt - [(*»_})*, (*«>-^)«n,
("•5.7)
«11 - IC"-*»)*.(«*-*)«]. *u = [(f»-3*)9.(*»-&)9\
(11.5,8)
and [¢, \jr] is defined (cf. (4.4,7) by
"*[*, M • / / M W * A («.S. 9)
If <f> and r/r are both even or both odd functions of the momenta mC, J*",
an argument similar to that leading to (11.3,4), based on' reverse' collisions,
can be used to establish that in this casef
«w. *i - tillIlimn*+*i-?- &)
xty+fl-&'-P1)gbdbdedcldfl1dcd& (11.5, 10)
* The UK of such integrated equationt it equivalent to smoothing at in 11.13.
f If one of «1 and ^ i* even in the momenta, and the other it odd, by symmetry

where, to avoid convergence difficulties, the integral i* cut off where x < ' , as in 3.6. By
using reverse collisions, it follows that in this case [f, ¢) - - [ I , «>].
11.5) MOLECULES WITH INTERNAL ENERGY 213
(cf. (4.4, 8)), so that [<f>, \[r] = [t/r, <j>]: hence alt = an. The factor TV—3 in
(11.5, 6) comes from (11.33, " ) •
The thermal flux-vector is given by
q = - jjf>\A. V In T) ECdcdSl
= - J V In TfffWEA. CdcdSl
by (11.5, 4) and (1.42, 4). Since q = — A.VT, the first approximation to A is
found to be . .
Wi = jf)yEACdcda
= - ( 5 ^ 1 + (^-3)¾)

3 *»r 2 5 a „ - io(N-3)alt + (N-3ran

8m ana„-a\t < " * ">
on using (11.5,4-6) and (11.33, " ) •
In the bracket integrals an, alt, an the velocities* c, cl are now replaced by
&o> 9> defined as in (9.2, 6), the symmetric expression (11.5, 10) being used
(since E is an even function of momenta). An integration with respect to V9
then follows; the resulting expressions are simplified by use of the relation

which expresses the conservation of the total energy at collision. The final
results have the form
a
«u = f» + «ii. n = -f». «tt = l« + «ii- (u-5. " )
Here
a=
i* (^fjjjJj^pi-^-^-^^-n^dbdedada.df,

x{f*-2f*f'*coatx+9'*}gbdbdedSldSl1df, (11.5,14)
I A3
_ n*n

- 2ff' costfd™ - *n°) (#« -#F) + f'W -WWgbdbdedadi^df;

("•5. »5)
also x is the deflection of the relative velocity at collision, so that
9-9' - W ' c o s x .
* As in 4.1, the coordinates, velocities and momenta of the two colliding molecules are
distinguished by being written aa Q, c, P and Q„ c„ P, respectively.
214 MOLECULES WITH INTERNAL ENERGY (11.51
Symmetry arguments similar to those used in establishing the identity of
(11.5, 9, 10) imply that the integrals are unaltered if, in the integrands,
?'« and /•f/W-ZVO* are replaced by y« and t\£™-#{>)*.

11.51. Viscosity: volume viscosity

By (11.4, 8) and (11.41, 1), the equations satisfied by B and B for a simple
gas are 0
J(B)-/"W, (ii-Si.O
m =/^(^-1)-^(/-/)). (11.5». *)
The first approximations to B and B are taken to be

»B=»Ai*V, (ii.5*»3)

-^-4.(,(^-1)-(/-/)). (11.51,4)
The first of these corresponds to the first approximation of Chapter 7. The
second can be shown (using (11.33, 10 )) to satisfy equations (11.41,4,7),
and it is the simplest expression that does so.
The coefficients bv bt are determined from the equations derived by
multiplying (11.51,1 and 2) by

Wdcda and [j(«* - } ) - ^ ( / - / ) ) dcdO.

and integrating. This gives
Mu-I, btbn-(N-3)/(,N, (ii.Si.5)
where 6U = [«%\ *&],

*M - [i(«*-i)-^(/-/). i(«"-i)-^(/-/)].
In these bracket integrals, expressed in the symmetric form of (11.5,10), the
velocities c„ c, are replaced by 9 „ f, as in 11.5. Integration over all values
of V0 then gives , , , , . , ,,
0
* *u-«u-f«> *M=*K (ii-5'.6)
where a and a'n are the integrals of (11.5,13,14).
The second approximation to/makes a contribution p(U to the pressure
tensor, where ?m _ _2UrmCC{*:Vc, + BV.c,)dcda.
0
This depends on space-derivatives of c0 only in the combinations Vc0 ( z 8)
and V.c 0 . Since integration cannot introduce further tensors other than the
unit tensor U, the integration yields a result of the form
p(u = _ 2 / t 8 - m U V . c 0 . (11.51,7)
11.511] MOLECULES WITH INTERNAL ENERGY 215
Here (i denotes the usual 'shear' viscosity; m is the 'volume' viscosity of
i I . I , measuring the resistance of the gas to changes in density. Using
(1.42, 2) and the theorem of 1.421, the first approximations to /1 and m
corresponding to (11.51, 3, 4) are found to be
Mi = ikTffjwW: BdcdSl
= 5*7726,,, (11.51,8)

("•51.9)
11.511. Volume viscosity and relaxation
The volume viscosity introduced above may be interpreted as a relaxation
phenomenon. It arises because in an expansion or contraction the work done
by the pressure alters the translatory energy immediately, but affects the
internal energy (through inelastic collisions) only after a certain time-lag.
We can introduce temperatures Te, T{ corresponding to the translatory
and internal energies, differing slightly from T; these are such that ^mC*
and £"*have the same values as in a gas in a uniform steady state at tempera-
ture Tt or T(. The specific heats c'v, c"r corresponding to the translatory and
internal energies are 3^/2»« and (N—2)kJ2m; hence if the temperatures Te
and 7] are to give the correct total energy for the overall temperature T,
3(Te - r ) + (N- 3) (Tt - r ) = o, (u.511,1)
..DT DTe ... .DTt .

Collisions may be assumed to transfer energy from translatory to internal

motions at a rate proportional to Tt — Tt. Thus, ignoring the effects of heat
conduction and shear viscosity, the equations of change of Tt, Tt can be
written down separately in the forms

ikn^j = -pV.c0-~lk»(Te-T(), (n.511,3)

3 3 3 3
^ T * " ^ - ^ * ^ ^ - ' (»-5».4)

wherep == nkTt. The quantity r is a time of relaxation; as (11.511, 4) implies,

in a non-normal state in which Tt is maintained constant, T— Tt decays
exponentially, with decay time r.
Equations (11.511, 3, 4) can be combined to give

^(7.-33-.-KV.Cb-^(3-.-¾).
216 MOLECULES WITH INTERNAL ENERGY [11.52
In a normal expansion, the time-derivative on the left is small compared
with the terms on the right; hence

or, using (11.511, 1), and neglecting the small difference between Te and T
in the already small term on the right,

Tt-T=-^^rTV.c0. (II.SH.S)
Thus the effect of relaxation is to alter/) from^, s nkT to

/, = ^ r e = / > 0 ( , - ^ 3 ) r V . c 0 ) ,

i.e. to introduce a volume viscosity m such that

m
" N* PoT' («-S».6)
Agreement with (11.51,9, 6) is obtained by assuming that, to a first approxi-
mation, T is given by ., x,
1 7V-3 N-2 , ,
Formulae like (11.511, 5) or (11.511,7) are valid only when Tt — 7J is small,
i.e. when r is small compared with the time-scale of expansion. For very
rapid temperature-changes it is necessary to revert to equations like
(11.511, 3) and (11.511, 4), which still remain approximately true. If r is
very large compared with the time-scale of expansion, (11.511,3) approxi-
mates to the temperature equation for a monatomic gas.
Equation (11.511, 6) can be generalized to apply when there are several
different parts of the internal energy, each with its own relaxation time.
If each part of the internal energy is supposed to interact only with the
translatory energy, the generalized form is

•-j$s(tfi*i+JVi+->- (11.511.8)
Here TV,, JV„ ... are the contributions to N from the separate parts of
£">, and r„ r t , ... are the corresponding relaxation times. As before, the
interpretation of the relaxation effect simply as a volume viscosity is justified
only if all the relaxation times are small compared with the time-scale of
expansion.

11.52. Diffusion
The coefficient of diffusion in a binary mixture depends on the vectors D, of
(11.41,1, 2). These satisfy the equations
./,(0) -xr l /rc„ UD) =• -xsVTC,. (11.52.0
11.6] MOLECULES WITH INTERNAL ENERGY 217
A first approximation [£>lt]i is obtained by putting
xtDi = </0C„ x,Z>, = -d„Ct, (11.52, 2)
where d0 is a constant; these forms satisfy (t 1.41, 6). The value of </0 is found
by substituting these expressions into the equation

x,-1 ffcj.y^Dj^Cirfnj-x,-1 JTc2. y,(z>) rfc.^n,

= 2x.-,JJ/J°»Clrfc.rfn.) (11.52, 3)

which replaces (11.52, 1). The right-hand side of (11.52,3) reduces to

2kTpnsJ(pip1) on integration. On the left, the conservation of momentum
secures the vanishing of the contributions from the parts of Ji(D), JJ.D)
corresponding to the collisions of like molecules. In the remaining integrals
on the left, a symmetry relation like (4.4, 10) is used, and the integration with
respect to "S0 is then performed, as in 11.5 and 11.51. Then (11.52,3)
reduces to
d
° = St< (»-5*4)
°P «00
where

x(9*-2f9'cosx+f'^gbdbdedSlidSlttif. (11.52, 5)
The velocity of diffusion Ci — Ct corresponding to (11.52, 5) is readily
found to be —npd0kT'dyjpxpt. Hence (cf. (8.4, 7))
(1I52 6)
i^-SS-siSfT-- '
The coefficient of self-diffusion in a simple gas is obtained from (11.52,
5, 6) by setting m1 = mt% Z, = Z t , etc.
The formulae governing thermal diffusion for molecules with internal
energy are very complicated, even to a first approximation. The theory for
molecules without internal energy is usually assumed (with a measure of
justification) to apply approximately to molecules with internal energy also.
11.6. Rough spheres
These results are first applied to the rough elastic spherical molecules of
Bryan and Pidduck, these being by definition such that at a collision the
relative velocity of the spheres at their point of impact is exactly reversed.
It can readily be verified that the total energy is conserved at collisions of
such molecules.
The motion of a molecule m, can now be specified completely by c„ <•>„
where c, is the velocity of its centre and u , its angular velocity. The
 Copyrighted Material

218 MOLECULES WITH INTERNAL ENERGY [11.6

components of u>, are not themselves Lagrangian velocities. Nevertheless,
a theory exactly analogous to that given above can be constructed in terms
of a generalized velocity-distribution function fjp, c„ u»,), such that the
number of molecules m, per unit volume with velocities and angular
velocities in the ranges dc„ da>t is ftdcadta,. The Boltzmann equation
satisfied by/, in a binary gas-mixture takes the form

|+c,.g'+F f .^ = •LJflifJt-fJtKgt.tdkdctiut.
(11.6, i)
H e r e ^ = ct—c„ ar* = 4(<rs+«rt), where tr, is the diameter of a molecule
mt; also/, =fj[c't,bi',), f't =/t(C{,wJ), where c'„ wj and c't, <a't are the
velocities and angular velocities before a collision of a pair of molecules
whose velocities and angular velocities after collision are c„ w, and ct, <ot;
and k denotes a unit vector in the direction of the line from the centre of the
molecule m, to that of the molecule mt at collision, so that#t,.k > o.
Let Ilt It be the moments of inertia of molecules mv ml about their
diameters. We write
jq = 4 / ^ 0 - 5 , Kt = ^IJmta%, m0K0 = m1K1+miKt. (11.6,2)
The values of Klt Kt are 0-4 for uniform spheres. As the effective radius of an
actual molecule is determined by its outworks, and the moment of inertia
by the disposition of the atomic nuclei, Kv Kt are likely to be less than 0-4
in all practical cases. In terms of K0, Kv Kt it can be shown that, in a collision
of molecules m1 and n^,
c[ = ^ + 2 ^ , ( ^ ^ + ^ ^ . ^ ) / ( ^ ^ 2 + ¾ ) . (11.6, 3)
c, = c,-2M1{/c1/i:,V+Kok(k.V)}/(K1K8+K0), (n.6, 4)
to>J =. <o1+4ilf,K1kAV/(r1(/i:1/ft+K0), (11.6, 5)
u>i = wt+4Af1/c1k AV/O-J^ATJ+XO). (11.6, 6)
In these equations V is the relative velocity after collision of the points of the
spheres which come into contact, i.e.
V = c,+$<r,k A u>» - ct+i^k A tov (11.6,7)
In the uniform steady state,

/. = «.(^^exp{-KC*+/,W;)/2Ar}, (11.6, 8)

andso(cf. (11.34)),^ =• N, = 6,andy «= $. This value is closely approached

by methane and ammonia, for which y is 1-310 and 1*318 at ordinary
temperatures; methane is well represented by a spherical molecule, because
of the symmetric arrangement of its hydrogen atoms.

Copyrighted Material
11.61] MOLECULES WITH INTERNAL ENERGY 219
In the limit as K„ K2 (and in consequence also / l f It) tend to zero, the
energies of translation and rotation cease to be interconvertible. For,
according to the equipartition principle, \ntaC*= \I,b)\ = \m,K,o*<i)\;
thus in the limit as K, -*• o we must regard c, as small compared with icr,to,,
but large compared with ^ t r ^ w , . Thus, after substitution from ( n . 6 , 7),
equations (11.6, 3-6) approximate to
^ = ^ + 2M,k(k .gtl), c't = ct- 2M l k(k .gn),
&>;=<>>, + (2MtKs/K0 o-j) k A [k A (o-j w t + <rta»,)],
wj = Wj + (2Mt KJK0 cr2) k A [k A ( a , <«>! + <rtw,)],

showing that the linear and angular velocities have become uncoupled.
However, the changes in angular velocity remain significant, even though
they do not affect the linear motions.

11.61. Transport coefficients for rough spheres

The collision integrals a'n, a, a'2i and dw of 11.5 and 11.52 can readily be
calculated for rough spheres,* using (11.6, 3-6), with gbdbde replaced by
c2#2i-k</k or <r{2gtl.kdk. Their values are

a u a
- \ i r ) (i+K)*' "-* \m ) (!+*)••
' _ *lnftT\l 6 ( l + K + 2/C2) . _ , (27lkTmt\i Kt + ZKiKj
"" - * \-JZ-J {1+K)* • "oo - *,, \-j^-J ^ + ^•
The corresponding first approximations to the transport coefficients are
giVCnby
[A], _^(^il±*m±lS!*±X*!}t (,,6,,,)
1
" ib<r2\Jim) 12 + 7 SK+101 K*+i02Ka' v
'
r . 15 /Arm\t(i+*r)« . . ,

r . 1 /kTm\i (I+JC)1 , , .
(II 6l,3)
w»-ji?(irj * • -
v^-dr (££)'££&• (»•*«. 4)
8/I<rfa \2 ITM]OTj/tf0+ 2K t Xj
The viscosity given by (11.61,2) equals that for smooth spheres if K = 0
(when the translatory and rotational motions become independent); it is
1 -05 times as great for /c = 0-4, but the two differ by less than one per cent if
K< 02. The thermal conductivity given by (11.61, 1) is always greater than
that for smooth spheres, because of the transport of internal energy; the
* In practice it is normally best to use the unsymmetric expression (11.5. 9) in calculating
the bracket integrals.
220 MOLECULES WITH INTERNAL ENERGY 11.62]
ratio of the two increases steadily from i -48 to 1 -533 as K increases from zero
to 0-4. The increase comes from the increased transport of rotational energy,
due to the tendency of rough spheres to knock back molecules with which
they collide, and at the same time to transfer part of their rotational energy
to them. In fact, the part of [A], due to the transport of translatory energy
(corresponding to the a, term in (11.5, 4)) is found to decrease in the ratio
0*916 as K increases from zero to 0-4, whereas the part due to the rotational
energy increases in the ratio 1-28. For rough spheres N — 6 and so cv =» 3/-/01
(cf. (11.33, 6)). Using this, the ratio [f], - [A],/{[/c],cv} is found to equal 1-85
for K •« o, and it lies between 1-87 and 1-825 f° r values of K between zero
and 0-4.
As Kohler* first pointed out, the volume viscosity [TO], given by (11.61, 3)
is not small compared with f/t],; it is 7[/<]i/'5 when K has its greatest value
0-4, and becomes large compared with [/»], when K is small. This agrees
with the relaxation expression (11.511,6), because the interchange between
rotational and translatory energies becomes indefinitely slow as K-+ 0.
Finally, the diffusion coefficient given by (11.61, 4) agrees with that for
smooth spheres if either x, or Kt is zero, but becomes smaller as #r, and Kt
increase; if K, =» /c, » K0 = 0-4, it is only seven-ninths as large. Its decrease
is due to the increased deflections of the relative velocity at collision as
it, and Kt increase.
The terms involving h,sC, and h,A(A,AC,) in (11.41, 3) and similar
relations may appreciably increase the calculated values of the transport
coefficients above the first approximations (11.61,1-4). Condiff, Lu and
Dahler.-f using an approximation to A involving terms up to the fifth degree
in C and <i>, found values of A ranging from i-o66[A], to about 11 [A], as
K increases from 0 to 0-4. Similarly, taking into account terms of fourth
degree in B, they found values of /i ranging from 1-015(^, to about 1-06(/(),
as K increases from 0 to 0-4. They also calculated second and third approxi-
mations to to; the third approximation is equal to [ID], when K is negligibly
small, but about 3 per cent greater when K — 0-4. They remark, however,
that the effect of the h, A C, and h, A (h, A C,) terms is likely to be greater for
rough spheres than for more realistic models.

11.62. Defects of the model

The Bryan-Pidduck rough-sphere model is unsatisfactory in more than one
respect. First, as with all rigid models, the deflection of the relative velocity
at the collision of two molecules depends only on the ratio of their velocities,
and not on the actual speeds. Secondly, because of the reversal of the relative
• M. Kohler. Z. Pkyt. 114. 757 («947); » $ • TS 0949)-
t D. W. Condiff. W.-K.Lu and J. S. Dahler, J. CHtm. Phyt. 42.3443 (1063). L. Waldmann
(Z. Naturf. 18a, 103} (1963)), retaining only the term h, A C r found for a Lorentzian
gae a decreaae in D„ below [D,,]x; however, thia ia replaced by an increaae if the
A, A (*, A CJ term ia alto retained.
11.7] MOLECULES WITH INTERNAL ENERGY 221

velocity of the points coming into contact, even a grazing collision can
produce a large deflection. Finally, experiment suggests a relatively slow
interchange between translatory and rotational energies; such a slow inter-
change is obtained by taking KV Kt small, but (contrary to what one would
expect of actual molecules) this still permits free interchange of rotational
energy between the molecules.
Chapman and Hainsworth* attempted to modify the model to meet the
first of these objections, but their attempt was itself open to serious objection.
The third objection can be met by assuming the molecules to be rough in
only a fraction of the collisions, and smooth in the rest; but this assumption
is artificial. The second objection appears to be insuperable. Thus, despite
its interest as giving almost the only exact results for gases whose molecules
possess internal energy, the rough-sphere model cannot represent the
behaviour of actual gases at all closely.

11.7. Spherocylinders
Curtiss and Muckenfussf have given formulae from which the transport
coefficients can be calculated for any smooth rigid convex form of molecule
without an external field of force. They worked out explicit expressions for
spherocylinders—cylinders of length L and diameter <r, with hemispheres,
also of diameter <r, fixed to the plane ends. Their formulae involve two
parameters, fi a L\<r and K a ^I/mo*, where I is the total moment of
inertia about a line through the centre of the cylinder and perpendicular to
its axis. When fi = o the molecules are spherical.
The formulae, though involving only elementary functions, are too com-
plicated in form to be quoted here, and indeed Curtiss and Muckenfuss did
not give them all explicitly. Their results can be summarized roughly as
follows. The effective collision cross-sections in integrals like the a'u, a£t of
11.5 are larger than those for spherical molecules of the same volume, unless
K is small, by an amount which increases with increasing p (i.e. with
increasing departure from the spherical form). However, if K is small the
impulsive forces between molecules at collision can produce large changes in
the angular velocity without appreciably affecting die motion of the centre,
and so the transport of momentum and translatory energy are both increased.
Thus the transport coefficients are less than those for smooth spheres unless
K is small, but greater if K is small (for A the comparison is with the modified
Eucken formula (11.8,4)). The volume viscosity w is always greater than
the shear viscosity for smooth spheres of the same volume. The ratio of the
* S. Chapman and W. Hainsworth. Phil. Mag. 48, 593 (1024). The effective diameter was
supposed to vary with g\ the relative velocity of approach of colliding molecules.
However, since g' is not conserved at collisions, this means that direct collisions cannot
be balanced against reverse collisions, ss in it.3.
t C. F. Curtiss, J. Chem. Phyi. 24. 115 (1956); C. F. Curtiss and C. Muckenfuss, Aid.
a6, 1619 (1957). and »9, 1157 (1958).
222 MOLECULES WITH INTERNAL ENERGY [11.71

two is very large if K is small, and also if ft is small, i.e. if the molecules are
nearly spheres; in either case the interchange between rotational and
translatory energies is slow.

11.71. Loaded spheres

For the loaded sphere model Dahler and his collaborators* obtained results
depending on the one parameter a = m^jzl, where £ is the distance between
the mass-centre G and the geometrical centre O, and I is the moment of
inertia about an axis through G perpendicular to OG. They found that the
ratio f a A//tc„ has approximately the value 1-98 given by the modified
Eucken formula (see 11.8, below) for all physically admissible values of a,
a decrease in the part contributed by translational energy being nearly
balanced by an increase in the contribution of internal energy. They also
found that different eccentricities of mass-distribution can produce a non-
vanishing thermal diffusion factor in a mixture of two gases whose molecules
are identical in mass and radius.

11.8. Nearly smooth molecules: Eucken's formula

Nearly smooth molecules are defined as such that a large number of collisions
are needed in order to produce any considerable change in the internal
energies.
Euckenf suggested an approximate expression for the thermal con-
ductivity, which strictly is applicable only to a gas of nearly smooth mole-
cules. He divided A into two non-interacting parts A' and A", which are
respectively the conductivities due to the transport of translatory and
internal energy; similarly he divided c„ into parts c'„ and c"v. For a monatomic
gas, A is very nearly equal to §(ur (cf. (9.7, 4)); Eucken assumed by analogy
that A' — $ftc'v. On the other hand, since there is little correlation between
the speed of a molecule and its internal energy, the argument given at the
end of 6.3 suggests that A" — /tc^, and Eucken assumed this; the assumption
is equivalent to supposing that the mean free paths effective in the transport
of momentum and internal energy are equal. Thus Eucken wrote

A-MK+O- ("-8.0
Now (cf. (11.33, 6, 7)) c„ c'v and c"r are given by
Nk , 3* . (W-3)A , Q ,
c c c
» = ™> i ™- «~ , m » (118,2)

• J. S. Dahler and N. F. Sather, J. Chem. Phy$. 33, 2029 (1961). and 38, 2363 (196a);
S. I. Sandler and J. S. Dahler, J. Chem. Phyt. 43, 1750 (1965). 46, 3520 (1967), and
47, 2621 (1967).
t A. Eucken, Phyt. Z. 14, 324 (1913).
 Copyrighted Material

11.8] MOLECULES WITH INTERNAL ENERGY 223

and if y is the ratio of specific heats, N = 2/(7 — 1). Hence (11.8, 1) is
equivalent* A = ^(97-5)^, (,,.8,3)
This is known as the Eucken formula.
However, for nearly smooth molecules the transport of internal energy
takes place by the diffusion of molecules from one part of the gas to another,
carrying with them the mean internal energy of the region in which they
originate. Thus the free path effective in the transport is that effective in
diffusion, and one should write A* = pDnc"v. Hence (11.8, ,) should be
replacedby A - ^ + ^ 4 (,,.8,4)
Putting pDu — u'nfi (cf. (6.4, 4)) one thenfindsinstead of (,1.8, 3)
A = i { i 5 ( y - 0 + 2Un(5-37)}/"V ("-8. 5)
The numerical factor u'u in general exceeds unity. For smooth rigid elastic
spheres it is ,-20, for Maxwellian molecules ,-55, and for most of the usual
molecular models it lies between these values. Hence (,,.8, 5) indicates a
somewhat greater value of f = \l/icv than does (,,.8, 3).
The modified Eucken formula (,,.8, 4) can also be derived from the
general expressions given in 11.5 and 11.52. For nearly smooth molecules we
can approximate by writing f = f\ 6® = && in the collision-integrals; also
we may average functions like {8^-6^ over the internal coordinates
and momenta, because ^ ° is not correlated with the translational velocities.
Hence the integral a of (,1.5, 13) vanishes, and so by (,,.5, ,2) and
(,,.5,,6,8), ^ = ^ = 5 ^ / 2 ^ , flll = o. (,,.8,6)
Also a22 = a'22t and by (,,.5, ,5),

xiP-zn'cosx+ftebdbdedfdSidSly (,,.8,7)
But, since &® and 8^ are uncorrected,
(g®-6fY = 2 ( ( ^ - ( ^ 0 ) 1 } = JV-3,
by (1,.33, 11); also the integral in (11.8, 7) is the same as that in the diffusion
expression dw of (,,.52, 5, 6). Hence
«« = a'2i = ( ^ - 3 ) ( 3 ^ 7 7 4 ^ 0 ^ . (,,.8, 8)
Using these values of an, an in (, ,.5, , 1) we find

qnt 20t

^Mi+^IM, (11.8,9)
by (,,.8, 2). This agrees with (11.8, 4), to thefirstapproximation.
Copyrighted Material
224 MOLECULES WITH INTERNAL ENERGY [11.81

11.81. The Mason-Monchlck theory

Mason and Monchick* have attempted to improve the modified Eucken
formula, assuming that the interchange between translatory and internal
energy is slow, but not completely negligible. The integral a of (i 1.5,13) is
then small; by (11.511,7), it is related to the first approximation [r], to the
relaxation time of 11.511 by
«-"V-3)/4»Mi. (11.81,1)
According to (11.5,12) and (11.51, 6, 8)

an.6ll+VaD^:+?s(^i3). (II.8l,3)
,J/
2[/t]i 24 rt[T]t
In the equation an = \a+a'tt, the expression (11.8,8) for a'n is still used,
but with [Dull replaced by £>lnt; D,nt is interpreted as the first approxima-
tion to a coefficient of diffusion of internal energy. Thus
_3(N-3)kT ,3(AT-3) , „ « . .,
a
» " 4pDlBt + 8n[r]r- {
"*l'»
Using these values, and substituting for c'„ c"v from (11.8, 2) we find from
(11.5, 11) after considerable simplification,

(II.8I,S)
Apart from the replacement of [Dn]x by Z>lBt, (11.81, 5) differs from the
modified Eucken formula (11.8,4) only through the appearance of the last
term on theright.It implies that, if DiBt is identified with [Du]u then [A],
and the ratio [f]t • [A]i/[/t],c„ are always smaller than is indicated by
(11.8,4)-
It is questionable how far D,n, can be identified with [Dn]v so that aj, can
still be given by (11.8, 8). As the form of (11.5,13) indicates, the collision-
integral a depends on the square of the energy-interchange between the
translatory and internal motions at collision, and it is not clear that (11.8, 8)
is correct to terms of second degree in this energy-interchange. In the
integral (11.5,15) the mean values of (^° - ^ 0 ) 1 and (f® - /IP*)* are equal,
because to every direct collision corresponds a reverse collision; but since
( ^ 0 - ^ - £ W — ^ ' ) ' ) , is essentially positive, the mean value of
(<f<>-*«>)(rfW-oD
* E. A. Muon tnd L. Monchick, J. Chem. Phyt. 36, i6aa (106a).
11.81] MOLECULES WITH INTERNAL ENERGY 225
must be less than that of (&*>-#<>)*. Thus ( u . 8 , 8) gives an adequate
approximation to (11.5, 15) only if, for those collisions in which £& — #}*
differs appreciably from S^ — Sf*, the average value of gg'coax *9 nearly
zero, i.e. the mean value of x is about \n.
For most of the realistic molecular models, encounters with X< \ n a r e
more important for diffusion than those with x > in- On the other hand,
collisions producing appreciable changes in internal energy are likely to be
weighted towards the larger values of x- Hence, in default of detailed analysis,
using £>,nt = [Z7lt], in (11.81, 5) may in many cases be a reasonable approxi-
mation. The rough spherical model indicates a slower decrease of [A]t with
increasing K than would be consistent with taking Dtni = [Dn]v This is
because of the peculiar properties of the model, which make the mean value
of gg' cos x for colliding molecules negative (corresponding to a mean x
greater than In), whereas the mean value of (<?">—^){^ w —f^lgg'cosx
is positive. Such peculiarities are not likely to appear with more realistic
models.
Even so, systematic deviations from DlBt = [D^ are likely with certain
molecular models. For example, as Mason and Monchick suggest, molecules
with permanent dipole moments may be able to interchange internal energy
without the dipole moments appreciably affecting the translatory motion at
collision. This would imply a reduction in DiBt below [Djjj, leading to
systematically lower values of A.
Mason and Monchick in their original paper approximated by assuming
both that D, n , = [£)„]„ and that [r], is so large that terms of order [r]f *can
be neglected. In this case (11.81,5) is replaced by
Wi = i[/']i<+p[Dn]1c;-ic^M1-p[Du)1y/p[T]l. (11.81,6)
Equation (11.81, 6) can be generalized to the case when there are several
different forms of internal energy, each interacting with translatory energy
with its own relaxation time. The effect is simply to replace c«/[r]t in the last
term on the right by k ( Nx Nt \

where, as in 11.511, N^/im, Ntk/2tn,... denote the contributions to c"v

from the separate forms of internal energy. Energy with a large relaxation
time makes only a small contribution to this sum, whereas it may make a
large contribution to the volume viscosity.
 Copyrighted Material

12
VISCOSITY: COMPARISON OF
THEORY WITH EXPERIMENT

12.1. Formulae for ft for different molecular models

The various formulae obtained for the coefficient of viscosity /i are grouped
together here for convenience of reference. The formulae for \ji]v the first
approximation to fi, are as follows:
(i) Smooth rigid elastic spherical molecules of diameter tr, (10.21, 1),

(ii) Molecules repelling each other with a force Af/f, (10.32, 1),

(iii) Attracting spheres, diameter a (Sutherland's model, (10.41, 8)),

w.-^(^7(- + f)- <•>->

(iv) Lennard-Jones and exp;6 models ((10.1, 4) and (10.4,7)),
w
»-si*(^7 #m (i2i,4)
where if^\z) is a function of kT/e, given numerically in Table 6 (p. 185)
for the Lennard-Jones 12,6 model, and illustrated in Fig. 8 (p. 185) for this
and the exp;6 model.
(v) Rough elastic spheres of diameter <r (the Bryan-Pidduck model,
(11.61,2)), /kmT\i(i + K)* . ,

where K is given by (11.6, 2).

Further approximations have been determined for most of these models.
For the first (cf. (10.21, 4)), the value of ft, correct to three decimal places of
the numerical factor, is
/1 = 1-016(/1^ = 0-1792(^17^/0-8 = o-^ggpCl (12.1, 6)
(cf. (4.11, 2), (5-2i, 4) and (6.2, 1)), and for the second (cf. (10.32, 2))

fi = M i I ,+ fc£ +
2(1/-1)(101^-113)
1 226 ]
Copyrighted Material
12.11] COMPARISON OF THEORY WITH EXPERIMENT 227
Thus for rigid elastic spheres /i is 1-6 per cent in excess of [/*]„ while for
Maxwellian molecules (v = 5) /1 is identical with [/*]t: for values of v between
5 and 13, fi is in excess of [/t], by less than 0-7 per cent. For other smooth
models the correction to [/«], is normally a fraction of 1 per cent. We may
therefore expect that no great error will be incurred for most smooth models
if the first approximation [//], is used in place of the true value ft.
On comparing (12.1, 1) with (12.1, 5) it appears, as in 11.61, that the
possession of internal energy by the molecules does not seriously affect the
rate at which they transport momentum, since the ratio of (12.1, 5) to
(12.1, 1) varies only between 0-994 and 1-05 as K ranges from 0 to the
uniform-sphere value 0-4. The effect of later approximations is, however,
more important for rough molecules than for smooth.

12.11. The dependence of viscosity on the density

Each of the formulae of 12.1 predicts that the coefficient of viscosity of a
simple gas is independent of its density: this is a general result independent of
the nature of the interaction between molecules (cf. 7.52). Thus the capacity
of a gas for transmitting momentum, and so for retarding the motion of a
body moving in it, is not decreased when its density is diminished. This
surprising law was first announced by Maxwell* on theoretical grounds, and
afterwards verified experimentally by Maxwell and others.
The consequences of this law are interesting. For example, in so far as the
damping of the oscillations of a pendulum is due to the viscous resistance
of the gas in which it moves, the degree of damping will be independent of
the density of the gas, and the oscillations will die away as rapidly in a
rarefied gas as in a dense gas. This was first noted by Boyle, f
Again, according to Stokes's law, the velocity of a sphere of mass M and
radius a falling under gravity (g) in a viscous fluid tends to the limiting value
g(M-M0)l67ra/i,
where A/0 is the mass of fluid displaced by the sphere. When the fluid is a
gas, M0 is negligible compared with M, so that the limiting velocity of fall
should be independent of the density.
When, however, a body moves at a high velocity in a viscous fluid, the
ordinary laminar motion of the fluid is unstable, and gives place to a tur-
bulent motion. In this case the conduction of momentum away from the
body is by eddies, not by the ordinary viscosity of the fluid. Thus the
resistance of a gas to a body moving at a high velocity is not the ordinary
viscous resistance, and may depend on the density.
Even when no turbulence is present, the viscous resistance of a gas shows
some dependence on the density. The volume of the molecules, though a
• J. C. Maxwell. Phil. Mag. 19. 19 (i860); 20. xi (i860); Collected Works, vol. 1, p. 391.
t See S. G. Brush, Kinetic Theory, vol. 1, p. 4 (Pergamon, 1965).
 Copyrighted Material

228 VISCOSITY: [12.11

small fraction of the volume occupied by the gas, is not wholly negligible;
this implies a correction to the preceding theory, which is investigated in
Chapter 16. Also in a vapour near to condensation, the molecules form
aggregations consisting of several molecules, and again the preceding theory
and its results need modification. A dependence of /t upon/3 for a number of
gases has, in fact, been determined experimentally* as a series

ji = a+bp+cp*+....
For a gas at ordinary densities the terms of this series after the first are small,
and in what follows we shall ignore the dependence of fi on p. The pheno-
menon of slip at the walls at low densities, noted in 6.21, may need to be
taken into account in interpreting experimental results, but it does not
represent a true dependence of/t on the density.

Table 11. Viscosities and molecular diameters

of gases at S.T.P.

Molecular I07X/t to* x<r

Gas weight (poise)t (cm.)

Hydrogen 2016 84S a-745

Deuterium 4029 1191 2-75«
Helium 4003 1865 2193
Methane 16-043 1024 4-187
Ammonia 17031 9*3 4-477
Neon 20-183 2975 2-002
Acetylene 26038 948 4912
Carbon monoxide 28-011 163S 3810
Nitrogen 28-013 1656 3784
Ethylene 28-054 927 5-062
Air — 1719 —
Nitric oxide 30007 1774 3720
Ethane 30070 858 5353
Oxygen 3» 1919 3636
Hydrogen sulphide 34-080 1163 4743
Hydrogen chloride 36461 1328 4-514
Argon 39-944 2117 3-659
Carbon dioxide 44011 1380 4-643
Nitrous oxide 44-013 135» 4-692
Methyl chloride 50-488 968 5737
Sulphur dioxide 64-064 1164 5-551
Chlorine 70-906 "33 5534
Krypton 83-80 2328 4199
Xenon 131-30 2107 4939

* See, for example, reference (i) in the list at the end of this chapter. Later references to
this list will be indicated by figures in the text, as (i).
f 1 poise = 1 g./cm. sec

Copyrighted Material
12.3] COMPARISON OF THEORY WITH EXPERIMENT 229

12.2. Viscosities and equivalent molecular diameters

The viscosity of a gas can be measured absolutely, or relative to that of a
standard gas. The absolute determination is the more difficult, and conse-
quently subject to greater experimental error; on the other hand, a relative
determination is affected by any inaccuracy of the value adopted for the
standard gas. Dry air is commonly used as a standard gas, and its viscosity
has been determined with great care by a number of workers. However, even
for air the probable error of a weighted mean of values of ft is about o-1 per
cent of the true value. For other gases the probable error may be several times
as large. This should be borne in mind in considering the comparisons with
individual runs of experimental values given in this chapter.
Table 11 gives experimental values* of /i for a number of the common
gases at S.T.P., and also equivalent diameters at o °C. derived from them,
using (12. i, 6). The table shows that the diameter increases somewhat with
increasing molecular weight, and also with increasing complexity of the
molecules. The viscosity itself increases much more slowly with molecular
weight than would be suggested by the mi factor in (12.1, 6).
The molecular diameter deduced from /1 in general differs from the
diameter derived from other properties of a gas, such as the equation of state.

THE DEPENDENCE OF THE VISCOSITY

OF THE TEMPERATURE
12.3. Rigid elastic spheres
The various formulae of 12.1 give different relations between the coefficient
of viscosity and the temperature. If the molecules are rigid elastic spheres,
/i°cTi, (12.3, 1)
while all the other models imply a greater degree of variation of viscosity
with temperature. Experiment shows that for all gases the actual variation of
viscosity with temperature is more rapid than that given by (12.3,1), as is to
be expected. Thus if equation (12.1,6) for the coefficient of viscosity is used,
the diameter a of a molecule must be supposed to vary with the temperature,
decreasing as the temperature increases. To illustrate the order of magnitude
of the variation, we give the diameters of helium molecules at different tem-
peratures, calculated from experimental! values (5, 6) of ft, using (12.1, 6).
* The experimental values, save for those for carbon monoxide and helium, are mean
values taken from Thtrmophysical Properties Research Center, Data Book 2 (Wright-
Patterson Air Force Base, Ohio, 1964). As regards carbon monoxide, the value quoted is
derived from those used in the Data Book with the exception of those of Johnston and
Grilly (2), which appear inconsistent with those of other workers. For helium, chief
reliance is placed on the results of Kestin and his co-workers (1) and of Wobaer and
MUller (3).
f The experimental values of/1 used in Tables 12 and 13 are 1 or 2 percent higher than
more modern ones, the corresponding value at 0 °C. being 1887x10"' poise; the
 Copyrighted Material

230 VISCOSITY: [12.31

Table 12. Values offi and cr for helium

Temperature 10' x /1 io'x<r
(°C.) (poise) (cm.)
-2581 «94-6 267
-197-6 8176 *-37
— 102-6 1392 2-23
17-6 1967 214
1837 2681 203
39» 3388 2'OOS
815 4703 1-97

The variation of the apparent radius with temperature receives a simple

explanation on the hypothesis that the molecules are centres of force, not
hard spheres. The motion of the centre B of one molecule relative to the
centre A of another at their encounter is then represented diagrammatically
by Fig. 3 a (p. 56). The deflection of the relative velocity £ at encounter is the
same as if, in thatfigure,the motion of B relative to A were along the lines PO,
OQ instead of along the curve LMN; that is, it is the same as if the molecules
were elastic spheres, and collided when B reached O: their effective diameter
at the encounter is thus AO. If the elastic-sphere theory is applied to such
molecules, the value of cr deduced from the experimental values of /i is the
mean of the values of this effective diameter for all encounters, weighted so
as to give greatest importance to encounters producing the largest deflections.
The deflection % of the relative velocity at encounter is given in terms of
the encounter-variable b and the effective diameter OA by the equation
(cf. (3.44, 2)) 6-O^cosix-
If g increases, b remaining constant, the time during which each molecule
is under the influence of the other's field diminishes, and % wiU also diminish:
thus OA, too, must decrease. Hence an increase of temperature, which
implies an increase in the average value of g at encounter, results in a
decrease in the apparent value of cr.

12.31. Point-centres of force

The law of variation of viscosity with temperature for molecules repelling
each other as the inverse vth power of the distance (cf. (12.1, 2)) is
/1 oc T; (12.31,1)
f=
where* 2+i^r (".31,2)
temperature variations which they indicate are, however, consistent with the more
modern data.
* This formula was inferred from dimensional considerations by Rayleigh (J. W. Strutt,
3rd Lord Rayleigh), Proe. Roy. Soc. A, 66, 68 (1900); Collected Papers, vol. 4, p. 452.

Copyrighted Material
12.31] COMPARISON OF THEORY WITH EXPERIMENT 231
Comparing this with (i 2.1, 6) we see that for this model the apparent radius
of a molecule varies inversely as the \j(v- i)th power of the temperature;
cf. (10.31, 8).
Equation (12.31, 1) can also be written in the form
it = fi'(Tiry, (12-31.3)
where ft' is the coefficient of viscosity at an assigned temperature 7". If the
value of ft is known for a second temperature, the value of s can be found.
Hence the formula (12.31, 3) can always be satisfied by the experimental
values of ft for two temperatures, if/*' and s are suitably chosen.
For several gases, notably hydrogen and helium, the experimental values
of /1 conform to an equation of the type (12.31, 3) over a large range of
temperature (4, 5). The results for helium are given in Table 13; experi-
mental values (5) of /1 are given in the second column, and in the third, the
corresponding values of ft calculated from (12.31, 3), taking s = 0-647, a r | d
adopting the value ft = 1887 x 10 -7 for the viscosity at o°C. The fourth
column gives the values calculated from Sutherland's formula (cf. 12.32).

Table 13. Viscosity of helium

Temperature i o ' x / t (calc.) i o ' x / < (c

<°C.) i o ' x /1 ( e x p . ) (force-centres) (Sutherla

-258-1 2946 2887 92

-253° 349-8 348-9 • 35
-1984 813-2 8155 621
-1976 8176 8213 628
-«833 9186 9185 74S
— 1026 '392 • 389 »3'7
-78-5 1506 '5'5 1460
—700 1564 1558 »5«3
—609 1587 1603 '5*3
-228 1788 1783 '77«
176 1967 1965 '974
187 1980 1970 »979
998 2337 2309 234S
I8J-7 2681 2632 2682

For other gases, however, equation (12.31, 3) does not agree so well with
the experimental results, and the value of s for which (12.31, 3) best fits
observation varies with the range of temperatures considered. Even for
helium a slight variation is found; for temperatures above 0 °C. the best fit
with the above data is obtained by taking s = 0-661 (5), and a recent study
(1) makes s in the range from — 2400 to 800 °C. fluctuate about a mean value
0-6567. For most gases s decreases markedly as the temperature increases;
thus Johnston and Grilly (2) found values of J for nitrogen, carbon monoxide,
air, methane, nitric oxide, oxygen and argon, in a range from about —1800
232 VISCOSITY: (12.32
to 20 °C, which are greater by about 01 than those found by Wobser and
Miiller(3) for the range 20-100 °C.
The decrease in s as the temperature increases implies an increase in v,
which may be interpreted as implying that the molecular repulsion varies
according to a smaller power of i/r when r is large than when r is small
(because at low temperatures the molecules do not penetrate so far into each
other's repulsive fields as at high temperatures). This is, however, not the
only possible interpretation; in the derivation of (12.31,1) no account is
taken of the fact that the interaction of molecules at large distances may be an
attraction; such an attraction would influence the molecular motions more
at low temperatures, when the velocities of molecules are small, than at high.
Table 14 lists the values of s and v for several gases, mostly for the
temperature-range 20-100 °C. The values of v range from about 13 for
helium, neon and hydrogen, to values around 5 for molecules of complex
structure. If v is large, the mutual repulsion between pairs of molecules at
encounter increases very rapidly as they approach, and the encounter
approximates to a sharp impact: in this case the molecules are said to be hard,
while for small values of v, in the neighbourhood of 5, they are said to be soft.

Table 14. Values of s and v

Temperature
Ga» 1 V range (°C.)
Hydrogen 0-668 (3) 129 20-100
Deuterium 0 6 9 9 (7) III -183-22
Helium 0-657(1) 13-7 — 230-800
Methane 0 8 3 6 (3) 7o\
Ammonia 1 10 (3) 43
Neon 0661 (3) 13*
Acetylene 0 9 9 8 (3) 50
Carbon monoxide 0-734 (3) 95 20-100
Nitrogen 0 738(3) 94
Air 0-77 (3) 8-4
Nitric oxide 0 7 8 8 (3) 79
Oxygen 0-773 (3) 8-4^
Hydrogen chloride «03 (8) 48 20-09
Argon 0811(3) 75
Carbon dioxide 0-933 (3) 5-6
Nitroua oxide 0-943 (3) 55 20-100
Chlorine J-oi (3) 49
Sulphur dioxide 1-05(3) 4-6J

12.32. Sutherland's formula

The viscosity of a gas whose molecules are rigid attracting spheres is given,
to a first approximation, by
,+ (,2 32,,)
"-i6S»pir)y( ?) -
12.32] COMPARISON OF THEORY WITH EXPERIMENT 233
(cf. (12.1, 3)). Comparing this with (12.1, 1) we see that the effect of the
attractive field is to increase the apparent diameter of the molecules in the
ratio *J(i+S/T): t. The increase is largest if the temperature is small, and
becomes negligible at very large temperatures. It results from the deflection
of the molecules by the attractive fields before collision, which makes the
collisions more nearly head-on than in the absence of such fields. The
Sutherland constant S is a measure of the strength of the attractive fields,
being proportional to the mutual potential energy of two molecules when in
contact.
Equation (12.32, 1) may be written in the form
,/T\iT' +S , .
P*/1 \Y') Y+J' (I2 32 2
' ' *
where /1 = /1' when T = 7". If ft is known at a second temperature T" as
well as at 7", a value of S can be found such that (12.32, 2) is satisfied at
both 7" and T". If 7" and T" are sufficiently close together, (12.32, 2) may be

Table 15. Values of Sutherland's constant

Sutherland's Temp, range

Gas constant (°C.)

Hydrogen 6 6 8 • (3) 20-100

Helium 7 2 9 > (3) 20-100
Methane 169 (3) 20-100
Ammonia 5<>3 (9) 25 1-300
Neon 64-1 (3) 20-100
Acetylene 3*0 (3) 20-100
Carbon monoxide 10a (3) ao-ioo
Ethylene "S (9) 20-250
Nitrogen 1047 (6) 198-825
Air "3 (3) 20-100
Nitric oxide '33 (3) 20-100
Ethane 2 5 2 (10) 20-250
Oxygen »25 (6) 148-829
Hydrogen sulphide 3 3 ' (") 17-100
Hydrogen chloride 3 6 2 (12) 21-250
Argon 148 (3) 20-100
Carbon dioxide 253 (3) 20-100
Nitrous oxide 263 (3) 20-100
Methyl chloride 441 (•3) 19-6-300-4
Sulphur dioxide 404 (3) 20-100
Chlorine 34S (3) 20-I00
Krypton 188 (14) I6-3-IOO
Xenon 252 (14) 15-3-IOO-I

expected to fit the experimental values well at intermediate temperatures.

Values of 5 for several gases at ordinary temperatures are given in Table 15.
Actually (12.32, 2) closely represents the variation of fi with temperature
for several gases over fairly wide ranges of temperature, as Table 16
9 CUT
234 VISCOSITY: (12.32
illustrates. The calculated values for nitrogen in this table correspond to
/t = 1746 x io - ' at 0 °C, with 5 = «047; those for ammonia correspond
to /i = ioo2 x io - ' at 25-1 °C, with S «= 503. The Sutherland formula also
represents the viscosity of other gases, such as oxygen, argon and carbon
monoxide, to a similar degree of accuracy; the same is true of air, though this
is a gas-mixture.
Table 16. Viscosity of nitrogen and ammonia

Nitrogen (6) Ammonia (9)

Temp. io'x/« io'x/i Temp. 10'x/t io'x/i

CC.) (exp.) (calc.) (°C.) (exp.) (calc.)

19-8 1746 1746 ao 981 98a

399 »797 a8oi 50 109a «095
408 3«4« 3134 too 1879 1282
490 3374 3366 ISO 1463 146$
600 3664 3656 200 1645 1644
7'3 3930 3934 aso 1813 1818
8a 5 419a 4193 300 1986 1987

The success of (12.32, 2) for many gases does not establish the validity of
Sutherland's molecular model for those gases. The formulae (12.32,1, 2)
are approximate formulae, valid only if the attractive fields are small; if this
is not the case the expression 1 +S/T in (12.32, 1) must be replaced by a
series of the form „ ™
1 + ^ + ^ + ..., (12.32,3)

and it is scarcely to be expected that the terms after the second will be
negligible if S/T is not small compared with unity. As Table 15 shows,
S/T often exceeds unity in the temperature-range in which (12.32, 2) fits
the experimental values, whereas for gases like hydrogen and helium, for
which S/T is fairly small, Sutherland's formula does not fit the experimental
values at all well. This is illustrated by the figures for helium given in the
fourth column of Table 13; S is taken as 78-2. For these gases the value of
S for which (12.32, 2) best fits the observations is found to increase with the
temperatures considered.
In general it is not adequate to represent the core of a molecule as a rigid
sphere, or to take molecular attractions into account to a first order only. The
greater rapidity of the experimental increase of ft with T, as compared with
that for non-attracting rigid spheres, has to be explained as due partly to the
'softness* of the repulsive field at small distances, and partly to attractive
forces which have more than a first-order effect. The chief value of Suther-
land's formula seems to be as a simple interpolation formula over restricted
ranges of temperature.
12.33] COMPARISON OF THEORY W I T H EXPERIMENT 235

12.33. The Lennard-Jones 12, 6 model

The Lennard-Jones model (cf. 10.4) takes into account both the softness of
the molecules and their mutual attraction at large distances. Most attention
has been paid to the 12,6 model, for which the mutual potential energy of
two molecules at a distance r is taken to be
6
K(r) = 4e {(<r/r)"-(Wr) } (12.33, 0
12
= e{(rm/r) -2(rm/r)«}. (".33,2)

The viscosity of a simple gas on this model is given by an equation of the

form
(kmT)i . .
(I2 33,3)
^a*I\kW) '
(cf. 10.4). Equation (12.33, 3) implies that, if log(///Tl) is plotted against
log T, curves for different values of or and e differ only by a change in origin.
Thus it is possible to determine <r and e by fitting a standard log(ft(Ti),
log T curve to a corresponding experimental curve, and determining the
relative displacement in origin of the two curves.
A comparison of theory with experiment has been made for a number of
gases for this model. The agreement for each gas is found to be at least as
good as for the models of 12.31 and 12.32, which likewise involve two
adjustable parameters.* This confirms that the 12,6 model approximates to
physical reality by providing repulsive and attractive fields varying with
distance roughly in the right manner, at least for non-polar gases. For polar
gases the attractive part of the molecular potential energy includes a term
proportional to r~3 (cf. (10.43, 3))> n o t taken into account in (12.33, ')> an(^
the fit with experiment is less close.
Figure 10 compares the experimental and calculated viscosities for a
number of gases. The values of e and <r used in this figure are taken
from Table 17 (below) save for carbon dioxide, for which the value
a = 3-96 x io~ 8 cm. is used. Temperatures below about 30 °K., at which
quantum effects are important for helium and hydrogen (cf. 17.4 and
17.41) are not represented in the experimental values. Part of the dis-
crepancies between experiment and the theoretical curve is due to incom-
patibility between the results of different experimenters.
Values of cr, rm and e/k obtained (mainly by Hirschfelder, Bird and Spotzf)
from the viscosities of various gases are given in Table 17. They apply to
* Helium ha8 so weak an attractive field that effectively the model attempt! to fit it to
an r - 1 1 repulsive potential energy,
t J. O. Hirschfelder, R. B. Bird and E. L. Spotz, J. Chem. Phys. 16, 068 (1948): Chem.
Rev. 44. 205 (1949)- The values for D, are those given by J. Kestin and A. Nagaahima,
Phys. Fluids, 7, 730 (1964). Values of e/A and <r given by other authors often differ
appreciably from those of Table 17.
9-s
236 VISCOSITY: [12.33
ranges of temperatures including o °C. The ' low-velocity diameter' <r gives
the distance of closest approach of two molecules with negligible initial
relative velocity; r m ( s 2»cr) is the distance at which the molecular inter-
action changes from a repulsion to an attraction, and - e is the corresponding
energy. Thus neither cr nor rm is exactly comparable with the effective
diameter of 12.2 (Table 11), which measures the distance near which most
of the molecular interaction takes place. For hydrogen, helium and neon,

log T* (upper curve)

0 0-5 10 15 20
I 1 1 1 1 1

-05 " ,^--—

-0-6 - *>P*"^^

-07- J^**^ ^ ^ ^

_ M . 1 1 1
0 0-5 10
log T* (lower curve)

Fig. 10. Companion for the la,6 model of the theoretical curve for log (/<•/v'7"•) with
experiment, where T* = kT/e, ft* = /»o-'/(me)'. Upper curve; • •= helium, + «• neon,
x « hydrogen. Lower curve (displaced one unit to the right); ® = argon, A = nitrogen,
• m carbon dioxide. Experimental data in the ranges 30-80, 80-290 and 290-1000 °K.
are taken from the work of van Itterbeek (15), Johnston (a) and Trautz (6), (8), (16)
and their respective co-workers.

whose molecules attract only weakly, the effective diameter of 12.2 is less
than <r; for all other gases it is greater, and for gases with strong attractive
fields (large e/k) it even exceeds rm. This indicates the importance to the
viscosity of the attractive part of the field.
Different ranges of temperature may give somewhat different values of a
and e. Values can also be calculated from other physical data; such values
are given in the last columns of Table 17. They are mainly obtained from
12.34] COMPARISON OF THEORY WITH EXPERIMENT 237
virial coefficients, but in the last five cases they are less accurate values
derived from liquefaction data. The discrepancies arise from the inaccuracy
of the physical data as well as from the limitations of the model.*

Table 17. Values ofejk, a and rmfor the 12, 6 model

Viscosity values Other values

«/* 0- x t o ' 1r » x i » ' «7* <rx t o '

Gaa (°K.) (cm.) (cm.) CK.) (cm.)

Hydrogen 333 a-97 333 37o* 292

Deuterium 295
Helium
35*
6-03 270
3'3« —
6-03
—
303 263
Methane •3°-5 3-8» 4*9 1427 381
Neon 357 280 314 357 a-74
Carbon monoxide 1103 359 403 9533 365
Nitrogen 01 46 368 4'3 959 37*
Nitric oxide 119 347 389 «3» 3«7
406
Air 97 3 62 — —
Oxygen 113a 343 3-85 1175 358
Argon «»4 34* 384 "95 341
Carbon dioxide 190 400 449 • 85 457
Nitrous oxide 220 388 435 189 459
Hydrogen chloride 360 3305 3-71 261 •*43 369
Sulphur dioxide »5* 4*9 4-82 3 6 6 .3*3 4«5
Chlorine 357 4115 462 33* •3'3 4»5
Krypton 190 361 405 169 • 158 396
Xenon 230 405 4-55 228 , 1 9 * 404

12.34. The exp; 6 and polar-gas models

In the exp;6 model, the mutual potential energy of two molecules is
taken to be r t . i i \» / \ «n

*>=(7^[H«(-rJ}-(7-)]- (••*.•)
Here e and rm have the same meanings as for the Lennard-Jones 12,6 model;
a is a pure number normally between 12 and 16. For a given a, the model
conforms to a law of corresponding states of the form (12.33, 3)> a n ^ >° a
graphical method similar to that described for the 12,6 model can be used
to determine e and rm.
Mason and Ricef have compared experimental viscosities with results
• R. Gibert and A. Dognin (C.R. 346. 2607 (1958)) point out that, over the range
0-9 < kT/e < 5, the function F(kTl€) in (12.33, 3)'» W*H approximated by
C(i + io77t/A7/),
where C is a constant. This fact may explain the success of the Sutherland formula for
many gases; it also suggests that an approximate value of e/k for such gases is .S/1-077.
t E. A. Mason and W. E. Rice, J. Chem. Phys. 22, 522, 843 (1954); aee also E. A. Mason,
J. Chem. Phys. 32, 1832 (i960), for a rediacussion of the krypton data.
238 VISCOSITY: (12.4
calculated from the exp; 6 potential for a number of gases. They find rather
better agreement than is given by the 12,6 model. This is not surprising,
since (12.34, 0 involves an extra adjustable parameter a; however, intro-
duction of the extra parameter improves the fit surprisingly little, except for
hydrogen and helium. Where marked deviations from measured viscosities
are found, they are generally in the same direction for the 12,6 and exp;6
models. Values of e and rm, together with corresponding values of a and of a,
where <r is such that v{&) = 0 , are given in Table 18; except for ethane and
carbon dioxide, they are due to Mason and Rice. Considering that the
experimental data used are not always identical with those used in Table 17,
the discrepancies between the values of e and <x for the 12, 6 and exp;6
models are not large.

Table 18. Values ofejk, <r, rm and a, for the exp; 6 model

e/A <rx i o ' rm x to*

Gaa (°K.) (cm.) (cm.) a

Hydrogen 373 3005 3-34 »4

Helium 916 »75 3»35 124
Methane 152-8 369 4-ai 12-3
Neon 38 a-8t 3«5 »45
Carbon monoxide 119-: 355 394 «7
Nitrogen 101-3 362 401 '7
Ethane (17) 219-1 44" 4905 16
Argon 113* 344 3-87 •4
Carbon dioxide (1) 26a 3-73 418 >4
Krypton 1583 356 406 12-3
Xenon 231 a 3-935 445 »3-0

Monchick and Mason* have applied the Stockmayer potential (cf.

(10.43, 3)) t 0 discuss the viscosity of polar gases, using the method described
in 10.43. For such gases the r~3 term in the potential leads to a more rapid
variation of ft with T at low temperatures. The gases considered include
ammonia, hydrogen chloride, sulphur dioxide and hydrogen sulphide; the
authors state that' the overall agreement between experiment and this model
for polar gases is comparable to that of the Lennard-Jones 12,6 model for
non-polar gases'.

GAS MIXTURES

12.4. The viscosity of a gas-mixture

The first approximation to the coefficient of viscosity of a gas-mixture is
» * X J R J + XJRJ + X|XJRJJ . .
Wi = ( 4
>4«JDhli+«l«t/Wi+xi«i«M *''
* L. Monchick and E. A. Maaon, J. Chtm. Phyt. 35. 1676 (1961)-
12.41| COMPARISON OF THEORY WITH EXPERIMENT 239
(cf. (9.84, 2)) where
Rt = § + m, A/>«2, R2 = § + m2 A/OT1( (12.4, 2)

(12.4. 3)
Here [/«t]„ [/'2]i are the first approximations to the coefficients of viscosity
of the constituent gases, and A, G are quantities depending only on the
interaction of molecules of different kinds. By (9.81, 1) the quantity E is
equal to §nm0[Z),t],, where [Z?tt], is the first approximation to the coefficient
of diffusion of the two gases; A is a dimensionless quantity, values of which
are listed, for molecules behaving like centres of a repulsive force KT~" in
Table 3 (p. 172), and for the 12,6 model in Table 6 (p. 185).
In 12.1 it has been shown that for a simple gas the first approximation [/*]t
to the coefficient of viscosity ft is in general only slightly in error. Hence the
error of (12.4, 1) is certainly small when either x, or x2 tends to zero. One
may expect a similar degree of accuracy when neither x t nor x, is small, if
A and E are correctly chosen, at least provided that the molecules of the two
gases are not too dissimilar.* The error is, however, fairly considerable for
inverse-square interaction (cf. 10.34).
If x2 is very small we get the case of a gas in which a small quantity of a
second gas is present as an impurity. In this case (12.4, 1) approximates to

Mi = W. + ^ K ^ + f-A)'-^}. (.M. 4)

12.41. Variation of the viscosity with composition

In comparing experimental results with the theoretical variation of /t with
«,/«2, equation (12.4, 1) is normally employed, ignoring the difference
between the first approximations and the true values of the viscosities.
Values have to be assigned for the constants A, E. The constant A may be
determined from the tables for the 12,6 and exp;6 models, taking the
energy ¢,2 as(e1e2)i. The 12,6 model makes A about 0 4 4 for most of the
important range of kT/elt, while the exp;6 model indicates a slow increase
of A from 0-44 to 0-46 as kTjelt increases in this range; thus the value of A is
insensitive to the precise value of e „ adopted. The constant E can be deter-
mined by a number of methods.
(I) By (9.81, 1), [£>i2]l = 3E/2«w0; hence E can be determined directly
from experimental values of D12, neglecting the difference between this and
[Dlt]v However, the error of the first approximation [Dlt]t is sometimes
* The results for a Lorentzian gas (p. 193, Table 10) are not representative of the case
when ffii/m, is large, since they refer only to the small part of the viscosity due to the
light gas.
240 VISCOSITY: (12.41
appreciable, and moreover the experimental determinations of coefficients
of diffusion are not always very reliable.
(II) We may calculate E from viscosity data for the simple gases. For
example, put
Q?>(2) - 2*i(*A7ym,)», Q S W - ^ ^ r / m . M . M , ) * , (12.41,1)
with a similar relation for Qi*'(2); then (cf. (9.7, 3), (9.8, 8))

iwk-APs-T- < - & ( ^ ) ' - <•>•<•• *>

and sv st and su denote effective viscosity diameters for the interaction of
like or unlike molecules. Assuming that *„ = $(*i + *t)> w e c a n calculate E
from experimental values of //,, fit, ignoring the difference between these
and [/«,]„ [/<Jj. However, though slt is likely to lie between sl and sv the
relation *„ = {(sl + st) applies strictly only for rigid spheres.
(III) A more sophisticated method is to calculate E from viscosity data,
using the 12,6 model, with the combination rules eu = foe,)* and
0-,, = }(<r, + <rt) or <rlt -= (<r, <rs)i. Similar methods can also be applied
to the exp;6 model. However, once again the combination rules, though
reasonable, cannot be exact.
(IV) We may simply determine E to give the best fit of theory to
experiment.
The method followed here is (IV); the values of E thus found are then
checked against diffusion coefficients and the values obtained by calculation.
The mixtures considered are mixtures of hydrogen and helium, and of
helium and argon.

Mixtures of hydrogen and helium at 0 "C.

The value adopted for A is 0-458; E is then determined to be 2-48 x io~* poise,
corresponding to [Z),t], = i^cm.'/sec. at 1 atmosphere pressure. For
comparison, experimental values of Dlt, reduced to 0 °C., are 1 -313 (Bunde)
and 1-21 (Suetin, Shchegolev and Klestov). The values of E calculated as
indicated in (II) and (III) are 247 x io~* and 2-44 x io - 4 poise.

Mixtures of helium and argon at 20 "C.

The value adopted for A is 0-452; E is found to be 8-70 x io~* poise,
corresponding to [Z)n], = 0715 cm.*/sec. at a pressure of 1 atmosphere.
Recent experimental values of Dlt are 0697 at 14-8 CC. (Strehlow), 0-76 at
26-9°C. (Walker and Westenberg), and 0731 at i6°C, 0-744 a t 2 3 ° c -
(Saxena and Mason). Values of E calculated by methods II and III are
8-125 x I0~* an<* 8-84 x io~*poise.
The calculated and experimental* values of the coefficient of viscosity are
* For helium-argon mixture*, aee ref. (18); for hydrogen-helium mixtures, see (19).
12.41] COMPARISON OF THEORY WITH EXPERIMENT 241

Table 19. The viscosities of two gas-mixtures

Hydrogen-helium Helium-argon

H, IO'x/» 10'x/t He io'x/» io'x/»

(%) (exp.) (calc.) (%) (exp.) (calc.)

0 18925 [1892-5] 0 2227-5 [22275]

39o6 1850-0 1846 199 22707 2271
10-43» 1759-6 1767 37-1 23095 1305
13-60 I73I-7 1730 63-4 2316-1 2319
*49'3 1603-» 1600 807 2252-8 "5S
40184 1430-6 •431 863 2202-7 2207
60143 1 »»6-7 1224 942 2090-2 »093
8I-I03 1016-5 1017 1000 1960-4 [1960-4]
100-0 841 [841] — — —

given in Table 19. As can be seen, the formula (12.4, 1) represents the
variation of /t very closely.
The results for helium-argon mixtures illustrate the curious fact, first
noticed by Graham,* that the addition of a moderate amount of a light and
relatively inviscid gas(hydrogen, in the case mentioned by Graham) to a more
viscous and heavy gas (carbon dioxide) may actually increase the viscosity of
the latter. The explanation seems to be somewhat as follows. The addition of
a quantity of the same gas to the heavy gas reduces the mean free path, but
increases the number of carriers of momentum; these two effects just balance.
When, however, a small quantity of a light gas is added to the heavy gas, the
mean free path of the molecules of the latter is hardly affected, because of the
large persistence of velocities after collisions with the lighter molecules (5.5):
and the small additional transport of momentum by the lighter molecules
may outweigh the effect of the decrease in the mean free path of the heavier.
More precisely, if px > /«», fi has a maximum as x,/x, varies if in (12.4, 4)
the coefficient of x, is positive. By (12.4, 2) this condition is equivalent to

or, in terms of the equivalent viscosity diameters *„ st, slt used in (12.41, 2)
A(m,+wt,)! s\ |/, m,\ I. , mt\\i . . .
Mj(2»t,)» s\t \\3 mtJ\3 mji \ •*>*/

Thus for given m1( mt and A, J,/*,, must exceed a certain lower limit. An
effective upper limit also exists, since sn can be expected not to differ too
widely from i(*i+*i) and, to ensure that /«, > //,, sjs3 must be less than
(mjmjl. Take A = f, in rough agreement with the results for the 12,6 and
• T. Graham, Phil. Trans. R. Soe. 136, 573 (1846).
242 VISCOSITY: (12.42
are
exp; 6 models; then the bounds placed on sjslt and *i/i(*i + *t) **
follows, for different values of mjmt:
m m
ll t - A i i 1 4 9 16

> 0-618 0-696 0-811 1 IIS 1-205 I-13S

«•/»(»!+Il) < 0-667 0-73» 0828 1 1-172 1-268 «•333

These figures show that a maximum of /1 is likely to occur only if sjslt

lies in a relatively restricted range. It is most likely when m1/mt is either
large or small compared with unity, but the viscosities of the two gases are
not too different (so that sl/^(sl + st) is not much less than its upper bound).
When ml and m, are nearly equal, a maximum can occur only when
*it < *i < sv 1 h ' s >s m o 8 t likely to be the case when *, and tt are nearly
equal, and the mixture is composed of one non-polar and one polar gas; in
this case the dipolar interaction increases s for the polar gas, but not x„.
A maximum of /t as x,/x, varies is observed in some mixtures (like
He-N t ) in which /*, > /it but mt < m„ as well as in those displaying the
original Graham phenomenon (in which fix > /it,ml > mt). A minimum of fi,
though theoretically not impossible, is improbable and does not seem to have
been actually observed at normal densities.

12.42. Variation of the viscosity with temperature

The complicated nature of the formula (12.4, 1) renders it difficult to
predict from it the precise law of the temperature variation of the viscosity
of a gas-mixture. Normally, however, the force-law between unlike molecules
can be expected to be intermediate between those for like molecules of the
two gases. Thus the variation of E with temperature can be expected in
general to be intermediate between those of j / t j | and \jit]1; the same is true
also of [//],.
The viscosities of mixtures at different temperatures can be used to infer
values of E at these temperatures, and from these the law of interaction
between unlike molecules can in principle be found. In practice, however,
because E is not determined too certainly, the process is normally inverted;
the law of interaction for unlike molecules is inferred from those for like
molecules, and the values of E and A calculated from it are used in (12.4,1)
to give theoretical values of the viscosity of mixtures. This method is usually
employed with the 12,6 or exp;6 models, using appropriate combination
rules to deduce the force constants (cf. (Ill) of 12.41); reasonable agreement
between theory and experiment is normally obtained.
As noted in 12.41, the viscosity of a mixture of a light gas with a heavy one
often has a maximum when the proportions are varied at a given temperature.
It is found that, as the temperature increases, this maximum tends to
disappear. Figure 11 illustrates this tendency for mixtures of hydrogen and
hydrogen chloride (12).
12.43] COMPARISON OF THEORY WITH EXPERIMENT 243
An analytical explanation of the tendency is as follows. Suppose, as in
12.41, that the first gas is the heavier; then, as the temperature increases, /*,
in general increases more rapidly than //2> and the variation of E is inter-
mediate between those of /*, and /it, so that E / / / , decreases. T h u s the
inequality (12.41, 3) may cease to be true for sufficiently high temperatures.

26 T = 250' C

24 ^ - \ ^
22
T=154°C ^ \
20
T»99' C ^~~~~\ \
<b 18 ~""~~^^ N . \
X
="• 16
r = 2i'c \ \ \
14 -
~~^\ \ V
12
>?v
10

a 1 1 1 1 1 1
20 40 60 80 100
% Hydrogen
Fig. n

12.43. Approximate formulae

Because of the relatively complicated form of (12.4, 1) a variety of simpler
approximate formulae have been suggested for the viscosity of a mixture.
These are in general of the Sutherland form (cf. 6.63); that is, fi is expressed
as the sum of contributions //<u, ^ (2) arising from the two gases in the mixture,
//(jj and fi(t) being given by expressions of the form
/*<D = x ^ O * , ) - 1 + x^KiV1}, Mm = W W / / ; , ) - 1 + XjOtt,)"1}.
(12-43. 0
Here filt pt are, as usual, the viscosities of the two pure gases; //J„ /i'n are
'mutual viscosities' whose values depend on the particular approximation
introduced.
The theory of Chapters 8 and 9 can be shown to imply that the parts
Mw< /'(1) ' n t 0 which ft is divided actually satisfy, to the first approximation,
the relation r . . .
 Copyrighted Material

244 VISCOSITY: (12.5

and a similar equation. These equations differ from (12.43,J) by reason of
the terms involving f—A ; these terms represent the transfer of momentum-
transport from one gas to the other at collision. Since these terms are
relatively small, a crude approximation is to replace /½ in (12.43, 2 ) by
XilHvlx-x, just as if the momenta transported by molecules of the two gases
were proportional only to the numbers of those molecules. Then (12.43, 2 )
becomes . .
/*ti)'

which is of the form (12.43, x)> w ' ^ ft'u given by

/t[t = &MJA. (12.43,4)

In terms of Dlt, ^ = x . / j ^ +_Jg_), ( I2 . 43 , 5 )

where p t is the density of the first gas, when pure, at the pressure and
temperature of the actual mixture. The formula (12.43,5)> w ^ 3A given the
rather high value 1-385, and the first approximations replaced by actual
values, gives the approximate formula of Buddenberg and Wilke.* The
total viscosity is always somewhat overestimated by using (12.43, 5) an( * ^ e
corresponding equation for //(8); this is partly compensated by the high
value of 3A.
A variety of other approximate formulae have been proposed, some based
on further transformations of (12.43, 2 ) o r (I2-4> 0» a n ^ some more
empiricaLf The complication of certain of these is such that it is simpler,
as well as more accurate, to use the formula (12.4, 1). However, (12.43, 0
has considerable value as a formula to represent experimental results,
regarding /t[t and /i'n as adjustable quantities.

VOLUME VISCOSITY
12.5. As was noted in 11.51, gases whose molecules possess internal energy
are subject to a volume viscosity w opposing expansion or contraction; the
effect of this volume viscosity is to increase the hydrostatic pressure by
—mV. c0, or raprxDp\Dt. The volume viscosity is a relaxation phenomenon;
if the internal energy is the sum of parts independently relaxing into trans-
latory energy with relaxation times T^T,, ...
w = 2p-ZNrTr/N* (12.5,1)
r
(cf. (11.511, 8)), where Nk/zm is the total specific heat cv of the gas, and
NTkJ7m is the part of this which corresponds to the relaxation time r r
• J. W. Buddenberg and C R. Wilke, Ind. Bng. Chem. 41, 1345 0949)-
t R- S. Brokaw, J. Chem. Phyt. 29, 391 (1958), and 43, J140 (1965); W. E. Francis,
Tram. Faraday Soc. 54, 1492 (1958); T. G. Cowling. P. Gray and P. G. Wright, Proc.
R. Soc. A, 376, 69 (1963); D. Burnett, J. Chem. Pkyt. 42, 2533 (1965).

Copyrighted Material
12.51 COMPARISON OF THEORY WITH EXPERIMENT 245
The absorption coefficient (per unit length travelled) in a plane sound
wave with angular frequency to and velocity V can be shown to be*
°>s L 7 — i •* 1 i •.

Thus the dissipative effect of volume viscosity, like that of fi and A, is

especially important when the density varies very rapidly, as in ultrasonic
waves (or shock waves). However, a description simply in terms of a volume
viscosity is inadequate if the time-scale of density-variations is too short.
If in ultrasonic waves arrr becomes large, the internal energy corresponding
to r r does not share in the temperature-changes, nor contribute appreciably
to the absorption; the effective specific heat of the gas is correspondingly
less, and the sound velocity increases.
In(i2.5, 2)/tand A/c,, are comparable with/>r, where r is an ideal collision-
frequency (cf. (6.6, 5, 7)). Volume viscosity is normally found to increase the
absorption coefficient considerably above that due to p and A alone, which
implies that certain of the rT are large compared with T. The prime interest
in ultrasonic measurements is in the determination of the rr One writes
t, = £rr, (i2-5. 3)
so that Cr is, roughly speaking, the number of collisions required to secure
energy-balance between the internal mode considered and the translational
motions. To make r precise, the formulae r = l/C, fi = 0 - 4 9 9 ^ / appropriate
to elastic spheres are used (cf. 5.21 and (12.1, 6)); thus
ft = o-wwiCfr = 1-271/)7 (12.5, 4)
by (4.11, 2).
Experiments at low frequencies, where the relaxation effects are adequately
represented by a volume viscosity, determine the sum T,NTrT in (12.5,1);
r
they do not suffice to separate the effects of the different energy-modes. To
determine the separate r r it is necessary to carry the experiments past the
point where (orr = 1; the change in effective specific heat enables one to
identify the corresponding Nr, and a maximum of absorption occurs roughly
at o)Tr - 1. Experiments at such high frequencies show that the major part
of the volume viscosity (12.5, 1) is normally due to vibrational relaxation.
It is usually sufficient to use a single relaxation time for all the vibrational
energy-modes. At ordinary temperatures ^. for vibrational energy takes
values ranging from above io 7 , for oxygen and nitrogen, through 10*, for
other common diatomic and polyatomic gases, down to 100 or less for more
complicated and polar molecules.
Hydrogen at ordinary temperatures has too little vibrational energy for
this to contribute appreciably to the absorption. The rotational relaxation of
* Sec, e.g. K. F. Herzfeld and T. A. Litovitz, Absorption and Dispersion of Ultrasonic
Waves, chapter 1 (Academic Press, 1959).
246 VISCOSITY [12.S
hydrogen and deuterium is relatively slow, £,. being about 300 for hydrogen
and 200 for deuterium at ordinary temperatures. For other gases £. for
rotational energy is relatively small and difficult to measure; values in the
range from 3 to 20 are normal.
For full details, and for the theory of the relaxation processes, reference
should be made to specialist books on the subject.*

REFERENCES
(1) J. Kestin and W. Leidenfrost, Physica, 35, 537, 1033 (1959); J. Kestin and
J. H. Whitelaw, Physica, 29, 335 (1963).
(2) H. L. Johnston and E. R. Grilly, J. Phyt. Chem. 46,948(1942); H. L. Johnston
and K. E. McCloskey, J. Phyt. Chem. 44, 1038 (1940).
(3) R. Wobser and Fr. Muller, Kolloidbeihtfte, 53, 165 (1941).
(4) H. K. Onnes, C. Dorsman and S. Weber, Vers. Km. Akad. Wet. Amst. ai,
»37S. »9»3-
(5) H. K. Onnes and S. Weber, Proc. Sect. Set. K. ned. Akad. Wet. 21, 1385
(»9»3)«
(6) M. Trautz and R. Zink, Annln Phys. 7, 427 (1930).
(7) A. B. van Cleave and O. Maass, Can. J. Res. 13 B, 384 (1935).
(8) M. Trautz and H. E. Binkele, Annln Phys. 5, 561 (1930).
(9) M. Trautz and R. Heberling, Annln Phys. 10, 155 (1931).
(10) M. Trautz and K. Sorg, Annln Phys. io, 81 (1931).
(it) A. O. Rankine and C. J. Smith, Phil. Mag. 42, 615 (1921).
(12) M. Trautz and A. Narath, Annln Phys. 79, 637 (1926).
(13) H. Braune and R. Linke, Z. Phys. Chem. A, 148, 195 (1930).
(14) A. O. Rankine, Proc. R. Soc. A, 84, 188 (1910).
(15) A. van Itterbeek and O. van Paemel, Physica, 5, 1009 (1938); J. M. J.
Coremans, A. van Itterbeek, J. J. M. Beenakker, H. F. P. Knaap and P.
Zandbergen, Physica, 24, 557 (1958).
(16) M. Trautz and P. B. Baumann, Annln Phys. 2, 733 (1929); M. Trautz
and F. Kurz, Annln Phys. 9, 981 (1931).
(17) A. G. de Rocco and J. O. Holford, J. Chem. Phys. 28, 1152 (1958).
(18) H. Iwasaki and J. Kestin, Physica, 29, 1345 (1963).
(19) A. Gille, Annln Phys. 48, 799 (1915).
* K. F. Herzfeld and T. A. Litovitz, op. eit., chapters 6 and 7; T. L. Cottrell and J. C.
McCoubrey, Molecular Energy Transfer in Cases, chapters 5 and 6 (Butterworths, 1961).
 13
THERMAL CONDUCTIVITY:
C O M P A R I S O N OF THEORY WITH
EXPERIMENT

13.1. Summary of the formulae

Various formulae for the coefficient of conduction of heat, A, in a simple gas
have been derived in Chapters 10 and 11. It has been found that A is con-
nected with the coefficient of viscosity, ft, by an equation of the form
A = (/ic„ (13.1, 1)
(cf. (6.3, 3)), where cv is the specific heat at constant volume, and f is a pure
number. The first approximations, [A], and [fi]i, to A and /1 are similarly
connected (cf. 9.7) by a relation
w ^ t n . [/0.'*.
where [f] t is the first approximation to f.
The value of [f], has been found equal to 2-5 for all smooth spherically
symmetrical molecules (cf. (9.7, 4)). Further approximations to f have been
determined in a number of cases, It was shown in (10.21, 5) that for rigid
elastic spheres f = 2-522; for molecules repelling each other with a force
varying inversely as the rth power of the distance (cf. (10.32, 2, 3)), f = 2-5
exactly if v = 5 (Maxwellian molecules), whereas for values of v between 5
and 13 (the important range in Table 14, p. 232), f lies between 2-5 and
about 2-511. Similar results apply to the 12,6 and exp;6 models, for which
f is nearly 2-5 when kT/e is comparable with unity, and rises to about 2-51
when kT/e is large. We may therefore expect that

<* = !/<Cv («3-». 2 )

is very nearly true for all smooth spherically symmetric molecules.
A variety of approximate expressions for [f], have been suggested for
diatomic and polyatomic gases. Assuming that the transport of momentum
and translational energy is unaffected by the internal molecular motions,
and that internal energy is transported at the same rate as momentum,
Eucken derived the formula
[01 = 1 ( 9 7 - 5 ) ('3-1.3)
(cf. (11.8, 3)), where y is the ratio of specific heats. The modified Eucken
formula ( n . 8 , 5) postulates that, when interconversion between internal
and translational energy is very slow, the transport of internal energy is by
1247)
248 THERMAL CONDUCTIVITY! (13.2
a self-diffusion process. If [Z)lt]i is the self-diffusion coefficient, and
u
ii - P[0iJi/[/*L. ( i 3 ' . 3)«» replaced by
[Hi = i { ' 5 ( 7 - 0 + *Un(5-37)}. (131.4)
Both (13.1, 3) and (13.1, 4) ignore interchange between the internal and
translational energies at collision. The Mason-Monchick formula (11.81,5),
which takes this interchange into account, is equivalent to
[f]i-i{is(r-0+2Un(5-3r)}
-Ks-srJd-un^/^+K.+f^-J). (13.1.s)
Here r is a time of relaxation of internal energy into translational, and
Un » pD^Jlfi]^, D ln , (not now assumed identical with [£>iji) is the first
approximation to the coefficient of diffusion of internal energy. If the inter-
change between internal and translational energies is slow, r is large and
£>,„„ uj, approximate to [£>,J„ u^; hence (cf. (11,81, 6)) (13.1, 5) approxi-
mates to
[Hi = 1(15(7-1) + 2 ^ , ( 5 - 3 7 ) ) - 1 ( 5 - 3 7 ) ( 1 - ^ ) ^ 1 , / ^ ] . . (13.1.6)
If different parts of the internal energy, making contributions /V,A/2m,
NtkJ2m,... to the specific heat cv(aNk/2m), possess different relaxation
times Tj,rt,..., (13.1, 6) is replaced by
[Hi- Histy-O+auiits-syM-KI-uW'tMi/f^osiv./ia. (13-1,7)
V

The results for rough spherical molecules agree with (i 3.1, 6) to a first
approximation. However, they suggest that higher approximations to f may
appreciably exceed [f ], for molecules with internal energy.

13.2. Thermal conductivities of gases at 0 °C.

Theory predicts that A, like ft, is independent of the pressure at a given
temperature. As with ft, this result is valid only to the extent that the volume
of the molecules themselves is a negligible fraction of the total volume
occupied by the gas; however, the variation of A with p is relatively insig-
nificant at ordinary pressures, in a gas not close to liquefaction. An apparent
variation with p may result from the onset of convection, or from the
temperature-jump at a wall (cf. 6.31), and these need to be taken into account
in any experimental determination of A. Corrections for losses of heat by
radiation or by conduction in the solid parts of the apparatus also have to
be made, and altogether the experimental determination of A is a matter of
considerable difficulty. The probable error (as distinct from internal in-
consistencies) of a set of measured values can be estimated as not less than
about one per cent
13.2] COMPARISON OF THEORY WITH EXPERIMENT 249
Table 20 gives experimental values* of A for gases at S.T.P., and values off,
or \/ftcv, derived from them using the values of fi and cv given in Tables 11
(p. 228) and 1 (p. 43). Eucken's value \{<)y — 5) for f is also tabulated for com-
parison. The gases in the table are listed in three groups, first the monatomic
gases, then the diatomic and polyatomic gases whose molecules possess
either no permanent dipole moment or a weak one, and finally certain more
strongly polar gases. The data for gases in these three groups will be
discussed separately.

Table 20. Thermal conductivities ofgases at 0 ° C

Ax 10'
Gas (cal./cm. sec. deg.) f m XIfit, i{9Y~S)

Monatomic gases
Helium 34>o a-45 250
Neonf mo 2-5* 250
Argon 392 2-48 2-50
Kryptonf 210 »535 250
Xenon 124 »58 250

Gases with weak or no dipole moment

Hydrogen 4130 2'02 I'92
Deuterium 3°5o 2 07 I 00
Methane 733 1-77 1-695
Acetylene 470 "54 ••54
Carbon monoxide 553 rot 1-90
Nitrogen 575 1-96 I-9I
Ethylene 403 148 156
Air 577 1-96 100
Nitric oxide 562 100 too
Ethane 426 '•44 1 45
Oxygen 585 104 too
Carbon dioxide 348 164 1-68
Nitrous oxide 362 •73 1-68
Chlorine 189 1-825 1-80
Strongly polar gases
Ammonia 524 ''455 i'7>5
Hydrogen sulphide 308 1-49 1-77
10
Hydrogen chloride 301 1-655 °
Methyl chloride 215 1-53 1-64
Sulphur dioxide 204 1-50 1-64

* In general, the values quoted are mean values derived from Thtrmophysical Properties
Research Center, Data Book 2 (Wright-Patterson Air Force Base, Ohio, 1064 and 1966).
However, extensive reference to the original papers has also been made. In consequence,
the values do not in all cases agree with those in the Data Book, though differing from
them only by amounts which its authors suggest are permissible.
t The values of c, for the monatomic gases neon and krypton are calculated from
e, — 3*/2mJ.
250 THERMAL CONDUCTIVITY: (13.3

13.3. Monatomic gases

The relation A •» \(tcv, (13.1, 2), can be expected to apply with reasonable
accuracy to monatomic gases, whose molecules have no convertible internal
energy and are to a high degree spherically symmetric. The testing of this
relation for such gases is of special interest, since here theory affords a definite
numerical relation between three quantities all determinable directly by
experiment. Moreover, the agreement of theory with experiment for the
monatomic gases is crucial to the theory of other gases; the formulae
(13.1, 3-5) are all based on the assumption that the transport of translational
energy conforms, to a first approximation, to the formula A' = f/wj,.
Table 20 indicates that the values of \//icv for the inert gases are indeed in
the neighbourhood of 2-5, but the deviations from this value are appreciable.
For neon, argon and krypton the deviations are within the bounds of
experimental error, and the difficulties of working with xenon are such that
the deviation in this case cannot be regarded as serious. The deviation for
helium is 2 per cent, which is more than would be expected for a gas so
frequently studied; improved accuracy of measurement may remove the
discrepancy.
Experiments on gas-flow through a nozzle have been used by O'Neal and
Brokaw (i) # to determine directly the Prandtl number ftcp/A. a y/f for
a number of gases. They found f = 2-503 for helium and f => 2-506 for
argon, in satisfactory agreement with theory. However, though their method
avoids the possibility of errors in the determinations of A and /* separately,
the possible sources of error in the method itself have not been fully explored.
Some runs of experiments suggest a variation of f with temperature, but
different runs do not always agree even as to its sense.f It is doubtful whether
this variation is real for monatomic gases. At very low temperatures (below
about 20 °K.) values of A for helium calculated from experimental values of
/1 using A •* \ficv still agree reasonably well with direct experimental values
(3), (4), though naturally the precision is less at such low temperatures.
For mercury vapour Schleiermacher (5) gave 3-15 as the value of f.
Zaitseva (6) has pointed out that if Schleiermacher's value of A is combined
with more modern values of ft, the value f » 2-69 is obtained, in rough
agreement with her own results.
For hydrogen the value of the ratio y of specific heats begins to increase
appreciably as the temperature falls below 0 °C, and near - 200 °C. it
approaches the value J appropriate to a monatomic gas. This indicates that
at such temperatures the molecular energy is almost wholly translatory. It is
therefore to be expected that f will rise towards 2-5 below — 200 °C.
Eucken (7) and Ubbink (8) found that this is actually so; Eucken found that
* At in Chapter ia, numbers thui written refer to a list of experimental papers given at
the end of this chapter,
f An extensive review of the experimental values of A, /t and f has been made by Keyes (a).
 Copyrighted Material

13.31] COMPARISON OF THEORY WITH EXPERIMENT 251

f = 2-25 at - 1 9 2 ° C , while Ubbink found that f is between 2*45 and 2-5 at
temperatures below 20 °K. For deuterium Ubbink similarly found a value
near 2-3 at low temperatures.
We may therefore conclude that for monatomic gases f is approximately
2*5, as theory predicts.

13.31. Non-polar gases

For diatomic and polyatomic gases, equations (13.1, 3-7) provide approxi-
mations to [f ]j. As is shown by the values quoted in Table 20, the Eucken
formula (13.1, 3) gives only a crude approximation to f.
If encounters leave the internal energies of molecules virtually unaltered,
(13.1, 4) is applicable; in this case [D^ approximates to its value 2K\j']JP
for molecules without internal energy (cf. (10.6, 4)), and uj x becomes 3A.
For the 12,6 and exp; 6 models 3A is about 1-32 or a little larger, and here we
shall use the mean value 3A = f. Then (13.1, 4) becomes
ff3i = i V ( 2 i r - s ) - (13-3^.0
This value of [f] t is always greater than the Eucken value £(97 — 5).
Equations (13.1, 5-7) take into account the interchange between internal
and translational energies. This interchange increases the part of A due to
the transport of internal energy, but decreases that due to transport of
translational energy to a greater extent, so that [i\ is decreased below the
value given by (13.31,1). Equation(i3.1,6) is asimplifiedversion of (13.1,5),
valid only if the interchange between internal and translational energies is
fairly slow. Equation (13.1, 7) is preferable to (13.1, 6), since it allows for the
possibility of different relaxation times for vibrational and rotational energy.
The relaxation of vibrational energy is normally so slow that its effect on
[f] t is negligible; hence, since y = 1 + 2/N, (13.1, 7) reduces to
CTi = ^ 5 ( 7 - 1 ) + ^ ( 5 - 3 7 ) - ( 1 - ^ 1 ^ ( 7 - 1 ) ^ 1 ^ / ^ ^ ,
(I3-3 1 . 2)
where Nr, [ T ^ now refer to the rotational energy. For diatomic and other
linear molecules NT = 2, for non-linear polyatomic molecules Nr = 3; also
we may write , , „, , . , .
M i = ZMJi'*7ip, (I3-3 1 . 3)
where (cf. (12.5, 3)) £. is the average number of' collisions' occurring in the
time [TT]t.
In (13.31, 2), the interchange of rotational and translational energy is still
supposed to be fairly slow, so that u'u ~ | . If the interchange is not slow,
u i i ( s P[#u]i/l>]i) m u s t b e replacedbyuj(= ^1),^010:also(cf.(13.1,5))
p[Tr]t is replaced by/>[Tr]i + t/i]i(Juu + ^Ar r ), so that (13.31, 2) becomes
[f]i = J { i 5 ( 7 - i ) + 2Un(S-37)
- ( t - u ; 0 f ( 7 - O^U>[T,yMi + iuI1 + AiVP)}. (13.31, 4)
Copyrighted Material
252 THERMAL CONDUCTIVITY: [13.32
For hydrogen and deuterium (13.31,4) and (13.31,2) are virtually
indistinguishable, since £, ~ 300 for these gases. From (13.31, 1) we find
the values [f]t « 2-05 and [f], = 2-03 for the two, agreeing satisfactorily
with the experimental values f — 2-02 and f — 2-07 of Table 20.
For other non-polar gases the values of f in Table 20 often agree better
with the Eucken expression $(97-5) than with (13.31, 1). This corresponds
to a relatively low value of [r r ]„ and so of the collision-number £r: thus
(13.31, 4) must be used rather than (13.31, 2). The value of [f], given by
(13.31, 4) agrees with the Eucken expression if

— + l u „ + A ; v r = 2 ( u . i r i ) -j-^y. (13.31.5)
For a diatomic gas whose internal energy is wholly rotational (Nr — 2,
y = 1 '4), the following table indicates the dependence of £, on u u implied by
(»33».5):
itf, ... ! 13 t»s 12
Cr ... 0688 1-165 3'11 3°55

The values of £, for a polyatomic gas, or one whose molecules have vibra-
tional energ}', are even lower. Since & is likely to be well above unity, values
of U|i as high as 1-3 are improbable.
For methane, nitrogen and oxygen the values of f given in Table 20 are
appreciably above the corresponding Eucken values; even for these, the
values of £, corresponding to u n = 1-3 are between 2 and 4. Ultrasonic
experiments give distinctly higher values* (though not very accordant),
indicating smaller values of u n . Thus the coefficient of diffusion of internal
energy, /),„„ is likely to be appreciably less than the self-diffusion coefficient
[*>„],.
The values in Table 20 refer to a temperature of 0 °C. Theoryf suggests
that £, may increase considerably with increasing temperature; corre-
spondingly Uu may also increase towards }. Thus at high temperatures f
may approximate more closely to the value given by the modified Eucken
formula (13.31, 1) or (13.1,4). Mason and MonchickJ give evidence that
this is actually so for a number of gases.

13.32. Polar gases

As appears from Table 20, for polar gases f is well below the Eucken value.
Mason and Monchick§ ascribed this to rotation changes during encounters,
caused by dipole moment interactions. They calculated that for certain
gases, notably hydrogen chloride and ammonia, this may reduce the diffusion
* See, for example, the valuea quoted by O'Neal and Brokaw (1).
t See J. G. Parker, Phyt. Fluids, a, 449 (1959); alao the reference! quoted in ia.s.
X E. A. Maaon and L. Monchick, J. Chtrn. Phyt. 36, 162a (1962).
i Loc.cit.
13.4] COMPARISON OF THEORY WITH EXPERIMENT 253
coefficient Dlnt for internal energy to make u j , in (13.1, 5) equal to about
unity, as well as decreasing the collision-number £,. For details the reader is
referred to the original paper.

13.4. Monatotnlc gas-mixtures

The first approximation to A for a binary mixture of gases without internal
energy is (cf. 9.82)

1 Jl Ki
x!Q1 + x i o i + x 1 x t Q l t ' + '
1
where Qlt = 3 ^ - ^ / , ) (5-48) + 4 ^ M , A ( I I - 4 B ) + 2P1P„ (13.4,2)
Qi« = 2F{Pi + P 2 + (1 I - 4B - 8A) Afj M,}, (13.4,3)
Q, = p 1 {6M| + (s- 4 B)Aff + 8M l M,A}, (13.4,4)
with a similar formula for Qr Here (cf. (9.82, 2)) P ^ P , and F are given by
FSISAE/4WO, P,-F/[AJ„ P, = F/[AJ!. (13.4,5)
The constants A, B are given for force-centres by (10.31, 10) and Table 3 ;
for the 12,6 model they are given in Table 6 (10.42). For most of the signi-
ficant range o(eltlkT for the 12,6 and exp;6 models, A lies between about
0-44 and 0-46, and B between 0-72 and 0-66; (13.4, 1) is not very sensitive to
the precise values adopted for them.
If x, is small, (13.4, 1) becomes the formula giving the effect of an impurity
on the conductivity; it then approximates to

Wt = [ A 1 ] 1 ( I + ^ [ { 2 P 1+ A / 1 M I ( I I - 4 B - 8 A ) } » - ^ | ] ) ,

(13-4,6)
an equation which can readily be seen not to involve [AJ,.
As in the case of a simple gas, (13.4, 1) is applicable only to mixtures of
monatomic gases. All pairs from among the five inert gases have been
studied, with reasonable agreement between theory and experiment. This
is illustrated in Table 21 by results for helium-argon mixtures at 29 ° C ,
based on the experimental data of von Ubisch (9). In applying (13.4, 1) the
values A = 0-452, B = 0-667 a r e adopted, the value of A being that used in
12.41 in discussing the^.viscosity of the same mixture. The value of E giving
the best fit with experiment is found to be 9-103 x i o - 4 poise, as against the
value 8-70 x 10-* poise (at 15 °C.) used in 12.41. About half of the difference
between the two may be ascribed to the different temperatures involved; the
remainder is due partly to experimental error (including possible errors in the
values of A, and A,) and partly to the fact that the formulae used are not
exact, but are first approximations to the exact formulae.
Table 21 gives a second set of calculated values obtained by Mason and
von Ubisch* from (13.4, 1); they used the same values of A, B, but different
* E. A. Mason and H. von Ubiach, Phyt. Fluids, 3, 355 (i960).
254 THERMAL CONDUCTIVITY: (13.41
values of E, [AJI, [AJJ, derived on the exp; 6 model with the force constants
for unlike molecules determined by empirical combination rules. As can be
seen, the three sets of values are in reasonable agreement.

Table 21. Thermal conductivity of mixtures of

helium and argon at 29 °C

Axio'
(cic.)
Helium Axio' Axio' (Mason and
(%) (exp.) (calc.) von Ubiach)

0 434 434 426

29-0 843 857 854-5
459 I2M 1205 1205
7*4 aoio 2019 2030
894 2920 2889 2897
100 3670 3670 3702

A discussion of possible maxima and minima of [A], as Xj/x, varies,

parallel to the corresponding discussion for \ji]lt can be based on (13.4, 1).
However, such maxima and minima are not observed for mixtures of
monatomic gases.

13.41. Mixtures of gases with Internal energy

Hirschfelder* has generalized the modified Eucken formula (13.1, 4) to
apply to mixtures of gases whose molecules possess internal energy. Trans-
lational and internal energies are regarded as conducted independently. The
translational conductivity A' is then the monatomic gas conductivity given
by (13.4, 1) with [AJ, [AJt replaced by if/tJiC^,, ilji^c'^, where c'^, c'^ are
the translational specific heats. The contribution of internal energy to A is
assumed to arise from the diffusion of molecules through the mixture,
carrying with them their internal energy. The final formula (which can readily
be derived by applying to mixtures the approximations made in deriving
("•8.9))» oc- oe-

where Dn, Dlt, Dn are the usual coefficients of diffusion and self-diffusion,
at the pressure of the actual gas.
The significance of the expression XjDf^ + XjDf,1 in (13.41, 1) is that
Dhl> D{t denote the resistances to the diffusion of a molecule m, in otherwise
pure gases of molecules m, or m,; x, Di* + x, Dft1 gives a weighted mean of
these resistances appropriate to the actual mixture. The conductivity AJ due
* J. O. Hirschfetder, Sixth International Combustion Symposium, p. 351 (Reinhold
Publication Corporation, New York, 1957). See also E. A. Mason and S. C. Saxena,
Phys. Fluids, 1, 361 (1958).
13.42] COMPARISON OF THEORY W I T H EXPERIMENT 255

to internal energy in a pure gas of molecules ffit at the same pressure is

c
li(pJ*i) Dn, since the density of the gas m, in the actual mixture is x t times
that in the pure gas. Hence (13.41, 1) can also be written
A" A"
A = A+
i + x s Z ) I l 1 / x 1 D l t + il-x 1 D M /x 2 Z) I t - ^3-41.2)
This equation can readily be generalized to apply to a mixture of any number
of constituents.
Equation (13.41, 2) is strictly applicable only to mixtures like H , - D „ in
which the interaction between translational and internal energies is very
slow. Monchick, Pereira and Mason* have given formulae valid when this
interaction is not slow, which generalize (13.1, 5) to apply togas-mixtures.
The formulae, even after a number of approximations, are highly com-
plicated. In general they involve four relaxation times, and coefRcients of
diffusion for internal energy which, as in simple gases, may deviate con-
siderably from the ordinary diffusion and self-diffusion coefficients.
Monchick, Pereira and Mason had reasonable success in fitting the experi-
mental data, both by their own formula and by the Hirschfelder formula
(13.41, 2), for a number of gases. They suggest that it is usually sufficient to
use (13.41, 2), with Dn, Dlt, Dtt replaced by adjustable diffusion coefficients
for internal energy. In calculating A' from (13.4, 1), however, they prefer,
when the pure-gas relaxation times are known, to replace [AJj, [A,]j in that
equation by translational conductivities AJ, A, given by

K = fL«i]i ( 4 - K i d O i l i -PilDxAiVPiTi) (1341.3)

and a similar relation; (13.41, 3) gives the translational contribution to the
total conductivity derived from (11.81, 6).

13.42. Approximate formulae for mixtures

As with viscosity, the complication of the formulae for A for mixtures has
led to a search for simple approximate formulae. These are normally of the
Wassiljewa form (cf. 6.63), expressing A as the sum of parts A ^ A<^ arising
from the two gases in the mixture, where
*»-xi/{*i<^.)-1+^1,)-1}. A (1) = X 2 / { X 2 ( A , ) - » + X 1 ( A ; 1 ) - ' } , (.3-42.1)
and A;„ AJ, are 'mutual conductivities' of the gases considered.
Such a formula can be derived for monatomic gases from (13.4, 1) by
making the (unphysical) assumptions B = f, A = 1; then this equation
becomes
W i = x,/{x,[AJr» + xJAiJ T »} + x.ftxJAJl» + xJAJJlT, (1342. 2)
1 1
where [AJJf = ( 3 M 1 + A/ 2 )/F, [A,,]," = ( 3 M a + M I )/F, (13.42,3)
* L. Monchick, A. N. G. Pereira and A. E. Mason, J. Chem. Phys. 4a, 3141 (1065). See
also L. Monchick, K. S. Yun and A. E. Maaon, J. Chem. Phyt. 39, 654 (1963).
256 THERMAL CONDUCTIVITY [13.42
and F - 15^/4«, - SPlD^JzT. However, the error involved in making
these substitutions is not likely to be small. A rather better approximation is
obtained by applying to the conductivity the argument employed with the
viscosity in deriving (12.43, 3); this gives an equation of form (13.42, 2), with
Wal7» - J P f , - m [6Af, - (S -4B)MJ + i6M, Mt A}/F (13.42.4)
and a similar equation for [A4JY1. This equation in general overestimates
[A]„ but not greatly if the molecules of the two gases are not too unlike.*
Formulae of the Wassiljewa type have been used, with fair success, to
represent the variation of A with composition when the molecules possess
internal energy. If no account is paid to physical interpretation, AJt and AJ,
can be regarded as adjustable parameters. More usually it is desired to
predict conductivities for mixtures from those of the pure gases. The pro-
cedure normally followed in this case is to start with a formula of type
A - ^xJfa+etaXiDJDiJI + A,x1/{xt+atlx1Dt»/Z)lt}, (13.42, 5)
where a„, an are regarded as functions of mjmt, which are determined
semi-empirically. In support of the assumption (13.42, 5) we may note that
substitution from (13.42, 3) or (13.42,4) into (13.42, 2) yields an expression
of the form (13.42, 5), since A, is a numerical multiple of pDn/T; so too is
the contribution to (13.41, 2) from the internal energy.
The Lindsay-Bromley formulaf is one of the best known of this type. It
is based on the Sutherland model, and replaces <t\tDxlIDlt in (13.42, 5) by

To conform exactly to the Sutherland model with a„ — a^ — 1, the power

of mt/ml in this should be (m^/m,)*, and the whole expression should be
multiplied by {zmi/(ml + m,)}*; the changes made are empirical corrections.
Lindsay and Bromley further suggest that the Sutherland constant Sn can
be taken as (S,5 t )i for non-polar gases, but if one gas is strongly polar it may
be only 0733 times as large.
REFERENCES
(1) C. O'Neal and R. S. Brokaw, Phys. Fluids, 5, 567 (1962), and 6, 1675 (1963).
(2) F. G. Keyes, Trans. Am. Soc. Mech. Eng. 73, 589 (1951).
(3) J. B. Ubbink, Physita, 13, 629 (1947).
(4) K. Fokkens, W. Vermeer, K. W. Taconis and R. de Bruyn Ouboter, Phytica,
30, 2153 (1964).
(5) A. Schleiermacher, Annln Phys. 36, 346 (1889).
(6) L. S. ZaiUeva, Sov. Phys. Tech. Phys. 4, 444 (1959).
(7) A. Eucken, Phys. Zeit. 14, 324 (1913).
(8) J. B. Ubbink, Physiea, 14, 165 (1948).
(9) H. von Ubisch, Ark. Fys. 16, 93 (1959).
• T. G. Cowling. P. Gray and P. G. Wright, Proc. R. Sot. A, 376, 69 (1963).
t A. L. Lindsay and L. A. Bromley, ltd. Eng. Chtm. 4a. 1508 (1950).
 14
DIFFUSION: COMPARISON OF
THEORY WITH EXPERIMENT

14.1. Causes of diffusion

The general equation of diffusion for a binary mixture, (8.4,7), can be put in
the form
W >,
Q - C , = -(x1xt)-^1,{vx1 + ' ^-W»)Vlnj>

- & £ ( * , - J U + krVln7-), (,4.1,,)

showing, as noted in 8.4, that the velocity of diffusion has components due
to non-uniformity of the composition, the pressure and the temperature
of the gas, and another component due to the different accelerative effects
of the external forces on the molecules of the two constituent gases. In
experimental determinations of the coefficient of diffusion, Dlt, it is usually
the diffusion due to non-uniformity of composition which is measured.
In a gas at rest (zero total number-Row of molecules) the equation of
diffusion becomes (cf. (8.3, 13))
x,c, = - x , c , = -Z) 18 {/)-»(V/. 1 -p l F,) + k r V l n T). (14.1, 2)
The first term on the right corresponds to diffusion set up by a lack of balance
between the forces on the molecules of the first gas, and the gradient of its
partial pressure. A similar equation exists in terms of the corresponding
quantities for the second gas.
As an example of pressure diffusion, corresponding to the second term on
the right of (14.1, 1), we may cite the process of diffusion in the atmosphere;
by reason of the variation of pressure with height the various constituents
tend to separate out, the heavier elements tending to sink to lower levels, and
the lighter elements to rise to higher levels. This process is not immediately
due to the force of gravity on the molecules, as the accelerative effect of this
on all molecules is the same; it is an indirect effect, due to the pressure
gradient which gravity sets up. (Actually but little variation of composition
with height is observed in the atmosphere up to 100 km. height, since the
mixing effect of wind currents and eddy motion counteracts the separative
tendency due to diffusion.)
Pressure diffusion also occurs in a gas that is made to rotate about an axis;
the heavy molecules then tend to the parts most distant from the axis, the
density-distribution of each constituent in the steady state being similar to
that given by (4.14, 9, 11), putting ¥ = 0.
[ 257)
258 DIFFUSION: [14.2
The most important example of forced diffusion, corresponding to the
third term on the right of (i 4.1, i) is the diffusion of electrically charged
particles in a partially ionized gas under the action of an electricfield,giving
rise to an electric current. This case is considered in detail in Chapter 19.
The steady state of a gas-mixture in a conservative field of force, repre-
sented by (4.3, 5) can be regarded as due to the attainment of a balance
between the velocities of diffusion due to non-uniformity of composition
and pressure, and to the applied forces. Thus, for example, in an isothermal
atmosphere under gravity the steady state is that in which pressure diffusion
exactly balances the diffusion due to non-uniformity of composition. It is
possible, in fact, to derive (4.3, 5) from the condition that there is no
diffusion, using (14.1, 2).
We defer consideration of thermal diffusion till 14.6.

14.2. The first approximation to Dlt

Observations of the mutual diffusion of pairs of gases are difficult to make,
and such observations as have been made are liable to a fairly large experi-
mental error. This should be borne in mind when comparing theory with
experiment.
Several formulae for the first approximation [O1J1 to the coefficient of
diffusion have been derived in Chapters 9 and 10; they are quoted here for
convenience of reference. For rigid elastic spheres of diameters <rv <rt, by
(10.1,5) and (10.2,1),

where trlt • tfoi + Ot).

For molecules repelling each other with a force «r„r*, by (10.1, 5) and
(10.31,8,9),

w }-n (¾^)1 (£T*.

(142. *)
For attracting spheres (Sutherland's model), by (10.41,9),

^.-,4(^)7(-^)- (14-2. 3)

For the 12,6 and exp;6 models, by (10.1, 5) and 10.4,

where TJ^YiXO is » function of kTjen, given in Table 6 (p. 185) for the 12,6
model, and illustrated in Fig. 8 (p. 185) for this and the exp;6 model.
14.21] COMPARISON OF THEORY WITH EXPERIMENT 259
For rough elastic spheres, by (i 1.61, 4),

[L>ul1
~ 8na\t\ 2nmlml ) ^ + 2/^,1 (I42
* 5)

where *r„ Kt, K0 arc related to the radii of gyration by (11.6, 2).
This indicates that interaction between rotation and translation decreases
[DjJi somewhat. However, the effect of such interaction is not large;
normally it is ignored in what follows, the molecules being treated as if
smooth.
The first approximation to the coefficient of diffusion does not depend
on the proportions of the mixture in which diffusion occurs: also it depends
only on encounters between molecules of opposite types. Thus, to a first
approximation, encounters between like molecules do not affect diffusion.

14.21. The second approximation to Dlt

The second approximation obtained in (9.81, 3) is of the form

[fliJi-PiJi/O-a). (i4-«. 0
, ,A/fp,x5 + M|PjXl + p 1 2 x,x. . ,
where A = 5c*—-—\ V li-l_* (14.21,2)
and P! = MjE/t/*,],, p 2 = M2n/[/it]v ('4-2J.3)
8
P « = 3(*A - ^ 2 ) + ^MtMt A, (14-21.4)
Q, = p , ( 6 ^ + s M f - 4 M j B + 8Af1A/2A), (14.21, 5)

Qi* = 3(Mi~M*i*(5-4B) + 4MiMtA(*>-4B) + 2 p i p «. (142 1 . 6 )

with a similar relation for Q2; the quantities A, B, C, E, are defined by (9.8, 7, 8).
For Maxwellian molecules (for which v = 5)0 = o(cf.(io.3i, io));hencein
this case & = 0 and the second (and every later) approximation is identical
with the first, in agreement with the results of 10.33.
Further approximations to Dl2 are larger than [Z)12]2. In the special case
in which the ratio m1/m2 of the molecular masses is very large and ns/nl is
very small it has been shown (cf. Table 7, 10.52) that the true value of Dlt
bears to [£>12]2 the ratio 1-132:1-083 = 1*046 for rigid elastic spheres; for
other more realistic types of interaction the ratio is usually between 1 and
1 -02. In other less special cases the effect of later approximations is similar in
order of magnitude, but smaller.
Kihara's approximation to [Z>12]t is obtained by setting B = j in (14.21,5,6).
This normally increases [Z)lt]i towards the exact Dlt; Kihara's [Dlt]z is, in
fact, often nearly as accurate as the more usual [Dlt]a, and is sufficiently
accurate as an approximation to DJ2 for most purposes.*
• E. A. Mason, J. Chem. Phys. rj, 75, 78» (1957).
260 DIFFUSION: {14.3

14.3. The variation of Dlt with the pressure and concentration-ratio

All the formulae of 14.2 and 14.21 indicate that D„ is inversely proportional
to n, i.e. to the pressure of the gas, if the temperature and composition are
constant. This proportionality was first observed by Loschmidt (1),* and
has been generally confirmed at moderate pressures.
As can be seen from (14.21, 2), the quantity A appearing in [Z>„], varies
with x,/x„ i.e. with the composition of the mixture in which diffusion is
taking place. For example, the values of A in the limiting cases in which
Xj -+ 1 and Xj -+ 0 are A„ A2, where
A, = 5«fcV{(5 -4B)mf+ 6^1 + 8 ^ ^ , } (14.3,1)
with a similar relation for A,. Hence A, > A2 ifml > mt; also it can be shown
that A increases from A, to A, as x, increases unless mt and m, are nearly equal
(in which case A does not vary greatly). Thus by (14.21, 1), when m, and m,
are not nearly equal D„ can be expected to increase with increasing concen-
tration of the heavy molecules; the increase is greatest when the molecules
of the two gases have very unequal masses.
For rigid clastic spheres (14.3, 1) becomes
*i = m?/(i3w! + 3omJ+ i6m,m,). (14.3, 2)
Hence in this case, to a second approximation
PIJLJ-O _ i j j i , _ ^.^1/(13^1 + 3^1.+ 16^13)
[o,j ni -o i - ^ i T-w!/(i3W«+yoOTi+16^,^,)-
If Mg/m, -*• o, A|, At tend to the limits 1/13 and zero, and the ratio of the
end-values of [Dlt]t becomes equal to 13/12. Thus to a second approximation
the maximum variation in the coefficient of diffusion for rigid elastic spheres,
as the proportions vary, is 8J per cent. Further approximations increase this
toi3-2percent(if«^m 1 ->o,[D„] Bi . 0 = [Dlt]v and [Z>WL,„0 = »-i32[^i»]i
—cf. the results for the Lorentz case already quoted). The Kihara second
approximation gives a variation of 1/9, or 1 I-I per cent.
For more realistic molecular models, the factor c* in A is less than half as
great as for clastic spheres: the other terms in A vary relatively little with the
model. Thus for actual gases the variation of/),, with composition is unlikely
to exceed about 6 per cent. This statement refers to uncharged molecules;
for inverse-square interaction, a two- or threefold variation in Du is possible
(cf. Table 4, p. 179).
14.31. Comparison with experiment for different concentration-ratios
The theoretically predicted variation of D, t with composition is not much
greater than the experimental errors, and is not easy to establish. Evidence
of such a variation was first provided by a series of experiments made for this
* As in Chapters 12 and 13, figures in parentheses refer to a list of experimental papers
at the end of the chapter.
14.311 COMPARISON OF THEORY WITH EXPERIMENT 261
purpose at Halle about the year 1908, and reported by Lonius (2). In that
series, the longest runs of experiments were for H,-C0 2 and He-A mixtures;
these indicated a variation of the correct sign, similar in magnitude to that
predicted by (14.21, 1, 2) assuming rigid elastic spherical molecules. How-
ever, the experimental variation was greater than the theoretical; the reverse
could be expected because, even though the theoretical variation is increased
by taking later approximations, this is more than offset by taking realistic
values of c* (about 0 4 times the rigid sphere value for He-A mixtures, and
about 0-25 times for Hj-COj). More recent experiments, using radioactive

0 20 40 00 80 100
".. He
Fig. 12. Variation of D„HDit)^, with composition in H,-N t mixtures at 294-8 °K. (upper
curve) and He-A mixtures at 295 °K. (lower curve); (D„)„, is the value of On for equimolar
mixtures. The curves give theoretical values, crosses experimental values taken from ref.
(6). (We wish to acknowledge the kindness of Professor Beenakker and Dr van Heijningen,
who made available their original data, used by us in constructing the figure.)

tracers (3, 4, 5), indicate a variation in Z)„ for these mixtures which is
definitely smaller than that found in the Halle series.
The variation found for other mixtures, apart from discrepancies that can
fairly be ascribed to experimental errors, similarly agrees with theory as
regards sign and order of magnitude. The magnitude of the variation, and
the experimental uncertainties, are illustrated in Fig. 12. This gives the
ratio of the measured Dlt to its value for equimolar mixtures (x t = x, = J)
for the gas-pairs H 2 -N 2 , He-A, found in a careful series of experiments at
Leiden (6). For comparison, theoretical curves are also given. These
are based on (14.21, 1, 2), using the Kihara approximation B = J; A is
given the value (~ 0-44) appropriate to the 12, 6 model, and c is regarded
262 DIFFUSION: [14.32
as adjustable. For N,-H, mixtures the value of c agrees with that calculated
on the 12,6 model, determining the force constant el3 either from the com-
bination rule e„ •= (e,e,)i or from the variation of Dlt with the temperature
(see 14.4, below). However, for He-A mixtures (and for other mixtures
containing helium) the Leiden workers found that the value of c required to
fit the observations was at least 20 per cent greater than that similarly
calculated.
Since the error in the first approximation to the coefficient of diffusion is
not very large, it is ignored in the subsequent discussion. This implies that
the variation of Dlt with the proportions of a mixture is ignored. The
discussion of the experimental results is thereby the less seriously affected,
since in some cases the proportions of the mixture were either not recorded,
or not kept constant, by the experimenter. However, it should be remembered
that the error involved in neglecting the difference between [ 0 l t ] | and Dit
may be decidedly greater than the corresponding errors for /1 and A, if mjmt
is very different from unity.

14.32. Molecular radii calculated from D„

Table 22 gives the experimental values of D„ for several pairs of gases,
reduced to 0 °C.a From these the values of <rlt given by (14.2, 1) have been
calculated; these are effective values s'lt of <r„ for diffusion. The quantity s[t
is not identical with slt, the corresponding effective value of <7„ in the
viscosity of mixtures (cf. 12.41), but is related to it by s[\ = (2/5A) s\t; taking
A = 4/9, we see that s'u is about 5 per cent smaller than slt. For comparison,
the table also includes values of i{cr1 + (Tt), where <JX, <rt are the viscosity
diameters of Table 11 (p/228).
It can be seen that the values of i(<r, + o-,) derived from viscosity are in
general larger by about 10 per cent than those of <r„ obtained from Dlt. This
discrepancy is largely due to the difference between t'lt and *,s, and to the
fact that s'lt is derived by using values of Dlt in a formula involving only the
first approximation [Dlt]lt whereas crv <rt are derived from an exact formula
for (t. However, it may also arise in part from the imperfect representation of
the molecules by rigid elastic spheres, and from a difference in nature
between the interactions of like and unlike molecules.
The values of Dlt quoted in Table 22 refer, where possible, to mixtures
in which each constituent is present in moderate proportions. The accuracy
of D l t measurements is not high, and the values quoted may, in unfavourable
cases, be in error by several per cent.
* Here (and throughout this book) values of D„ are expressed in cm.'/sec.
«Q>>>OOOQQQZZZZZZZZZZZZZZZOnxss XSXZXOXXXXSXXSXXSXS

uiiiiiiiiiiH^iyitiiiiiiiiiiiHiiiii^iiiji!^
'i M o 6>vi 6 -0 * bo* •* 6 vi ui * <
N>5 < - • -

ui o^ui *o — fi - 6tfc>vj -* <J t i O v j u i > cib*j ^ 6 > H \ i i

264 DIFFUSION: [14.4

14.4. The dependence of Dlt on the temperature; the intermolecular

force-law
The theoretical variation of Z)„ with temperature depends on the special
molecular model employed. If the pressure is constant, so that nT is constant,
by (9.81,1), (9.71,1) and (9.8, 7),

2
d\nT d\nT * ^ U4-4.»)
Now c = 0 for Maxwellian molecules, c = J for rigid elastic spheres; for
physically satisfactory models of electrically neutral molecules, c in general
lies between 0 and -fg. Hence in a small temperature-range [£„], varies as
T1+t, where t ( s 1 — jc) lies between about f and 1. If unlike molecules repel
each other with a force varying as I-"", by (14.2, 2), [Z>iJi varies exactly as
a power T1*', where
* - i +r^-:; (144.2)

this is precisely similar in form to (12.31, 2). In many cases the number s can
be expected to be intermediate between the two values of the * of 12.31,
derived from the viscosities of the constituent gases of the mixture.
Experimental values of Dlt over a range of temperatures can be analysed
to determine force constants for unlike molecules according to various
molecular models, much as viscosity data were used in Chapter 12 for like
molecules. The models most commonly used are the repulsive inverse-power
model, and the 12,6 and exp;6 models. Experimental data covering a wide
range of temperatures are rare, especially below 0 °C. The constants derived
from experiment for a number of gases are given in Table 23. Walker and
Westenberg (11) and Saxena and Mason (3) suggest that at temperatures
above 0 °C. the experimental values are as well represented by a T1*'
dependence as by more complicated expressions; however, the 12,6 and
exp; 6 models can be expected to give a much better representation of the
actual force-law.
The values quoted for the force-constants do not always lie between those
for the separate pure gases. Part of the discrepancy is likely to be experi-
mental in origin, part may arise from indeterminacy in the molecular
constants; values of D„ calculated on the exp; 6 or 12,6 models, using force-
constants derived from those for the pure gases by the combination rules of
12.41, in general agree fairly well with the experimental values. Where the
agreement is not good, the discrepancy may be attributed to a defective
molecular model, or to a difference between the interactions of like and
unlike molecules.
14.5] COMPARISON OF THEORY WITH EXPERIMENT 265

Table 23. Variation of Dlt with temperature*

Exp;;6 model 12,6 model r~'" repulsion

Gases Ref. a «„/* «"11 » "it

co,-o, (11) *5S 3616 17 213 3-365 0-805 7S4

H.-0, (n) 143 3*34 14 15* 2825 0780 8l3
CH.-0, (11) 220 3-612 17 182 3367 0793 7-83
CO-O, (•I) no 3733 «7 91 3480 0-710 10-52
He-N, (11) 85 3069 17 69 287 0691 "45
CO.-N, (•I) 184 3782 17 154 35* 0753 891
He-A (6,11) 298 351 14 402 2-98 0725 99
He-Ne (6) 205 301 '3 237 264 — —
Ne-A (6) 609 3-50 15 617 311 — —
NH.-A (16) — — — 22465 3286 — —
NH,-Kr (16) — — — 264-95 3349 — —
Ne-CO, (3) — — — 824 339» 083 705
H.-CO, (3) — — — 94 330 084 69
He-CO, (3) — — — 615 319 0-84 69

* This list does not attempt to be exhaustive.

14.5. The coefficient of self-diffusion Dn

If the molecules of the two gases whose mutual diffusion is considered are
identical, the coefficient of diffusion becomes the coefficient of self-diffusion
of a simple gas. By (14.2, 1) the first approximation to this for rigid elastic
spheres of diameter <x is given by

lU ll
" ~8n<r*\nm) ~ 5 p ' 1*4-5. »)
while in general, by (10.6, 4), the first approximation is given by
[OiJi = 3AL"]i/P. ('4-5. 2)
[/<]t being the first approximation to the coefficient of viscosity, and A
a numerical factor, denned in (9.8, 7). If we write
pDn = 3Afi, (14.5, 3)
ignoring the difference between Dn, /1, and their first approximations, then
(cf. 10.6) for rigid elastic spheres the error incurred is only J per cent. The
error with more realistic molecular models is at most half as great, and can be
neglected.
Equation (14.5, 3), like the equation A = |//c r , provides a crucial test of
the theory, and of the correctness of assumed molecular models. For
Maxwellian molecules 3A = 1-55; for rigid elastic spheres 3A = 1-2; and for
molecules repelling like r~", 3A varies between these values as v increases
from 5 to 00. On the other hand, 3A is close to 1-32 for the 12,6 model, and
between about 1-32 and 1-36 for the exp;6 model, \ikTje lies between 0-5
and 10. Hence (14.5, 3) provides a clear test for distinguishing between
IO CUT
266 DIFFUSION: [14.S
models with or without attractive fields. However, the results for rough
spheres suggest that unduly low values of pDu//t may result if interchange
can occur between internal and translational energy.
Naturally it is not possible to follow the motions of individual molecules,
possessing no feature distinguishing them from other molecules of the gas,
so that no actual experimental determination of the self-diffusion coefficient
can be made. For this reason Lord Kelvin* proceeded by calculating values
of Z)„, using the elastic-sphere model. His method was to calculate <rxt from
Dlt, as in Table 22 (14.32). If <rlt is assumed to equal the mean of the dia-
meters of unlike molecules, from the coefficients of mutual diffusion Dlt,
/),3, Dn of the three pairs which can be chosen out of three gases the
diameters o*„ crt, a, of molecules of the gases can be found. On substitution
in (14.5, 1), the values of Dn are found for each of the three gases. A check
on the method is afforded by the agreement or otherwise between the values
of Dn for the same gas, derived from different triads of gas including it.
By this method, the following values of the coefficients of self-diffusion of
four gases, hydrogen, oxygen, carbon monoxide and carbon dioxide, have
been found, using the values of Dlt given in Table 22:

The three gam uaed Du for H, Du for O, Z>„ for CO D„ for CO,
H.-0.-CO I »4 0-191 0179 —
H,-0,-CO, I 30 0-182 0-107
H.-CO-CO,
—
1-34 — 0-:70 0-105
Or-CO-CO, — 0193 0-176 0-101
Mean I »9 0-189 0-175 0-104

The agreement between the values of D „ obtained from the different

triads of gases is reasonable. From the means of the values of Du values
of pDxJ/i have been calculated; they are given in Table 24. They do not
agree with the value 1-2 appropriate to rigid elastic spheres; the reason
is the systematic difference between the effective radii for viscosity and
diffusion, already mentioned in 14.32. Since the values were calculated on
the assumption of elastic spheres, it is not legitimate to attempt to derive
from them information about the actual force-law.

Table 24. Coefficients oftelf-diffusion

Gaa fin px. io* /ix 10' D„pl/t

H, 129 899 850 «37
O, 0-189 1429 1926 140
CO 0175 1250 1665 1-3«
CO, 0-104 «977 1380 t-49

* Lord Kelvin, Baltimore Lectures, p. 295. The formula for D,t uaed by Kelvin, however,
differed from (14.2, 1) by a conatant factor.
14.51] COMPARISON OF THEORY WITH EXPERIMENT 267

14.51. Mutual diffusion of isotopes and like molecules

The mutual diffusion of isotopes of the same gas closely resembles self-
diffusion, especially when, as is the case with the heavier gases, the ratio
mjmt of the molecular masses is nearly equal to unity. The diffusion is
relatively easy to follow when the molecules of one isotope are radioactive.
In deriving Dn from the mutual diffusion coefficient Dlt of isotopes it is
necessary to remember that Dn is proportional to the mass factor
[(»«1 + »J,)/(2MJ»I 2 )]J, whereas Dn is proportional to mji.
Values of pDu/fi derived from isotopic diffusion experiments for a
number of gases are given in Table 25. The majority lie between 1-32 and
1-38, corresponding to values of A between 0-44 and 0-46. On the simple
inverse-power model of 10.31, this would imply a force-index v between
about 13 and 17—irreconcilable with the viscosity results. However, values of
pDxxJn between 1-32 and 1-36 are to be expected on the 12,6 and exp;6
models. The rather higher values also found may be due partly to defects
in these models, but may also be explained by a slightly smaller effective
collision cross-section for unlike isotopes than for like ones. There is no
evidence of a decrease in pDu//i due to interchange between internal and
translational energy.

Table 25. Self-diffusion coefficients from isotopic

diffusion experiments

Gas Temp. (°C.) D„ pD,,l/i Ref.

Hydrogen »3 '455 '•37 (7)

Helium *3 »555 '3* (7)
Methane 0 C206 '4* (ao)
25 0223 '33 ('9)
Neon 0 0451 '35 (20)
-'955 00492 '35 (20)
Carbon monoxide 0 OIOO I 42 ('8)
Nitrogen 0 0-:78 '34 (20, 2 1 )
-'955 00168 '43 (*>)
Oxygen 0 0181 '35 (20, 2 1 )
-195 5 00153 '38 (JO)
Hydrogen chloride 22 01246 '•33 (a2)
Argon O 0157 »33 (»0, 2 3 )
m

Carbon dioxide O '375 (18, 2 0 , 21)

895 136 (18)
-78-3 131 (18, 3 0 )
Krypton 0 00795 I 29 (*4)
200 0214 I 27 (24)
Xenon 0 00480 '33 Us)
105 00900 '•3' (*5)

IO-3
268 DIFFUSION: [14.6
The values in Table 25 mostly refer to 0 °C. or to room temperatures, but
a few values at other temperatures are included for comparison. On the 12,6
and exp; 6 models, pDtJ/i should increase slightly with T when k T/elt > 2;
in Table 25 any such increase is masked by random errors. Values of Dn
calculated from^>Dn — 1-34/( in general agree with experiment as closely as
any calculated by more elaborate procedures, and have as much theoretical
justification.
The value of Dn quoted for hydrogen is derived from the mutual diffusion
of hydrogen and deuterium, and of para- and ortho-hydrogen. In both cases
there is reason to believe that the molecular fields of the gases involved are
slightly different. Similar but smaller differences can be expected for other
isotopic fields.
Values of Dn have also been inferred from the mutual diffusion of gases
whose molecules, though chemically different, may be expected to behave
similarly. For example, carbon monoxide and nitrogen have almost identical
molecular weights and similar molecular structure; also their transport
properties are very similar (cf. Chapter 12, Tables 11 and 17, and Chapter 13,
Table 20).* Taking Du = 0-192, ft = 1-645 x l0~* ( m e mean of the values
for the two gases), we find that pDn//t = 1 -46. In the same way, for the
similar gases carbon dioxide and nitrous oxide, pDn/fi = 1-39. However,
values derived from isotopic diffusion are to be preferred to these, because
different, though similar, molecules are likely to exert forces on each other
different from those which each exerts on another of its own kind. In fact,
experiment indicates that Dlt for CO,-N,0 is two per cent greater than D„
for CO,, while Dlt for N,-CO is less than Dn for carbon monoxide (18).

THERMAL DIFFUSION, A N D THE THERMO-DIFFUSION

EFFECT

14.6. As noted in 14.1, the velocity of diffusion has a component due to non-
uniformity of the temperature. This depends on the coefficients k r , a„—
the thermal diffusion ratio and the thermal diffusion factor, connected by
k r E xlxtalt. Thermal diffusion was first experimentally demonstrated by
Chapman and Dootson in 1916.1 It provides a sensitive test of intermolecular
force-laws, J and in the separation column, developed by Clusius and Dickelf
in 1938, it has supplied a powerful method of separating gases, particularly
in isotopic mixtures.
* L. E. Boardman and N. E. Wild (Proe. R. Soc. A, l6a, 511 (1937)) first calculated
pD,Jfi for the gas-pairs CO, N, and CO,, N.O. Attention was drawn to the similarity
of properties of these pairs by C. J. Smith, Proe. Phyt. Soc. 34, 162 (1922).
t S. Chapman and F. W. Dootson, Phil. Mag. 33, 248 (1917).
t Cf. K. E. Grew and T. L. Ibbs, Thermal Diffusion in Gates (Cambridge University
Press, 193»)-
i K. Clusius and G. Dickel, Z. Phyt. Chtm. B, 44. 397 (1939)-
14.6] COMPARISON OF THEORY WITH EXPERIMENT 269
The separation column relies on the combined effects of thermal diffusion
and a slow thermal convection. The theory of the column is complicated,*
depending on the viscosity of the gas-mixture as well as on a „ , and it involves
approximations which preclude the use of the column for exact determina-
tions of a „ . The column gives the sign of «12, and a rough estimate of its
magnitude. For more accurate determinations, a steady-state method has
been employed.
In a vessel containing a mixed gas, if different parts are maintained at
different temperatures, thermal diffusion results, and tends to make the
composition non-uniform. The concentration-gradients thus set up are
opposed by the ordinary process of diffusion tending to equalize the com-
position, and in time a steady state is reached, in which the opposing
influences of thermal and ordinary diffusion balance. Then, by (14.1, 1),
Vx, = - x , x , a , 2 V I n T; (14.6, 1)
the thermal diffusion factor ait is used instead of k T because of its much
smaller variation with composition. If the dependence of ait on composition
is ignored, integration of (14.6, 1) gives
In (x;/x2) - In (x,/x2) = - J r altdln T (14.6, 2)

= - a w l n (777), (14.6,3)
where xj, x2 and x„ x2 refer to the composition of the gas at temperatures 7",
T, and a, 2 is a mean value of <xu between the temperatures 7" and T.
Equation (14.6,3) can be used to determine 5 , t from the steady concentra-
tion-ratios xj/x2, x t /x 2 ultimately reached in two vessels in free communica-
tion with each other, and maintained at different temperatures 7", T. In a
series of experiments the temperature 7" of one vessel is often maintained
constant and the temperature T of the other is varied; by taking the dif-
ference of two equations (14.6, 2), the value of alt between two different
values of T can be found.
The degree of separation produced by thermal diffusion in an experiment
of this type is not large. For non-ionized gases, values of alt as large as 0-5
are rare; they are met only when a light gas has a small proportion of a much
heavier gas mixed with it.f The following table relates the equilibrium con-
centrations x„ \[ in the two vessels when the temperature T of the first is
half the temperature 7" of the second, and 5 W = 0-5.

x, = 0 1 0-3 0-5 0-7 0-9

x| = 0073 0-133 0-414 o-6a3 0-864

• L. Waldmann, Z. Phys. 114, 53 (1939); W. H. Furry, R . C . J o n e s and L. Onsager,

Phys. Rev. 35, 1083 (1939); R. C. Jones and W . H. Furry, Rev. Mod. Pkyt. 18, 151
(1946)-
t See K. E. Grew and T. L. Ibbs, op. eit.
270 DIFFUSION: [14.6
Because of the smallness of the separation, it was earlier thought necessary
to work between widely different temperatures T, T, with a corresponding
uncertainty regarding the temperature to which alt corresponds. This
difficulty was overcome in the swing separator,* consisting of a number of
separation tubes in series, in which gas in the cold section of one tube could
be interchanged with gas in the hot section of the next by drawing gas to and
fro through a connecting capillary. The swing separator operates effectively
as a number of separation tubes in series, and requires only a small tempera-
ture-difference T' — T for its operation. _ _
As noted in 8.41, a diffusion velocity Cl — Ct is accompanied by a heat-
flow px^a^Cx-Ct). This diffusion thermo-effect has been investigated
both theoretically and experimentally by Waldmann.f His experiments
were of two kinds. In the first (the' stationary' effect), mixtures of the same
two gases in different proportions were made to flow along two parallel
tubes, connected through a slot cut along their lengths; changes in the gas
temperatures on the two sides of the slot were measured by platinum wires
parallel to the slot, used as resistance thermometers. In the second (the
'non-stationary' effect), two vertical cylinders filled with mixtures in
different proportions were placed end to end, and diffusion was permitted
between them; temperature changes in the two cylinders were measured
throughout the duration of the diffusion, again by resistance thermometers.
The recorded temperature changes were small (comparable with 0-5 °C),
but sufficient to be recorded with fair accuracy. As Waldmann recognized,
the smallness of the temperature-range is in one respect an advantage, since
the results are practically independent of any variation of <xlt with the
temperature.
Although the thermo-effect involves diffusion, a „ can be determined
from it without prior knowledge of the diffusion coefficient Dlt, by using the
equation of conduction of heat, and the continuity equation for each of the
gases. The thermal conductivity A must, however, be known, and a heat
conduction equation has to be solved, in both the 'stationary* and 'non-
stationary ' cases. In both cases, not only can alt be determined, but values
of the diffusion coefficient can be obtained as a by-product, from the
ultimately exponential decrease of the temperature-deviations (with time in
the 'non-stationary' case, with distance along the tubes in the 'stationary'
case).
Values of a, t obtained from the diffusion thermo-effect are subject to
several possible sources of error. They depend on approximate solutions of
the heat-conduction equation, derived on the assumption that A is constant.
Again, A for many gas-mixtures is not known accurately; sometimes no

* K. Clusius «nd M. Huber, Z. Naturf. 10a, »30 (1955).

t L. Waldmann, Z. Phyt. lax, 501 (1943); 114. 2, 30, 175 (1947-8); Z. Naturfortchung,
I, 59, 483 (1946); 4a, 105 (1949):50, 3 " . 3*7. 399 («95o); L. Waldmann and E. W.
Becker, Z. Naturfortchung, 30, 180 (1948).
14.7] COMPARISON OF THEORY W I T H EXPERIMENT 271

experimental values are known, and A may have to be estimated by an

approximate formula from the values for the corresponding pure gases.
Nevertheless, Waldmann has been able to use the thermo-effect to obtain
a number of interesting results about an, particularly at low temperatures.

14.7. The thermal diffusion factor alt

The first approximation to alt, given in (9.83, 1), is
, x, s, x«s« . .
[
^^^xfQ^xSg^xix^,- ^4-7.0
Here st = MlP1 + 3Mi(Ml-M2)-^MiMt\, ('47> 2)
Q l = p,{6MJ + ( 5 - 4 B ) M\ + 8M, Mt A}, (14.7, 3)
Qn = 3 ( M i - ^ ) * ( 5 - 4 B ) + 4 ^ l ^ i A ( n - 4 B ) + 2PiP«. (14-7.4)
with similar relations for s, and Q2; also Pj = M,E/[/ti]j, P, = Af a E/[//j!, and
the quantities A, B, c, E are as denned in 9.8. The Kihara first approximation
[aj8]i is obtained by putting B = J in the expressions for Qlt Q2 and Qlt. In a
mixture of isotopes of nearly equal mass (cf. 10.61)
_ ISC(I+A) ml-m2 150^-m, .
[0Citii [CCitii KH1 5)
~ AIFI - 4 B + 8 A ) ^ 7 + ^ - *° "8T ia^rt- '
Further approximations have been derived by Mason ;• they are given
by very complicated algebraic expressions. Mason calculated successive
approximations for a Lorentzian gas (10.5), a quasi-Lorentzian gas (10.53)
and an isotopic mixture, in each case considering power-law repulsion and
the exp;6 model. On the power law model he found that [a| 2 ] t is a better
approximation than [«H]I; ' n t n e Lorentz case [ajj], ' s exact, whereas [<x1I]I
is less than the true value by an amount increasing from zero to about 15 per
cent as v increases from 5 to 13; in the quasi-Lorentzian case [aj2]i and [ajJi
are equal, each being in defect by an amount which increases from zero to
about 3¾ per cent as v increases from 5 to 13; in the isotopic case [a,,], and
[al2]t are in defect by amounts increasing from zero to 4$ and 1J per cent
respectively in the same range of v.
For the exp;6 model [oc'n]i has less advantage. In the Lorentzian case
[<xlt]t is a good approximation when kT/elt < 2, but is in defect by 8 per cent
whenA77e u = 5, and 13 per cent when k T/eit = 2o;[aJ2]1( on the other hand,
is always in excess, by about 15 per cent when kT/elt is between 2 and 3, but
by a much smaller amount when kTjeu is either large or small. In the quasi-
Lorentzian case [«J2]i and [al2]i are again equal, being in defect by an
amount which increases from zero to about 4 per cent as kT/elt increases
• E. A. Mason. J. Chtm. Phys. vj. 75, 782 (1057); see also S. C. Saxena and E. A. Mason,
J. Chem. Phys. 38, 623 (1957). Saxena and his associates have, in a number of papers,
considered the case when m,lm, is small, giving expansions for [a l s ], in powers of
(m,/m,)*; see, e.g., S. C. Saxena and S. M. Dave, Rev. Mod. Phyt. 33, 148 (196:).
272 DIFFUSION: [14.71
from 2 to 20. In the isotopic case [a,,], is in defect by an amount which
increases from zero to about 4 per cent in the same range of kT/elt; [a'^ is
in excess by 2 per cent when kT/elt is between 2 and 3, in defect by about
1J per cent when kT/elt « 20.
Thus for inverse-power repulsion, sufficient accuracy can usually be
obtained by using [aJJ„ but this is not true for the more important exp;6
model. The approximate nature of the formula (14.7, 1) must be borne in
mind in any detailed comparison with experiment.*

14.71. The sign and composition-dependence of [a lt ] t

The denominator of the fraction on the right of (14.7, 1) is essentially
positive; also C is usually positive for electrically neutral molecules. Assume
that c > o; then die sign of fa lt ], is mat of XJSJ —XjS,. As in (12.41, 2) we

so that sj is an equivalent viscosity diameter for molecules mlt while slt has
a similar meaning for unlike molecules. Then
Sl = 4Aft Mt A{(2Afs)-l tys\t -1} + 3M^Mt - Mt). (14.71, 2)
The quantities *„ slt are in general not widely different. Hence if m, and m,
are not too nearly equal, the sign of s, is that of m1-mi (if m1 > m„
Mj > J > A/,). The sign of s, is similarly that of « , - « , ; thus if mt > m„
x,s, — x,s, is positive, and a „ is positive. That is, (cf. (14.1, 1)), the heavier
molecules diffuse towards the cooler regions.
If mx and m, are nearly equal, the quantities s]/s]t, 4/*?i may be more
important than mi]ml in determining the sign of [a ]t ]|. If
S, - A(s\/S\t- I), S, = A(JS/*}, - I).

Thus s t > o, s, < o if *, > *„ > *,; in this case [a, J , > 0. That is, roughly,
if m, and m, are nearly equal, the larger molecules tend to diffuse into the
cooler regions.
In more detail,! suppose, as we may, that m, > m%, and assume
A = { = 0-444
(the argument is not sensitive to the value of A adopted). Then ifmjmf > 2-5,
s, cannot be negative, and s, can be positive only if ^ 1 1 n a s a n improbably
* If in (14.7. 3). (14.7, 4) we replace B by B \ where
B' = 2 B - j + c(i + tc),
value* of [«i,]i calculated from (14.7, 1) are found to agree cloaely with Maaon'a values
for the exp ;6 model for iaotopic and Lorentzian molecule*, and are identical with thoae
of [aiilt for the power law model. This tuggest* that error* may in general be reduced
by substituting B' for B.
t A detailed discussion of the characteristics of thermal diffusion was given by S. Chapman,
Prot. R. Sot. A, 177, 38 (1040).
14.71] COMPARISON OF THEORY WITH EXPERIMENT 273
high value, greater than 2-4. For values of m1/mt nearer to unity, the following
table gives the greatest value of st/sls for which s, can be negative, and the
least value of sjslt for which s, can be positive. The table illustrates the
importance of the mass-ratio as compared with sjslt, sjsu in determining
the signs of s1( Sj.

Af, ... 05 0-55 06 0-65 07

1 •J it 2i
Max. »,/»„ 1 0771 0-577 0-362 OI35
Min. *,/«„ 1 1 261 1 557 1-90 2-30

In special circumstances, when W|/m, is not too different from unity and
the heavier molecules are the smaller, s, and s, may have the same sign. For
example, the table shows that this can occur when mjmt = f if sjslt > 0-577
and sjslt > 1*557 (this gives slt < i(*i-M»), which is more probable on
physical grounds than the reverse inequality). In this case [a^h changes its
sign as x t increases from 0 to 1; however, sx, sa and consequently [<*it]i are
then small.
We assumed in the above that c > 0. On the 12,6 and exp;6 models (on
the latter, if a < 15), the quantity c may just become negative at low
temperatures, for values of kT/eJt between about $ and 1; at such tempera-
tures at least one of the gases in the mixture is well below its critical tempera-
ture. For inverse-square interaction—e.g. in a fully ionized gas—c is
strongly negative. In either case, thermal diffusion is reversed in direction;
that is, the heavier (or larger) particles diffuse towards the hotter regions.
In particular, in the sun and the stars thermal diffusion reinforces pressure
diffusion in tending to concentrate heavy ions in the hot central regions. This
effect is not important in a star as a whole, because of the enormous times
required for diffusion to produce appreciable changes of composition in so
large a volume; in the outer atmosphere, however, it may be significant.
When tnjmt is large, [ a u]i may increase threefold or more as the propor-
tion x, of the light molecules increases from zero to unity. In this case the
thermal diffusion is due almost entirely to the motions of the light mole-
cules ma. In terms of Mt, regarded as a small quantity, by (14.71,1)
W,=0(i), Mx = CXMj), E = 0(MJ*).
Hence the coefficients apq in 9.8 are such that
«11 - O ( 0 . «-1-1 = 0(M-ti), a„ = 0(Mt), a,, = 0(M\), * _ , = 0(i).
Thus if quantities of order M\ are neglected, the general expression for [k T ] t
given in 9.83 reduces to
P<r]i = I x t A ^ a o - i / a - i - i .
indicating the dependence on the molecules mt. In terms of the quantities
P, Q, this gives [ < X i J i _ 5 c / { X i ( Q i / P i ) + 2 X a Pf}> (14.7I> 3 )
274 DIFFUSION: [14.711
1
so that [ocjtll is a linear function of the proportions x„ x,. This linear
relationship was first found by Laranjeira* by a different approximate
argument. He found that experimental results for a number of gas-pairs
conform closely to such a relationship, sometimes even when Wj/m, is not
very different from unity.
When tnjntf is large, Q,/P, reduces to 5 - 4B, which normally is about 2-2;
2Pt is 2^J2(As\js\t), or 1 -2^s\js\t approximately, where *,, su are the viscosity
diameters. Since normally sjslt is decidedly less than unity when mjmt is
large, (14.71, 3) indicates a considerable increase in [a l t ] t as x, increases
from zero to unity.

14.711. Experiment and the sign of <xlt

Experimental values of 04, in general agree with the rules of sign derived
from the expression (14.7,1) for [<*n]v That is, the heavier molecules or, when
ml/mt ~ i, those of larger radius, tend to diffuse towards the cooler regions.
The change in sign of a „ at sufficiently low temperatures, due to the
change in sign of c, was first observed by Watson and Woernley (26) in
isotopic thermal diffusion in ammonia (WNH,-14NH,); here it occurred at
about 20 °C. Waldmann (27), using the diffusion thermo-effect, found a
similar change in a l t for N,-A, O t -A and N,-CO, at low temperatures. The
same effect has been found for other gas-pairs, e.g. H,-D,, H t -He at liquid
hydrogen temperatures; but Grew and Mundy (28) found a positive mini-
mum for Kr-A, without change of sign. This behaviour is what is to be
expected from the temperature-variation of c on the exp;6 model, when
a. > 15; however, Grew and Mundy find that, though giving a qualitatively
correct explanation, the exp;6 results require the minimum to occur at a
much lower temperature than that observed.
A change in sign of a l t as the composition is varied was first observed by
Grew (29) in Ne-NH, mixtures. The heavier but smaller neon molecules
travel to the cold side when their concentration exceeds 75 per cent; other-
wise ammonia seeks the cold side. Clusius and Huber (30) confirmed and
extended Grew's results by investigating the three mixtures "Ne-NH,,
M
Ne-NH, and "Ne-ND,. For the "Ne-NH, mixtures their results agreed
with those of Grew; for the "Ne-NH, mixtures a change in sign of <x„
occurred when the Ne concentration was about 58 per cent. In "Ne-ND,
mixtures, for which the molecular masses are very nearly equal, no change in
sign is found—the larger N D, molecules always diffuse towards the cold side.
Assuming that *„ *„ slt are identical in the two Ne-NH, mixtures, we can
determine their ratios from the concentrations at which a u vanishes, using
(14.71, 2). Taking A = 4/9, wefindthat sSe:sNHt:slt = 0-941:1-515:1.Thus
slt is intermediate between t, and st; if the same j-values apply to Ne-ND,
mixtures, this explains why a „ does not reverse its sign in the latter. Also
* M. F. Laranjeira, Phytica, a6, 409, 417 (i960).
14.72] COMPARISON OF THEORY WITH EXPERIMENT 275
=
*Ne/*NH, 0-62; the similar ratio derived from viscosity (Table 11, 12.11)
is 0-59. However, sls found in this way is considerably less than the value
given in Table 22.
A change of sign of alt was also found (31) in "A-D^CI mixtures, studied
between 120 and 179 °C. Here the DC1 molecules diffuse towards the cool
side when their concentration exceeds 47 per cent, and in the opposite
direction when their concentration is smaller.
Effects have been noted which do not appear explicable in terms of the
theory of spherically symmetric molecules. When working between tem-
peratures of 65° and 117 ° C , Becker and Beyrich (32) found that the values
of a, 2 for "C^CWO-WO^MO and " O ^ C O - W O ^ ' O mixtures have
opposite signs; working between 98° and 180 ° C , they found the same sign
but significantly different values. Since the molecules 1 * 0 1 , C M 0 and
" O ' ^ ' O have equal masses, their differences must be due to differences in
intermolecular forces or in the distributions of mass. Similar effects have
been observed in mixtures of isotopes of carbon monoxide (33) and hydrogen
(34). Sandler and Dahler* gave an explanation of the phenomenon, based
on the loaded sphere model, in terms of the eccentric distribution of mass in
certain of the isotopes.

14.72. a „ and the intermolecular force-law: isotopic thermal diffusion

In principle, important information about the intermolecular force-law is
immediately derivable from the value of a12. In (147, 1) [«iJi is expressed
as the product of 5c into a factor which varies only slowly with the molecular
model (through the quantities A, B, and the s\ls\t, J|/JJ S of (14.71,1, 2)).
A reasonably accurate value of this second factor can readily be inferred
from viscosity data. Hence from an experimental value of a I t we can deter-
mine c; from c the value of v can be found for inverse-power interaction, or
that of e l a for the 12,6 or exp;6 model. The advantages of using a l t are
obvious; whereas the determination of c from Dlt depends on the accurate
measurement of its temperature-variation, here c is found from the value of
a l t itself.
The drawback is in the large error of the first approximation [<xlx]i; the
use of [a„], for alt leads to spuriously high values of c. However, in one case,
that of isotopic thermal diffusion, Mason and Saxenaf have given tables from
which reasonably exact values of a „ can be calculated for the inverse-power,
12,6 and exp; 6 models. The quantity actually tabulated by them is a 0 , related
t0 a y
" «»»= <*o(«i - >« 2 )/K+«») (14-72.»)
(cf. (14.7,5)); this is expressed as a function of v or k T/elt. Using their tables,
the value of v or e „ can be found for any gas for which a „ is known.
• S. I. Sandler and J. S. Dahler. J. Chem. Phys. 47, 2621 (1967).
t E. A. Mason, J. Chem. Phys. rj, 782 (1957); S. C. Saxena and E. A. Mason, J. Chem.
Phyt. 28, 623 (1058).
276 DIFFUSION: [14.73
In practice, because of the large probable errors in determining <x„ (which
is small in isotopic mixtures), the process is inverted; a theoretical a<, is
calculated from force-constants derived from viscosity, and this is compared
with the «0 found experimentally. The results are set out in Table 26, which
gives a, at about 25 °C. for a number of gases, together with theoretical
values derived using the inverse-power, 12,6 and exp;6 models. The
theoretical values show the same general trends as the experimental, but
tend to be somewhat larger; the exp;6 values agree rather better with
experiment than those for the other two models.

Table 26. The isotopic thermal diffusion factor a0 at 25 "C.

a , (calc)

Gaa Reference a, (ob».) r-' ia,6 exp;6

Hydrogen* (35) 048 o-SS o-5S 0-51

Helium* (33. 36) 043 o-55 058 »•43
Methane (35) o-aa o»55 018 019
Ammonia (35) 0 -o-i — —
Neon (35. 36) 0-46 0-56 o-55 05a
Nitrogen (35) 037 04a 040 0-4»
Oxygen (35) o-3S 0-36 0-34 —
Argon (35. 36) oa4 o-aos 031 03a
Krypton (37) o-ia — 017 oa3
Xenon (37) O'll
— O'lO 0-07

* In the caae of hydrogen (H,-D,) and to a leu extent helium (*He-'He) the introduction
of a, may not be juatified, aince (mt — mJHm, + mj ia not ao amall aa to juatify the
neglect of ita aquare.

The temperature-variation of a, is mainly due to that of c. In general

terms the observed variation follows the variation of c given by the 12,6 and
exp;6 models; a, is small for low temperatures, and reversals of sign have
been observed in hydrogen (at liquid-hydrogen temperatures) and krypton
(in the range 195-273 °K.) as well as in ammonia. As T increases above such
temperatures, a,, increases, becoming roughly constant at high temperatures.
However, though the general behaviour accords with the 12,6 and exp;6
models, there are anomalies; it is not clear whether these are due to defects
in the models, to differences in the force-law for different isotopes, or to
experimental inaccuracies.
14.73. <x„ and the Intermolecular force-law; unlike molecules
For non-isotopic mixtures, only the rather inadequate first approximation
[a,,], has in general been studied. Mason and Smith* have suggested the
following way of deriving better approximations when mjmt is large. They
point out that in this case alt approximates to the quasi-Lorentzian value
• E. A. Maaon and F. J. Smith, J. Chtm. Pkyt. 44, 3100 (1966).
14.73] COMPARISON OF THEORY WITH EXPERIMENT 277
when Xj -> o; they give better approximate formulae for calculating alt when
x, -+ o, and assume that for a general composition
«iV = Xi(*u)x,-o + xJ*Tt)Xl-» (I4-73. ' )
1
However, though afi does vary nearly linearly with x, and x, (cf. 14.71), it is
not clear that the linear relation which gives the best fit to experiment is
precisely of the form (14.73, ')•
Here we compare theory with experiment* by calculating [ a ^ h from
(14.7, 1) for equimolar mixtures (x t = x2 = J) at o°C. The 12,6 model is
used; the value A = 0-44 is used throughout, and B and c are found from
values of kT/e12, using the combination rule elt = foes)*- The expressions
for the quantities p„ p 2 appearing in (14.7, 2-4) are equivalent to
Pi = 5"»»t[Oidi/L«i]i. p« = l«»»i[Ai]i/L«Ji; (14-73.2)
in the calculations, [Dlt]lt [fi^ and [//t]j in these relations are replaced by
experimental values of Z) lt , /it and //¾.
The results of the calculations are given in Table 27; for comparison
values given by the Kihara first approximation (B = }) are also given. Where
the calculated values are given in brackets, either et (for C,H 4 -H t ) or Dlt
(for the other gas-pairs) has had to be roughly estimated, in the absence of
experimental values. The experimental values of oclt are those quoted by
Grew and Ibbs (35). Where the data are sufficient, values at 0 °C. have been
estimated; otherwise a mean temperature T = 7(^1¾ is given, where 7^, Tt
are the temperatures between which the apparatus is working. An increase
in temperature in general increases <xlt, a fact which should be borne in mind
in a comparison with the calculated values.
Since normally [a^ is less than the true <x19, the experimental values
would be expected to exceed [ a ^ . Actually the reverse is often the case in
Table 27. Part of the discrepancy may arise because the early experiments
tended to underestimate a „ ; but there appears to be an unexplained residual,
due either to the values of C given by the 12,6 model being too high, or to the
theory of spherical molecules applying only approximately to thermal
diffusion for gases with internal energy.f
The temperature variation of <x18 equally suggests that this theory is only
approximately applicable. The chief part of the temperature variation of
[a,i], is that of c; on the 12,6 and exp; 6 models, as the temperature decreases
below els/k, c passes through a minimum, which in certain cases may be
positive. A positive minimum of <xlt is actually observed for CO,-A mixtures,
* In earlier editions the comparison was made by determining Rr, the ratio of a„ to the
fa,,], calculated on the assumption that the molecules are rigid elastic spheres, and
using the viscosity diameters. The intermolecular force-index v in r~" interaction was
then determined by comparing fir with its Lorentzian value. However, the latter value
usually overestimates the error in [a,t]i, and leads to valuea of v which are spuriously
large.
f The effect on a„ of inelastic collisions was discussed by L. Monchick, R. J. Munn and
E. A. Mason, J. Chtm. Phyt. 45, 3051 (1966), who found that the effect is small.
278 DIFFUSION: [14.73

Table 27. Comparison of theory and experiment

for ctltfor unlike gases

««(exp) t*lt]l
G«KI and T (deg K.) («itli (Kih«r«)

He-H, 0-145 0-15 0-16

CH^H, 0-25 031 034
Ne-H, 0-42 (0-32) (035)
CO-H, 0-30 o-33 o-37
NrH, 0-30 0-345 0-38
C t H 4 -Hf 028 (031) (o-35)
0.-H, 0-19(162°) 032 o-3S
A-H, 027 032 0-3S5
COr-H, 0-28 0-315 o-35
NrD, o-31 (327*) (031) (o-34)
Ne-He o-33 0-335 O-38
NrHe 0 3 6 (317°) 0-38 042
A-He 038 0-38 0-425
CO,-He 0-44 (J66*) 0-375 0-42
Ki-He 044 039 0-44
Xe-He o-43 0-40 0-45
A-Ne 017 0-17 0-18
Kr-Ne 0-28 0-245 0-265
Xe-Ne 0-29 0-275 0-30
Or-N, 0-018 (293°) 0-020 0-021
A-N, 0071 (293*) 0-056 0-058
COr-N, 0 0 3 0 (339") 0-064 0-066
N.O-N, 0 048 (339°) (0064) (0066)
A-O, 0 0 5 0 (283°) 0033 0034
C<VO, 0 0 4 0 (293°) 0049 0-051
COr-A 0-019 (283 0 ) (0019) (o-oao)
Kr-A 0-070 0-072 0-076
Xe-A 0-084 0093 0-098
Xe-Kr 0-016 (327*) (0024) (0025)

but it occurs at a temperature well above that expected on the theoretical

grounds (38); the temperature-variation of a „ is likewise anomalous for
mixtures of carbon dioxide with other inert gases (38,39).

REFERENCES
(1)J. Loschmidt, Wien. Ber. 61, 367 (1870); 6a, 468 (1870).
(2)A. Lonius, Annin Pkys. 29, 664 (1909).
(3)S. C. Saxena and E. A. Mason, Molec. Phyt. a, 379 (1959)-
(4)J. C. Giddings and S. L. Seager, lnd. Eng. Chem. Fund. 1, 277 (1062).
(5)R. E. Walker and A. A. Westenberg, J. Chem. Phys. 31, 519 (1959).
(6)R. J. }. van Heijningen, A. Feberwee, A. vanOostenandJ. J. M. Beenakker,
Phytica, 32,1649 (1966); R. J. J. van Heijningen, J. P. Harpe and J. J.M.
Beenakker, Physica, 38, 1 (1968).
(7) P. J. Bendt, Pkys. Rev. n o , 85 (1958).
 Copyrighted Material

14.73J COMPARISON OF THEORY WITH EXPERIMENT 279

(8) R. E. Bunde, Rep. Univ. of Wisconsin Naval Res. Lab., CM-850 (1955).
(9) P. E. Suetin, G. T. Shchegolev and R. E. Klestov, Sov. Phys. Tech. Phys.
4,964(1960).
(10) C. A. Boyd, N. Stein, V. Steingrimsson and VV. F. Rumpel, J. Chem. Phys.
»9. 548 (1951)-
(11) R. E. Walker and A. A. Westenberg, J. Chem. Phys. »9, 1139, 1147 (1959),
32, 436 (i960); R. E. Walker, L. Monchick, A. A. Westenberg and
S. Favin, Plan. Space Sci. 3, 221 (1961).
(12) R. Strehlow, J. Chem. Phys. 21, 2101 (1953).
(13) S. Weissmann, S. C. Saxena and E. A. Mason, Phys. Fluids, 4, 643 (1961);
E. A. Mason, S. Weissmann and R. P. Wendt, Phys. Fluids, 7,174(1964).
(14) B. N. and I. B. Srivastava, J. Chem. Phys. 36, 2616 (1962); I. B. Srivastava,
Ind. J. Phys. 36, 193 (1962).
(15) B. N. and K. P. Srivastava, J. Chem. Phys. 30,984 (1959); K. P. Srivastava,
Physica, 25, 571 (1959); K. P. Srivastava and A. K. Barua, Ind. J. Phys.
33. 229 (1959)-
(16) R. Paul and I. B. Srivastava, Ind. J. Phys. 35, 465, 523 (1961).
(17) J. N. Holsen and M. R, Strunk, Ind. Eng. Chem. Fund. 3, 143 (1964).
(18) I. Amdur, J. W. Irvine, E. A. Mason and J. Ross, J. Chem. Phys. 20, 436
(1952); I. Amdur and L. M. Shuler, J. Chem. Phys. 38, 188 (1963).
(19) E. B. Winn and E. P. Ney, Phys. Rev. 72, 77 (1947).
(20) E. B. Winn, Phys. Rev., 80, 1024 (1950).
(21) E. R. S. Winter, Trans. Far. Soc. 47, 342 (1951).
(22) H. Braune and F. Zehle, Z. Phys. Chem. B 49, 247 (1941).
(23) F. Hutchinson, J. Chem. Phys. 17, 1081 (1949)-
(24) B. N. Srivastava and R. Paul, Physica, 28, 646 (1962).
(25) I. Amdur and T. F. Schatzki, J. Chem. Phys. 27, 1049 (1957).
(26) W. W. Watson and D. Woernley, Phys. Rev. 63, 181 (1943).
(27) L. Waldmann, Z. Naturf. 4a, 105 (1949).
(28) K. E. Grew and J. N. Mundy, Phys. Fluids, 4, 1325 (1961).
(29) K. E. Grew, Phil. Mag. 35, 30 (1944).
(30) K. Clusius and M. Huber, Z. Naturf. 10a, 556 (1955).
(31) K. Clusius and P. Flubacher, Heh. Chim. Acta, 41, 2323 (1958).
(32) E. W. Becker and W. Beyrich, J. Phys. Chem. 56, 911 (1952).
(33) A. E. de Vries, A. Haring and W. Slots, Physica, 22, 247 (1956).
(34) J. Schirdewahn, A. Klemm and L. Waldmann, Z. Naturf. 16a, 133 (1961).
(35) References quoted by K. E. Grew and T. L. Ibbs, Thermal Diffusion in
Gases, table VA, p. 128 (Cambridge University Press, 1952).
(36) S. C. Saxena, J. G. Kelley and W. W. Watson, Phys. Fluids, 4,1216 (1961).
(37) T. I. Moran and W. W. Watson, Phys. Rev. 109, 1184 (1958).
(38) J. R. Cozens and K. E. Grew, Phys. Fluids, 7, 1395 (1964).
(39) S. Weissmann, S. C. Saxena and E. A. Mason, Phys. Fluids, 4, 643 (1961).

Copyrighted Material
 15
T H E T H I R D A P P R O X I M A T I O N TO T H E
VELOCITY-DISTRIBUTION FUNCTION

15.1. Successive approximations t o /

In Chapter 7 it was shown how the velocity-distribution function/for a
simple gas can be determined by successive approximation to any desired
degree of accuracy; the first, second, third, ... approximations are f",
/«»+/<«,/«»+/<»+/»> In 7.15 it was pointed out that/ro,/™,/*),... are
respectively proportional to nl, rfi, n~l,..., where «is the number-density of
the molecules. Hence the later terms in the approximations become relatively
more important as the density decreases.
Thefirstapproximation,/"", is identical with Maxwell's function

'(iSr)1'-'
mC/UT
/ •

"(:
For /"> the following expressions can be derived from the results of 7.3
and 7.31:
yw =/(oxi><i) _ _Iy»)j/^\*A.VlnT+2&:Vc 0 )

(i5->» *)
and so we can write
/ » » A'(C)C.VT+ B'(C)CC:Vc» (15-1.3)
where A', B' denote functions of C and T, defined by this equation. In virtue
of (1.31,7,9) we can in (15.1, 2) and (15.1, 3) replace the velocity-gradient
tensor Vc0 by the rate-of-strain tensor e a Vc0 or by its non-divergent part 8,
without altering the relations.
We can readily estimate approximately the ratio /w/fm o r ¢0), i n the
notation of 7.51 and 7.52,firstapproximations to A and B are

B =fcjbro- bxS^(<P)rifo = 6,*&\

b
where a, = ajan = -Ww i " A/*u - 5/**ii-
(280]
15.1] THE VELOCITY-DISTRIBUTION FUNCTION 281

Now (cf. 9.7) au = 4,1, and the first approximation to the coefficient of
viscosity is

Hence to a first approximation

and A-lfenn-toV, B = ^«-V, (15.1,4)

whence it follows that (in agreement with (10.33, &))

*™ = "^ {~f {^-\)^.VT-2^:Vcv

We may write /1 = 0-499/90/, where / is an equivalent free path and
C = (8kT/nm)l (cf. (12.1, 6), (4.11, 2)). Let V be a scale mass-velocity, and
L„ L2 be scale lengths, such that the gradients of T and c 0 are comparable in
magnitude with TjLr and F / i j . Then if V is taken to be of unit order of
magnitude, the two terms in 4>(I) are comparable respectively with //L, and
IV\L% C. Now the free path / is comparable with io~*cm. in a gas at S.T.P.,
whereas experimental values of L,, Lt are normally comparable with the
dimensions of the apparatus—of order 1 cm., say; also V/C is normally small
or at most comparable with unity. Hence normally W° is small, and/ 0 * is
small compared with /101. However, at densities io~* times the normal
density, or in shock waves (where L, and Lt are very small), /> u becomes
comparable with/"', and it is necessary to examine whether/™ is not of the
same order. In so doing we confine attention for the present to a simple gas.
Results for diffusion in a binary mixture are described in 15.42.
Burnett* determined the complete third approximation to / for a
simple gas. Incomplete third approximations had already been given by
Maxwellf and by (Lennard-)Jones,J who included terms involving second
derivatives of T and c0, but ignored squares and products of the first deri-
vatives (thus avoiding some of the chief mathematical difficulties of the
problem); Enskog§ gave a solution which included products of the tempera-
ture gradients (but not those of c0).
Burnett's solution will not be given here in full; the equation to determine
/"' will be considered so far as to infer from it the nature of the various terms
in/ 1 ", but their exact expressions will not be derived. It will be shown that
the new terms introduced by/ f f i into the expressions for the thermal flux
q and the pressure tensor p can be determined without solving the equation
for/1*', and the magnitudes of the terms will be roughly estimated.
* D. Burnett, Proe. Land. Math. Soe. 40, 382 (1935).
t J. C.Maxwell. Mil. Trans. R. Soe. 170, »31 (1879); Collected Papers, vol. a, 681 (1890).
t J. E. (Lennard-)Jones, Phil. Trans. R. Soe. A, 223, 1 (1923).
f D. Enakog, Inaug. Dial. Uppaala, chapter 6 (1917).
282 THE THIRD APPROXIMATION TO [15.2

15.2. The Integral equation for/ 1 *

The equation to determine/>*' is (cf. (7.12, 1), (7.14, 19))

In this write / » =/"»<I>»; (15.2,1)

then it becomes, by (7.12, 3),
_„./(<!,«) =. ^ + ^ + c . ^ . > + F . ^ + y ( / < i y } i ) ) . (is.2i a)

The various terms on the right of (15.2, 2) are developed as follows. Using
(7.14,9,11, 13) we have*

^-/•^-'"SKi-'-a
.-^((Uf-.)(p»:^ + ^.«») + C^^). (.5*3)
If/1¾ is regarded as a function of C, r, r, the next three terms on the right
of (15.2, 2) are replaced by

Dt

"-Dr+c~o>+p^"ac~8cCoYc,»- (isa,4)

Now, using the expression (15.1, 3) for/lM,

dfm /0A' 81? 0 \ 0
?L- - ^ C . V r + ^ C C : V c 0 + ^'C.Vj Vr+B'V(CC:Vc 0 ),
(»5-2, 5)

|J! = 2 (|^C.Vr+|gcC:Vc0) C+^'Vr+2B'C.^,

(15.2, 6)
and, using (7.14,16),

+ ^ ' C . § ( V r ) + B'cfc:§(Vc 0 ). (15.2,7)

• In the derivation of equations (15.1. 3-9) uae it made of the tensor relations (i.a, 7, 8),
(1.3a, 6, 10). (1.33, 7, 8).
 Copyrighted Material

15.2] THE VELOCITY-DISTRIBUTION FUNCTION 283

The time-derivatives on the right-hand side of (15.2, 7) can be determined in
terms of space-derivatives; for, by (7.14, 15, 16),

g(vr)=(| + c 0 .v)vr
=v
(¥- + c °- v : r )" ( V c o ) - v r
= -tV(rV.c0)-(Vc0).V7\ (i S .a,8)

g(Vc0)=(|+c0.v)vc0

= v(^+(c0.V)c0)-(Vc0).(Vc0)

= V(F~V^-(VC0).(VC0). ( I 5 .2, 9 )
It is, however, more convenient for the present to retain the time-derivatives
in (15.2, 7).
Again (cf.(7.n, 2) and (15.1, 3), replacing Vc0 by 8)

•OTP) = JiA'C.VTA'&.VT^ + JiA'C.VTB&CsZ)

+J{B'CC:%A'1C1.VT)+J(B'CC:%B'1C1Cl:%). (15.2, 10)
The first and fourth of the expressions on the right of this equation, as
may readily be seen, are even functions of C; the second and third are odd
functions.
Combining the above results, we obtain an expression for — n2I(Q(Z>) as
the sum of numerous terms which can be divided into groups as follows.
First there is a part involving only the scalar C; this is

-'f°\^-i)(pa):Vc0+V.^) + -A'Vp.VT. (15.2, 11)

Next there is a part of odd degree in C, which after suitable rearrangement
takes the form

p
+ A'C. (g(VT)-(Vc0).Vr) +? g(CC:8)(C.Vp)

+ -B'CVp:% + B'C.V(CC:%)

+ (%-^)(CC:%HC.VT) + J{A'C.VTB'1C1C1:Z)

+ J(B'CC:%A'1C1.VT), (15-2,12)

Copyrighted Material
284 THE THIRD APPROXIMATION TO [15.3

where A = V.c 0 = * * + * • + £ , (15.2,13)

that is, A is the divergence of the mass-velocity of the gas.
Finally, there is a part of even degree in C, which may be expressed as
follows:

-K r ^S) ( C C : 8 ) + B ™# 8 ) - 2 8 V c ')
1. 8A'
+ A'CC:V(VT) + - —CC.VpVT
HA'
+ j^CC:VTVT+J(A'C.VTA'1Cl.VT)

-2^ i (CC:8)(CC:8) + y(B'CC:8B;C1C1:8). (15.2,14)

15.3. The third approximation to the thermal flux and the stress
tensor
The equation (15.2, 2) for/ w or QW was solved by Burnett by the methods
applied to obtain/1" from (7.3,7). His solution for W is in the form of a
"rie» ¢(8 „ 2 „„,$}?>(«*) WC)» L «,,%* (15.3. 0
P. J p. J
where ^"(C) is a spherical harmonic function of degree I in the components
of C; that is, a homogeneous polynomial in U, V, W, satisfying the equation
d 0^0> 0ty<o {fifth BhjS» _

In Burnett's case, only the values / < 3 were required. The unknown
coefficients a^ can be determined by methods similar to those of 7.51,7.52.
It is, however, possible to ascertain the new terms in the thermal flux q
and in the stress tensor p without obtainingfl*1';for
q*» - [f»\mC*Cdc

m kT ff)* f/OHDWiCSf* - \)<€dc

-p{£Pffr»l{A)dc

(f)'P\AH**)ic, (15.3.3)
15.3] THE VELOCITY-DISTRIBUTION FUNCTION 285
using (7.12, 4), (7.31, 2) and (4.4, 7, 8). Similarly
p» = fFa*tCCdc
= jfWmCCdc

= 7kT\fW»'ivdc
= 2/>/<D<»/(B)</c

= 2/>[<D<», B]

= 2/>/B/(<D<»)<fc, (15-3.4)
t n>
using the same relations, and (7.31,3). Since — n /(<I> ) is given as the sum of
(15.2, 11, 12, 14), expressions for g™ and p** can be derived.
Since 0
A-VAP), b = WeB{V) (15.3,5)
it follows from 1.42 that the groups of terms (15.2, 11, 14) will not contribute
to¢^, nor the terms (15.2,11, 12) top**. Again, the contribution tog™ from
the first term of (15.2, 12) (the only one of these terms that does not depend
on A' or B') is

which vanishes, by (7.31, 6). All the other terms of (15.2, 12) make contri-
butions to 5**'; thus we can write

**> = e^L vr+0,^g(vr)-(Vco).vr

+09^vP.%+et^v.(%)+et^VT.i, (15.3,6)
the vectors in this expression being the only ones which can be formed from
the elements involved in the terms of (15.2, 12); the coefficients of the vectors
are so chosen that dlt 8t, 63,6t and 0t are pure numbers. In the second term
we can substitute for the time-derivative from (15.2, 8), when the expression
in the bracket become, _ | V ( r A ) _ 2 ( V c o ) . v r . (,5.3, ? )

The coefficient of the last term in (15.3, 6), namely 0t, is the only one which
depends on the integrals J.
Since e and A both depend on the space-derivatives of e0, every term in
(15.3, 6) depends on such derivatives. If the gas is at rest, or in motion
uniform in space, ¢ ^ = o, and ¢ = - AVT not only to the second but also
to the third approximation. The third approximation, moreover, introduces
no terms depending on the second or higher space-derivatives of T, nor on
 Copyrighted Material

286 THE THIRD APPROXIMATION TO [15.4

squares or products of derivatives of T; it does, however, supply a thermal
flux even when T is uniform, if either p or A or I is not uniform.
In the same way, we can show that

p«) = roi^A8+m2^(g(8)-2V^l)+^vlf

0 (I5 3,8)
+ ^ ^ + % ^ ^ ^ '

where wv wv w3, mv xui and me are pure numbers; the tensors in this
expression are the only symmetrical and non-divergent tensors that can be
formed from the elements involved in the terms of (15.2,14), and (cf.
(15.3, 4, 5)) p w is both symmetrical and non-divergent. The coefficients wb
and«r(, and no others, depend on the integrals 7in (15.2, 14). The expression
in the bracket of the second term is, by (15.2,9), equal to

v(p-±Vp) -(Vc 0 ).(Vc 0 )-2Vc 0 .8. (15.3,9)

Thus the third approximation to the stress-system introduces terms

depending on (a) products of first-order derivatives of the mean motion c0;
1 Dc
(b) first-order derivatives of F—Vp or -yr^', (c) products of first-order
derivatives of the temperature with each other, and with the first-order
derivatives of the pressure; and (d) second-order derivatives of the tempera-
ture. The terms (a) depend not only on the elements of the rate-of-strain
tensor e, but also on the anti-symmetrical part of the velocity-gradient
tensor Vc0, and therefore they have a part depending on the vorticity of the
motion (£ curl c0). Similarly the contribution to if® from the second term of
(15.3, 6) includes a part proportional to VT A curlc0.

15.4. The terms in $*

Approximate values of the coefficients in the various parts of 5*» and p®
can be found without great difficulty.
By (15.1, 2-4), to a first approximation,

A = A(V)V = ^ ( « * - f ) # , B = B(f)W = ^ C T , (15.4,1)

m \i m

(
BXQ . - ( znkf)
m \i fim
1
- 5 - )(ATp*
^^. (15.4.3)

Copyrighted Material
15.4J THE VELOCITY-DISTRIBUTION FUNCTION 287
Hence, remembering that V* = m&lzkT,

1
BT ~ \znkT) \(kT)*( i, +
k*TdTI'
8£ BIT _ / m \ l ^
u
ac» ^ " \airAiv (*r)» •
Using these expressions, we can obtain approximations to the various
coefficients in (15.3, 6, 8).
As noted above, the first term of (15.2, 12) makes no contribution to g™.
By (15.3, 3), the second term contributes the expression

by (1.42, 4). Inserting the above values for A and A', we get

so that in (15.3,6) 0, = II g ^ ! ^ . (l5.4,4)

The coefficients 6t, d% and 0t are similarly determined, using (142, 4) and
1.421: we find a .. a a , »
* 0* = ¥> #3 = - 3 . 04 = 3- (154.5)
The coefficient 6t depends on the integrals J in (15.2, 12). Wefindthat the
contribution of the seventh term of (15.2, 12) to «7"' is, to the present
approximation, ^ / 3 5 . 1 ^ - , - .
PTU pdrr
The contribution of the terms involving the integrals J is

~ T fir)* \A{J(A'C V7,BiC1C1:8) + 7(B'CC: lA'^.VT^dc

=-T ( 2 -^7If^' (c)cvrB ' (c,)CiCi:8

+ ^'(C,)C 1 ^7'B'(C)CC:8-^'(C")C'.V7 , J5'(QC;Ci:8
- A\C[) Ci.VT B'(C) C'C: %}g*ldt'dcdcl.
288 THE THIRD APPROXIMATION TO [15.41
using the definition of 7, (7.11,2). Transforming by methods similar to those
of 3.54, this becomes

"T( ? x^)(Ai JJJ j4 (C)CV7,B (Cl)ClCi:8

' -A'^ga^t'dcdcy
+ A -A'
'
l

Inserting the above approximate expressions for A, A'(C), and B'(C), and
using the relation V+V1 — V + V[, expressing the conservation of momen-
tum, we get

x ( « * * + « ! * , -<€•"€' -VfV'Jgc^de'dVdV!.
Let 0O, f be the variables of (9.2, 6), so that for a simple gas
tf = 2-*(«,-y), V, - 2-»(»,+?)
and similarly for «", Vy Then
J(«'H?+«!«',-«"nr -«";n?;)altrfe'
= 2*ff{(V0.9),-(V0y)9'}bdbde

on integrating with respect to e as in 10.5.

Using this result and taking V0 and f as new variables of integration we
can (with some labour) complete the integration; the result is

In particular, for Maxwellian molecules (for which (15.4,1-3) are exact)

the expression vanishes; thus it may be neglected to a first approximation,
and we get - -, , .

<.-(H£)- (•«•«
15.41. The terms In p<*
The magnitudes of the various coefficients in the expression (15.3, 8) for p®
can be found similarly. The four coefficients which do not involve the
integrals 7 are, to a first approximation, given by
4/7 Tdp\ . .
Wl w
~l\2~p3t)' *"2' ro
»-3. w«-o. (15-41.1)
The contributions to p * from the fifth and seventh terms of (15.2,14) are
found to be 0
15.41| THE VELOCITY-DISTRIBUTION FUNCTION 289
respectively. In determining the second of these contributions it is necessary
to use the integral theorem
o
J e - t f ' t f V ^ : 8 ) ( W : Z ) d V = ^ 0 / ^ «"•<*«• (15.41, 2)
(cf. (1.421, 1)). The contributions to p^' from the sixth and eighth terms of
(15.2, 14), which involve integrals J, are found in a way similar to that used
in determining the contribution to tf9 from the integrals J; in integrating,
we use

in the second integral we must also use (15.41, 2). The contributions are
thus ultimately found to be

-%&n WW - 9nP(3)+MM vrvr

lopkT31
0

and _i|^ { ^»( 3 )_|^ (2 ) } i.I

respectively. Since both of these expressions vanish for Maxwellian mole-
cules, they are neglected to a first approximation, and so

Wi= W =
pdf' ' 05-41.3)
The values of the coefficients mr, 6, given in this and the preceding section
are exact for Maxwellian molecules, and it may be expected that they will
be not far from the true values for other molecular models. The coefficients
Wj, m3 were calculated more exactly by (Lennard-)Jones for rigid spherical
molecules, using third approximations to A, B (and A'(C), B'(C)) in place
of the first approximations used above; his values are respectively 1-013
and 0-800 times the values given by (15.41, 1). Burnett gave a general
formula for the terms in p (,) for the case of molecules repelling as r~* (in-
cluding, as a limiting case, rigid elastic spheres), and determined the
coefficients for Maxwellian molecules and for rigid elastic spheres, using
fourth approximations to A and B. For Maxwellian molecules his results
agree with (15.41, 1) and (15.41, 3); for rigid elastic spheres his results are
equivalent to
4 / 7 Tdp\ Q , )
Wi= 1 0 1 4 x 2 | i - -ff\, ro, = 1-014x2, ros = 0-806x3,

m, = 0-681, w5 = 0806 x 3— - - 0 - 9 9 0 , mt = 0-928x8,

(i5-4».4)
290 THE THIRD APPROXIMATION TO [15.42
expressions which indicate the factors by which the various first approxi-
mations are multiplied.
For molecules repelling like r- the expressions for g*" and p™ are slightly
simpler than the general expressions (15.3, 6, 8). In this case, because the
integrals Qfp(s) for difFerent values of r and s all involve the same power of
T(cf. 10.31), we can prove that A'(C) and B'(C) can each be expressed as the
product of /tjTi and a function of V; thus

1
dT + C DC* [2 ndT)A'
8R SB h TdA
8
T-dT+^da'-b-fdr) '
and the second and third of the terms of (15.2,12) involve identical functions
of C, and so also do the first and second of the terms of (15.2, 14). It follows

This explains the appearance of the same factor 1-014 in the expressions
(15.41, 4) for Wj and mt\ the reason for the appearance of 0806 in the
expressions for mt andros is similar.

15.42. The velocity of diffusion

So far only simple gases have been discussed. The third approximation to
the velocity distribution for a mixed gas has been considered,* in order to
determine the corresponding contribution to the velocity of diffusion. Here
we shall simply quote the result, that the third approximation adds to the
velocity of diffusion C, - C, nine terms, given by

where ea,eb,...,et are numerical quantities of order unity, which may,

however, vary with the temperature and the properties of the mixture.
Certain of the quantities e have been determined correct to a first approxi-
* S. Chipmin and T. G. Cowling, Proe. R. Soc. A, 179, 159 (1941).
 THE
15.5J VELOCITY-DISTRIBUTION FUNCTION 291
mation similar to that of 15.41: their values, correct to this approximation,
are
e = 2 ( 7 _ Z ?RT\ e _ 2 h_I_ ^i?\
e
" JU DT df)> * _ i U Dlt dT}'
All the new terms in the diffusion velocity can be shown to depend on the
velocity-gradient, and to vanish if the mass-velocity c 0 is uniform. They are
in general small compared with the diffusion velocities considered earlier,
and are liable to be masked by turbulence when they are large. However, the
last term in (15.42, 1) may be comparable with pressure diffusion in flow
along a capillary tube.

15.5. The orders of magnitude of g"1 and p™

Consider now the magnitudes of the different terms in the expressions for
qP\ p(2) for a simple gas. Because (Tl/i) {dftjdT) is a pure number whose value
for ordinary gases lies between J and 1 (cf. 12.31), all the coefficients 0r and
mr are less than *£.
Let V, Lv Lt, I have the same meanings as in 15.1, so that the first- and
second-order space gradients of T, c0 are respectively comparable with TjLv
V/Lt and TjL\, VjL\. In motions involving adiabatic expansions, Vp is
comparable with p/L2; in nearly steady motion it is normally smaller, being
comparable with pV2/Lt or />F 2 /L 2 C*. Thus the first and fifth terms of
(15.3, 6) are comparable with fi*VlpLxLt\ the fourth term is comparable
with /iW/pLl; and the remaining terms are comparable with or smaller than
these quantities. On the other hand, the thermal flux g™ given by the second
approximation t o / i s
-AVT=- Iftc.VT = - ¥ * ( * / * ) V7\
which is comparable with fikT/mL^ or ftp/pLv Since /1 = o-^ggplC, and/> is
comparable with pC2, the terms of g*2' bear ratios to g*1' comparable with
Vl/CL2 or VlLjCL\. This means (cf. 15.1) that in ordinary conditions, in
which //L, is small, the thermal flux g'2' is negligible compared with g 0 '. The
flux g*2) is in any case the less important because it depends essentially on
the motion of the gas, and vanishes when the gas is at rest. In normal gas-
flows, the transport of heat by convection is more important than that by
conduction.
Similarly in the expression (15.3, 8) for pm the first, second and last
terms are comparable with p*V*lpl\. Since p a ) is comparable with fiVjLt,
the ratio of these terms to p(1> is similar to that of p(1) to p; in particular, at
normal pressures these terms are small compared with pft), and may be
neglected. The thermal stresses given by the third and fifth terms of (15.3, 8),
and also, in special circumstances, the stresses given by the fourth term, may
be more important. They are comparable with fi*/pL\, and bear to p°> ratios
292 THE THIRD APPROXIMATION TO [15.51

comparable with fiLJpVL\, i.e. with ClLJVL\. Thus when C/Vis large and
L, and Lt are comparable, the temperature stresses may not be completely
negligible compared with the ordinary viscous stresses, even at ordinary
pressures at which //£,, is very small.

15.51. The range of validity of the third approximation

The regime in which fm +/111 gives a fully adequate approximation t o / i s
known as the Navier-Stokes regime. In this regime the transport properties
of a simple gas are adequately described in terms of the viscous stress p(1>
and the heat flux g">.
The cases in which g™, p (S are also likely to be important are those in
which /is large (low pressures, of order io~*atmos.) or Lu Lt are very small
(as in shock-wave fronts) or intermediate cases (capillaryflowor ultrasonic
waves at fairly low pressures). However, whenever g**' and p m are important
there is every reason to believe that higher-order approximations also need
to be taken into account. The expansion/ = fm+fl" +/•*+... is, in effect,
an expansion in powers of //L,, l/Lt;/m is linear in them, /<J) quadratic, and
so on. Normally //£,, ljLt are small, and only one or two terms of the expan-
sion need be considered. But if g<2> and pm become important, //L, and l/Lt
are comparable with unity; the successive terms of the expansion decrease
slowly, and the expansion may cease to converge. That is, the higher
approximations can only be used when the Navier-Stokes approximation is
already good, in order to give an ultra-refined fluid description.*
For example, Wang-Changf attempted to apply the results of the third
(Burnett) approximation to the propagation and attenuation of ultrasonic
waves, and even gave equations corresponding to the fourth (super-Burnett)
approximation. However, a comparison of these results with experiment
shows that, whereas the Burnett and super-Burnett approximations may
give somewhat improved agreement when ljLt is still fairly small (where L2
here denotes the wave-length), they give no agreement when IILt is com-
parable with or larger than unity. J This is to be expected; the first approxi-
mations to 8n/dt and dT/8t correspond to adiabatic expansion, and when
ljLt is large the motion is nearly isothermal.
When //Lj and l/Lt are not small, a new method of approximating t o /
becomes necessary. Generalizing the methods given earlier, a form is
assumed for/involving unknown parameters, equations to determine these
• Cf. H. Grad, Phys. Fluid!, 6,147 (1963). Grod also suggests that generally the expansion
/<»>+/u> +/i«> + ... m a y be asymptotic rather than convergent; the difference is unim-
portant in normal conditions, when it would become apparent only at very high orders
of approximation, but important when //1.,, IIL, cease to be small.
t C. S. Wang-Chang, Univ. of Michigan, Dept. of Engineering Rep. UMH-3-F (APL/
JHU CH-467), 1948. The first to use the Burnett equations in hydrodynamics was
L. H. Thomas (J. Chem. Phyt. 12, 449 (1944)). in studying the profile of shock fronts.
t M. Greenspan, J. Acoutt. Soc. Am. aa, 569 (1950), and a8, 644 (1956); E. Meyer and
G. Sessler, Z. Phyt. 149, 15 (1957).
15.6] THE VELOCITY-DISTRIBUTION FUNCTION 293

being then found by using an appropriate number of equations of transfer

(cf. 3.11). The success of this method depends on the satisfactory choice of
the form for/. Alternatively, numerical methods of solving the Boltzmann
equation have been tried; these are usually applied to an approximate form
of the equation. The problem has especially been studied in connection with
shock-wave fronts.

15.6. The method of Grad

The method developed by Grad* is the closest to the Enskog approach, from
which it differed mainly in its discussion of the time-derivatives. Because
the rapidity of changes in a shock front may prevent a 'normal' solution
(cf. 7.2) of the Boltzmann equation from being attained, Grad sought
solutions of a more general type. He assumed an approximate expression
for/, of the form , .,„,, , _, , , , .
1
/=^(-+1¾¾}. ('5-6,1)
p. 1
are a
where the functions v p , ( finite number of) the functions of (15.3, 1),
and the quantities ap, are variable parameters. Arbitrary initial values of the
parameters a;j, can be assumed; the time variation of the parameters, as well
as that of n, c 0 and T, is then determined from the equations of change of
appropriate molecular properties (cf. (3.11, 1)), so that the values of the
parameters at any subsequent time can be found. By taking a sufficiently
large number of functions and parameters Op,,/can be determined to any
desired order of approximation.
A simple approximation, but one giving good results, is

/ = / « » { i + a . C ( ^ - | ) + b : CC}. (.5.6,2)
Here the vector a and the tensor b are the unknown parameters; they are
related to the heat flow q and the non-hydrostatic part p' = p — U/> of the
pressure tensor by the equations
b
-!?• = 2>' <'5.6.3>
The time-derivatives of a and b are determined from the equations of change
0
of the molecular properties (Sf2 — $)C and CC. This leads to equations
equivalent, in our notation, to
n ' 0 o_ 0
^._ 2 V C o .p' + 4g.p< + jAp' + |V 9 + 2/>! + Jp' = o, (15.6,4)
5f-Vc0.*+y-?.*+5A9+^v.p'+^V7\P'

- t . r o + V.pO + i ^ V r + U , - * (.5.6,5)
• H. Grad, Communs pure appl. Maths, a, 331 (1949); 5. 357 (195*).
294 THE THIRD APPROXIMATION TO (15.6
where ft is the first approximation to the coefficient of viscosity. These,
together with the equations of continuity, motion and energy, form a com-
plete set of differential equations giving the variation of all the physical
variables. The approximation (15.6, 2) is known as the 13-moment approxi-
mation; in order to get the correct values of n, T and the 11 independent
components of c„ q and p', we need the values of 13 moment-integrals
jfXpidc
When the parameters l/Lu ljLt of 15.5 are small, (15.6, 4) and (15-6, 5)
yield results equivalent to those of 15.3. As can be seen by comparing the
first and last terms of each equation, p' and q may at first vary rapidly,
approaching normal values (depending only on n, c0 and T and their deri-
vatives) in a relaxation time comparable with ji/p, i.e. with the mean collision-
interval. After the 'normal* state is attained, the last two terms of each
equation become dominant; thus a first approximation to the solution is
p' -. p<w, £ - gWt where

p™ = -2/«l, <,«> = -1£* ftVT'-AVT,

4?ft

A being the first approximation to the thermal conductivity. These are

formally equivalent to the first approximations to p' and q derived in
Chapter 7; second approximations are then derived by substituting p' = pa),
q - q™ in the earlier terms of (15.6, 4) and (15.6, 5), and p' = p<1)+p<*,
q « qOi+qVi m the last term of each, and replacing D/Dt by DJDt. The
expressions so found for p<* and q™ are identical with those of 15.3, with the
approximate values for the coefficients 6 and w given in 15.4 and 15.41.
Grad sought to apply (15.6, 4) and (15.6, 5) to the case when l/Llt l/Lt are
not small, and DJDt may not be a good approximation to D/Dt. He was able
to discuss moderately strong shocks, and his theory marked a definite
advance on that of Wang-Chang, based on the Burnett approximation.*
However, his equations, based on the 13-moment approximation, possessed
no solution when the Mach number of the inflowing gas exceeded 1 -65. An
approximation with further terms can extend the range of solubility, but,
as Holway showed,! no solution of Boltzmann's equation of Grad's type can
converge if applied to a shock-wave with Mach number greater than 1-851.
This is not a wholly decisive objection to Grad's method; it is not required
of an approximate solution that it should in the limit converge, but simply
that it should approximate t o / sufficiently well to yield meaningful results.
Unfortunately, however, with increasing shock strength a polynomial ex-
pansion like (15.6,1) or (15.6, 2) becomes progressively less able to give an
adequate approximation to/.
• C. S. Wang-Chang, Univ. of Michigan, Dept. of Engineering Rep. UMH-3-F (APL/
JHU CM-503), 1948.
t L . E Holway, Phys. Fluids. 7, 911 (1964).
 TH
15.62) E VELOCITY-DISTRIBUTION FUNCTION 295

15.61. The Mott-Smith approach

Because of the extreme rapidity of the transition across a shock front, Mott-
Smith* assumed that at any point within the transition zone/is the sum of
two Maxwellian functions/,,/ 8 corresponding to the temperatures and mean
velocities on the two sides of the shock. The number-densities «,, n3 oifvft
were assumed to vary across the transition zone, nx vanishing on one side and
n2 on the other. The variation of «,/n2 across the zone was determined by
using the equation of transfer of u% or of u3, where nis the velocity-component
normal to the shock front; the equations of transfer of the summational
invariants are in any case also used. Mott-Smith obtained shock-wave
thicknesses agreeing moderately well with experiment for strong shocks.
For weak shocks, in which the transition across the shock front is relatively
slow, his results agree less well with experiment than do those obtained from
the Navier-Stokes equations.
An extension of the Mott-Smith method was made by Salwen, Grosch and
Ziering.f who expressed/in a shock front as the sum of parts proportional
to a number of assigned functions. The proportions arising from the separate
functions varied from point to point of the front, in a manner found from a
number of equations of transfer. They considered in detail the case of three
functions, / „ / 2 , / 3 ; the first two were, as before, Maxwellian functions
corresponding to the T and c 0 on the two sides of the shock, and the third
was the derivative of an intermediate Maxwellian function with respect to
the velocity-component u. They found that the extra function strikingly
improved the agreement with experiment, especially for weak shocks.

15.62. Numerical solutions

Liepmann, Narasimha and ChahineJ sought to avoid the convergence
difficulties by using numerical methods. To simplify the problem, they
used the relaxation approach of 6.6 (often called the BGK, or Bhatnagar,
Gross and Krook approximation; see p. 104, first footnote), in which the
Boltzmann equation takes the form

Ft+C£ =
—T-- C*-62'1)
The relaxation time r is given by nip, and so is a determinate function of
density and temperature for any gas. §
Take axes relative to which the shock front is at rest and parallel to x = 0.
• H. M. Mott-Smith, Phys. Rev. 82, 885 (1951).
t H. Salwen, C. E. Grosch and S. Ziering. Phys. Fluids. 7, 180 (1964).
{ H. W. Liepmann. R. Narasimha and M. T. Chahine. Phys. Fluids, 5. 1313 (1962).
| Note, however, that the relaxation time for temperature changes is j/i/ip for a mon-
atomic gas(cf. (6.6,7)). The assumption of a single relaxation time is a possible source
of errors.
296 THE THIRD APPROXIMATION (15.62
Then dfjdt » o, and (15.62, 1) reduces to

urfx-f*»-f. (15.62,2)

The solution of this, satisfying the boundary conditions that/-»-/[ B ,/i 0) at

The value of/is found from this by iteration. Values of/"" for all x, satisfying
the boundary conditions at * =» + 00, are assumed known at the start of each
iterative step, and/(c, x) is deduced by numerical integration; a new/"" for
each x is then derived corresponding to the values of n, c0, T given by
n = \fdc, nc0 = ffcdc, \nkT = \f™\mC*dc.
The process is then repeated; if a reasonable first approximation to/^° is
assumed, it is rapidly convergent.
By this method Liepmann and his colleagues were able to determine
shock profiles in argon for Mach numbers up to 10. They found that the
Navier-Stokes terms p*1', 9*" were dominant throughout the profile for
Mach numbers less than 1-5. For higher Mach numbers, however, though
the Navier-Stokes terms are still dominant on the high-density side, on
the low-density side there is a considerable departure from the Navier-
Stokes profile, a long ' tail' of the profile appearing upstream of the main
shock. They suggest that any tendency towards a bi-modal distribution of
the Mott-Smith type is mainly confined to this tail.
Their discussion rests on a simplified collision model, and to that extent
is inexact. Earlier approaches were based on a more exact discussion of
collisions, followed by truncation and other approximations. None can be
regarded as altogether satisfactory; the discussion based on the BGK
approximation has at least the advantage that it is capable of giving shock
profiles valid for all Mach numbers.
The BGK approximation has also been applied with some success to
solve numerically problems of the formation of shock fronts,* the efflux of
gas through a small aperture at low pressures, f and the flow of rarefied gas
along a cylindrical tube. J The last problem has also been solved more
exactly by Simons § in terms of the eigenvalues of the collision operator.
• C. K. Chu. Phys. Fluids, 8. ix and 1430 (1965).
t H. W. Liepmann. J. Fluid Mtch. 10, 65 (1961); R. Naratimha, ibid. 10. 371 (1961).
I C. Cercignani and F. Sernagiotto. Phys. Fluids, 9. 40 (1966).
§ S. Simons. Proc. R. Soe. A. 301. 387, 401 (1967). See also C. Cercignani. Malhemalieal
Methods in Kinetic Theory, chapter 7 (New York: Plenum Prets, 1969).
 16
DENSE GASES

16.1. Collislonal transfer of molecular properties

In a gas at S.T.P., the mean free path of a molecule is large compared with the
molecular dimensions. This disparity is much reduced if the gas is com-
pressed to a density a hundred times as great, and in consequence an
additional mechanism for the transfer of momentum and energy, negligible
at ordinary gaseous densities, then becomes important. Hitherto the transfer
of molecular properties has been regarded as due solely to the free motions
of molecules between collisions; there is also, however, a transfer at collisions,
over the distance separating the centres of the two colliding molecules,
during the brief time of encounter. The extreme example of this is the
instan taneous transfer of energy and momentum at the collision of two smooth
elastic spherical molecules, across the distance o~lt between their centres.
In dense gases this collisional transfer requires consideration; it was first
studied by Enskog,* for rigid spherical molecules.
The rigid spherical model has the mathematical advantage in this con-
nection that collisions are instantaneous, and multiple encounters can be
neglected. This is not, however, the case with actual molecules; in a gas at
high pressure a molecule is in the field of force of others during a large part
of its motion, and multiple encounters are not rare. Enskog preferred the
rigid spherical model also because he believed, on the basis of analysis by
Jeans.f t n a t f° r r '8'd elastic spheres the assumption of molecular chaos
made in 3.5 remains valid at large densities. However, Jeans's argument
applies strictly only to a gas in a uniform steady state. In a dense gas, even
one composed of rigid spherical molecules, there may be some correlation
between the velocities of neighbouring molecules, because of their recent
interaction with each other or with the same neighbours. This may be
important in a gas not in a uniform steady state. Thus what Enskog gained
in mathematical simplicity is partially offset by inadequacy in the repre-
sentation of physical reality.
An adequate discussion of dense gases for general molecular models must
allow for multiple and repeated encounters, and also for the finite number of
molecules which at any instant are actually undergoing encounters. Several
attempts have been made to provide such a discussion. In certain respects
these parallel Enskog's theory, though they also reveal inadequacies in
it. They have been able to provide formal expressions for the different
• D. Enskog, K. Svemk. Vet.-Akod. Hand/. 63, no. 4 (1921).
t J. H. Jeans, Dynamical Theory of Gates, 4th ed„ p. 54 (1925).
11 [ 297 1 CMT
298 DENSE OASES (16.2
transport coefficients, but not to supply exact formulae that can be compared
with experiment. We here describe Enskog's theory in detail; a brief indi-
cation of the more recent work is given in sections 16.7-16.73.

16.2. The probability of collision

Consider a simple gas composed of rigid spherical molecules of diameter <r.
In discussing a collision between molecules with velocities c, cx and centres
O, Ox we specify the direction of the line of centres OxO at collision by the
unit vector k (cf. 3.43). Then <r*dk denotes a surface element on the sphere,
of radius <r and centre O, on which Ox must lie at the collision. The elemen-
tary area bdbde of 3.5 is the projection of a*dk on a plane normal tog, where
g "ex — c, and if 0 is the angle between g and k,
gbdbde ~ g^dkcoaO —g.k.^dk.
Hence (cf. (3.5,1)), for a gas at ordinary pressure, the probability per unit
time of a collision such that O lies in a volume dr, the velocities of the two
molecules lie in ranges dc, dex and k lies in </k is given by
/(«. r )/(ci. r ) 0*8.kdkdcdcxdr.
This expression requires correction when the gas is dense, even though
we still make the assumption of molecular chaos. First, because O is at r, Ox
must be at r — ok, so that/(cx, r) must be replaced by/(c lt r—ok). Secondly,
in a dense gas the volume of the molecules is comparable with the volume
occupied by the gas. The effect is to reduce the volume in which the centre of
any one molecule can lie, and so to increase the probability of a collision.
Thus the above expression must be multiplied by a factor x< which is a
function of position but, by the assumption of molecular chaos, not of
velocity. The function % must be evaluated at the point r—J ok at which
the spheres actually collide: hence the corrected form of the above expres-
sion is
X(r - Jo-k)/(c, r)/(c„ r - <rk) o*g .kdkdedcxdr
-y{r-\&k)fir)f l (r-(rlL)a*g\d\dcde l dr t (i6.a, 1)
using a notation similar to that of 3.5.

16.21. The factor x

The quantity x •» equal to unity for a rare gas, and increases with increasing
density, becoming infinite as the gas approaches the state in which the
molecules are packed so closely together that motion is impossible. Approxi-
mations to its value for fairly dense gases may be obtained as follows.
When two molecules collide, the distance between their centres is <r.
Let a sphere of radius <r be described about the centre of each molecule;
then at a collision the centre of one molecule lies on the sphere associated
with the other, and the centre of one molecule can never lie within the
16.21] DENSE OASES 299
associated sphere of another. If the gas is fairly rare, the number of the
associated spheres which intersect each other will form a small fraction of
the whole, and so the volume which cannot be occupied by the centre of a
molecule will be approximately the whole volume of the associated spheres,
which is \imo* per unit volume. Hence the volume in which the centre of a
molecule can lie is reduced in the ratio i — zbp, where
bp = ITT/KT8, b = \na*lm. (16.21,1)
Correspondingly the probability of molecular collisions is increased in the
ratio 1 / ( 1 - zbp).
A second factor, of comparable effect, operates to reduce the probability
of collisions: this is the shielding of one molecule by another. If an area S of
the sphere associated with a given molecule lies within the sphere associated
with a second, no other molecule can collide with the first in such a way that
at collision its centre lies on S.
If the gas is fairly rare, the number of cases in which more than two
associated spheres have a common volume can be neglected. Let two
spheres of radii x and x+dx, where x lies between <x and 2<r, be drawn con-
centric with a given molecule. Then the probability that the centre of
another molecule should lie within the shell of volume ^nxtdx between
these two spheres is ^nnx*dx. The sphere associated with such a molecule
would cut off from that associated with the given molecule a cap of height
cr—\x, and area Z7r<r(cr — \x). Hence the total probable area cut off from
the sphere associated with the given molecule by those associated with other
molecules is .,,
I zn<r(<r - $x) ^nnxadx = if-nhio*.

Since the whole area of the sphere is ^ncr*, the part on which the centres of
molecules can lie at collision is a fraction 1 — tynno* o n — tybp of the whole.
Hence the effect of shielding by other molecules is to reduce the probability
of a collision in this ratio.
On combining these results, the value of x for a fairly rare uniform gas is
found to be % _ ( l -X^bp)/(i - zbp)
-i+fbp, (16.21,2)
correct to the first order in bp. Boltzmann and Clausius* carried the calcula-
tion as far as the second order in bp, by taking into account the volume
common to more than two associated spheres; further terms have been
obtained by numerical methods, f The result is
X " i + | b / 0 + o-2869(bp) , +o-iiO3(tvt))*+o-O386(V) 4 +.... (16.21,3)
* R. Clauaiua, Mich. WOrmetheorie, vol. 3 (2nd ed.), 57 (1879). L. Boltzmann, Proe.
Amsterdam, p. 403 (1899); Witt. Abhandl. 3, 663.
t M. N. Roienbluth and A. W. Rosenbluth, J. Chem. Phyt. aa, 881 (1954); B. J. Alder
and T. E. Wainwright, J. Chem. Phys. 33. 1439 (i960); F. H. Ree and W. G. Hoover,
J. Chem. Phyt. 46, 4181 (1967).
11-a
 Copyrighted Material

300 DENSE OASES [16.3

In a non-uniform gas the value of x may be expected to involve the space-
derivatives of the density. Since, however, no invariant function of these
derivatives can be constructed which does not involve either products of
first derivatives or derivatives of liigher order than the first, in a study of
the first and second approximations the above value for X will be used.*

16.3. Bottzmann's equation; djjdt

Boltzmann's equation for a dense gas, as for a normal gas, can be written in

The expression for def/dt, however, is not quite the same as before. Consider
the probability per unit time of an inverse collision, such that the centre of
thefirstmolecule lies in a volume dr, and the velocities of the two molecules
after collision lie in ranges dc, dcx, while the direction of the line of centres is
—k, where k lies in the range dk. In such a collision the centre of the second
molecule is at r+crk, while the molecules actually impinge at r+£<rk.
Hence, by analogy with (3.52, 7) and (16.2,1), the probability in question
will be
X(r+i<rk)/(c', r)f{c[, r+<rk) <r*g .kdkdcdc,, dr
= X(r+i<rk)/'(r)/i(r + ak)^g.kdkdcdc.dr, (16.3, 2)
where c', c[ denote the velocities before collision of the two molecules, so
that, as in (3.43, 2)
c' = c+k(*.k), c; = C l -k(tf.k). (16.3,3)
The evaluation of djjdt now proceeds as in 3.52, using (16.3, 2) and
(16.2, 1) in place of the corresponding expressions of that section. The
result is .f ..
|/=JJ{x(r+i(rk)/'(r)/i(r + «rk)

-X{r-\<Tk)f(r)Ur-<Tk)}<r*g.kdkdCl. (16.3,4)
Combining this equation with (16.3, 1), we derive the equation for/.
Since in this relation the expressions xJ'\>f\aie evaluated at points other
than r, it is not possible to show by a transformation similar to that of 3.53
that the expression c a f
„A$sj^-ddc
vanishes when <f> is a summational invariant. This, in fact, is not true in
general. For though the total amount of <j> possessed by molecules is con-
served at collision, the effect of a collision is to transfer part of this total
* The possibility that x should also include scalar combinations of k and of the gradients
of p, T and c, is not considered.

Copyrighted Material
16.31] DENSE OASES 301
through a distance er from one molecule to another. Hence, since »A{S
denotes the total change by collisions in the total amount of <f> possessed by
molecules per unit volume at a given point, in general it will not vanish.
The derivation of the equations of momentum and energy is therefore
deferred till later.

16.31. The equation for/™

If the gas is uniform the expressions x and/do not depend on r; hence in
this case
^-xJJa'/i-Z/O^r.kA^
This expression differs from that obtained earlier only by the presence of the
factor x- The arguments of 4.1 can accordingly be used to show that in the
uniform steady state/again takes the Maxwellian form

When the gas is not in a uniform steady state, a first approximation to the
value of/is given by (16.31, 1). A second approximation is
/<°>(i+**»), (16.31,2)
1
where ¢^ is a linear function of the first derivatives of n, T and the mass-
velocity c0; the equation satisfied by <t>(1> is obtained from that satisfied by/
by neglecting all terms involving products of derivatives of these quantities,
or derivatives of higher order than the first. Thus in substituting from
(16.31, 2) into the left-hand side of (16.3, 1), the terms involving «P<1) are
to be neglected. This equation may therefore be written in the form

M,
et •/•(£+*«+*•«) w *
. . , . D d d
where, asm (3.12, 1), Dts 8t + e"'9r'
In this we substitute the value of/(0> from (16.31, 1); then, since C = e—e0
and c„ is a function of r, t, it becomes

+ C.(vin(Anr) + ( ^ - f ) V l n r + j ^ ( ^ - F ) ) ] , (16.31,3)

where, as before, i? = (mjzkT)* C.

302 DENSE OASES (16.32

16.3Z The second approximation to d,f/dt

The approximate form of equation (16.3,4) is obtained by expanding x>fvfi
in powers of <rk by Taylor's theorem, and retaining only diefirstderivatives;
it is
*j£- xjf(fTi-ff,)<T*g.kdkJcl

+X JJk.(rV/i+/V/i> ^g.kdkdc,
1
+ -jjk.VX(fT1+f/^<T*g.kikdc1, (16.32,1)
where in this relation all the quantities are evaluated at the point r.
In substituting from (16.31, 2) into the first term on the right in this
equation we neglect the productsflX^O?',$&&}*, as before. Hence, since

this term becomes

xflfWyW+•?" - «*" - OF1) <r*8 • kdkdev (16.32, 3)
The second and third terms on therightalready involve space-derivatives;
thus in substituting from (16.31, 2) into these we may neglect all the terms
involving Q"', and so write/"" in place of/. Using (16.32, 2), these two terms
then become xjjf%<»k.Vln(f?Wx)a*g.kdkdcl
or, after substituting the values of/P , ,/f v

X<r»JJ/«7i«k. jvin (^) +£p(Cl+CnVT

+Vc0.jJ(C1+Ci))ir.krfk&1.
The integration with respect to k, using results proved later (16.8, 2, 5,6),
givts . r / 1\

m
+ 5 g,vr.{aq # -Ks # f jr .<y+Ci < ^+|rtr}
+j^Fc9-.izClg-lgg)-\fSI.c.)\dcv
Substituting g ™ Ct — C and integrating with respect to Cv we get
-bpxT{C.(V]n(k'xT)+l(<in-t)VlnT)
+ K a W : V e , + ( * » - | ) V.c,)}, (16.32.4)
since (cf. (16.21,1)) bp -> f »r»<7*. By (16.32,1), dj/dt is the sum of
(16.32, 3 and 4).
 Copyrighted Material

16.33] DENSE OASES 303

16.33. The value of /•»

The equation satisfied by the function ¢ ^ is found by substituting the value
of dj/dt in (16.31, 3). It is

= ->"[;g*<--»}£*«-(&(S-')
+(«Ar)-1v/>0+(i+fb/>X)(^-|)vinrj
+ tb/>x(^-f)V.c 0 + 2(i + fbptffftf:Vc0], (16.33,1)
where p0 = knT(i+bpx). (16.33,2)
If (16.33, 0 1S multiplied by i/rdc, where ijr is a summational invariant, and
integrated over all values of c, the left-hand side of the resulting equation
vanishes, by virtue of the transformation used in deriving (4.4, 8).* Thus
when \/r = 1 we get the result
Dn
^•+«V.Co = o, (16.33,3)
Dc 1
and when ^ = mC,f -=£ - F+- Vp0 = 0. (16.33,4)
Finally, \]r = \mC* gives
ikTDn ikDT .,„. ,, ._
+ +Wl+fb
i^5F 2^r ^ -Co=0
v

or, using (16.33,2,3), T5i~~ynkV'C* (l6-33» 5)

Equations (16.3, 3-5) are first approximations to the equations of con-
tinuity, momentum and energy; they replace (7.14,14-16), which would
reduce to them if the hydrostatic pressure was taken to be po, given by
(16.33, 2)> instead of knT. Using the values of Dn/Dt, DTjDt and DcJDt
• The proof of this transformation differs slightly from that given in 4.4, since account
is taken in this chapter of the difference in position of the centres of colliding molecules.
This difficulty is overcome by considering an ideal uniform gas, whose velocity-
distribution function at all points is equal to /(c, r) at the point considered. The
proof is then readily derived as in 4.4, provided that the summational invariant ^ is
not itself a function of position.
f The function mC is not a summational invariant if C is defined as c—e„, and if
c0 • c„(r) for one of the colliding molecules, and c, «• c^r+crk) for the other. To
overcome thia difficulty we put C = c—c, for each molecule, where c0 is taken to be
a vector independent of position, which coincides with the mass-velocity at the special
point r under consideration. This ensures that /nC is not a function of position. Similar
remarks apply to the summational invariant ImC

Copyrighted Material
304 DENSE OASES [16.34
we
given by (16.33,3~5)> &*& *"•* (^-33,«) reduces to

- -r'f^i +fbpx)(<f»-{)Vln r + 2 ( i +fbpx)«%:Vc 0 }. (16.33, 6)

This differs from the corresponding equation for a gas at ordinary densities
only in that the term involving Vc0 is multiplied by (1 + lbpx)lx< an<^ &**
involving V In T by (1 + $bpx)/X- Hence its solution can be given at once in
terms of the functions A, B of 7.31; it is

• « - - ^ { ( ' + fbpx) ( ^ ) * i l . V l n T+2(i + |lv>x)B:Vc0)).

(»6-33.7)
16.34. The mean values otpCC and \pC*C
Approximations to the mean values of certain functions of the velocity can
now be obtained. For example, the first approximations to the values of
pCC and ipC*C, as for a normal gas, are
ptfCF-knTU, MCtCT-o, (16-34.1)
where U is the unit tensor. The second approximations are obtained by
adding to these the quantities p(CCp, J/^C'C)0', which, by (16.33, 7)> a r e
respectively (1 + ibpx)lx and (' + V>PX)IX times those for a normal gas;
that is, 0
rtCCp = -r\* + 1*>PX)• */*Vc^ (16.34, 2)
i/KC*cy5--;r'(i + ibpx).Avr, (16.34,3)
where A and fi are the conductivity and viscosity of the gas at normal
densities.
The quantities pCC and \pC*C give the parts of the pressure tensor
p and the thermal flux q that arise from the transport of momentum and
energy by the motion of molecules from point to point. To these must be
added the contributions arising from the transport of momentum and
energy by molecular collisions. These will now be evaluated, correct to the
second approximation.

16.4. The colHsional transfer of molecular properties

Consider the collisional transfer of a molecular property xjr across an element
of area dS at the point r. It will be supposed that ^ is a summational
invariant, and does not depend on r; in this case a collision does not alter the
total amount of \[r possessed by the molecules, but transfers part from one
molecule to the other, so that aflowof tfr results. If a collision is to produce
a transfer oftfracross dS, the centres of the colliding molecules must lie on
opposite sides of dS, and their join must cut it.
16.4) DENSE OASES 305
Let the normal to dS, drawn from its negative side to its positive, have a
direction specified by the unit vector n. In a collision between molecules
with velocities c and cv of which the first lies on the positive side of dS and
the second on the negative, since k is the direction of the line drawn from the
centre of the second to that of the first, k . n is positive. If the line of centres
at collision, of length <r, is to cut dS, the centre of the first molecule must lie
within a cylinder on dS as base, with generators parallel to k and of length <r;
the volume of this cylinder is <r(k. n) dS. Again, the mean positions of the
centres of the two molecules are the points r + Jerk, while the mean position
of the point of impact is the point r . Hence the probable number of such
collisions per unit time, in which c, c 1( k lie respectively in ranges dc, delt
</k, is, by analogy with (16.2, 1),
XiDfir + J e r k ) / , ( r - J<rk) o*g.kdkdcdcx{k.n)dS.
Each such collision causes a molecule on the positive side of dS to gain a
quantity ijr'—\jr of the property \{r at the expense of a molecule on the
negative side. Thus the total transfer across dS by collisions of this type is
(r-^)x{r)f(r + WWi(r-iak)<7»(g.k)(k.B)dkdcdcldS,
and the total rate of transfer of \jr across dS by collisions, per unit time and
area,is
^ r ) J J J ( ^ - ^ ) / ( r + J<7k)/ l (r-itrk)(^.k)(k.n)A</crfc 1 ,
the integration being over all values of the variables such that g.k and
k . n are positive. In this expression let the variables c and c, be inter-
changed ; this is equivalent to interchanging the roles of the two colliding
molecules. Thus k is to be replaced by — k, g by —g, and \jr' — ijf by ^f\ — ^ , ,
which is equal to — ffi — yjf). On performing the interchange, we obtain an
expression identical in form with the original one, but with the integration
now taken over all values of the variables such that g.k is positive and
k . n negative. The rate of transfer in question is therefore equal to
i*»X(r) Mir -*)/<r + i<rk)Mr- \<rk)(g.k) (k.n)dkdcdcv
taken over all values of the variables such that g.k is positive. This expres-
sion is the scalar product of n and another vector which, by analogy with
2.3, represents the contribution of collisions to the vector of flow of &.
As in 16.32,/and/, are expanded by Taylor's theorem, and derivatives of
higher order than the first are neglected; the collisional part of the vector of
flow of v^ is then found to be

+ J^///(^'-^)//1k.Vln(///i)k^.k)A</c</c1, (16.4,1)
where all the quantities are evaluated at the point r. Since the second term
of this already involves space-derivatives,/and/, can here be replaced by
/w and/1°'.
306 DENSE OASES [16.41
16.41. The viscosity of a dense gas
Consider first the transfer of momentum. Let
jjr -mC*»m(c-c 0 ),
where e0 is evaluated at the special point r under consideration, and does
not vary with the position of the molecule. Then (16.4,1) becomes
lX^HJm{C'-C)ff1kig.k)dkdcdel
+ iX<*flfaC'-C)f^\.Vln(f<»lfl»)k(g.k)dkdcdcv
After integration with respect to k, using (16.8, 7, 8), this expression becomes

Ts X^jjfM2**+ U **) dcdc *

+ ^X<*»jffmm{g-Vln(fmIA,n)}(gg+gtU)lg
+^Vln(/»W)+Vln(/ (0, //[ 0) )#}]«fc«/ci- (16.41,1)
In the first integral we use the relation £ •= ¢^-C; then since, for any
function ¢,^ /#*-//,**.-«*
and Cx — o, C «• o, the value of this integral is
Jb^aCC+CMJ). (16.41,2)
In the second integral, since
vincry/[«) - ^(c^-cflvr+^c.Mc-c,),
(16.41,3)
the terms involving VT yield odd functions of C, Clt which vanish on
integration; hence the integral is equal to

-^^jj^A'Wco-ggHgB+i^Vg
+g{<?Co g)g +*(Vc e -g)}] dcdev
This can readily be evaluated by changing the variables of integration from
c and cx to G9 and g, where G„ = \(C+ Ct). Its value is found to be

-m{ff^+UV.c 0 } - -07{J8+UV.C0}, (16.41, 4)

4 ,
where w B |irtx»*<r (mA7 )t - gbV^mArp/jrlor*. (16.41,5)
Alternatively, in terms of the coefficient of viscosity at ordinary pressures,
namely ^ _ I0i6 x 5 ( ^ 7 ^ 5 , ^

w is given by w =• 1 •ooifix^P1- (16.41, 6)

16.42] DENSE OASES 307
Adding (16.41, 2) and (16.41, 4), we obtain for the value of (16.41, 1)
ibp'x(2CC+ CnJ) - m ( f 8 + UV.c 0 ).
This gives the part of the pressure tensor which arises from the collisional
transport of momentum. To this we must add pCC, which gives the rate of
transport by molecular motions; then the whole pressure tensor is found
t0
** p = p(i + Ibpx)CC+ Jbp^CHJ -t*(|8 + UV.c.).
On substituting in this the value of pCC found in 16.34, i* becomes
p = (p0-wV.c0) U - ( 2 / ^ 1 + Ibptf+i*)8, (16.41, 7)
wherep 0 = knT(i + bpx), as in (16.33, 2).
For a gas in a uniform steady state, p reduces to a hydrostatic pressurep 0 —
a result already implicit in the forms of equations (16.33, 4' 5)- For a gas not
in a uniform steady state, the hydrostatic pressure is p0 — wV. c0. The extra
term — roV.e0 differs from zero whenever the density of the gas is varying
(cf. (16.33, 3))! li represents a volume viscosity similar to that of 11.51,
opposing expansion or contraction of the gas. The deviation of the pressure
system from the hydrostatic is given by the last term of (16.41, 7), which
represents the usual shear viscosity. Equating this term to — 2/t'§, we see
that in a dense gas the coefficient of viscosity /1' is given in terms of /i, the
viscosity at ordinary densities (at the same temperature), by the relation
A ' - A A T H i + f b P ^ + fw
= /* b P{(Vx) _1 + 0 8 +o^f)t\bpx} (16.41, 8)
using (16.41, 6).

16.42. The thermal conductivity of a dense gas

To evaluate the thermal flow, we write \jr = \mC* in (16.4, 1), where again
C is taken as the velocity of a molecule relative to the mass-velocity of the
gas at the special point r. Then the expression becomes
ixm<T*ffj(C"- C^ffMS-^^dcdc,
+ iXm<r*jjj(C'i-C*)f^^k.Vln(f^fnk(g.k)dkdcdc1.
The two integrals in this can be evaluated in the same way as those in 16.41,
using (16.8, 9, 10). On carrying out the integration, the collisional transport
of energy is found to be ^ptj^C-c.wVT, (16.42,1)
where m is given by (16.41, 5), and c„ = 3*/2w. On adding to (16.42, 1)
the transport \pC*C by molecular motions, the total flux-vector of energy is
found to be given by
q = \pC*C{i + \bpX) -cv w VT
or, using the value of $pC*C found in 16.34, by
9 = -fcrHi+Jb / o^)»A + c l ,w}Vr. (16.42,2)
308 DENSE OASES [16.S
Thus at any temperature the true conductivity, A', is given in terms of that
at ordinary densities, A, by the equation

- bpA{(bpX)-1 +1 2+o-7S74l¥>X}- (i6-4*. 3)

16.5. Comparison with experiment

Equations (16.41, 8), (16.42, 3), which give the coefficients of viscosity and
heat conduction /1', A' for a dense gas in terms of the corresponding coeffi-
cients /t, A for a normal gas, may be written in the forms
/t'/bp - /»{(bp^)-1+08 + 07614VX}. (16.5,1)
l
A'/bp - A{(bpx)" + i-2 + 07574bpx}. (16.5, 2)
Thus the theory predicts that, as p varies at a given temperature, fi'jp and
A'/p should have minima respectively corresponding to
bpx " (07614)-* = 1 146, bpX - (0 7574)"* - " 4 9 . («6-5,3)
i.e. at practically the same density. These minima are actually observed;
qualitatively, at least, Enskog's formulafitsthe experimental variations with
density at any temperature well above the critical temperature.
For quantitative comparisons with experiment we require values of bp
and x- If the molecules were strictlyrigidspheres, these could be calculated
from viscosity diameters, using (16.21,1,3). Sengers* used viscosity
diameters at ordinary pressures at the temperature concerned in a com-
parison of (16.5,1, 2) with experiment. He found that the calculated
values of jt' and A' give moderate agreement with experiment for values of
bp less than 0-4, but are too high for greater values.
Enskog preferred a different procedure, based on the close relation
between bpx and the compressibility (cf. (16.33, *))• He observed that, if the
(spherical) molecules were surrounded by weak attractive fields of force, the
equation of state , , . ^ , +1 ^ (,6.5,4)
would be modified to the form
p,+ap» - k*T(i +bpx), (16.5, 5)
where a is independent of the temperature (the term ap* is the well-known
van der Waak correction). He therefore suggested that bpx should be
determined from compressibility experiments, using the relation

7^-An7Xi+bpx) (16.5,6)

* J. V. Scngcrt, Int. J. Htat Man Tranrftr, 8, 1103 (196s).

16.5] DENSE OASES 309
instead of (16.5, 4). This means that p0 in (16.5, 4) is replaced by the
'thermal pressure'* TdpJdT. For spherical molecules without attractive
fields of force, (16.5, 4) and (16.5, 6) are equivalent: for actual molecules
this is no longer the case, because the values calculated for bpx depend on T.
When bpx is determined from (16.5, 6), b can be found from the limit of
b\ as p tends to zero. That is, if the experimental p0, p relation is expressed
in the form p0 = k»T(i+PB(T)+p*C(T)+...), (16.5,7)
b is given by b = d{TB)/dT. (16.5, 8)
Alternatively, a direct determination of b can be avoided by using the
minima of fi'/p and X'/p. If the minimum values are denoted by (/*'/p)min
and (A7/>)mln, (16.5, 1, 2) are equivalent to
2-S4S/*'/P = (/«7P)O.ID ((bpx)-1 + o-8 + o-76i4b/>x). (16.5,9)
,
2-94iA'/p = (A'/p)mlI1((bpA:)- + i-2 + o-7574b/ox). (16.5,10)
Values of /1' and A' calculated from these equations can be expected to fit
the experimental results best near the minima of /i'/p and A'/p, i.e. for large
pressures; values calculated from (16.5, 1, 2), using (16.5, 8), can be
expected to give a betterfitat lower pressures.
The following table gives the experimental values of the coefficient of
viscosity of nitrogen at 50 °C. obtained by Michels and Gibson.f together
with values calculated using (16.5,1 and 9). The values of bpx were derived

Table 28. Viscosity of nitrogen at Ugh pressures

/t'xio* /I'XIO*
Pressure (calc. (calc. fl' x to* P M'lP
(atmos.) bpx O6.5. 1)) (16.5. 9)) (exp) (g./cm.») (exp.)

15-37 0-031 191 181 io>-3 0-01623 0-01179

57-6o 0-119 201 too 1981 0-06049 0003274
"°4-5 0-215 217 205 208-8 0-1083 0-001928
2124 0491 »37 224 »373 0-2067 0001148
3*o-4 0717 281 266 «737 02875 0-000952
43»« 0-920 3»6 308 3129 0-35*8 0-000887
S4I-7 I-III 368 348 350-9 04053 0-000866
6304 1247 401 380 3786 0-4409 0-000859
74*-1 1413 442 4«8 4163 0-4786 0-000870
854' I-S76 481 4SS 4550 0-5117 0-000889
965-8 1-732 S20 492 49» 3 05404 0-000909

* The first approximation to the equation for DTjDt for a general dense gas is
DT ^?p.„
Dt iT
so that the thermal pressure replaces the total pressure p, in this equation. On the other
hand, it i»p, which appears in the equation of motion,
t A. Michels and R. O. Gibson, Proe. R. Sot. A, 134, 288 (1931).
310 DENSE OASES [16.5
from the experimental results of Deming and Shupe.* The values calculated
from (16.5,1), which are adjusted to agree with experiment at moderate
pressures, are nearly 6 per cent too high at large pressures; those calculated
from (16.5, 9) agree with experiment very well at high pressures, but are too
low at normal pressures. The minimum value of /t'/p (last column) corre-
sponds roughly to bpx » 1*2.
A similarly rough agreement with experiment was quoted by Sengersf
for a number of other gases; for neon and hydrogen the theoretical ft'
increases faster with density than the experimental; for helium, argon and
xenon the situation is reversed. Sengers quoted similar results for A'; for
both neon and argon A' calculated from (16.5, 2) is less than the experi-
mental A' at high pressures, by small amounts at pressures below 250
atmospheres, but by 6 per cent for neon and 15 per cent for argon at pressures
of order 2000 atmospheres.
Michels and BotzenJ attempted to apply (16.5, 2) to the thermal con-
ductivity of nitrogen, obtaining calculated values 15 per cent higher than
the experimental at 2000 atmospheres. This is not surprising if the transport
of internal energy is by a self-diffusion process, as in 11.8, since at high
densities the self-diffusion coefficient is cut down roughly in the same ratio
as i/px (cf. 16.6). However, Sengers, taking the transport of translatory
energy to be given by (16.5, 2) and that of internal energy to be by self-
diffusion, obtained calculated values of A' for nitrogen well below the
experimental ones, suggesting that the self-diffusion approximation is only
roughly valid.
To summarize, Enskog's theory approximately represents the observed
variation of/*' and A' with p at a given temperature. It is unable to represent
their variation with temperature, because of the rigid sphere model; it is
necessary to determine bpx at any temperature (and density) by a semi-
empirical rule. The dependence of ft' and A' on temperature has its own
features of interest; for a given density, A'/A and ft'/ft appear to decrease
with increasing temperature, and for some gases it has been suggested that
ft'—ftia nearly a function of p, independent of the temperature. § Enskog's
theory cannot explain such behaviour.
Madigosky|| in ultrasonic experiments with compressed argon, found
evidence of a volume viscosity w similar to that of (16.41,6,7). When
X and b were determined from compressibility, as in (16.5, 6), a moderate
agreement with (16.41,6) was obtained.
* W. E. Deming and L. E. Shupe, Phys. Rev. 37, 638 (1931).
t Lae. eit.
t A. Michelt and A. Botzen, Phytica, 19, 585 (1953).
I See. e.g., G. P. Flynn, R. V. Hankt. N. A. I^maire and J. Rots, J. Chem. Pkyt. 38, 154
(1063). However, the retult quoted it only very approximate.
II W . M . Madigoaky, J. Chtm. Pkyt. 46. 4441 (1967).
16.6] DENSE OASES 311

16.6. Extension to mixed dense gases

Enskog's methods were generalized by the late H. H. Thorne (of the
University of Sydney, Australia), to apply to a binary mixture. We give
here only his results as they affect the first approximation to the coefficient
of diffusion.*
Let m, and m, be the masses of molecules of the two gases, crl and o", their
radii. Then b^ b„ b[ and b't are defined by
blp1 = innlai, b,p, = J7rn,oi bJ/Oj = fff^oi,, b'tpt = \mta\t.
(16.6,1)
Corresponding to the factor x of 16.2, three factors Xv Xi an d Xi» are intro-
duced, related respectively to collisions between pairs of molecules w„ pairs
of molecules m„ and pairs of dissimilar molecules. It is found that

X i = 1+finni<rl + ~ » s (oi+i6ai,-120-5,0-1) + ...,

(16.6, 2)
^ = 1+^^0-5(8-30-^0-1,) + - ^ 0 1 ( 8 - 3 0 - ^ 0 - , ^ + ...,

with a corresponding equation for Xr In terms of these, the first approxima-

tion to the hydrostatic pressure is found to bep0 = Pw+Pta, where
/>io = k"i ?"(» + *>iPiXi + b'tPiXu). P» = *»tT{i + btPtXt + b'iPrfit)-
(16.6, 3)
The Boltzmann equations for the two gases are of obvious forms. First
approximations to their solutions are the Maxwellian functions/,(0) and/}0);
second approximations are /l 0, (i + 4 ^ ) , /T(i + W"), where ¢ ^ andfl^1'
are found to satisfy the equations

+xJfA0,Aamu+W)-Wy-WY)*Ug*i*<nLdct
= - / H O + faPiXi + ¥**» MJbiPtXu) (*! -1) Ci. V In T
+ xr l du. Ci + 2( 1 + f b,p, Xl + \Mt b'tPtXu) * f o : Vc0
+ l*t{\PiXi ~ hPtXt + 2Xi,(Mtpt b't - MlPlb[)} {V\ - \) V. c,]
(16.6.4)
and a similar equation; </lt andrf81are given by
d
» = "rf« = pm{F>-F>+kVp»-kVp»)
PiPt

+ x,b;piX„(Vln(x,/x1) + (A/ 1 -M,)Vln T). (16.6, 5)

For a gas at rest at uniform temperature under no forces, </,, — Vx, for
* A fuller account was given in earlier editions of this book.
312 DENSE OASES [16.7
self-diffusion, as in 8.4; however, du may differ considerably from Vx,
for a general mixed gas.
The velocity of diffusion is found to be
Q - C , - -(x^-'DIM.+krVlnD,
where D'lt is, to a first approximation, given by
pD'it-Po[DiMXi* (16.6,6)
f/7|t]i being the first approximation to the coefficient of diffusion for the
same gas at a moderate density p0. Thus the effect of high pressure is simply
to decrease Dlt in the ratio/>,//%,, as a consequence of the increased number
of collisions between unlike molecules. For self-diffusion similarly
pD'n-pJD^Jx. (16.6,7)
The experimental results give, at best, only moderate agreement with
(16.6, 6, 7). They suggest that dense-gas corrections are considerably
overestimated by these formulae, though much depends on the method of
calculation of the correction factors x,„ x*

THE BB3KY EQUATIONS

16.7. A brief sketch will now be given of a more general theory of dense
gases, based on the concept of multiple velocity-distribution functions. We
consider only a simple gas whose molecules do not possess internal energy.
A generalized velocity-distribution functionf/"' is defined as such that
the probability of finding t molecules whose centres are respectively in small
volume elements drv drt, ..., dv„ and whose respective velocities lie in the
ranges dex, dc* ..., de„ is
f*(ci>c» >c„rurt, ...,r„t)drv..dr,del...det.
A generalized number-density «"* is similarly defined; it satisfies the
equation «<•>-/...//»>&,...&.. (16.7,1)
The quantities tfl*, /M are identical with the usual number-density n and
velocity-distribution function/. In a gas at ordinary densities one could write,
to a sufficient approximation,
/*-f{c»rM(ct,rt), f» -/(ci.rO/fci.rj/fcfcr,),
and so on; but in a dense gas this is no longer adequate, since any one
molecule perturbs the distribution of others in its vicinity. J
• K. D. Timmerhau* and H. G. Drickamer, J. Chtm. Phyt. so, 081 (195a); T. R. Mifflin
and C. O. Bennett, J. Chtm. Phyt. ao, 075 (1958); M. Lipaic**. J. Chtm. Phyt. 36, 1x35
(196a); L. Durbin and R. Kobayaahi, J. Chtm. Phyt. 37. 1643 (196a).
t The change of notation abould be noted ;/"* does not here refer to auccetarve approxi-
mation* to/.
t Note that the volume element* dr„ 4rv ..., ir, u*ed in defining/"1 mutt be aaaumed
•mall compared with the dimention* of the molecularfield*of force.
16.7) DENSE OASES 313
The functions/"' can be used to construct mean values in accordance with
the formula n(.fy.) = f...jf)$dCl...dc,. (16.7,2)
n) m
A function 0(cj) may have several different mean values ^ , ^ , .... For
example, ci1' (or c,) is the mean velocity of all molecules near r „ c**' is the
mean velocity of a molecule in drx when it is known that a second molecule
is \ndrt; when r t — r, is small, cV' and c\J) may differ appreciably .The mean
value 0(c) is understood to be ^{cfl).
Each function n**' satisfies an equation of continuity

^ + 2 ^ . ( ^ ^ ) = 0- (16.7,3)
Similarly /"' satisfies a Boltzmann equation

where T)'*' is the average acceleration of the ith molecule of the group of s,
the average being taken over all possible positions and velocities of molecules
not belonging to the group. Let F{, X(] be the accelerations of the molecule 1
due respectively to external forces and to the force exerted by the molecule/;
X{j is taken to be in the direction of the vector rtj a r{ — rt, and is given by
»>X„ = - g » , (16.7, 5)
where vit ( s v^r^) is the mutual potential energy of the molecules i,j. Then*

the last term on the right represents the average acceleration due to mole-
cules outside the group of*. Hence (16.7, 4) becomes

CC • 0/»+»
= -J J S ^ M + l ^ - * " ' ^ 1 , (l6-7 6)
'
The equations (16.7,6) were given in 1946 independently by Bogoliubov.f
by Born and Green, J and by Kirkwood ;§ they had already been introduced
in 1935 by Yvon, | but attracted little attention at the time. In recognition of
the multiple authorship, they are frequently styled the B B G K Y equations.
* Here and later, the symbol E' it used to indicate that meaningless terms representing
the interaction of a molecule with itself are omitted from the sum.
t N. N. Bogoliubov, Problems of a Dynamical Theory in Statistical Physics (Moscow:
State Technical Press, 1946). English translation by E. K. Gora in Studies m Statistical
Mechanics, vol. I (ed. J. de Boer and G. E. Uhlenbeck, 196a).
X M. Bom and H. S. Green, Proc. R. Soc. A, 188, 10 (1946); 189, 103 (1947); 190, 455
(i947).
I J. G. Kirkwood. J. Chem. Phys. 14. 180 (1946); 15, 72 (1947).
II J. Yvon, La TMorie desftuides et de Viquation d'etat (Paris: Hermann et Cie, 1935).
314 DENSE OASES {16.71

16.71. The equations of transfer

In a dense gas, the pressure tensor p can be expressed in the form
p=/>CC+p'. (16.71,1)
Here pCC denotes, as usual, the transport of momentum relative to the
mass-velocity c,„ and p' gives the effect of the intermolecular forces mXif.
By an appropriate generalization of the argument of 16.4, the value of p' at
the point r can be shown to be

P' = iwJ'jJ^ri.rJAtJruXuA',,, (16.71, 2)

where r1 = r + xrli, rt = r-(i-x)rlt. (16.71,3)

The contribution of p' to the hydrostatic pressure is

P' = J«JjJ"' ""(r,, r%)rf*j r„.X l t dr u , (16.71, 4)

and the force per unit volume at r due to p' is*

- ^ . p' = w J W . r , ) * „ * • „ . (16.71, 5)
The heat energy of the gas includes a part corresponding to intermole-
cular potential energy. Half of the energy vtj is arbitrarily supposed to
reside in each of the molecules /',;'; thus the average heat energy of a molecule
at r is E, where _ ~ .
nE = \PC*+ Jji^r.r,) v(rxt)drlv (16.71, 6)
The thermalflux-vectorq is given by
q = JpC*C+ ij^r, rt) v(rlt) C?drxt+q'. (16.71, 7)
Thefirsttwo terms on the right of (16.71, 7) give the transport of kinetic and
potential energies arising from the molecular motions, and q' represents
work done by intermolecular forces. In calculating q' it has to be remembered
that work done in altering c„ is taken into account as work done by p'. It is
found that
«' = ^*-11{Jo'ii'«(r1,r1)«Xu.ci*><fajrfr11-p'.c0(r), (16.71, 8)

where r,, rt are given by (16.71, 3). The rate of heating per unit volume at
r arising from q' is
9 = w ,+
-fr ' 1 J^''«) *»(^ 3Vru + £ ( P ^
(16.71, 9)
* In (16.71, 5) (and (16.71, 6, 7, 9)) the molecule 1 ii taken to be at the point r, and to
poaaeat velocity c.
16.72] DENSE GASES 315
The equation of transfer of momentum is obtained by multiplying by
mcx the equation for/ (1) ((16.7, &)» w ' t n * = ' ) a n t * integrating with respect
to ex. The energy equation is found by multiplying the equation for/ a > by
\mc\ and integrating with respect to c „ and adding the equation got by
multiplying the equation of continuity (16.7, 3) for n*1' by \vu and inte-
grating with respect to rs. Using (16.71, 1-8), the momentum and energy
equations can be reduced to the usual forms
Dc0 „ 3 dE 8 8 , , ,
P n
Dt -"*-*•* di = -*rq-*-*C>- (l6
-7I',0)
The expressions p', q' correspond to the collisional transport of momen-
tum and energy in Enskog's theory. The second term on the right of
(16.71, 6), however, has no analogue in Enskog's work. For a gas in a
uniform steady state, (16.71, 4) reduces to the expression given by the
virial theorem.

16.72. The uniform steady state

In the uniform steady state under no forces, the BBGKY equations (16.7, 6)
are satisfied by*

provided that, for all i < s,

Wt-%r**& X" = ifSXt'+^'+l)dr'+v (l6 72 2)

- '
Since mXti = - 8v(1/dr{ = - V( vit, we write

X.(?) = e x p | - J > „ / A 7 J (q>s), (16.72,3)

* » s «%( 2 )^(3)...X...W JV». (16.72, 4)

Then (16.72, 2) becomes
VtN<» = nJN<*"ViX,(s+ i)dr,+l. (16.72, 5)

This equation has to be solved subject to the boundary condition that, when
the molecules 1,2 s are all sufficiently far apart, «*" = «* and so Nw = 1
(since then vtj = o, i,j = 1,2, ...,s).
If the gas density is not too large, (16.72,5) can be solved by successive
approximations. The first approximation is got by neglecting the right-hand
side; this gives iV(,) = const. = 1, using the boundary condition.The second
approximation is then got by substituting N('+l> = 1 on the right of (16.72, 5)
and solving for Nu): the third approximation by substituting the second
* No complete proof of the uniqueness of this solution ha* been given, though 1 number
of partial proofs exist.
316 DENSE OASES (16.73
v
approximation for W+ , and again solving: and so on. Thus, for example,
***** z,, . e x p ( - v(,/kT) - i (16.72, 6)
it can be shown that
n» - n*(i +z„){i +njz„zttdr,
2 2 2
+ K / J t " * " ^ + 2Z14 + 2Z„ + ZHZ,,)
+ Z I » Z » Z U Z « ] ^ » * 4 + ...}. (16.72, 7)
Using (16.71,4,6), the contributions p'0, E' from the intermolecular
forces to the hydrostatic pressure and the mean heat energy per molecule
in the uniform steady state can be shown to be

pl--kn*T8£, E'-kT*8?, (16.7a, 8)

where*
T a injzndrt + ^jjzuz^zndrtdr,
+A^/JJzitZtsSsiZwfo+6z„+zwz14)</rIrfr1«/r4+.... (16.72,9)
The expression — n ffY/dn in the equation for p't is a series of powers of the
density, equivalent to die expression bpx in (16.33, 2 )- ^ w e whc
pi = kn1\pB(T)+p*C(T)+...) (16.72, 10)
tne
(cf. (16.5,7)), coefficients B, C,... give the second, third, ... virial
coefficients. The coefficients arise respectively from the interactions of
molecules in groups of two, three,....

16.73. The transport phenomena

For a gas not in a uniform steady state, the local temperature T is defined as
in 2.41, i.e. as the temperature at which the same gas, when in a uniform
steady state at the same density, would have the same mean energy E per
molecule.
For simplicity we consider a gas subject to no external forces (F = 0). In
order to determine the transport coefficients, we need to solve the BBGKY
equations for/1", correct to terms of first order in the space gradients of T
and c0. Bogoliubov proposed the following generalization of the method of
successive approximations used in 16.72 to determine «**' in the uniform
steady state.
A first approximation to/ 1 ", where s > 1, is obtained from (16.7, 6) by
neglecting the integral on the right. The equation then asserts that/* 1 is
* The remits (16.7a, 8,9) are usually derived by a rather different method, depending
on ideas of statistical mechanics. See. e.g., J. E. Mayer and M. G. Mayer, Statistical
Mtdutmct, chapter 13 (Wiley, 1940).
16.73] DENSE OASES 317
constant along a trajectory of the 5 molecules as they interact; since their
velocities were uncorrelated before they began to interact, we may write to
this approximation
/ » = P\c[, ri)/<«(ci, rj)...f l \c'„ r',). (16.73,')
However, we cannot simply assert that cj, rj, etc., are the velocities and
positions 'before encounter', since the precise position is important, and
'before encounter' has no precise meaning. Bogoliubov assumed that c[, r[
could be obtained byfirstfollowing the motion backward for a time r, until
an instant is reached before the molecules began to interact, and then
following the motion forward with constant velocity for the same time T.
Then thefirstapproximation to/1*' is known in terms of/™.
When the first approximation has been found, a second approximation to
/"• is got by substituting thefirstapproximation to /w- n on the right of
(16.7, 6) and solving for/**'. The equation now indicates how fast/*' varies
along a trajectory as a consequence of (s+ i)-molecule interactions; once
more it is integrated using the boundary condition (16.73, 1), with c[, r[,
etc., obtained by the same process of backward and forward projection of the
j-molecule motion. Further approximations are similarly derived. Thus
each/"' (J > 1) is in principle determinate to any order of approximation in
terms of/1', the successive approximations introducing terms depending
respectively on the encounters of s, s+i, s+2, ... molecules. When/™ is
known in terms of/fl> to any order of approximation, / ° 1 can in principle be
determined to a similar order from thefirstof equations (16.7, 6).
Essentially, Bogoliubov's approach replaces thefirstof equations (16.7, 6)
by an equation of the form

where J, is a functional depending on products of s functions f*> (with

different arguments). In afirstapproximation we retain only Jt, which now
reduces to .7¾01. the Boltzmann collision-integral for dcf/dt. In the second
approximation we retain 7,, which represents the effect of ternary encounters,
to a first approximation TJ"; also Jt becomes Jjf" + J\?\ where ^ ' represents
the effect of differences of position of molecules undergoing binary en-
counters (cf. 16.3,2). We may write
/Pi_yiu» + yai> t (16.73,3)
0> n,,)
where /°- is the usual Maxwellian function, and / is linear in the
gradients of T and c0. Correct to terms offirstorder in these gradients, we
can replace fv on the left of (16.73, 2)an<* m AX) by/(1'e), and neglect terms
involving products of functions /^-*>. Hence (16.73, 2 ) becomes, to the
318 DENSE OASES (16.73

second approximation,

BT+c • \~=JM0, iyn,a>)+

TO 0 , 1 * 0 ,l))

+JfX/n.yi.iyii.«)+4^/a«ya.^a.ii), (16.73,4)
W 01
since J| , ./¾ vanish if f® - / n - * . This equation may be compared with
(16.31, 3), (16.32, 3, 4); the terms in J},1* in (16.73, 4) correspond to those
appearing in (16.32, 4), taking x - 1.
Further approximations to (16.73, 2 ) can be similarly obtained. However,
most workers have not proceeded beyond (16.73, 4)- This '* n o t °nly
because of difficulties in discussing many-body encounters; there is also
a doubt whether the later approximations are correct Indeed, (16.73,4)
itself may not be exact.
The expression Ji 0) in (16.73,4) represents all types of ternary effects,
including (cf. 16.21) the effect of one molecule being shielded from colliding
with a second by the intervention of a third. Also, in view of uncertainty
where exactly an encounter begins and ends, and of the degree of correlation
of the velocities of two molecules after encounter, it is necessary to regard as
a ternary encounter three successive binary encounters of three molecules
1, 2,3, e.g. the successions (1,2)-(2,3)-(1. 2 ) o r ( i . * H 2 . 3 M i » 3)- In such
successions, the later encounters may occur at points well separated from
the first. In practice, such distant successions are ruled out because at least
one of the molecules would collide with a fourth molecule before the suc-
cession is complete; but this correction cannot be made until one introduces
Jt. The correction is not large, since three successive binary encounters
involving three molecules are very unlikely when the distance between them
is large. But when one considers successions of four binary encounters the
case is altered; distant encounters become so numerous that the integral
giving Jt is found to diverge. Clearly this divergence does not correspond to
any physical reality, but its removal poses a question of considerable
difficulty.*
The difficulties of a discussion of three body encounters have so far pre-
vented the derivation of exact formulae permitting a detailed comparison
with experiment. For rigid elastic spheres, results equivalent to those of
Enskog can be obtained from (16.73, 4). but o n ' v by making additional
assumptions.! SengersJ has found that, if no such assumptions are made,
• The first discussion of equation (16.73,4) w»» 8>ven by S. T. Choh and G. E. Uhlen-
beck, The Kinetic Theory of Dense Gases (Univ. of Michigan Report, 1958). See also
M. S. Green, J. Chem. Phys. »5, 836 (:9s 6 ): E- ° - D - Cohen, Physica. 18, 1025, 1045.
1060 (1962); L. S. Gircla-Colln. M. S. Green and F. Chaos, Physica, 3a, 450 (1966);
J. R. Dorfman and E. G. D. Cohen, J. Math. Phys. 8. 282 (1967)-
t J. T. O'Toole and J. S. Dahler. J. Chem. Phys. 3», 1097 (i960); •*• F- Snider and
C. F. Curtiss, Phys. Fluids, 3, 903 (i960).
t J. V. Sengera, Phys. Fluids, 9, 1333 (1966).
16.81 DENSE OASES 319
mav er
the results derived from (16.73, 4) diff appreciably from those given
by Enskog's theory, retaining in that theory only the terms linear in bp.
Curtiss and his co-workers* have discussed more general molecular
models by a method based on (16.73, 4)> but making an extra assumption
similar to that made in deducing Enskog's theory. Even after this assump-
tion, the results obtained are extremely complicated, and they are not given
here. The authors state that they give satisfactory agreement with experiment.

16.8. The evaluation of certain integrals

Certain results of integration with respect to k, assumed earlier in this
chapter, will be proved here.
Let h, i, j denote three mutually perpendicular unit vectors, h being in
the direction of g. Let 0, 9 be the polar angles of k with respect to h as axis
and the plane of h and i as initial plane; then

k = hcos0 + isin0cos<p+jsin0sin<p, (16.8, 1)

so that g. k = g cos 0, and the element of solid angle rfk can be expressed in
the form j. a ,a,
dk = sin ddOdy.
In all the integrals to be evaluated the integration is over all values of k for
which g.k is positive. Thus the limits of integration are, for 0, 0 and n/2,
and for 9, 0 and 2TT, and so all terms of the integrand containing odd powers
of sinip or cos<p may be neglected. Hence in particular

ik(g.U)dk= f j hcos0.gcos0.sin0d0df

=
7 7*' (16.8,2)
Again,

fkk(*.k)»rfk = f f(hhcos*0 + iisin*0cos»9

+j} sin* 6 sin*<p)£2 cos* 0 sin 0dOd<f

= — ^(3hh + n+jj)

^ * ( 2 h h + U) = \n-(2gg+g*U), (16.8, 3)
using (1.3, 9).
• R. F. Snider and C. F. Curtiss. Phys. Fluids. 1, 122 (1958). and 3, 903 (i960); R. F.
Snider and F. R. McCourt. Phys. Fluids, 6, 1020 (1963); D. K. Hoffmann and C. F.
Curtiss, Phys. Fluids. 7, 1887 (1964), and 8, 667 (1965).
320 DENSE OASES [16.8
Also, if v is any vector independent of 0, 9,

J kk(v.k)(f .k)*rfk = | kk(v.hcos0+v,isin0cos9

+v. j sin 0 sin?) (g. k)*dk
• IT {v.hcos^Chhcos^+iisJ^ecos1?

+jjsin*0 sin*<p)+v.i(hi + il»)cos0 sin^cos*?

+v. j(hj + jh) cos 8 sin* 8 sin* <p}g* cos* 8 sin 8d8d<f

^ { v . h(4hh+u + jj)+v. i(hi+ih)+v. j(hj + jh)}

^ » { v . h(hh+U)+v. (hh+U + jj) h

+h(hh+a+jj).v}
= —j*{v.h(hh+U)+vh+hv}

- j-2{"•g{gg+gtU)lg+g(-*g+tr')}- (16.8,4)
Now, by (16.3, 3),
C'-C+k(k.g), d-d-VJk.g).
Hence C» = (C+k(k.j)}.{C+k(k.j)}
~C* + 2(k.C)(k.g) + (k.g)\

and similarly CJ» - CJ-2(k.C 1 )(k.^)+(k.^)«.

Thus, using (16.8, 2-4),

JHC\+C'1*)g.kdk=jk{2C\-2(k.Cl)(k.g)+(k.g)*}(g.k)dk

- jl^Clg-Uzgig.C^+gH:^lg*g], (16.8, s)

Jk(Cx+Ci)*.k<fk - Jk{2C1-k(k.^)}(^.k)A

y{aC^-«2W+^U)}, (16.8,6)
16.8] DENSE OASES 321

j(C'-C)k(g.k)dk = jkk(g.k)'Jk

-^(W+M. ('6.8,7)

j(C'-C)k(k.v)(g.k)dk=(kk(v.k)(g.k)1dk

= ^fv.ffof+^/f+rfvj+jrv)}, (16.8, 8)

j(C'-C*)k(g.k)dk = jk.(g+2C)k(g.k)*dk

= 2J(k.G0)k(g.k)*dk

-^{vb-GJ+gHtJ, (16.8,9)

j(C'*-C*)kk(g.k)dk = 2j(k.C0)kk(^.k)«ik

I {(Gog) (gg +g'V)lg +g(gG0 + Go*)}-

(16.8, io)
 17
Q U A N T U M THEORY A N D
THE TRANSPORT PHENOMENA
THE QUANTUM THEORY OF MOLECULAR
COLLISIONS

17.1. The wave fields of molecules

When quantum methods are applied to molecular encounters some diver-
gence from the classical results is found. The deflection of the relative
velocity g, resulting from the encounter of a pair of molecules of masses Wj
and m,, is much the same as if, on the classical theory, each molecule were
supposed surrounded by a' wave'fieldwhose linear extension is of the order
of the'wave-length' ., .,

h being Planck's constant, equal to 6-626 xio~ng.cm.*/aec. Thus, for

example, rigid spherical molecules behave as if such wavefieldsproduce a
deflection even when the spheres do not actually collide; in general, the
effective diameters of rigid spherical molecules are larger according to
quantum theory than according to classical theory. For more realistic mole-
cular models, on the other hand, the wavefieldsmay decrease the effective
diameters. The quantum effects are appreciable if the extension of the wave
fields is not small compared with the molecular collision-distance <rlt; that
is, using a mean value for g, if
h{mJ + mi)i/2rr(T1^mlmtkT)i (>7-i» 0
is not small compared with unity. This quantity is largest for light molecules
and low temperatures; for helium at 0 °C. it is 0*13, and for hydrogen it is
somewhat larger; hence for these two gases the quantum correction is
appreciable even at ordinary temperatures, and becomes very important
at low temperatures. Molecules of other gases have much larger masses, and
their radii are in most cases larger; moreover, it is not necessary to consider
such low temperatures for the other gases, because they liquefy more readily.
In consequence, the modifications introduced by the use of quantum methods
are relatively unimportant except for hydrogen and helium.
A second modification of the classical theory may be required at very
low temperatures, when the wavefieldsassociated with the molecules may
increase the effective size of the molecules, so that there may be a state of
congestion similar to that considered in the last chapter, even when the total
volume of the molecules is small compared with that of the containing vessel.
(322]
17.3J QUANTUM THEORY 323
A gas in such a state of congestion is said to be degenerate. In ordinary gases
the congestion is very slight even at the lowest temperatures considered,
but in an electron gas, such as is encountered in a metal or in a dense star,
the congestion is extreme.
Naturally it is not possible in this book to give a full account of the relevant
sections of quantum theory; we shall simply indicate its applications to our
subject.

17.2. Interaction of two molecular streams

It is necessary first to put certain of the earlier results into forms suitable for
use in connection with the quantum theory of collisions.* The uncertainty
principle states that the precise position and velocity of a molecule cannot be
determined simultaneously. It is therefore not possible to determine exactly
the deflection x of the relative velocity at the encounter of two molecules mv
m, of velocities elt ct. It is, however, possible, when the set of molecules »4
with velocities in the range cv dcl streams through the set of molecules m,
with velocities in the range ct, dc%, to determine the probable number of
encounters per unit volume and time, such that the unit vector e' in the
direction of the relative velocity g'n after collision lies in the solid angle d€.
This number is proportional to the number-densities/, dcu / , dct of the two
sets, and depends on x an< i °n the magnitude; of the relative velocity. If it is
written as r y / \ j / j j / \
fJtg«\t(g>X)<ie dcidc, (17.2,1)
the quantity <xlt is identical with that introduced in 3.5, save that the
variables on which it depends are now; and x instead of; and (. Since in 3.5
bdbde = altde' = altsinxdxde
the earlier theory has simply to be modified by replacing bdb by
<*«(*. X) sin X<&
in all integrals involving collision variables.

17.3. The distribution of molecular deflections

We proceed to obtain an expression for a„(;, x)A Let the mutual potential
energy of a pair of molecules m„ mt at a distance r be v. Then the wave
equation governing the relative motion when a molecule m1 of velocity ct
encounters a molecule mt of velocity c, is the same as that governing the
motion of a molecule of mass mlmtl{mx + m^ and initial velocity gn under
* This modification was first msde in gas-theory by E. A. Uehling and G. E. Uhtenbeck,
Phys. Rev. 43, 552 (1933). »nd H. S. W. Massey and C. B. O. Mohr, Proc. R. Soc. A,
141,434 (1933). Similsr changes are, however, implicit in earlier discussions of electron-
theory.
f The results of this section are due to H. Faxen and J. Holtsmark, Z. Phyt. 45, 307
09*7).
324 QUANTUM THEORY AND THE (17.3
the action of a stationary centre of force whose potential energy is v at a
distance r. Suppose that the force-centre is at the origin, and that a beam of
molecules of velocity/ and of unit number-density is incident on it in the
direction of Oz. Then the number of scattered molecules crossing a large
sphere whose centre is the origin, with velocities whose directions lie in a
small solid angle e', dt', »g(^^g,x)dt' per unit time, where x is the angle
between e' and Oz.
The wave equation governing the motion is
h%m
* xn.t, J. / ¾ ¾ ^ _
otihftt
where h is Planck's constant and In this we write
lnm m
/n \ tg
(173. *)
when it reduce, to 8nhn1mtv\ .
Vf+\j* j S » T r "°" 07-3.3)
The solution we require, representing the beam of particles scattered by the
force-centre, is symmetrical about Oz andfiniteat the origin. If * — r cos 5,
one solution possessing these properties is
* - »n(r)Pn(c(*0)/r, (17.3, 4)
where P„(cos0) is the Legendre polynomial of order n, and un(r) is a real
function of r, satisfying the equation
d*un I _ B7i*mlmtv «(« + i)\ , .

and such that «m(o) » 0. These conditions determine un save for an arbitrary
factor; if this is suitably adjusted, the asymptotic expression for un corre-
sponding to large values of r is
un ~ sin (jr-\nn+8n), (17.3, 6)
where <J„ is a constant depending on the form of the function v, and on the
values of/ and n.
The solution of (17.3,3) appropriate to the present problem is a combina-
tion of solutions of the type (17.3,4), say

f -1 £ i.n.MPJcoiS), (17.3.7)
" »-0
where the quantities t)n are constants, real or complex. This may be divided
into two parts, ft' and \jr"; the first represents the incident beam, of unit
number-density, advancing parallel to Oz; for this we have
^ ' - , . * « . . ««*«<»« ( t - V - 0 . (17-3.8)
17.3] TRANSPORT PHENOMENA 325

which, as can readily be seen, is asymptotic to a solution of (17.3,3) for large r.

Using known results in the theory of Bessel functions,* this becomes

f = J o {*n+i)L«J(^JH+i(jr)Pn{«Hd)

and so, for large values of r,

t' ~ I S (2»+ i)t"8in(;V-J»»r)P„(co9<?). (17.3, 9)

/'n-0
Since ^ is a solution of (17.3, 1), and i}r' is asymptotic to a solution, the
remaining part t/r" of \jr must also be asymptotic to a solution. It represents
the scattered molecules, which at a large distance r from the origin are
moving radially outward; hence it must be such that dfi'/dr ~ ij^r" when r is
large. These conditions are satisfied if

r~l',irXv"nPn(«*d), (17.3, 10)

' n~0
where the quantities r]"n are constants.
Combining the results (17.3, 6, 7, 9, 10) we have

£ {(2«+ i)t"sinOr-i»7r)/;> + 7;e''Vr}PB(cose)

n-0
I "
= - 2 Vn »«n (Jr -\nn+Sn)Pn(coa 0),
• n-0
whence, equating the coefficients of Pn(cos0), we obtain
(2«+ i)j"sin(/r- \n7t)lj+r)"ne',r = r)„ sin {jr-\nn+Sn).
This equation is an identity in r. Suppose that
jr-\nn + 8n = o.
Then it is found that
rf-=7.(^+0(^-0.
and so (17.3, 10) becomes

V ~ S S (»•+0(«*»— 0P„(cot«).
The number-density of scattered molecules is l^'l*. At a large distance
from the origin these are moving radially outward with velocity g. Hence
the number per unit time which cross a sphere of large radius r with velocities
* G. N. Watson, Beuel Function!, pp. 56, 138, 368 (Cambridge, 1944).
 Copyrighted Material

326 QUANTUM THEORY AND THE [17.31

%
whose directions lie in a solid angle e', de' is \t/r"\ gr*de'. Since this is the
samt as galtde',
m 1

«iste.X) = 77, S ( 2 » + l)(«*"»-l)P„(C0SX) (»7-3. " )

4J n-0
This expression gives the distribution of scattered molecules when
encounters of unlike molecules are considered. If the molecules are identical,
their interchangeability requires the replacement of (17.3,8) by
p = z-i(e'»± e-'K). (i7-3»")
This corresponds to superposed beams moving with velocities £, —g parallel
to Ox; the numerical factor is chosen so as to make the average number-
density unity. The upper and lower signs in (17.3, 12) correspond re-
spectively to Bose-Einstein and Fermi-Dirac molecules, i.e. molecules
composed respectively of even and odd numbers of elementary particles
(electrons, protons and neutrons). From (17.3,12),

where the sum £' is one taken over even values of n (including zero) or odd
values, according as Bose-Einstein or Fermi-Dirac molecules are con-
sidered. We must therefore replace (17.3,11) by

«i(*a) = ^ | s ' ( w + ! ) ( « • * - i)Pw(co.*) (17-3. J3)

According to quantum theory,* molecules of the same substance cannot
be regarded as identical unless they possess the same nuclear spins and the
same rotational quantum numbers. In consequence, even in helium the
formula for unlike molecules must be used for the collisions of two mole-
cules of oppositely directed spins; and in the majority of collisions occurring
in even a simple diatomic or polyatomic gas, the scattering probability is
that for unlike molecules.

17.31. The collision-probability and mean free path

The total probability per unit time that a molecule of velocity cv moving in
a stream of unlike molecules of velocity c, whose number-density is unity,
should undergo a collision, is

gj*u(*,X)<fe' = 27¾^ a^g,x)sinxdx

- I f \i(2n+i)(*»—i)Pn{x)'dx
*J J - i | » - o

= 43? S(2«+i)sin**„ 07-3^)

J n-0
• O. Halpern and E. Gwtthmey, Phys. Rev. 53, 944 (1937).
Copyrighted Material
 Copyrighted Material

17.31] TRANSPORT PHENOMENA 327

using results in the theory of Legendre polynomials.* It is found that this
series converges (so that there is a finite probability of collision), when V
tends to zero, as r-*- co, like r*, where v > 3. Thus when thefieldsatisfies
this condition there is a natural definition of a free path in quantum theory,
whereas on the classical theory we can define a free path, when the molecules
are not rigid, only by making arbitrary assumptions as to the limits of the
molecularfields.The molecules undergo as many collisions as if they had a
definite joint collision-distance a-n and collision cross-section Q given by

Q = no*n = % £ (2«+ i)sin^ n . (17.31, 2)

J n-0
The similar relation for identical molecules involves a sum S' over either
even or odd values of n, multiplied by the factor Sn/jK
The free path corresponding to the cross-section Q is, however, very
different from that appropriate to viscosity and the other transport pheno-
mena. Even on the quantum theory the majority of the 'collisions' are
between relatively distant molecules, and result in comparatively small
deflections of the relative velocity. The small attractive forces at large
distances are found to affect the free path more than the larger repulsions at
close encounters. Hence the persistence of velocities at collision is nearly
unity; in fact the free path corresponding to (17.31, 2) is important only in
experiments in which one can study the motion of individual molecules, and
distinguish whether they undergo a deflection or not in a given distance.
Such experiments have not been carried out with uncharged molecules, but
measurements of free paths of electrons and positive ions in gases have
been made. These give values for the collision-radius of molecules which are
often many times as large as the values appropriate for viscosity, but which
vary widely with the speed of the ions.
The first approximations to the coefficients of viscosity, thermal con-
duction and diffusion depend on the integrals <f$, ¢$, now given by

^ = 27rJ (i-cos'x)a1^g,x)sinxdx

(cf. (9.33, 4)). Thus for unlike particlesf

*-£/>-> 2(2« +!)(**"»-!)P (*) dx

n-0
n

- § £ (n + i)sin*(o>n+1-«U (»7.3i. 3)
J n-0
• See Massey and Mohr, he. cit.
f The forms (17.31, 3, 4) (simpler than those originally given by Matsey and Mohr) are
due to H. A. Kramers; see J. de Boer and E. G. D. Cohen, Pkyrica, 17, 993 (1951).

Copyrighted Material
328 QUANTUM THEORY AND THE [17.32
using the properties of Legendre polynomials. Similarly
,,., \n " (»+i)(» + 2) . . . . .. . .
^iV-V £ " , — «n'(*«+»-*,)• (i7-3«.4)
For identical molecules,*
J12> 8 » „ , ( » + 1 ) ( 1 1 + 2) . . . . ,. . .
& - jr s x — ^ — w ^ - t y (17.31, s)
(the sum 2' being over even or odd values of »), with a corresponding
expression for ^f'.f
17.32. The phase-angles Sn
It is in general not possible to give an exact expression for the phases Sn.
Approximate expressions have, however, been given by MottJ and Jeffreys §
for the cases in which 8n is small or large compared with unity. It is found
that in is small if %nhnlmtvjh\ml + m^ is small compared with »(« + i)/r*
for all values of r such that j> is not small compared with n. In this case Mott
has shown by a perturbation method that

approximately. When 8H is large, we can use Jeffreys's approximation for

un, which is , fP .
« . - «n (7 +jj/(*)}»<£xj,

where Jr) -y.-^^-JUL r ( r ) - - ^ J ,

and r, is the largest zero of/(r); this approximation is valid if/'//* is small in
the range of integration. The corresponding value of <J„ is

2
4 Jr.
* Matsey »nd Mohr give an expression for the cross-section of molecules appropriate to
self-diffusion; they find that, even at high temperatures, this cross-section is twice the
classical one. This fact is. however, not of direct importance, since, on account of the
interchangeability of the molecules, the two molecules lose their identity at an encounter;
it is therefore not possible to trace the diffusion of a selected set of the molecules through
the rest, i.e. it is impossible to measure self-diffusion. The closest we can get to self-
diffusion is by considering the mutual diffusion of two sets of similar, but distinguishable,
molecules.
t For an elegant proof that, in the limit when/ is large, (17.31, 3-5) reduce to the corre-
sponding classical expressions, see L. Waldmann, Handbuch dtr Physik, vol. ta, 460
(•958).
} N. F. Mott, Prot. Comb. Phil. Soc. 35, 304 (1929).
f H. Jeffreys, Proc. Land. Math. Soc. »3, 418 (1914). In books on quantum theory
Jeffreys' expression is often known as the WKB (Wentzct-Kramers-Brillouin)
approximation.
17.4] TRANSPORT PHENOMENA 329
When £n is neither small nor large compared with unity, neither of the
approximations (17.32, 1, 2) is valid. Normally only a few values of 8n are
near unity, and these can be found with tolerable accuracy by interpolation.
More exact values can be obtained by using electronic computers.
For one model it is possible to obtain an explicit expression for Sn. Suppose
that the molecules are rigid elastic spheres of diameter cr. Then in the wave
equation (17.3, 3) we have v = 0 if r > 0 and v = 00 if r < 0. This is
equivalent to taking un as the solution of the equation

dhin /., tt(n+i)\

such that u„ = 0 if r = 0. The solution is of the form

Since for large values ofjr

•/»+ *W~ (^r)*""^-?)' 7-"-i0>)~ (4) ic08 ( ;> + 7)'

a comparison with (17.3, 6) shows that
Sn = tan-' {(-1)-+«7.+,0^)//-»0'»». ('7-32. 3)
In particular, S„ = —jo.
For more realistic models, numerical methods have to be used throughout.
The values of Sn, and hence of the integrals cV>, have first to be found for a
variety of values of/, i.e. of g (cf. (17.3, 2)); the integrals OPty) are then
determined by numerical integration.
17.4. Comparison with experiment for helium
As noted in 17.1, quantum corrections to the classical formulae for the
transport coefficients are unimportant for gases other than hydrogen and
helium. Detailed calculations by de Boer and Bird* show that for other gases
the corrections do not in general exceed the experimental errors. Even for
hydrogen and helium, Mason and Ricef estimate that the corrections to the
viscosity do not exceed o-6 per cent for helium above 2000 K. and for
hydrogen above 2500 K.
Massey and MohrJ compared theory with experiment for helium and
hydrogen, supposing the molecules to be rigid elastic spheres. As the general
considerations of 17.1 suggest, the effect of quantum interaction is to increase
the apparent radius of the molecules; the increase is fairly small at ordinary
temperatures, but becomes large as the temperature decreases; as the
• J. de Boer and R. B. Bird, Phyrica, ao, 185 (1954).
t E. A. Mtson and W. E. Rice, J. Chem. Phyi. 23, s " 0954)-
t H. S. W. Maaaey and C. B. O. Mohr. Proc. R. Soc. A, 141, 434 (1933).
 330 QUANTUM THEORY AND THE [17.4

temperature approaches the absolute zero, the apparent radius for viscosity
is four times the actual radius. In Table 29 values calculated by the classical
and quantum theories for rigid elastic spheres with suitably chosen radii are
compared with the experimental values for helium.* The improved agree-
ment with experiment at low temperatures as compared with classical theory
is striking. However, this cannot be regarded as justifying the use of the
rigid spherical model; the decrease in apparent molecular radius inferred
from experiment on classical theory continues to high temperatures, where
quantum effects are inappreciable.
Several alternative comparisons of quantum results with experiment have
been made for helium, using the 12,6 and exp;6 models, and the full
Buckingham (exp;8,6) model, f These refer both to normal helium (*He)
and to the 3He isotope. On classical theory, assuming the same force-law for
molecular interactions in the two, the viscosity of 3He should be VO/4)
times that of 'He, the molecular masses being in the ratio 3/4. Quantum
theory predicts deviations from this simple law, both because the smaller
mass makes quantum effects in 'He more important than in 'He, and
because 'He and 'He are respectively Fermi-Dirac and Bose-Einstein
particles.
Let at and a„ denote the values of a, given by (17.3,13), when the sum is
taken over even and odd values of n respectively. Since 'He consists of
Bose-Einstein molecules without nuclear spin, all its molecules are identical,
and for it at =• a,. On the other hand, 'He consists of Fermi-Dirac molecules
with nuclear spin ± $, and half of the encounters in it are between molecules
of opposite spin; for these (17.3. 11) must be used, corresponding to
a
i ~ i(««+ao)- Hence the average value of a.l for all types of encounter in
8Hc
" *i = ««.+««.+«.)} = K + K ('7-4. 0
with a corresponding result for the effective molecular cross-section.
The calculated viscosities of *He and 'He at very low temperatures are
compared with the experimental values in Fig. 13 and in the last column of
Table 29. The calculated values are those of Keller J based on the 12,6
model, using values of the constants e, <r derived by the virial method for
more normal temperatures. Similar results are obtained using the exp; 6 and
Buckingham models. As Fig. 13 shows, the agreement with experiment is
• E. A. Uehling {.Phyt. Rev. 46, 917 (1934)) recalculated theae value* and found a con-
siderable divergence from the results of Massey and Mohr.
t H. S. W. Maasey and C. B. O. Mohr, Proe. R. Soc. A, 144. 188 (1934); R- A. Bucking-
ham, J. Hamilton and H. S. W. Masaey, Prof. R. Soc. A. 179. 103 (1041); J. de Boer,
Phyiica, 10, 348 (1943); J. de Boer and E. G. D. Cohen, Phyrica, 17, 993 (1951);
O. Halpern and R. A. Buckingham, Phyt. Rev. 98,1616 (1953); W. E. Keller, Phyi. Rev.
105, 41 (1956).
t W. E. Keller, Phyt. Rev. 105, 41 (1956). The value) in the last column of Table 19 are
obtained from Keller's tables by interpolation. At 40 °K., the highest temperature
considered by Keller, the quantum values of the viscosity exceed the clstiicsl by less
than 4 per cent
17.41 TRANSPORT PHENOMENA 331

Table 29. Viscosity of helium on the quantum theory

H x 10* (calculated)

Rigid elastic spheres 12,6 model

t * (Keller,
Absolute Classical Quantum quantum
temperature H x 1 0 ' (ob«.) theory theory theory)

»945 "994 2000 1850 —

1870
»73 •> 1930 1770 —
»50-3 1788 1840 1600 —
203-1 1670 1500
1564 —
1520
»705
898
139*
9.8 • 100
1350
920
—
—
751 8l5 1010 815 —
20'2 35°-3 520 3S5 354
1704 3«o ( 0 — —
300
317
291
'5 2946 450
1412 280 (1) 279-5
42
— — 108-6
1078 (2) — —
2-68 66-0 (2) — — 66-0
1-29 34-0 (2) 370
— —
Experimental references: (1) E. W. Becker and R. Misenta, Z. Phyt. 140, 535 (1935);
(2) E. W. Becker. R. Miienta and F. Schmeissner, Z. Phys. 137, 126 (19S4); the other
values are, as in Table 13, from H. K. Onnes and S. Weber, Proc. Stct. Set. K. tud.
Akad. Wtt. 21, 1385 (1913)-

16

1-4 -

-5:1-2
&
1? 10
-

08

06

0-4
02 0-4 0-6 0-8 10 1-2 1-4 1-6
togr
Fig. 13. Keller's quantum calculations of /t for 'He and 'He, based on the 12,6 model,
compared with observation. Continuous curves, theory; x , A, observed value* for *He
and •He respectively; +, values calculated for 'He from obaerved valuea of A, uaing
A = tfcf. Observed values of /i from Becker and Misenta and from Becker, Miaenta and
Schmeissner; observed values of A from J. B. Ubbink and W. J. de Haas, Phytiea, xo,
465 (1943)-
332 QUANTUM THEORY AND THE (17.41
remarkably good considering the wide temperature-range over which the
force-law is extrapolated. Contrary to classical theory, the viscosity of 'He
actually exceeds that of 'He at very low temperatures. Close agreement
between theory and experiment can be obtained for 'He only when the
combination rule (17.4, 1) is used in determining the effective cross-section.
A comparison of theory and experiment for the coefficient of diffusion in
mixtures of 'He and 4He at low temperatures has been made by Bendt,*
likewise with satisfactory results.

17.41. Hydrogen at low temperatures

The theory of this chapter, though strictly applicable only to spherical
molecules, may be expected also to apply moderately well to hydrogen at
low temperatures, at which the rotational degrees of freedom are not excited
(cf. 17.63). However, hydrogen exists in two forms, para- and ortho-
hydrogen, which have to be regarded as independent gases; in normal
hydrogen these are present in the ratio 3:1. A hydrogen molecule is a
Bose-Einstein particle; however, because of nuclear spins orthohydrogen
has nine independent quantum states, even at low temperatures. In conse-
quence the relation, similar to (17.4, 1), giving the mean value of a t for
orthohydrogen is _ _ R + j [ 1 ( a < + a < ( ) ] . K + K . (l?.4I> l}

Parahydrogen has only one quantum state at low temperatures, and a, — a,

in it; for encounters between molecules of different modifications,
*n = *(««+«<.)•
The quantities at, a, may conceivably differ for para- and orthohydrogen,
but this possibility is not normally considered.
Calculated viscosities for para- and orthohydrogen, and for normal
hydrogen, regarded as a binary mixture, show satisfactory agreement with
experiment at low temperatures.t The calculations were made on the 12,6
and Buckingham models, using values of the force-constants derived from
experiments at more normal temperatures. Theory predicts that, because of
the difference between the quantum values of Si, parahydrogen should
possess a rather greater viscosity and thermal conductivity than ortho-
hydrogen, the difference increasing to about 1 per cent near T = 10 °K.
This is in rough agreement with experiment,^ though part of the experi-
mental difference may be due to other effects.
A comparison of H, with HD and D, is not straightforward, because the
HD molecule, unlike the others, is a Fermi-Dirac particle; also para- and
• P. J. Bendt, Phyt. Rtv. n o , 85 (1958).
t E. G. D. Cohen, M. J. Ofierhaua, J. M. J. vui Leeuwen. B. W. Root and J. de Boer,
Phytita, as, 791 (1956); R. A. Buckingham, A. R. Daviet tnd D. C. Gillet, Proc. Phyt.
Soe. 71. 457 (1958).
X E. W. Becker tnd O. Steht, Z. Phyt. 133, 615 (195a); K. Heinzinger, A. Klemm and
L. Wtldmann, Z. Natwrf. 16a, 1338 (1961).
17.5J TRANSPORT PHENOMENA 333
orthodeuterium, differing from the corresponding H , modifications in
nuclear spin, have in place of (17.41, 1)
«i = rs<*«r + T*«o (ortho), a, = K + K (para). (17.41, 2)
We may note, however, that on classical theory, if the force-law is the same
for D , and H„ their viscosities should be in the ratio ,/2, or 1-414. In fact
the ratio, though not far from ^2 at ordinary temperatures, falls to 1-37 at
71-5 °K., and 125 at 14-4 °K.;* this can fairly be attributed to quantum
effects.
17.5. Degeneracy for Ferml-Dirac particles
As noted in 17.1, a state of congestion results when the mean distance
between neighbouring molecules is comparable with the size of the
quantum wave fields with which molecules are surrounded. Consider,
for example, a gas composed of electrons or other Fermi-Dirac particles.
According to quantum theory, not more than Vfidc(m/h)* such particles in
a volume V can possess velocities in the range dc, where m and fi are
the mass and the 'statistical weight' of a particle,f and h is Planck's
constant. Hence if in the velocity-range there is already this number of
particles, and the result of a collision would be that a further particle would
enter the range, the collision cannot happen. More generally, if Vf(c)dc
particles have velocities in this range, the probability of a collision which
would result in a particle entering this range is reduced in the ratio
1 —f(c)h3lm3fl. Accordingly (3.52, 9) is to be replaced by

^g^g<X)^xdxdedct. (17.5,1)
The uniform steady state for a gas-mixture is given by

w(-$a(-s*)-w(-&)(-®
and similar equations, whence it follows that

'"('•/(-$,))• *{'7(-iS)
• A. O. Rietveld, A. van Itterbeek and C. A. Velds, Physica, a$, 205 (1959). A small
difference in the molecular force-law between D , and H , is probable; see, e.g., D. E.
Diller and E. A. Mason, J. Chem. Phyt. 44, 2604 (1966).
t T h e statistical weight is the number of independent quantum states in which the
particle can possess the same internal energy: for an electron it is 2, corresponding
to the two possible values of the spin. According to quantum mechanics, not more than
one electron in an atom can occupy a given quantum state. It is also a fundamental
assumption of statistical mechanics that the a priori probability that an electron should
occupy any one quantum state is the same as the probability that it should occupy a
volume h* of the six-dimensional space whose coordinates are the space-coordinates and
components of momentum of the electron. T h u s not more than one electron with a
given spin, and not more than two in all, can occupy the volume A*. T h e result is
generalized to apply to all Fermi-Dirac particles.
334 QUANTUM THEORY AND THE [17.5
are summational invariants for encounters. This gives
/ I - « f A / * , ( i + i i i « - » c l ) , / . - m i / W i + ^ e " - 0 * ) , (17.5.2)
where Av At, a are related to the number-densities and temperature of the
gases, and Cv C, are, as before, the velocities of the particles relative to the
mean motion. Large values of AU At correspond, for a given a, to small
number-densities of the gas, and vice versa.
If Alt At are large, (17.5, 2) reduces to the usual (Maxwellian) form

h
h»At' * J* k>A, '
««d so a - -^,. (17.5, 3)
If the velocity-distribution of one of the gases is independent of the density
of the other, as is true in the classical theory, a will have the same value if
Ax is large and At is not large, or, again, by the same argument, if neither AY
nor At is large. Statistical mechanics shows that this is actually the case:
thus (17.5, 3) is valid for all densities.
The relations connecting the quantities Alt At with the corresponding
number-densities nx, n, are of the same form as for a simple gas, whose
velocity-distribution function is
/ - «V/A»(i +Ae'»*l»T). (17.5, 4)
If A is large compared with unity,/approximates to the Maxwellian form
with n given by m*fi I m \i
TfiA"",\^kT) •
Hence the condition (A > 1) that there is no appreciable departure from the
Maxwellian distribution implies that
hhtlftznmkT)* < 1. (17.5, s)
When this condition is satisfied, the gas is said to be non-degenerate.
If A is small compared with unity,/is nearly equal to m*/?/A* when C is
less than the quantity w defined by
mwx . (i\
2A7- = ,n
W'
but falls off very rapidly as C increases beyond this value. In this case the gas
is said to be degenerate. To a first approximation we can write
/ - ^ if C<w,
/-o if C>w.
This gives n - -r- | 4irC*dC

(I7S 6)
T"F"l m )' '
17.511 TRANSPORT PHENOMENA 335
Also, using the same approximation, we find

5 A» \ m )

-,&r> <•»•'>
Hence the energy of the molecules is, to this approximation, independent of
the temperature. The pressure of the gas, which is given by

is also, to this approximation, independent of the temperature.

Using the fact that A is small compared with unity, in conjunction with
equation (17.5, 6), we obtain as the condition for degeneracy
Pn/fanrnkT)* > I. (17.5, 9)
It follows from this condition that the mean translational energy of the
molecules, and the pressure, are large compared with the values 3A/2W and
knT which they would have in the absence of degeneracy. The relations
(17.5, 5) and (17.5, 9) constitute Sommerfeld's degeneracy criterion.* It is
clear from them that degeneracy is most probable if m is small: thus it is
most likely to occur in an electron gas. For other gases it is most likely to
occur at great densities and low temperatures. It is found, however, that
even for the lightest gases at temperatures as low as 15 °K. the congestion
produced by degeneracy is less than that due to the finite size of the mole-
cules, considered in Chapter 16, and so the effect of degeneracy can normally
be neglected, save for an electron-gas (cf. p. 336, second footnote).f
17.51. Degeneracy for Bose-Einstein particles
Rather different results apply to Bose-Einstein particles. For these the
presence of a like particle in the velocity-range dc increases the probability
that a particle will enter that range; the presence of/(c) dc particles per unit
volume increases this probability in the ratio 1 +/(c) h?lm?fi. The analysis in
this case is similar to that of 17.5, save that the sign of P/m3/? is changed: in
the uniform steady state it is found that
f=trflfi/^(AemC,i*kT-1). (17.51, 1)
The constant A, which is related to the number-density, is in this case
always greater than unity; the limiting case of extreme degeneracy is
approached as A -*• 1. The molecular energy and pressure corresponding to
(17.51, 1) are less than those given by the classical formulae. It is, however,
unnecessary to go into details, since the results for ordinary gases are not
appreciably affected by degeneracy.
* A Sommerfetd, Z. Phyt. 47, 1, 43 (1928).
t In « dense star, because of the increased electron energies due to degeneracy,
relativistic effects often are important.
336 QUANTUM THEORY AND THE [17.6

17.52. Transport phenomena in a degenerate gas

The equation (17.5,1) and the corresponding equation for Bose-Einstein
particles were used by Uehling and Uhlenbeck* and later authors to
construct a theory of the transport phenomena in a degenerate gas. However,
(17.5,1) is at most only approximately valid; it is based on the assumption of
binary encounters, which is not valid in a degenerate gas. In fact, since the
wave-length h/rrmg of the quantum wave field surrounding a pair of mole-
cules is comparable with hJ2it(mkT)^t the condition (17.3.9) for degeneracy
implies that this wave-length considerably exceeds the mean distance n~l
between neighbouring particles.
Transport phenomena in ordinary gases are not appreciably affected by
degeneracy, even at low temperatures, f The situation is different for an
electron-gas; however, an exact discussion for such a gas needs to take into
account the interaction of electron-waves with the whole of the neighbouring
particles, rather than binary encounters. An elementary discussion in terms
of free path ideas was given by Sommerfeld;J he was able to account for
some of the anomalies encountered in the classical theory of an electron-gas.
However, a full discussion of (say) the electron-gas in a metal demands far
more sophisticated concepts than those embodied in (17.5, 1).

INTERNAL ENERGY

17.6. Quantized Internal energies

In discussing molecules with internal energy, a simple semi-classical
approach is here employed. The translatory motions are described classically,
the rotations and internal vibrations in terms of quantum states: for sim-
plicity these states are assumed to be non-degenerate, so that to each energy
level corresponds one and only one quantum state. A correspondence
principle is invoked to derive quantum results from the corresponding
classical ones. § Such an approach is simpler than a full quantum treatment
but it is incomplete; in particular, it is unable to take into account the
* E. A. Uehling ind G. E. Uhlenbeck, Phyt. Rev. 43, 552 (1933).
t E. A. Uehling, Phyt. Rev. 46, 917 (1934). The onaet of degeneracy ihould be accom-
panied by a considerable variation of viscosity with pressure. The observed variation for
helium at very low temperatures is much more modest—only i per cent per atmosphere
(J. M. ). Coremans, J. J. M. Beenakker, A. van Itterbeek and P. Zandbergen, Bull.
Inst. Int. Froid, Annexe 1958-1, p. 289).
t A. Sommerfeld, Z. Phys. 47, 1, 43 (1928).
I The approach is essentially that of C. S. Wang-Chang and G. E. Uhlenbeck, Univ. of
Michigan report CM-681, 1951; see also C.S.Wang-Chang, G. E. Uhlenbeck and
J. de Boer, in Studies in Statistical Mechanics (ed. J. de Boer and G. E. Uhlenbeck), vol. 2,
249 (North-Holland, 1964). Its defects have been discussed by L. Waldmann. who
has attempted a more exact discussion in a special case (Handbuch ier Physik, vol. 12,
469-93 (Springer, 1958); Proc. Int. Seminar on Transport Properties of Gases,
pp. 59-80 (Brown University, Providence, R.I., 1964).
 Copyrighted Material

17.61] TRANSPORT PHENOMENA 337

correlation between linear and angular momentum indicated by classical
theory (in the second and third terms on the right of (11.41, 3), etc.).
The state of a molecule m, is specified by the velocity c, of its mass-centre,
and by a set of internal quantum numbers denoted collectively by Kt.
A separate velocity-distribution function/,(c„ K„ r, t) has to be defined for
molecules in each of the quantum states K,\ it is, as usual, such that the
probable number of molecules m, in the state K, at time t, with positions
and velocities in the ranges r, dr and c„ dc„ is f,(c„K„r,t)drdcr Each
internal state Kt corresponds to an elementary volume A*-8 of the internal
phase-space of Chapter 11 (R-3 is the number of internal coordinates);*
an integration over the internal phase-space is to be replaced by a sum over
all values of Kt. For brevity,/,(c,, K„ r, t) is written/.^; likewise 2 ¾ is used
to denote the internal energy of a molecule m, in the state Kt (measured from
the state of least internal energy as zero).

17.61. Encounter probabilities

An encounter between a molecule mt and a molecule mt may alter their
internal states as well as their translatory velocities. Let / ¾ . . £$<, g'u and
JEJJ^, £ $ , gu denote the internal energies and relative velocities before and
after the encounter; we write g'^ = g't\ gu = ge, where e, e' are unit vectors.
From the conservation of energy relative to the mass-centre of the two
molecules, 1 m nu
(l7 6l>l)
1 ^ ^ ^ ^ = ^% -
where A(£(<)) denotes the increase of internal energy at the encounter, i.e.
A(£W) « £ & + « 8 ! - . E i £ - 3 2 - (17-61, 2)
When the values of K„ Lt, K't, L't are assigned, equation (17.61, 1) gives
either of g and g' in terms of the other. In order to specify an encounter
completely, one needs to know the initial states c'„ K't and c'„ L',, the final
internal states K„ 1¾. and the final direction e (or alternatively c„ K„ ct, Lt
K'„ L't and e').
The number of encounters per unit volume and time between pairs of
molecules m„ mt in the sets c'„ dc'„ K', and c't, dc't, L\ which result in the
molecules entering the internal states K„ Lt, and having the direction e of
the final relative velocity in the solid angle de, is proportional to de and to
the numbers f'lK-dc', and f\i;dc\ of molecules involved. We write it as
farftv ^I(sl8fs)g^dc',de't. (17.61,3)
In this af:'£'de, like the earlier altde', denotes an elementary collision cross-
section. If G is the velocity of the mass-centre of the pair of molecules, by
(3-52. S). gdedc',dc't = gdeJGdg'u = gdedGg'*dg'de'
= g'dt'dGdgu = g'de'dc,dct, (17.61,4)
• The function /,(c„ JC„ r, t) does not correspond exactly to the function /^Q r P„ t)
of Chapter 11. but to **-*/,«&, J»„ ()•

Copyrighted Material
338 Q U A N T U M THEORY A N D THE (17.62
since, by (17.61, i),g'dg' - gdg. Hence (17.61, 3) can also be written
f.KfU.^t^Bu)g'dt:,de.dct. (17.61, 5)

17.62. The Boltzmann equation

The state of a molecule m, is assumed to change only at binary encounters.
Then/,jf satisfies a Boltzmann equation of the type

dt + < V dr + # V de. dt • K7
*' '
where djdt denotes a rate of change due to encounters.
The total number of molecules m, entering the velocity-range c„ dc, and
the internal state K, as a result of encounters per unit volume and time is, by
(I? 61 5>
* ' ' dc.Y.222 f\f.K•r,L^'£Xg'u,gu)g'<l*'det. (17.62,2)
t K', LjL, J J
In this, c'„ e't are determined, for given c„ et, e', from the vectors G, g\,\
G is known directly from c„c«, and f^is;'e', where;' is given by (17.61,1).
Similarly the total number of molecules m, leaving the state c„ dc„ K, per
unit volume and time is (cf. (17.61, 3))
dc, S S S S ff/«/, t afffig^g^g'd^det, (17.62, 3)
in which ATJ, Z,J, f J„ ;", e" refer to the state of the molecules after encounter.
Taking the difference of (17.62, 2) and (17.62, 3), and dividing by dc„ we get

^ - 2 2 ^ 2 2 \r.KrtL°$Etf»gu)g'M
- 2 2 (uftLtfh.igtogDg'to'W (17.62,4)
x'.u'J '
Equation (17.62, 4) differs in form from the classical expression for
Bf/Jdt, which involves only the combination/,'/}—fjt. For (17.62, 4) to be
reducible to the classical form it is necessary that
2 2 {*&L<gu,gl)g"d*' » 2 2 te'tfctt.te'de'. (17.62, s)
This condition can be shown to be satisfied;* hence
^ - 2 2 2 2 fftfKftL-frtfti.)**fg'dt'dct. (17.62,6)
* In order to justify (17.6a, 5) by the correspondence principle, one seeka the quantum
equivalent of (n.xi, 1), i.e.
gdtb'db'dt'drdc',dc',dU',da.', « gdtbdbd*drdc,dc,dCl,dCl,.
This equation implies that, for etch phase-element occupied by molecules before
encounters occurring in dr during dt, and carrying them into ranges dc,JSl, and dc,dSl„
there it an equal phase-dement occupied by molecules before encounters which carry
them out of these ranges. In replacing the internal phase elements by quantum states,
17.63] TRANSPORT PHENOMENA 339

17.63. The uniform steady state

The derivation of the velocity-distribution function in the uniform steady
state follows much the same lines as in 11.3 and 11.32. One finds that lnf,K
is a summational invariant; it follows that (cf. (11.33,1))
f.K = «.Z;1exp(-E,K/kT) (17-63.1)
= «.Z,- 1 exp{-(Jm,Cl + ^ ) / * r } , (17.63, 2)
where E,K is the total heat-energy of a molecule (energy relative to the mass-
velocity c 0 ). Equation (17.63, 1) expresses the quantum form of the Boltz-
mann distribution. The equation giving the partition function Z, is now

Z,= z(cxp(-E,K/kT)dc,
K,J

= f e x p ( - m , C ; / 2 * 7 V c , x Z e x p ( - £ & / * 7")
J K.
= (2;rA7ym.)IZ<", (17-63.3)
where Z<« = £ exp( - E $ / k T ) . (17-63, 4)

The sum ZJ*> is a sum over all possible internal states. To a crude first
approximation, £J& can often be divided into independent parts
£ & = E,R + E„,+E,r.+.... (17-63. 5)
where E,R is the energy of rotation and E,r, E,r are the energies of the
various internal vibrations. To the same approximation,
Z." = ( £ ¢ - ^ ^ ) ( 2 ^ ^ ) ( - - - )
F. V,
sZ,RZ,y.... (1763.6)
For actual molecules, the rotational states are degenerate, i.e. there is a
number w,R of such states all possessing the same rotational energy E,R. This
degeneracy is the quantum equivalent of the fact that in classical mechanics
the angular momentum can vary in direction as well as magnitude. The
analysis of the preceding sections is strictly not applicable to such degenerate
states. However, the formulae of this section are valid even when there is
degeneracy; one may then write
Z.R=Sw.R«-J5'"*r, (17-63.7)

dSl', and XI, have to be replaced by equal powera of h, and similarly for dtl\ and dtlf.
Maog'b'db de' and gbdbd* correspond to
S E<*jr?'(jk #«.)*•* «nd S EaJrVC*!..*!)**"*''
V Li' ••" V
the summations being introduced to allow for the microscopic indeterminacy inherent
in quantum mechanics. Hence, on using (17.61, 4), and integrating over all <fo' and <fo",
we recover equation (17.62, 5).
340 QUANTUM THEORY AND THE (17.63
the sum being over all possible rotational energies E,„; there is a similar
formula for Z<". The number w,H is the statistical weight attached to the
rotational energy E,R.
The mean thermal energy of a molecule satisfies (11.33, 5)> ' e -

E.-kT*±(\nZ.)
- }kT+(ZJ'>)-» S £Ji exp(-B&lkT). (17.63,8)

In the approximation (17.63, 5) this becomes

E,-ikT+(Z.H)-*XE.ne-W
«.
+(Z.^)- 1 2 E„ e-W +.... (17.63, 9)
v.
The first term on the right of (17.63, 9) gives the mean translational energy;
the second and later terms give the mean energies of rotation and of the
various internal vibrations.
If any of the sums in (17.63, 9) contains a moderately large number of
terms such that the partial energy (E,n or E,y, etc.) is smaller than or compar-
able with k T, the sum approximates to an integral, and the corresponding con-
tribution to E, approximates to the equipartition value found in Chapter 11.
However, as can readily be shown, the contributions of any of the sums
to E„ and to the specific heat per molecule dEjdT, are both small when kT
is small compared with all the partial excitation energies (the non-zero
values of EtRt E,y, etc.) appearing in that sum. In this case the degrees of
freedom corresponding to that sum are effectively not excited.
For example, the possible excitation energies of molecular rotation are
multiples of k*l^n*I, where I is the moment of inertia involved; thus
equipartition obtains if ^nlIkTjh* is large, but the molecules effectively do
not rotate if it is small. Moments of inertia arising from the mass of electrons
alone are very small; because of this, molecular rotations in a monatomic gas
are not excited, and the same is true of the rotation of a diatomic molecule
about the line joining the two atomic nuclei (cf. 11.34). For molecules of
H, rotating about a line perpendicular to that joining the nuclei, I is such
that the rotational contribution to the specific heat per molecule has nearly
its equipartition value A when T > 300 °K. but is negligible when T < 50 °K.
For the hydrogen isotopes HD and D„ equipartition prevails above about
i5o"K. For other diatomic and polyatomic gases I is much larger, and at
all significant temperatures the rotational specific heat per molecule has
nearly its equipartition value (k for diatomic gases, \k for polyatomic).
However, equipartition is normally far from attainment for internal
vibrations. When it comes near to attainment, the internal motions are
becoming so violent that dissociation of the molecules is not far off. The
17.64] TRANSPORT PHENOMENA 341
vibrational contribution to the specific heat is of most importance for
polyatomic gases, e.g. carbon dioxide; the only common diatomic gas for
which it is important at ordinary temperatures is chlorine.

17.64. Internal energy and the transport phenomena

The quantum discussion for a non-uniform gas also follows much the same
lines as the classical discussion in Chapter n . The velocity-distribution
function f,K is, to a second approximation, given by
/•*-/»(!+*»). (17-64.0
where/1¾ is the Boltzmann distribution function of (17.63, 1), and 0 ¾ is
small. In a binary mixture <t>^, <t>& satisfy (cf. 11.4, 8)

-^• o 0-/»{xr , c 1 .* 1 +(^-/ l -i)Q.vinr

+ a « r « , : V c b + [ K » ! - | ) - ^ ( / . - ¾ ] V.e,J (17.64. 2)

and a similar equation. In these, as before,

*i = CAmJzkT)*, SXK = E1K/kT,
and $Nk is the specific heat per molecule; also

(1764.3)
The expression for ¢^¾ is of the form (cf. (11.41,1))
* t t = -AlK.VlnT-DlK.dlt-2*lK:Vc0-2B1KV.c0. (17.64,4)
From these relations the equations for JIK(.A), /12(D), etc., can readily be
obtained.
First approximations to the transport coefficients are deduced as in
Chapter 11. For example, a first approximation to AE for a simple gas is
taken to be ^ _^ - f t C + ^ - ^ G (17.64. 5)
where the coefficients av at are independent of the internal quantum state K;
they are determined from equations derived by multiplying the equation
for JK(A) in turn by («"* - f) C and (#$ - 4™) C, summing over all K, and
integrating with respect to c. Similar modifications in the argument of
Chapter 11 are required in discussing the quantities A, B, B.
The final formulae for ft, A, w and Dlt derived in this manner have the
same form as those of Chapter 11, in terms of quantities au, an, etc.,
differing only trivially from those of that chapter. The difference is essentially
that the element gbdbde is replaced by g'a%'£'dc', and the integrations over
342 QUANTUM THEORY [17.64
the internal phase spaces are replaced by summations; also the value of Z is
now given by (17.63, 4). Thus, for example, the a of (11.5,13) becomes

(17.64, 6)
The changes in (11.5, 14, 15), and (11.52, 5) are similar.
The general conclusions regarding volume viscosity and heat conduction
remain as in Chapter 11; the only serious modification is through the
quantum effect on the specific heat. In particular, the discussion of the
modified Eucken theory and of the Mason-Monchick formula (cf. 11.8,
11.81) remains unaffected.*
Once more it is necessary to emphasize that the theory outlined above is
strictly not applicable to molecular rotations, because it ignores the
degeneracy of the rotational states. As Waldmannf has stressed, the theory
is in consequence unable to take proper account of the vector properties of
the molecular angular momentum. The quantum equivalent of these vector
properties is that a molecule cannot be regarded as in a pure state, its
internal wave function between collisions being a linear combination of all
the wave functions corresponding to its rotational energy and total scalar
angular momentum. For a full discussion involving the vector properties of
the angular momentum it is not sufficient simply to introduce cross-sections
«§'£' for transitions between pure states.
The vector properties of the angular momentum are represented only by
terms like those involving h A C and h A {h A C) in (11.41,3). Since the ex-
pression finally assumed for A, in (11.5,4) as well as in equation (17.64, 5),
includes no terms of these types, one may expect that the first approxima-
tions of this section will not be affected by the neglect of degeneracy.
However, as noted in 11.61, the neglect of such terms in higher approxima-
tions may lead to appreciable errors in the transport coefficients.
In principle it should be possible to complete the evaluation of the trans-
port coefficients by a quantum determination of a$fc', followed by an
evaluation of integrals like (17.64, 6). In practice this has proved very
laborious, and no detailed quantum discussion has so far been given; however,
some progress has been made with the theory of diatomic molecules.^
* The work of Maton and Monchick was, in fact, originally baaed on the formulae of
Wang-Chang and Uhlenbeck.
t L. Waldmann, Proc. Int. Seminar on Transport Properties of Gates, pp. 59-80 (Brown
University, Providence, R.I.. 1964).
X A. M. Arthurs and A. Dalgarno, Proc. R. Soe. A, 356, 540 (i960); R. B. Bernstein,
A. Dalgarno, H. S. W. Maasey and I. C. Percival, Proc. R. Soe. A, 374, 427 (1963).
See also K. Takayanagi, Prog. Theor. Phyt., Osaka, II, 557 (1954).
 Copyrighted Material

18
MULTIPLE GAS MIXTURES

18.1. Mixtures of several constituents

The methods of Chapter 8 can be generalized to apply to a mixture of three
or more constituent gases.* Equations of type (8.i, 1,3, 4, 7, 9) hold for
each constituent, and the equations of continuity, momentum and energy
of the gas as a whole are given by (8.1, 5, 6, 8, 10), with the obvious modifica-
tions required to take account of the additional constituents.
Enskog's method of solution of the Boltzmann equation is evidently still
applicable. We discuss only the second approximation to the velocity-
distribution function. Formal results are given for a ^-constituent mixture;
detailed applications are, however, limited to the case K = 3.

18.2. The second approximation

The first approximation to/, is the Maxwellian function

The second approximation is fa°\i + <!>,), where ¢, satisfies the equation

a /(0) a/-(0) a/10)
M . +C . ^ . + F , . ^ - = -Siyi^O.+O,) (18.2, 2)

(cf. (8.3,1,2)). In the sum on the right of (18.2, 2) it is to be understood that

IJP*+®t) reduces to /,(*,) when t = s.
The left-hand side of (18.2, 2) can be transformed as in 8.3, using
generalized forms of (8.21, 2, 4, 6). The result (cf. (8.3, 8, 9)) is

/ H C . K - ^ + C B l - t ) V l n r ) + 2ff>,:Vc 0 ], (18.3, 3)
where xg = njn, and

*-J{VA-A(*.-T5?)} < I8 - 2 '4)

• For early discussions, see E. J. Hellund, Phys. Rev. 57, 319, 328, 737, 743 (1940);
L. Waldmann, Z. Phys. 134, 175 (1947); C. F. Curtiss and J. O. Hirschfelder, J. Chan.
Phys. 17, 550 (1949). An elegant account of the theory, given by L. Waldmann in
Handbuck der Physik, vol. 13, 402-26 (1958), is largely followed here.
[343]
Copyrighted Material
344 MULTIPLE GAS MIXTURES [18.2
This may also be written (cf. (8.3,12))
d. - Vx.+(x.-pJp)Vlnp-(pJpp){pF,-'ZptFt). (18.2,6)
Clearly, from (18.2, 5 or 6) Srf.-o. (18.2,7)

From (18.2,4), if the gas is at rest d, reduces to CVp,-p,F,)/p. However, in

general (18.2,7) does not imply any relation between the vectors Vp,-/>,F„
which are linearly independent.
As in 8.31, the functions O, must be linear in V In T, Vc0 and the vectors
dt. We write
¢ , - - * , . V l n 7'-2B,:Vc0-2Z%,>.</„ (18.2, 8)

where A„ B„ /¾ 0 are respectively of the forms

C.AJC,), C?C,BJiC.), C.DjC.).
f)
In view of (18.2, 7), the vectors D^ (for given t) possess a degree of in-
determinacy; this is removed by imposing the further condition
Sp,0i«> - o. (18.2,9)
Then, from (18.2,4),
SDi".rf« -p-'XWiVpt-PtF,),
so that (18.2, 8) becomes
4», - -A..V\n T-2B.:Vc9-p-W>.<ypt-PtFt). (18.2.10)
1
The equations satisfied by A„ B, and Di" are found by substituting from
(18.2, 3, 10) into (18.2, 2) and equating coefficients of Vln7\ Vc0 and
Vpt -ptFt. This gives
^.^1^.+At) -f?X*\-\)C„ (18.2,11)
1
S».»t^(B.+ Bt) -/.<•>*>„ (18.2,12)
1

xn.^uw+w1) - x.-1(*„-pjp)/i»>c„ (18.2,13)

where tn is the Kronecker delta. Equations (18.2,13) and (18.2,9) are
consistent, as can be seen by multiplying (18.2,13) by p„ and summing over
allu.

18.3. Diffusion
The velocity of diffusion C, of the rth constituent relative to the mass
motion is given by „,C,-jffV.C.de,. (18.3,1)
18.3] MULTIPLE GAS MIXTURES 345
On substituting for ¢ , from (18.2, 10) this becomes
C, = - S**rH?Pt -PtPt) - DT,V In T, (18.3, 2)
1

where **-±{f?>lKu.C.4cm DT, = ±-ff,<»A,.C,dc,. (18.3,3)

Here DT, is a thermal diffusion coefficient for therthconstituent; AK is a
generalized diffusion coefficient for the 5th constituent, measuring the
' induced' diffusion due to the forces acting on the rth constituent.*
The diffusion coefficients are not all independent. From (18.3, 3) and
,8 2
( - '9)> 2> t A„ = o. (18.3,4)
1
Again, the solutions of equations (18.2, 11, 13) (which are not unique, in
that multiples am,C„ Sutn,C, of the summational invariants tn,C, can be
added to each A, or &,u)) are to be fixed by imposing the further condition
SftC, = 0 (cf. (8.31, 7, 8)). It follows that
Z/>.Art = o, Y.p.Dj., - 0. (18.3, 5)
In virtue of (18.3, 4) and (18.2, 4), equation (18.3, 2) can be written in the
form
C, = -LA.«rf,-.D r .Vln T. (18.3, 6)
Multiply (18.2,13) in turn by D,v) and A„ integrate with respect to c„ and
sum over all values ofs. Then, using (18.3, 3) and (18.3, 5),
^.-MV*,!*}, (.8.3,7)
DTu = in{V«\A}, (18.3,8)
the bracket expressions being defined by obvious generalizations of (4.4,12).
From (18.3,7) follows the symmetry relation A„r = A n , so that the set of
equations (18.3, 4) are identical with the first set of (18.3, 5). Either set may
be used to express the coefficients A„ in terms of the coefficients Art (*+1).
In view of this and the symmetry relations, there are only \K(K— i)
independent coefficients A,,; similarly there are only K—i independent
thermal diffusion coefficients DTa.
An alternative way of writing (18.3, 6) is
C. - - S M ^ + k r t Vln T). (18.3,9)
t
Here the coefficients kTt are defined by
DT, = S A ^ k „ , (18.3,10)
<
S k r e = o, (18.3,11)
1
* Note that A,, doe* not reduce to the usual mutual diffusion coefficient for a binary
mixture. The above discussion differs from that given by Waldmann (HmcBmch dtr
Phytih, vol. ta, 406-11) in that he used diffusion velocities relative to the mean mole-
cular velocity c instead of those relative to c,.
346 MULTIPLE GAS MIXTURES [18.31
the supplementary equation (18.3,11) being required because (cf. (18.3, 5))
only AT—1 of the set (18.3, 10) are independent relations. Together with
(18.2, 7), this choice of the supplementary equation ensures that
Z(rf,+k T ,Vln T) - 0. (18.3,12)
*
Equations (18.3,6,9) again imply that relative diffusion of two of the con-
stituent gases may be induced by factors not directly affecting these gases,
such as forces acting on the molecules of another constituent gas.
Using (18.3, 12) together with (18.3, 9) (which also is equivalent to only
K-i independent equations, by (18.3,5)), we can express the vectors
d, + kT, V In T in terms of the diffusion velocities, in the form*
rf, + k r t Vln T = -2x.x,(C,-C t )/D r t , (18.3,13)

where Da = Dtt. In (18.3, 13), DH is closely analogous to the binary

diffusion coefficient of Chapter 8, to which it reduces for a binary gas
(AT = 2); also (see (18.41), below) the first approximation to DH is equal to
the binary gas [D^\x of 9.81. For these reasons, and because it does not
involve redundant diffusion coefficients, (18.3,13) is usually preferable to
(18.3, 6) as the basic diffusion equation.
If all the diffusion velocities vanish,
</, + k r , V l n r = o (18.3,14)
for all s. In a gas at rest under no forces, this becomes
Vx, + k r , V l n r = o. (18.3,15)
Hence k r „ like k r in a binary mixture, measures changes in equilibrium
concentrations produced by a temperature gradient.

18.31. Heat conduction

The heatflowq is given by

» »J
• The coefficient x,x,/D„ of C,-Ct in (18.3,13) can be thown to bt-9,,/9, where 9 it
the determinant of K+ 1 rowa and columns, obtained by bordering the determinant
whoae elementa are A„ with an extra row and column whoae common element is zero,
every other element being unity; 9M ia the cofactor of A„ in 9. Since "L9« « o,
(18.3,13) it equivalent to '
rf, + kr.V In T - £ ¢ , , 5 , / 3 - Lx,x,C,/D H ,
1 t
where £>„ it defined aa such that £x,/D„ — 0. Similarly (18.3, 9) can be derived by
_ i
solving for C, equations (18.3, 13) together with (18. a, 9). One finds
A„ - - 9',,I9',
where 9' it the determinant obtained by bordering that whose elements are x,x,/D,,
with an extra row and column whose common element is zero, the other elements being
p„ p„,..,pK;9',it the cofactor of x,xtIDu in 9'.
18.31] MULTIPLE GAS MIXTURES 347
(cf. (8.41, 2)). On substituting for ¢ , from (18.2, 8) this becomes

q/kT = f 5>.C,-JVln 7T \f?^.-\)A,.C.de.

* »J
-IZdtZJfl'W-toW.C.dcr (18.31,1)

Let a, be an arbitrary vector function of c,\ then, on multiplying (18.2, 11)

by a„ integrating with respect to c„ and summing over all values of s,

S j / W - I K - C . r f c . = n*{a,A}. (18.31, 2)

Thus, using (18.3, 8),

q/kT = %Xn,C,-ln*{A,A}V\nT-nj:DT.d,. (18.31,3)
< *
The thermal conductivity A, as usually defined, gives the heatflowin the
absence of diffusion. Since the vectors d, + kT,VlnT all vanish in the
absence of diffusion, we write (18.31, 3) in the form
q = $kTZn,C. -pXDT,(d, + kT,VlnT)-AVT, (18.31,4)
• *
where A = lkn*{A, A}-kn^ k r ,£> r ,. (18.31, 5)
f
In (18.31, 4) thefirstterm on the right represents the difference between the
heat flows relative to the mass velocity c0 and the mean velocity c of the
molecules. The second term is the heat flow, relative to c, due to diffusion;
by (18.3, 9, 10), it is equal to
- / • 2 k r , S ^ ( r f , + k r ,Vlnr)=/.Sk r t C». (18.31,6)
The last term gives the heat flow due to the thermal conductivity A.
An alternative expression for A is obtained by writing
^ = ^.-2krtDif>. (18.31,7)
t
Then by (18.31, 5) and (18.3, 8),
A-JfcffV.il}. (18.31,8)
Also by (18.3, 7, 8, 10),
{A,V»} = (3M(D r t t -£A„tM
1
= 0. (18.31,9)
Hence A = \kn*{A, A). (18.31,10)
The equality of the coefficients k r , appearing in (18.31, 6) and (18.3, 9)
(more correctly, the equality of the coefficients DTs in (18.31,4) and (18.3,6))
is an example of the symmetry relations met in the thermodynamics of irre-
versible processes; further examples are the symmetry conditions A,t =» At„
Drt = Dt, between the generalized diffusion coefficients in (18.3, 2, 13).
348 MULTIPLE GAS MIXTURES (18.32

18.32. 'Viscosity
The deviation p™ of the pressure tensor from the hydrostatic pressure is
given by /•
po>-sJ/!0)*.ctc,<Vcr
As in 8.42 it follows from (18.2,8) that
0

P»»-2^VC0,

where p - \kTSJ/J"***.: B.dc„ (18.32,1)

or, by (18.2,12), n - fAn*r{B, B}. (18.32,2)
18.4. Expressions for the gas coefficients
Explicit expressions for the gas coefficients introduced in the last three
sections are obtained by expanding A„ B/j,t} and B, in series of terms

Equations to determine the coefficients are obtained by multiplying

(18.2,11,13) by C.SMtfJ), or (18.2, 12) by cfes^Y*!). and integrating
with respect to c,. The mth order approximation is found by restricting p
and q to their first m permissible values. The mth approximations to A, ft
and A„ increase with increasing m; and they are best possible approxima-
tions in the sense used in the first footnote to p. 131.*
The mth approximation in general involves the solution of tnK simul-
taneous linear equations. In view of the complexity of the solution, and of
the fact that the first or second approximations for a simple or a binary gas
usually differ little from the true values, attention is here confined to the
lowest non-vanishing approximation in each case.

18.41. The diffusion coefficients

In the absence of temperature and velocity gradients, the equation for 4>(
becomes fj»C..d. - - * S * « , / * ( * . + • , ) (18.41,1)
<
(cf. (18.2, 2, 3)). The first approximation to ¢ , in this case is taken to be
<t>, = m,*,.CJkT, (18.41,2)
where a, is a vector independent of c,.
To determine «„ multiply (18.41,1) by m,C, and integrate over all values
of c,; this gives
nkTd,--n*[m,C„<t>A,- S **[*.<*•*•+*(]•»
* E. J. Hellund, Phyt. Rtv. 57. 319 (1940).
18.42] MULTIPLE OAS MIXTURES 349
where the bracket expressions are defined as in (4.4, 7, 9). The first term on
the right of this is zero, since tn,C, is a summational invariant. Thus, sub-
stituting from (18.41, 2),
nk*T*d, = - J 2 w,«.«f{«.[C„m,CJ^ + a t [C„m t CeL}.
i+ »

Using equations similar to (9.6, 1, 4) and (9.81, i), this becomes

^ = -^2^^--^°^ 1 )
3*i 1 m,+mt
= -Ex,xt(«.-««)/rz>*]i-
Here [Z),e], denotes the first approximation to the diffusion coefficient in a
binary mixture of therthand rth gases, at the pressure and temperature of
the actual multiple mixture. Also (cf. (18.3, 1)) the diffusion velocity C,
corresponding to (18.41, 2) is a,; hence to the present first approximation
d. = - S x.x t (C.-C t )/[DJ 1 . (18.41,3)
1
Equation (18.41, 3) shows that the generalized diffusion coefficients D*
of (18.3, 13) are identical, to a first approximation, with binary diffusion
coefficients at the same temperature and pressure. An interpretation of
(18.41, 3) as an (approximate) equation of motion of therthgas was given
in 6.63.
Equation (18.41, 3) is equivalent to K — 1 independent equations, from
which the K— 1 velocities of diffusion of the other constituents relative to a
given constituent can be calculated. For example, for a ternary mature,
from the first two of equations (18.41, 3),
Cl ( 4
°*-x 1 x 1 (x 1 [Z) tt ] 1 + xt[DM]1+x8[Z)1J1)- '4)
Similar expressions, in terms of ratios of determinants, can be given for
general values of K (cf. 18.3, footnote to p. 346). However, in view of the
difficulty of evaluating the determinants, it is normally simpler to retain the
equations (18.41, 3).

18.42. Thermal conductivity

By (18.31, 2, 8) the thermal conductivity A is given by 2A<,), where
Aw = §Aj/H«?-f)C..if.rfc, ' (18.42,1)
We may regard A,,, as the contribution of therthgas to the total conductivity.
From (18.2, 11, 13) and (18.31, 7), the equation satisfied by A, is
Xn.ntUAt + At) = / 3 0 , C . ( ^ - { - * . - l k r . ) . ('8-4*. 2)
1

since 2 k r t = 0.
1
350 MULTIPLE OAS MIXTURES (18.42
Equation (18.42, 2) can be used to determine A without first knowing kTr
From the definitions of A, and k r „ using (18.3, 3),
\fi"A..C.dc, = 31.,(0,..-Sk rt A„) - o. (18.42, 3)

Thus A, does not contribute to the diffusion velocity C, (as is natural,

because A, was introduced to represent heat conduction in the absence of
diffusion). Hence when A, is expressed in terms of the vectors SfX^C,
the term p = o, which alone contributes to C„ must be absent. The coeffi-
cients of the vectors for/> > 0 can be determined by multiplying (18.42, 2)
by the vectors S^\^)CP where q > o, and integrating with respect to c,\
when this is done, the term involving kT, is found to disappear.
The first approximation to A, can be written
A. - -{2mJSh*Tn,)a,S^\^C,; (18.42, 4)
the factor multiplying a, is introduced in order to make a, equal to the first
approximation to A^. To determine a, we multiply (18.42,2) by — S^\V!)Ct
and integrate with respect to c,. Then on substituting for A, from (18.42,4)
we get

+ n. 2 «,(«>.)» W W " S i W t L e = 75k*TnJ8me

1*» * "
(18.42, 5)
The bracket integrals are given by equations similar to (9.6, 3, 6, 8); the
expressions for them can be simplified by using equations similar to (9.7, 1),
(9.8,7,8) and (9.81,1). Using these equations, (18.42, 5) can be reduced to
the form
* M * J i + 2 (rx./S/'[DJ l )(6MS+(s-4BjAf. 1 t + 8Af-Mt.Art)}
<+»

- x , S «t(TM,tMJsp[DM(u -4B^-8A-) = x r (18.42, 6)

(•«

Here [AJ, and [D^]i denote, as before, first approximations to the thermal
conductivity of the ith gas (when pure) and the diffusion coefficient in a
binary s, t mixture, at pressure p and temperature T; Art, B^ are defined in
terms of collision-integrals for collisions of molecules m„ mt in the same way
as the A, B of (9.8, 7); and
Mf*mJ{m,+mt), Mt,-mtftm,+mt). (18.42,7)
The first approximation [£>,<], is used instead of Ert because of its closer
relation to experimental quantities.
In any actual calculation, the quantities a, are best determined by solving
the simultaneous equations (18.42,6) by numerical methods. The thermal
18.43) MULTIPLE GAS MIXTURES 351
conductivity is then given, to a first approximation, by
[A]i = Z[A«], = 2X- (18.42.8)
a a
The expression for [A], given by (18.42, 6, 8) may be compared with
the Wassiljewa expression given in 6.63. Whereas the latter expression
implies that the effect of collisions is solely to oppose the conduction of heat,
(18.42, 6) implies that there is also a secondary effect, the transport of heat
— AJOVT1 by the tth gas being in part converted into heat transport by the
5th gas at collisions between molecules m, and mt.*

18.43. Thermal diffusion

We here consider the thermal diffusion ratio k r , rather than DTt, because
of the physical significance of kT, in relations like (18.3, 15) and (18.31, 6).
In order to determine k r , we multiply (18.42, 2) by tn,C, and integrate with
respect to e,. Then the term arising from "^ — f on the right vanishes, and
-3/*r« = » ! K C „ i l j , + I,nsnl[m,C„A, + At]^. (18.43,1)
Thefirstterm on the right vanishes because m,C, is a summational invariant.
In the second term we substitute for A, the approximate form given by
(18.42, 4). Since [m,C,+mtCt,A,],t = o, this gives

-«.*[*«. Si"(*5*«W- (18.43,2)

The bracket integrals in (18.43, ) a r e g'ver> by formulae like (9.6, 2).
2

On substituting their values, and using equations similar to (9.81, i), we get
[k r Ji = K ^ W f l J . l f x . ^ M . - j i . a . M , , ) . (18.43, 3)
Here [Dtt]v MH, Mu are as defined in 18.42, and c rt is the equivalent, for
encounters of molecules m,, mt, of the c defined in (9.8, 7). The value of
[k T Ji is now obtained by substituting for a„ at their values found by solving
(18.42, 6). Since a„ being the first approximation to A<j), has x, as a factor,
we can write in generalization of (8.4, 8)
[ M i = 2x.xe« rt , (18.43, 4)
where a,t = - at„.
Equation (18.43, 3) indicates a close dependence of kT, on the parts A<() of
the thermal conductivity A. However, whereas A is appreciably affected by
the internal energy of the molecules, it is usually assumed that the internal
energy hardly affects the value of kT„. Thus (18.43, 3) should apply approxi-
mately to molecules possessing internal energy, provided that in (18.42, 6)
[A,]t is given its monatomic value i5*[/*,]i/4»*,.
• T. G. Cowling, P. Gray and P. G. Wright, Proc. R. Soc. A, «7«, 69 (1963). See also
J. M. Yos, AVCO Report, AVSSD-ona-67-RM (1967).
352 MULTIPLE OAS MIXTURES [18.5

18.44. The viscosity

The viscosity is given by (18.32,1), i.e.
/1 - £/<«, (18.44,»)
where /*&, the contribution of therthgas to /t, is given by
to = | * r J / 2 » » > . : B.cfc. (18.44, 2)
and B, satisfies (18.2,12). The first approximation to /* is derived by using
the approximate forms 0
Z.-{bJn,kT)V,V.; (18.44,3)
the factor n,kT is inserted in (18.44, 3) 8 0 M t 0 make b, equal to the first
0
approximation to /*(,>. To determine b, we multiply (18.2,12) by V,V, and
integrate with respect to c,. This gives, after substituting for B, from
(18.44. 3).

+ S M . [ * > „ *>J*+».*»[*>., *&]*)• (18.44.4)

are
The bracket integrals in (18.44, 4) determined from equations similar
to (9.6, 14-16), (9.7, 1), (9.8, 7) and (9.81, 1). Using these, we find from
(18.44,4)

In (18.44,5)> [/*Ji> [At]i> A* are as defined earlier; pi, pj denote the densities
of therthand rth gases when pure, at the pressure and temperature of the
actual gas. When the quantities b, have been found from (18.44, 5) the n r 8 t
approximation to ft is given by
M i - S t/*w]i - 2 *.. (18.44,6)
< «

APPROXIMATE VALUES IN SPBCIAL CASES

18.5. Isotonic mixtures

Consider a mixture of several isotopes of the same gas. The mean mass M of
molecules in the mixture is defined by
m - 2»<«<, (18.5,1)
4
and the quantities (m(-m)[m are supposed to be small enough for their
squares and products to be neglected. Then, assuming that the same force-
18.5] MULTIPLE GAS MIXTURES 353
law applies to all types of encounter in the mixture, the thermal conductivity
and viscosity are found to be those corresponding to a simple gas of mole-
cular mass m. The constants A, B, C are the same for each pair of isotopes; the
thermal diffusion ratio kTs is such that*

1 r,J1 v 3
2 A ( I I - 4 B + 8A) m '
The Kihara approximation is found by replacing B by }.
In the steady state in which thermal diffusion is balanced by diffusion due
to inhomogeneity, Vx, = — k r , V l n T. Thus, to a first approximation,

v
\x e / 2 A ( I I - 4 B + 8A) m •"J/
Since this is valid only to the first order in (m,—mt)lM, the quantity JR in the
denominator could be replaced by any other mean mass. Hence (18.5, 3)
shows that the relative concentration of molecules has the same
spatial variation as if the other isotopes were absent.
Equations (18.5, 2) and (18.5, 3) indicate that results for a mixture of
several isotopes can be used to determine the thermal diffusion ratio for a
pair of isotopes, and vice versa. A similar remark also applies to the diffusion
thermo-effect.t
The thermal separation of two isotopic gases (tnv mt) for which c u is small
can be greatly increased by the introduction, in excess, of a third gas (m,)
such that c u , cn are large. Suppose that the properties of the third gas are
otherwise intermediate between those of the two isotopes, so that in (18.43, 3)
aijxxm1 > ajx3ms > o^XjWj.
Then in the steady state in which thermal diffusion is balanced by
that due to inhomogeneity
Vln fc) = v r r ( x i + x»)mim»cn / f i ft_\

x 3 m 3 fM 1 8 c 1 8 / ax ^ \ Macni aa a, \Y\
P \[D»li\xi«, W [J>wli W " . xtmj\\- ^0,5, 4
'
Because the properties of molecules m3 are intermediate between those
of the two isotopes, the three quantities
ff!|ffig ffijfftg ftt»tn*
(Vnyf mt) [D^' (mi"+««,) [£>i»]i' K+^[^K
are nearly equal. Also C,„, c M are nearly equal; hence (18.5,4) approximates to

Vln (¾) = J ^ « l ^ l W l W , ( - ^ - ^ - ) VT. v(18.5,

3 5)
\xj />(«i + w„) [Dlt}t ' *\x1ml xtmj ' >'
• R. C. Jones, Phys. Rev. $t), 1019 (1941).
f L. Waldmann, Z. Naturfortchung, 1, u (1946).
354 MULTIPLE OAS MIXTURES [18.51
Suppose that ajxx, <^/x, (the contributions of molecules mlt m, to the thermal
conductivity, in proportion to their numbers) are not greatly altered by the
presence of the third gas. Then the effect of its presence is to increase the
expression on the right of (18.5, 5) in the ratio
{(' -xJCn+x.Cnl/c,, = 1 + x,(c u /c 11 -1).
This is large if c 13 /c lt is large and x, is not small.
Clusius and his associates have used the enhanced separation due to an
added gas in separating isotopes with the thermal diffusion column. A nearly
steady state is eventually reached in which the two isotopes, each containing
a small amount of the added gas, are separated by a mass of that gas in an
approximately pure state. The trace of the added gas is then removed from
the isotopes by a chemical process.*

18.51. One gas present as a trace

Consider a ternary mixture, with molecules m, present in trace only. The
viscosity and thermal conductivity are normally those of a binary gas of
molecules if the third gas were absent; the relative diffusion of
molecules likewise not appreciably affected by the molecules m,.
The diffusion of molecules m, through an otherwise homogeneous gas of
molecules m,, m, at rest under no forces is given, to the first approximation, by
Vx, - -x&xJiD^+xJlDM (18.51, 1)
(cf. (18.41, 3)). This is the same as for diffusion through a single gas, with
coefficient of diffusion Dltti, where
i/I^u, J , = x./fZ),,], + xJ[D„]v (18.51, 2)
Equation (18.51, 2) was verified experimentally by Walker, de Haas and
Westenberg in the diffusion of CO, through mixtures of He and N,."f It
should be noted, however, that the linear dependence indicated by (18.51, 2)
is correct to a first approximation only .J

18.52. Ternary mixture, including electrons

Consider a ternary mixture in which the molecules m, are electrons, so that
mjmx and mjma are small. We shall give approximate expressions, regarding
(fnjmjl and (mjmt)i as negligible, and assuming that x, is not small.
The quantities A„ D,„ Dn in (18.42,6) are all proportional to mj», and
so are large. Hence at is also proportional to « , ' , and is large compared with
* See, for example, K. Clusius and E. Schumacher, Httv. Chim. Acta, 36,969 (1953), for
the separation of MA, "A, "A using H"C1 and H"CI or D**CI and D»'C1 as auxiliary
gases.
f R. E. Walker, N. de Haas and A. A. Westenberg, J. Chem. Pkyt. 3a, 1314 (1960); see
also B. N. Srivastava and R. Paul, Pkytica, a*. 646 (196a); R. Paul, Ind. J. Pkyt. 36,
464 (196»)-
% S. I. Sandler and E. A. Mason, J. Chem. Pkyt. 48,1873 (1968).
18.52] MULTIPLE GAS MIXTURES 355
ax and at. This implies that the thermal conductivity is, to the present
approximation, due wholly to the electrons, being given by

The error involved in neglecting (mjmjl and (tnjmt)i may be 2 or 3 per

cent. This is likely to be smaller than that involved in proceeding only to a
first approximation [A]lf especially when the molecules m1 amd m, are both
ionized. If further approximations to A are required (still regarding (mjm^i
and (wg/w,)* as negligible), they can best be calculated by regarding the mole-
cules Wj and mt as inert targets, but taking into account the encounters of
electrons with other electrons as well as with these targets. In a wholly
ionized gas the thermal conductivity then is that of a binary electron-ion gas,
with a number-density «x + nt of heavy particles each possessing charge eu>
W Cfe
K + »a)*i« = «i«f + »»e|. (18.52,2)
In the viscosity formulae, [fi3]i, ">3[Z7u]t and mJD^ are each small, of
order my, thus from (18.44, 5) *a ' s a ' s o of order mf, and may be neglected.
Hence to the present order of approximation the viscosity is that of the two
heavy gases, unaffected by the presence of electrons.
In order to calculate the thermal diffusion ratios it is necessary to calculate
a, and a,; these have, to the present approximation, the same values as if
electrons were absent. In terms of au a,, «3, the thermal diffusion ratios are
8 i v e n by a T /c x c x \

with a similar relation for [kj-,],. Since a3, [Di3]v [D^i are all proportional
to m,}, all the terms in (18.52, 3, 4) are similar in order of magnitude.
The diffusion velocities are given, as in (18.3, 13), by
d', = rf, + k r ,Vln T = - Z x ^ C . - C O / D . , . (18.52, 5)

In this the coefficients k r , are similar in order of magnitude for all s; the
same may be assumed for d„ since in (18.2, 4) the various pressure-gradients
Vp, are similar in order of magnitude, and the volume force p,F, due to an
electric field is comparable in magnitude for electrons and heavy ions.
Solving (18.52, 5) for the diffusion velocities as in (18.41, 4) we get

Cl
" c «" iSSB£+M>«+*./>11> (l8s2,6)

and two similar relations. Thus, in view of the smallness of Dlt compared
with Dn and D„, the diffusion process falls into two parts. The first and
356 MULTIPLE GAS MIXTURES [18.S2
major part is the diffusion of electrons relative to heavy molecules, with
Vel0Clty
Ct-Ct + ^-^-DnDn&lxJXtDv+^Dn). (18.52,7)
This is the same as the velocity of diffusion of electrons relative to a single
heavy gas of number-density »i +»», with diffusion coefficient D1%t, where
(x 1 +x t )/D lt , i - xJDu+x,/Dn. (18.52, 8)
The second part consists of the (much smaller) relative diffusion of the
two heavy gases, given by
(l8sSi9)
^-^- i.shv^iy
Let d'v d'x be expressed in terms of d3 ( • - d[ - rfi) and <fj t , defined by
<*M - (P*d'i-Pi<t'JI(Pi+Pi); (18.52,10)
for a binary gas in the absence of a temperature gradient d'lt reduces to the
dlt of 8.3. Then (18.52, 9) becomes

This indicates the presence of 'induced' diffusion due to d't, as well as the
usual diffusion due to d'n.
Equations (18.52, 7, 9, 11) do not depend on any particular order of
approximation to D, t , Dlt and Dn. In general it is sufficient to replace D, t
by its first approximation [Z>lt]t; however, as with A it may be desirable,
especially with a fully ionized gas, to proceed to higher approximations in
calculating Dlt and Dn. Electron diffusion in a fully ionized gas can be
calculated from binary-gas formulae, using (18.52, 2).
If we are concerned only with thermal diffusion, d', = kT, V In T. Thus,
using (18.52,4) and a similar equation, we find from (18.52, 9) that, to a
first approximation,
Ci - C, = /rWflx.x,)- 1 c ^ x . ^ Mtl - x, a, M^
+ (c„-c 1 ,KD u /(x 1 Z) t l ,+x t D u )}. (18.52, 12)
Since av at, at denotefirstapproximations to the contributions of the three
gases to the thermal conductivity, the a» term in (18.52,12) represents an
induced thermal diffusion due to the presence of the electrons; it is normally
similar in order of magnitude to the other terms, but vanishes if c u « c„,
as is the case for example in a fully ionized gas (c u — c„ =» c„ •» — o-6).
Thermal diffusion is especially important in a fully ionized gas, because for
such a gas the quantities CH are several times as large as for a neutral gas.
A case of special interest is when molecules mx are protons, and molecules
m„ which are few in number, are ions of much greater mass and charge. In
18.S2] MULTIPLE GAS MIXTURES 357
this case M „ is small and Mn = i; also x1at/x2a1 is of order (Mit)l and at is
approximately [A,],. Hence (18.52, 12) becomes approximately
cx-ct - /.-^r'c.^A.ijVr. (18.52,13)
This does not depend on the charge et on the heavy ions, whereas the
coefficient of diffusion Dlt is proportional to ije\. Thus if «, is large, thermal
diffusion may be able to produce a greatly enhanced separation of the two
kinds of positive ion before being balanced by inhomogeneity diffusion,*
the heavy ions travelling towards the hotter regions.
* S. Chapman, Proc. Phyt. Soc. 7a, 353 (1958). A more exact diacutaion waa given by
J. M. Burgers, in Plasma Physici (ed. F. H. Clauaer), pp. »60-4 (Pergamon Press, i960).
This took into account electric fields of thermal origin, and obtained a rather larger
diffusion velocity.
 19
ELECTROMAGNETIC PHENOMENA
IN IONIZED GASES
19.1. Convection currents and conduction currents
In ordinary gases the molecules are electrically neutral; when some of them
carry electric charges, the gas is said to be ionized. In an ionized gas the
different types of particle (to all of which the general term' molecule' will be
applied) may include neutral molecules or atoms, atomic or molecular ions
of either sign, and electrons. The charged particles will carry an integral
multiple (positive or negative) of the charge e, where e denotes the numerical
value of the electronic charge.
Let »„ m„ e, denote the number-density, mass, and charge of the rth type
of molecule in an ionized gas (s = 1,2,...). It is convenient to enumerate
first the various kinds of neutral molecules, so that if, for example, there is
only one kind of neutral molecule present, 1 = 1 will refer to this.
The volume-density of charge, or charge per unit volume, may be denoted
by Pt; clearly ^ ^ ^ , j}
The current-intensity is defined as the vector in the direction of current-
flow at any point, with magnitude equal to the current per unit cross-section
normal to this direction: it is equal to
Zn,e,c, = 2n,e,c0 + 2n,e,C,
= Ptc,+JL (191.2)
where j & 2»,*,C,. (19.1, 3)
Clearly/ is the chargeflux-vectorof 2.3 or 2.5, corresponding to ¢, = e,.
In (19.1, 2) the first term is due to transport of the volume-charge with
velocity c 0 ; this part of the current will be called the convection current, and
the remaining part,.;', will be called the conduction current. The division here
made between the two is to some extent arbitrary, because, so far as electrical
considerations are concerned, we might equally well have written c, = c + C'„
where c denotes the mean molecular velocity (not the maw-velocity of
(2.5, 7)); in that case we might regard pec as the convection current, and
2n,e, C', as the conduction current. The choice actually made here, however,
is that indicated by (19.1, 2).
If either pe =» 0, or cB » o, there is no convection current. In many ionized
gases p ( » 0 to a high degree of approximation; when this is not so, an
electric field is set up which tends to dissipate the volume charge; the gas
retains an appreciable charge only if external agencies maintain it.
[ 3S8]
 Copyrighted Material

19.11] IONIZED OASES 359

19.11. The electric current in a binary mixture

In the case of a binary mixture, (19.1, 3) reduces to
j = »!«!(?! + n2e2C2

= («i"»(^iW2-Vi)(Ci-C2). (19.11, 1)
In this we substitute for C^ — C^ from (14.1,1), giving

- ^ ( F ^ F J +krVbrJ. (19.11,2)

If the forces mlF1 and »i 2 F 2 acting on the two types of molecule are due
to the presence of an electric field E, or include a part due to this cause, the
corresponding part ofj (the ohmic current) is

^-(^mj-^M,)8!),^^^, (19.11,3)

where # = ^ ^ (t l m 2 - etmxf Dlt. (19.11,4)

The coefficient #, which is essentially positive, is called the electrical con-

ductivity of the gas.
If the molecules m1 ire either ions or neutral molecules, and the others are
electrons, the ratio mi/ml is so small that to a high degree of approximation
p = nlmv and (19.11, 4) is equivalent to

*-aff D «- <•>••••'>
Since this involves e2, and not also ev it shows that the conductivity is nearly
all due to the electrons. If the gas is electrically neutral, so that n^^ — - nzet,
the convection current vanishes, and (19.11, 5) may also be written in the
more symmetric form
*--^SZ)„; (19.11,6)

this expression for # is still positive, because on the present hypothesis ex

and e2 are opposite in sign. If the ions are singly charged, (19.11,6) reduces to

=
* kTDl* 0 9 - " . 7)

Besides the conduction current due to an electric field, contributions to

the total conduction current also arise from gradients of concentration,

Copyrighted Material
360 ELECTROMAGNETIC PHENOMENA IN [19.12
pressure, and temperature, and also from any non-electrical forces which
tend to produce unequal accelerations of the two types of molecule; gravi-
tational or centrifugal forces tend to accelerate all particles equally, so that
they do not contribute directly to the electric current, though they may
contribute indirectly, by setting up a pressure or concentration gradient.
19.12. Electrical conductivity in a slightly ionized gas
Consider a gas, the great majority of whose molecules are neutral. If there is
only one kind of neutral molecule (el •> o) and one kind of charged molecule
(«, # o), the electric conductivity & is, by (19.11, 4) (since n and p are
approximately nt and «1>n1), n.t
*-**D
A w
~'kf'
as if each charged particle contributed an amount
&lnt-j±Du
to*.
In this case, by (19.11,1), using similar approximations,
c,-cj - C , - C , - / f a , * ,
->(&/ntet)E = (e,DltlkT)E.
This gives the mean velocity of the charged particles relative to the neutral
particles. When the gas is at atmospheric pressure, the coefficient
gfPii
kT
in this equation is called the mobility of the charged particles in the neutral
gas.
If more than one kind of charged particle is present, each in numbers so
small that its influence on the mean velocities of the others is negligible, we
have equations such as _ _ e D
C.-C^^E
for each type s. In this case the total current-intensity is
~Ln,t,c, - pec0+I,n,e,C,
-pJiCo+C1)^n,ej(C,-Cl)
-ft.^+2. kT E.
If pt - o, the current is &E, where now
»-(Zn,elDu)/kT;
thus each charged particle m, makes an average contribution to 9 equal to
e\Du
kT •
19.2) IONIZED OASES 361

19.13. Electrical conduction in a multiple mixture

If more than one kind of charged particle is present in the gas, and the
degree of ionization is not small, the electrical conductivity must be calcu-
lated in terms of the generalized diffusion coefficients AK or D, of 18.3. If
diffusion is due wholly to an electric field, and the gas as a whole is electrically
neutral, by (18.3, 2), _
j = 2».*.C,= 2 I,n,nte,etAMElp.
• » 1

Using the relation ZPtA* = o, one finds thaty = &E, where

t

p,+ttntmt
The sum in (19.13, 1) is over all unequal pairs s, t, each pair being taken once
only; # is positive,* because negative values of AH predominate if s * t.
This gives the appropriate generalization of (19.11, 4).
In the important case when electrons are present in numbers comparable
with those of the other charged particles, the contribution of the heavy
particles to & is negligible compared with that of the electrons (cf. 18.52). In
this case, if the suffix 2 refers to electrons, by (18.3, 13), approximately
j = ntetCt = -dte%nl{ 2 xJD^)
<+»

= EelnnM'L^IDtt).
(•a
Hence # = n,e|/AT( £ x£/D«), (19.13, 2)

indicating that the other gases in the mixture contribute additively to the
electrical resistance.

MAGNETIC FIELDS
19.2. Boltzmann's equation for an ionized gas in the presence of a
magnetic field
All the work of previous chapters has been based on the assumption that the
applied forces acting on the molecules are independent of the molecular
velocities. In the presence of a magnetic field of intensity H, the force on a
particle of mass m and charge e electromagnetic units, moving with velocity c,
includes a term ecsH (cf. p. 11 for this notation). The acceleration of a
molecule can therefore be divided into two parts, one, denoted by F, inde-
pendent of c, and the other, due to the magnetic field, equal to (e/m) CAH.
• By (18.3, 7).
» - (3*r)" l {S».*.D" - . S».«,©"'}> 0.
«3 CUT
362 ELECTROMAGNETIC PHENOMENA IN [19.2

The occurrence of the second term necessitates a re-examination of the

argument of this book, from 3.1 onwards.*
The first modification comes in the proof of Boltzmann's equation. In 3.1
it was taken as obvious that molecules in the velocity-range c, dc at the
beginning of an interval dt must occupy an equal velocity-range at the end of
this time; this is no longer self-evident. Since the molecular velocity
increases during dt from e to e + Fdt + (e/tn) c A Hdt, the range occupied by
these molecules at the end of dt is now of magnitudef
. S[c+{F+(elm)cAH)dt)
--Hydt
3(c) 171

1. -H.dt
m *
--Hxdt, 1
m *
which equals dc, since dt* is to be neglected. The velocity-range is therefore
still invariant in magnitude; the derivation of Boltzmann's equation proceeds
as before, and yields the result

As was noted in 10.34, a simple ionized gas (consisting of molecules with

charges all of one sign) does not exist in nature. However, in a gas-mixture
the Boltzmann equation for each/, is identical in form with (19.2,1). If/, is
regarded as a function of C„ r, t instead of c„ r, t, the equation becomes

I T " D i + c » * + \ F ' + «. c " sH —m) ac.

+ c
^ -^S.-ft c ' : ^ c - (19 - 2>a)
* The firtt theoretical work on the effect of magnetic fields on get phenomena seemt to
have been done by R. Cans (Annln Phys. ao, 203 (1906)), who considered the effect of
a magnetic field on the flow of heat and electricity in a Lorentzian gas. Further work
on a Lorentzian gas waa done by N. Bohr (thesis). The general theory of the transport
phenomena in an ionized gas subject to a magnetic field wasfirstdiscussed in the first
edition of this book; for a special case of electrical conductivity, ace T. G. Cowling,
Mon. Not. R. attr. Soc. 93, 90 (1932). The subject has, however, frequently been
considered by collision-intervsl methods, following J. S. Townsend (Proc. R. Soc. A,
S6, 571 (1913), and EltctrUily m Gaits, §§80-93). It ia of great importance in the theory
of radio propagation, in which an important part ia played by the ionized layers of the
upper atmosphere; for early work on this, see P. O. Pedersen, Tht Propagation of Radio
Watm (Copenhagen, 1917), chapters 6 and 7. The theory is also of importance for
solar physics, owing to the sun's general magnetic field and the much more intense
magnetic fields of sunspots; and for the physics of interplanetary and interstellar gas.
t The notation for Jacobians ia that of 1.411.
19.3] IONIZED OASES 363
The equation of change of a molecular property <j>$ can be deduced in the
form

(»9-2. 3)
The equations of continuity, mass motion and energy are obtained by
giving <f>, appropriate values in (19.2, 3) (cf. 8.1). They are
Dn, d 8 , -^^ Dp d ,
^ 7 + 1 . 3 ^ , + ^ . ( ^ . ) = 0, 5 7 + ^ ^ . = °. 09-2.4)
/>§7°= Xp.F. + CLn.e,)c0MI+jMl-£.p, (19.2, 5)

ffci-gj- = $kT^.Zn,C. + Xp,F,. C.+j.{c0*H)-p:^c0-^.q,

(19.2, 6)
where/ is given by (19.1, 3).
It may be proved by the methods of Chapter 4 that the distribution of
velocities in a stationary gas in a uniform steady state is of Maxwellian form,
provided only that the forces F, are derived from potential functions Y r
The density distribution is unaffected by the presence of the magnetic field,
being given by the equation

as in (4.14, 7).
Maxwell's formula is also valid for a gas rotating as if rigid, with angular
velocity <u, in the presence of a magnetic field whose lines of force are in
planes passing through the axis of rotation, and of fields of force whose
potentials T , are symmetric about this axis. Let Oz be the axis of rotation,
and let the vector potential of the magnetic field have components — yA,
xA,o. Then(cf. (4.14,11)) it is found that the variation of density of the rth
gas is the same as if it were at rest in afieldof force of potential
T , - « V +/) - (*>,) «(*• +/) A.

19.3. The motion of a charged particle in a magnetic field

As a preliminary to considering an elementary theory of the transport
phenomena in a magnetic field we first consider in detail the motion of a
single molecule of mass m1 and charge e, during a free path. The fields of
force—magnetic and otherwise—which act on the particle are supposed
uniform. Let Ox be taken parallel to the magnetic intensity H, and let Oz
be taken in the direction of the component of F, which is perpendicular
13-a
364 ELECTROMAGNETIC PHENOMENA IN [19.31
to H. Then if the components of F, are Xlt o, Z„ the equations of motion
of the molecule are
m,* « m, Xlt mrf = «-,//*, m,* - m,Z, -exH$.
Let wt a eyHlmv (19.3, 1)
Then these equations become
* = * „ ^ = w,i, i - Z . - w , ^ . (19.3,2)
Integrating, we obtain the following equations for the position r and
velocity c, of the molecule at any time /, in terms of the position r' and
velocity c[ at time / » o:
u, = «; + * , / ,
0, - (t>;-ZJwJcos w,/+t^sin« l f + ZJitfl,} (19.3, 3)
w, =• te| cos 0¾ / —(v[ — ZJ(ot) sin w, /,
x-ac' + MiZ + JA"!/*,
>,=>+
i{( t ' I "zrl) 8in<u,/+t( ' i(I ~ C08Wi/)+Zi ')' («9-3. 4)
*-«' + ^-{w;««n w , / - K - ^ j (1 - cosw,/)J.
The motion parallel to H or Ox is clearly unaffected by the magnetic field.
If there is no non-magnetic field, the molecule moves in this direction with
uniform velocity; in the transverse direction the motion is circular, so that
the resultant motion is spiral, the spiral-axes being lines of magnetic force;
the spiral-frequency is wj2n. When non-magnetic forces act, the motion
transverse to H includes, in addition to the circular motion, a mean drift in
the j-direction, i.e. perpendicular to H and F,; the drift velocity is given vec-
torially by m^F^ A Hje^H*. The resultant transverse motion is trochoidal.*
Because motion parallel to the lines of force is unaffected by the presence
of the magnetic field, it is natural to expect that the presence of a magnetic
field will not affect diffusion and heat conduction in the direction of the field.
The exact theory confirms that this is so.

19.31. Approximate theory of diffusion In a magnetic field

An approximate theory can be given for the transport phenomena in an
ionized gas in a magneticfield,f using a relaxation approach similar to that
of 6.6-6.63. We first consider diffusion. Since the relaxation approach is
* Here appeara one of the intereating paradoxes of the kinetic theory. If the interval
between two consecutive encounters ia very long, every molecule mi will possess an
average velocity Zi/oi, perpendicular to F, and H; yet a uniform steady state ia possible,
such that the gaa at any point ia at rest. A similar remark applies when F, — 0 and the
magnetic field varies from point to point. For a discussion of the paradox, see T. G.
Cowling, Mm. Not. R. atlr. Soe. 90. 140 (1929); 9a, 407 (193a).
t The theory given here is essentially equivalent to that given in earlier editiona of thia
book, in spite of considerable apparent differencea.
19.31] IONIZED OASES 365
unable to give a theory of thermal diffusion, we assume the temperature to
be uniform; velocity-gradients are also ignored.
The relaxation theory of diffusion (cf. (6.63, 4)) was based on a con-
sideration of the equation of mass motion of the rth constituent gas in a
mixture, under the influence of its own partial pressure gradient, of body
forces, and of drags exerted on it by the other gases. The effect of the magnetic
field is simply to introduce an extra body force n,e,c, A H per unit volume.
Thus assuming that the acceleration of the rth gas can be taken to equal
DcJDt, the acceleration of the mass as a whole, the equation of motion of
this gas becomes

P.-$ = P.F. + ».',c.*H-Vp.+ I,£l£t(ct-c.), (19.31, 1)

where D,t is the coefficient of diffusion in a binary s, t mixture at the pressure

and temperature of the actual mixture. The set of equations (19.31,1) gives
the diffusion velocities ct - c„ and also (on summation over all s) the equation
of motion of the gas as a whole. The term involving H does not affect the
component of (19.31,1) parallel to H; hence (in agreement with 19.3)
diffusion parallel to the magnetic field is unaffected by the field.
Consider now diffusion perpendicular to the field H. We consider first
a slightly ionized gas, in which encounters between pairs of charged particles
are negligible in importance. The gas as a whole is supposed to be at rest
relative to the axes used; thus c0 = o, DcJDt = 0. Because of the pre-
ponderance of neutral molecules, their mean velocity c t is effectively c0, i.e.
zero. Hence for any of the charged constituents, (19.31, 1) reduces to
0 = p,F,-Vp,+ n,efi,AH-n,m,c,Vlr„ (19.31, 2)
where T, is defined by ^ . ,^ „ , , .
J
* PiP.lpDu = n.mjr,. (i9-3»>3)
Thus (cf. (6.62, 7)) T, is (m, +»«,)/«! times the effective collision-interval for
a molecule m, in collisions with molecules ntl; if m, refers to electrons, mjm1
is negligible, and r, is the actual collision-interval.
As in 19.3, thefieldH is assumed parallel to Ox; in addition it is assumed
that F, and Vp, reduce to their components Z, and dpjdz in the ^-direction.
Then, taking the components of (19.31, 2) in the y and z directions,

0 = <D,W,-V,/T„ O = Z,-± -^-O),V,-WJT„

where <o, = e,H/mf, as before. Thus

^KI)'"-r*l(2-7,a-"»""
W,TJ
».-
The velocity of diffusion in the absence of a magnetic field is found by
putting to, = 0. The presence of thefieldaccordingly reduces the velocity of
366 ELECTROMAGNETIC PHENOMENA IN {19.31
diffusion in the ^-direction in the ratio 1/(1 + WJTJ). It also leads to aflowof
molecules perpendicular both to H and to the direction of the ordinary
diffusion; this may be called transverse diffusion.
The case when the diffusion is due to an electric force E is of special
interest It then appears that the electric current set up by an electric force
transverse to a magnetic field is not altogether in the direction of the electric
force; a component perpendicular to both H and E also exists, which is
known as the Hall current. Also theflowin the direction of the electric force
is less than it would be in the absence of the magneticfield,and so we can
regard the electrical conductivity in this direction as reduced by the magnetic
field. The currents in the * and ^ directions are given by_/, — ^lE,jw — 9llE,
where • • «

The quantities #*, &11 may be termed the direct (Pedersen) and transverse
(Hall) conductivities.
Equation (19.31, 5) implies that the direct conductivity is much reduced
when the gas is so rare, or the magnetic field so strong, that the collision-
frequencies 77x are not large compared with the spiral-frequencies <o,; under
these conditions the transverse current may be comparable with the direct
current. In a gas containing electrons and positive ions, when H is not too
great, the electric current is mainly due to the diffusion of electrons; but
owing to the dependence of w, on m„ when the magnetic field is so strong
that O),T, is large compared with unity for both electrons and ions, the
current is due mainly to the motion of ions. This can occur only for very
intense magnetic fields or very low densities. For example, in a gas of
atmospheric composition and temperature, <O,T, is large for electrons only
when Hip > 10s, and for ions only when H/p > 10', if H is measured in
gauss xa&p in atmospheres.
In metals, when the direction of the Hall current is perpendicular to the
boundary of the metal, an electricfieldis set up which prevents any further
flow of die Hall current. If the Hall current is similarly inhibited in a gas,
then, in addition to the applied electricfieldE in the x-direction, there is a
'polarization'fieldE' in the negatives-direction, such that
&i£' . WE.
The total current-density in the direction of £ is now {rlE+&nE' - &mE,
whcre
01" - #1 + (*n)»/»i. (19.31,6)
Thus &m (the Cowling conductivity*) is the conductivity effective in this
case. It is also in general the conductivity effective in the dissipation of the
* The conductivities #' »nd 9m, important in ionoapheric theory, were firit named the
'Pedenen' and 'Cowling' conductivitie* by S. Chapman. Nuovo Cim. X, 4 (Supp. 4),
«3*5 (1956). For their introduction, lee the reference* quoted in the footnote to p. 36a.
19.31] IONIZED OASES 367
energy of currents. When an electric field E produces direct and transverse
currents &lE, &llE, the rate of supply or dissipation of energy per unit
volume nj. E, i.e. 9lE*; this is also expressible asj*l&in.
As a second application of (19.31, 1), consider a fully ionized gas com-
posed of electrons of mass me and charge —e and an equal number of
positive ions of mass m( and charge e (the suffixes *, i here refer throughout
to the electronic and ionic constituents). Because of the smallness of me, the
mean velocity of the ions is effectively the mass-velocity c0 of the gas, and the
inertia term peDcJDt can be omitted* from the equation (19.31, 1) for
electrons. Thus (cf. (19.31, 3)) this equation becomes
0 -p t F t -n e ec 0 AH-Vp e -nJim e CjT e +eC e AH), (19.31, 7)
where r, is the effective mean collision-interval for an electron colliding with
ions. In place of the ion equation we may use the equation of motion of the
gas as a whole, n
p
15t " W . + i A » - v p , ( I9 . 3 i, 8)

where now 3 - - nee Ct. (19.31,9)

Equation (19.31, 7) is similar in form to (19.31, 2), save for the extra term
— neee0 A H due to the electricfieldC0AH acting on moving matter. Hence
the diffusion consists again of direct and transverse components, whose
magnitudes are in the ratio i/o>ere, where <ot = eHjme. One might expect also
to be able to introduce direct and transverse conductivities d1, &11, related
to the conductivity # 0 = nee?rjme in the absence of the field H by
&1-&<J(I+<O\T*), &" = &0(otrJ(i+u>li*). (19.31,10)
111
The third conductivity 9 would in this case be
0"i a fli+(*")*/*' - »0»
so that the magnetic field would neither increase the dissipation of currents,
nor alter the effective conductivity when the Hall current is prevented from
flowing.
It is, however, questionable whether the concept of conductivity has any
real significance when o)ere is large. If diffusion is due wholly to electric
fields, and (o^r,) -1 is negligible, (19.31,7) approximates to
0 - -nte(E+ce/\H). (19.31,11)
Hence the electrons move with a mean velocity ce such that no resultant
electric force acts on them; this velocity is just the drift velocity found
* This ominion it not justified in the caie of very rapid (plttmi) oscillation*, for which,
in »ny case, the acceleration of the electron gas cannot be taken as DcJDt.
 368 ELECTROMAGNETIC PHENOMENA IN [19.32
in 19.3. The ions move with velocity c„ where, by (19.31, 8,9),
De —
n mt
' ~Dt "J*1*" -«.<C„AH

Thus the motion of the ions relative to c, is the circular motion, with angular
velocity <af m eHjmt, of a single ion in the field H; this again agrees with
the results of 19.3. However, such a rapid rotation of ct-c„ and so of j
( a nee(c0 - dt)), is not normally contemplated in considering the conductivity.
When collisions are not completely negligible, c 0 - c, is found to tend to zero;
thus ultimately the ion gas and the electron gas move with a common mean
velocity, such that no resultant electric force acts on either. In this case there
are no electric currents. In the more general case in which pressure gradients
exist, electric currents can persist; but these can hardly be regarded as due
to a conductivity, in the normal sense.
The general problem of currents in a ternary mixture of neutral molecules,
ions and electrons has been considered by a number of authors.* A discussion
similar to those for the cases treated above can be given, and conductivities
&1, &u, &m can be introduced as before. The main new feature is the
possibility, especially when collisions are few, of ambipolar diffusion, i.e.
diffusion in which the relative motion of ions and electrons is small, the
main diffusion velocity being a common velocity of all the charged particles
relative to the neutral molecules.
Thermal diffusion is affected by the magnetic field in much the same way
as ordinary diffusion, as the general theory shows. The production of a
transverse electric current in this case is called the Nernst effect.
19.32. Approximate theory of heat conduction and viscosity
The relaxation theory of heat conduction and viscosity in a magnetic field
involves fewer difficulties of interpretation than that of diffusion.
We assume that the gas is in a steady state, uniform in composition, and
subject to no non-magnetic forces, and that c 0 (but not necessarily its
gradient) is zero at the point considered. Then the Boltzmann equation for
f. in a gas-mixture is a - ., a,
= <v
ir cV + «. ( c ' A l f ) -6V
The relaxation approximation replaces Sefjdt on the left by — (/, —/?")/T^
Also in the first term on the right/, is replaced by/J0>; in the second term/,
can be replaced by ft—f,0) without altering its value. Hence the equation
becomes t _/»> t <Y/-/!°>) a/10'
^ - + ^ ( C . A # ) . ¾ ^ - -¾.¾ . (1932. 1)
* A. SchlUter, Z. Naturforuhtmg, 6a, 73 (1951); ]. H. Piddington. Mm. Not. R. attt. Soe.
114, 638, 631 (195s); T. G. Cowling. Mm. Not. R. attt. Soc. tx6, 114 (1956).
19.32] IONIZED OASES 369
We first consider heat conduction perpendicular to H in a gas at rest.
In a steady state the pressure must be uniform, and so n,cc i/T. Hence in
(19.32, l) aflO) , OT

Suppose that H is parallel to Ox, and that the temperature is a function of

z alone. Then (19.32, 1) becomes
T_,dT
t 8z'
the solution of which is
r
x /to. ./T(«1-f), , 1 8T

The corresponding heat flux Jp,cjc, carried by the rth gas has components

ipjr.to, = - - 4 - j -5-- (19-32, 2)

in the direction of the temperature gradient, and

(I932,3)
***• aisrr+tRto
transverse to that direction. Thus when the temperature gradient is per-
pendicular to H the thermal flux down the temperature gradient is reduced,
and in addition there is a flux perpendicular to both H and VT (the Righi-
Leduc effect).
In discussing viscosity we consider a gas of uniform temperature; the
term 8flv)/dr depends on the gradient of c 0 , which does not vanish at the
point considered, though c 0 itself does. We find

c J
' br ' kTe,e* '

where e denotes the tensor Vc„, as before.

Suppose that H is parallel to Ox, and that either c„ is parallel to Ox and
depends only on z, or it is parallel to Oz and depends only on *; in either case
e reduces to the pair of equal components e„ and e„. Then (19.32,1) becomes

whose solution is
 Copyrighted Material

370 ELECTROMAGNETIC PHENOMENA IN [19.4

Thus the rth gas contributes to the components p„, p^ of the pressure
tensor the amounts given by
(I932 4)
^ = -¾¾ *
2p,tojr\ex. . .
PM* = ~ TVr.rx-
i +<#T (»«». 5)
The magnetic field accordingly reduces the contribution of the rth gas to
the viscosity, and produces a cross-stress perpendicular to the usual viscous
stress and to H.
Similarly if c0 is parallel to Oy and depends only on Os, so that e reduces
to the pair of equal components eyi and en, the rth gas makes viscous con-
tributions to pyj, pyy and pa given respectively by
jtfVV> 4P,0JtTfgw 4/>,«v!ci» /TA„ «

As in 6.6, the collision-intervals rt effective in heat conduction and

viscosity need not be identical.

19.4. Boltzmann's equation: the second approximation to/for an

ionized gas
We now return to the general theory. The solution of Boltzmann's equation
by successive approximation will be given for a binary gas-mixture. Boltz-
mann's equation for the first gas is

= - A ( / A W n ( / i / . ) . (194. i)
where the integrals /are as defined in (8.2, 2, 3). Because of the large mole-
cular velocities, the term on the left which involves Cx A His in general much
larger than the other terms on this side; the first approximation to (19.4, 1) is
accordingly taken to be

^ ( Q A tf)|P = -Ji(//i)-J'aCAA). (»9-4.a)

and the second approximation is

+iifCA^M!! 3S?c-i*
=-AWWit/w-^t/mwi^n (194,3)
Copyrighted Material
19.4] IONIZED OASES 371
0 1
The novel feature is the retention of terms involving/} ' and/J * respectively
in the differential parts of (19.4, 2, 3).
The general solution of (19.4, 2) is of the form

& = Wl (2¾)' CXP( ""»^ 2 ^ (I94, 4)

as can be shown by a generalized form of the //-theorem;», and T, which are

arbitrary constants in the solution, are taken to be identical with the number-
density of the first gas and the temperature. Writing
/ | " - / r » ^ , /P»-/P>«P». (19-4. S)
we can put (19.4, 3) in the form

" "m/° >(ClAM) • S ? " " ! W ^ ~ n i " a ' " ^ + ^ (I94,6)

the integrals I being as defined in (4.4, 3, 4).

The term involving,/in (19.2,5) corresponds to the term involving Cj A H
in (19.2, 3): hence, corresponding to the retention of/™ in the differential
part of (19.4, 3) we must substitute j a ) for/ when obtaining a first approxi-
mation to D0c0/Dt from (19.2, 5). On the other hand, the term involving,;*
in (19.2, 6) corresponds to the term involving c0 A H in (19.2, 3), and so in
obtaining the first approximation to DTjDt from (19.2,6) we put,/ =,/ ro »• o.*
The first approximations to equations (19.2,4-6) are thus

^ + « 4 . c , = o, ^ +„,^.c0«o, (19.4,7)

=
* ~5f ~PdrC° =
-*" r {£- c o- (»9-4. 9)
0>
Using these relations, and substituting for /i from (19.4, 4), equation
(19.4, 6) can be put in the form
/ H ( * ! - t ) C 1 . V l n r + x r 1 r f 1 , . C 1 + 2«> 1 :Vc 0 }

-nViim-ninthtW + On ('9-4. ")

• It may readily be verified that (19.4. 7-9) can be deduced by multiplying (19.4, 6) in
turn by 1, m, C,. and jm, C[< integrating over all values of c„ and adding to the corre-
sponding equation for the second gas.
372 ELECTROMAGNETIC PHENOMENA IN (19.4
where in the present problem

(19.4, 11)
Similarly from Boltzmann's equation for the second gas we obtain the
equation
fn«l-l)Ct.mnT-xildu.Ct+2V°tVt:Vc,}

-^/,(^)-^^/,,(^)+^). (,9.4, la)

1
Since / » - n^Cl ' + «,*,&'
- //W^iCi*!* J/W-.C,** (19.4.13)
each term on the right-hand sides of (19.4, 10, 12) involves the unknown
functions <I>|1) and <t>i" linearly. The left-hand sides are linear functions of
VT, Vct and </„; hence
•P»--ii 1 .Vbr-aB l :Vfp-«,.dUl
{,9
•,*> - -W,.Vlnr-aB,:Vc,-l» t .rf w / + '4)
where the quantities A and Z) are vectors and the quantities B are tensors,
which involve only the vectors C„ C„ If and scalars. Now H is a rotation-
vector in the sense of 1.11; in mechanical equations of translation it appears
only through its vector products with other vectors. Thus the only true
vectors (not rotation-vectors) on which Ax can depend are scalar multiples
of C„ Cx A H, (C, A H) A H, etc.; only the first three of this series need be
considered,since [{CIAH)AH]AH„ .jpc^M.
Hence we can write
Ax - A\(Cl,H)Cl+Ai1(CvH)ClAH+Al11{CuH)(ClAN)*H,
(«9-4, IS)
with a similar equation for At. Similarly
D, - DKCfcJiQQ+DPfCi.JSQC^A H+Otn(C1,H)(ClAH)AH.
(19.4,16)
The corresponding equations for B„ B, involve six independent non-
divergent symmetrical tensors, formed by combining the three vectors
CltCiAH,(CxAH)AH.
19.42] IONIZED OASES 373

19.41. Direct and transverse diffusion

Suppose for the moment that VT and Vc0 vanish. Then to the present
approximation

- - J / r C x ^ C j + Di'QAff+DPtdAffjAffJ.rf,,^
--//P , C l C 1 .{Dl^ i +DI I JfArf I 1 +D| u HA(JfArf I t )}db 1
- -ijfimC\{D\du + D[lHAdlt + D\"HA(HAdlt)}dcv (19.41,1)
Thus the velocity of diffusion has components, not only in the direction of
dlt, but also in the directions of HAdlt and H/\(H/\dlt). To see the
meaning of this result more clearly we consider certain special cases.
Suppose first that dlt is parallel to H; then H A «flt ~ o, and so
l
<Di> = -DlC^dn,
the terms in D]1, D\u vanishing. Thus C„ Ct and in consequence also,/0'
are parallel to dlt and H, and
/»Alf-o, (Clf,H).^-~-(ClAH).mdlt-o.
Hence equation (19.4, 10) reduces to
xrVfrfis-c, - - wf /tC^i^) - »t«, /i,^!1'+*«*»).
which does not involve H, and is the same as the equation obtained when the
magnetic field is absent. The rate of diffusion parallel to the magnetic field is
therefore unaltered by the presence of the field.
Suppose next that dlt is perpendicular to H; then
dit. [(Ct A H) A H] - - H*dlt. C„
and so 4>p can be expressed in the form
• P = -^.(DJC + DFQAff-DP/PCO,
or (as the term in D\n can now be absorbed into the term in D\) in the form
W - - ^ . ( D l C x + DFCxAH). (19.41.2)
Hence in this case the velocity of diffusion possesses two components, one
parallel to dlt, the other parallel to HA dxt< i.e. perpendicular to both dlt
and H. These are the velocities of direct and transverse diffusion of 19.31.

19.42. The coefficients of diffusion

In the general case, when dlt is neither parallel nor perpendicular to H, the
velocity of diffusion is the vector sum of the velocities of diffusion produced
by the components of dlt parallel and perpendicular to M when acting
 374 ELECTROMAGNETIC PHENOMENA IN [19.42
separately. Thus only the cases when dlt is parallel or perpendicular to H
need be considered; and as the first case is covered by the ordinary theory
of diffusion, it is necessary to discuss only the case in which dn is perpendi-
cular to H. Then G>il) and <t>i!) are given by (19.41,2) and a similar equation,
and/ 1 ' is expressible in the form
yw - -(Lirf lt +L»i!TA <*„), (19.42,1)
1 n
where L , L are constants. Also (19.4,10) takes the form

We substitute in this the known expressions for Qi1', Oi1' and j™; then,
equating coefficients of dlt and of H A dlt, we obtain the two equations

+»i»tA«fZ>|Ci+0iCt).

If Di + tKDF-Ci. Dj + ii/Z>Jr*-k, V+iHLllaZ, (19.4*.*)

these two equations can be combined into the single (complex) equation

(19-44.3)
Similarly from (19.4,12) we derive the equation

+^1/11(^,+^(¾. (1942.4)
Equations (19.42, 3 and 4) replace the equations (8.31, 5). It can be deduced
from (19.42, 1), (19.4,13) and (19.41, 2) that Z is connected with £„ £,
bytheequation z _ ^^dc^\et]fntC{dct. (19.42, S )
To solve (19.42, 3, 4) we use a method similar to that of 8.51. We
assume that £t, £, are expressible in series of the functions ai"', ajjp> of 8.51,
the coefficients being now complex; the equations from which the coefficients
are to be found are derived by multiplying (19.42, 3,4) by a^» and <4W,
integrating over all values of Cj and c, respectively, and adding. In particular,
a first approximation to the solution is
&Q-4«r. &<*-w. (19-42.6)
19.42] IONIZED OASES 375
where </t is determined from the equation

0,
+«/. tH I"J f/fWf». a/o'rfc, + £* J/i <>. <></c,l + «*</„{««», a™}.

It was shown in 8.51 that

lfPmiCl.afl>tkl + j/PmnCt.a!»det = o;
also, by (8.51,16), ,„
2i»n0lL>ltji
where [£^,], denotes the first approximation to the coefficient of diffusion.
Inserting these values and effecting the remaining integrations, we get

K«j?H).
wherc M a % ^ ^ A ) T .«L?^Eab. (I0 . 42(7)

Thus J^kTVpnT^w)

Combining this with (19.42, 2, 6) and (19.41, 2), we see that

The corresponding velocity of diffusion is given by

c-c = (^Q* 7 ')*"^ (rfM-(w/g)iyArf,,j
1
* 301/¾ I I+W*T« /

ppr (dlt-(wrlH)HMln) .
^7, F+W "• (»9-42.»)
Thus the direct diffusion is reduced in the ratio 1/(1+ w*r*), while the trans-
verse diffusion is on times the direct. Similar results hold for the direct and
transverse electric currents, and for the direct and transverse conductivities
of the gas.
376 ELECTROMAGNETIC PHENOMENA IN (19.43
In the case when mt/ml is negligibly small, equation (19.42, 8) is equivalent
to (19.31, 7). This is a special case of a general result; for a mixture of any
number of gases the formula (19.31, 1), with DeJDt replaced by D0cJDt,
agrees with the diffusion formula derived from the first approximation to
exact theory.*
The collision-interval r in (19.42, 7) is the one effective in diffusion; by
(14.2,1), for rigid spherical molecules it is
3 K + W,)/ «,«, \*
8p<r*lt {znkTXm^mt)) '
whereas in 5.21 the (differently defined) collision-intervals of molecules
for collisions with molecules of opposite types were found to be
T _ 1 / mlmt \* T _ _ i _ / «i«! \*

Thus r - l(PtTi+PiTt)/P> so that, apart from the factor {, r is a weighted

mean of T, and rt. In the limiting case when mjml is small, corresponding
to an electron-ion mixture, r - f T,. The electrical conductivity when there
is no magnetic field (given by (19.11,4)) is connected with T by the equation
& - >W'i««-etm i ) t T l l pm 1 m t . (19.4a, 9)
In 19.31 it was pointed out that, when a polarization electric field com-
pletely prevents transverse diffusion, direct diffusion in a binary gas is
increased to its value in the absence of a magnetic field. This result is only
approximate, since (19.42, 8) and (19.31, 7) are correct only to a first
approximation. If further approximations to £, and ¢, are used it is found
that the polarization field increases the direct diffusion, but not up to the
value when there is no magnetic field. For the Lorentz approximation and
for 'hard' molecules, the maximum reduction of the diffusion by the
magnetic field when this electric force acts is in the ratio 977/32 = 088,
corresponding to rigid elastic spheres; however, for electrostatic interaction
the reduction is in the ratio 377/32 • 0-29.

19.43. Thermal conduction

The general theory of thermal conduction and thermal diffusion in an
ionized gas subject to a magnetic field is similar to that of diffusion. Thus, for
example, the rate of thermal conduction and thermal diffusion parallel to
the magnetic field is not affected by the field; if VT is perpendicular toH,
on the other hand, the rates of thermal conduction and thermal diffusion in
the direction of the temperature gradient are reduced, and additional
conduction and diffusion occur in the direction of HsVT. The general
theory will not be given in detail, because of its complication; it involves
* T. G. Cowling, Proc. R. Soe. A, 183, 453 (1945)-
19.43] IONIZED OASES 377
an
combining the methods of 18.42, 18.43 d 19-42. We therefore merely
quote the results for a binary mixture.
Suppose that the gas is at rest, and that the magnetic field is perpendicular
to the temperature gradient. Then the contributions to the heat flux due to
the two constituents in the mixture may be written
-a\VT-a?H^T, -aJVr-aJ'HAVr. (19.43,1)
1
Here a\, J/a, are the contributions from the first gas to the direct and
transverse thermal conductivity, and similarly for a{, Hal1. If
a, = a} + iHci1, a, = a\ + ttfoj1 (i943> 2 )
the quantities a„ a, satisfy, to thefirstapproximation,
ai
fe + t w i i + ^ { 6 M ! + ( 5 -^ ) M J + 8 M i H
and a similar equation (cf. (18.42, 6)).
These equations have not exactly the same form as those derived by
relaxation theory, because of the appearance of a, in (19.43,3)>an^ aim *he
corresponding equation for the second gas (compare 18.42). When mtjml is
negligibly small, however, the results become similar to those given by the
relaxation equations. Normally in this case ¢4 is large compared with a,,
and is given by TX1(5-4B)1 2m,!
a
4ta^^x) ' wJ =I -+ w (I943,4)
The expression in the brackets {} in this case denotes i/[A0]„ where [A,],
is the first approximation to the thermal conductivity in the absence of a
magnetic field. Write ,, , . ,. , v
T, = 2114^,/5^,. (»943. 5)
Then a, = [AJ,/(i +io>trt), and
°\ = [A.],/(i + WJTJ), ffoj1 - - [AJ,o>,TJ{I + UJTJ) (19.43,6)
in agreement with (19.32, 2, 3).
The thermal diffusion velocity is of the form
C - C , = - (x,x,)-1/),,^¾. V In T+k1,!!* AVln77, (19.43,7)
where k j., //k" are direct and transverse thermal diffusion ratios. Write
k£. + t/JkJJ a KT. (19-43.8)
Then KT is given to afirstapproximation, in terms of thea„ a, of (19.43, 3),
by
["rli = rc(x,a, M, - x,a, Af.V/.fDJ, (19.43,9)
(cf. (18.43, 3))-
378 ELECTROMAGNETIC PHENOMENA IN [19.44

19.44. The stress tensor In a magnetic field

The effect of a magnetic field on the pressure distribution is more com-
plicated, and less interesting to the physicist because the corresponding
phenomena cannot be observed in metals. As with heat conduction, the
results do not agree exactly with those of relaxation theory, because of the
interaction of terms representing the viscous stresses of the different gases.
However, they do possess a general resemblance to the relaxation results;
and there is identity of form in certain special cases, notably that of a binary
mixture in which tnt/ml is negligibly small. In this case the viscous stresses
are due wholly to the first gas. Let the viscosity of this gas be /«„ and write
fix - pxrv If H is in the x-direction, the stress-component^, has the same
value as when the magneticfieldis absent. The other stress-components are
given to a first approximation by

where e • Vc0. These results generalize those of (19.32,4-6).

19.45. Transport phenomena In a Lorentzian gas in a magnetic field

The Lorentz approximation is of peculiar importance in considering the
motion of electrons in a slightly ionized gas. The modifications in the theory
when a magnetic field is present are readily made. Suppose that electrons
or light molecules constitute the second gas of a binary mixture at rest; then,
approximating as in 10.5, we can put (19.4,12) in the form
{(*i-t)C,.Vln T-x;ldu.CJ

thefirstterm in the bracket on the right of (19.4, 12) being omitted because of
the smallness of m,.
Substitute in this equation
•W" - -(^JC,+^i I C 1 Alf).Vlnr-(DiC 1 +Z)J I C l A»)<'if
 Copyrighted Material

19.5] IONIZED OASES 379

Then, integrating with respect to b and e as in 10.5, and equating coefScients
of V 71, # A vT, dlt and HA <*M, we find
<g\-$ = -u>iHA?+A\lTi,
o = (0tA\+HA¥lTt,

0 -* ^Dl+HD^JTa.
Here w2 = -etH(mt = eHjn^, -e being the electronic charge; also T, is a
function of C2 given by
u = {"A^VCC)}-1 = Kct)ict (IMS. »)
in terms of the' free path' l(Ct) of 10.5. From these equations it follows that
A\ = -HAV/toM = ( t f l - M i + a>?r§), (19.45,2)
0 | = -ffl^/w,T t = -TJ/X 2 (I + W!T|). (19.45,3)
Thus, for example, the rate of diffusion is given by
«*c2=j/wc2<fca

-^f/P^-Df-^-"^ 7 ^)^. (,94S,4,

If 72 is independent of C2, this shows that the direct diffusion is reduced in
the ratio 1/(1 +W|T|), and the transverse diffusion is &/2T2 times as large as
the direct. In general r 8 is not independent of Cs, and it is impossible to
evaluate the integrals in finite terms; for example, for rigid elastic spherical
molecules it is l(Ct) which is independent of Cv and T, OC I/C 2 . However,
the formulae are still reminiscent of those of 19.31; they still indicate that
for ordinary magnetic fields and normal densities the transverse diffusion is
proportional to H and the reduction in conductivity to H1, while for very
large magnetic fields or very small densities the direct diffusion is propor-
tional to i/H* and the transverse to i/H.*
Similar results hold for thermal conduction.
19.5. Alternating electric fields
In the theory given in earlier chapters it was assumed implicitly that the
acceleration F of a molecule is a relatively slowly varying function of the
time. When, as in the propagation of radio waves, an alternating electric
field acts, whose period is comparable with the collision-interval for electrons,
the theory requires considerable modification.
Consider a uniform gas under no forces other than the alternating electric
* For the evaluation of the integrals for rigid elastic spheres, see R. Gans, Annln Phys. 30,
203 (1906), and L. Tonks and W. P. Allis, Phyt. Rtv. 52, 710 (1937).

Copyrighted Material
380 ELECTROMAGNETIC PHENOMENA IN (19.5
field and a constant magnetic field H. To satisfy Maxwell's electromagnetic
equations, the magnetic field should actually have an alternating part; but
in practice this is small compared with the slowly varying terrestrial fields,
and it will accordingly be ignored.
The electric field can be divided into components parallel and perpendi-
cular to H; the diffusion due to the first of these can be shown, by an argu-
ment similar to that of 19.41, to be the same as if the magnetic field were
absent. Thus we have only to consider diffusion due to an alternating field
in the two cases when H is perpendicular to the field, and when there is no
magnetic field; and since the latter is a special case of the former, it is
sufficient to work out the theory when H is perpendicular to the electric
field. For reasons which will appear later, it is convenient to work with an
electricfieldnot always directed parallel to a given direction, but circularly
polarized in a plane perpendicular to H.
Consider the motion of a single particle of mass m, and charge *,. For
convenience let H be parallel to Ox, and let the electricfieldbe given by
JB •> (o, -Eunst, Ecottt). ('9-5.')
Then the part Fx of the acceleration of the molecule which is due to the
electric field is „ , -, . . „ ,.
F1 - (0, -FjSinrf, F,cosrt)

- lo, --*— smrf, -j—cosrf). (19.5. 2)

The equations of motion of a particle are

* - o , j)" Wji-Fjsinxr, * • -<a1^+Flco»st
(cf. (19.3, 2)), where wx - elH/ml. Ifj? - v[ andi - w[ initially, whent "O,
it follows by integration that
F
J* — cjcosw.f+w;tin«,< + ;—i—; (cos tr-cos <>>,{),
* * * * (<*»i+*)
F
t - -rJsinw1r+toJcosw1r+T——r(sinrt+sinw,f).
Now w'x and v[ denote the initial velocities of the particle in directions
respectively parallel to B, and perpendicular to both E and H (actually the
direction is that of £ A H). If the velocities parallel to E and to EA H after
time t are denoted by wx and v„ then
w1 - *cosrt-j>sin»r
- wicosK+x)*- ( ^ 1 - - i - J sin(wj+»)f,
t>! — j? cos ft+4 sin rf

" (•l"5^i)C0,(^»+,)*+"i"«>(«i+«)* + j ^ .
19.51 IONIZED OASES 381
Comparing these equations with (19.3, 4), we see that the component
velocities parallel to E and to E A H vary in the same way as if the acceleration
Fj was constant in direction and of constant magnitude elE/m1, and the
angular velocity of spiralling in the magnetic field was increased from G>, to
<i>x+s. Since the gas is uniform, diffusion in it depends on the velocities of
the molecules, not on their positions; thus the velocity of diffusion at any
instant is the same as if each molecule m, were subject to a force in a constant
direction producing an acceleration Fx = elE/m1 and to a perpendicular
magnetic field producing an angular velocityto,+ J of spiralling.
AH the results of 19.31, 19-41 and 19.42 can accordingly be modified to
apply to the present problem. In addition to the direct diffusion current in
the direction in which the electric field is at the moment acting, there is a
second transverse current in a direction perpendicular to both the electric
and the magnetic fields. By analogy with (19.31, 4) the relaxation method
indicates that in a lightly ionized gas the velocity of diffusion of charged
molecules m, relative to the whole mass of gas has components B>r V, in the
direct and transverse directions, where
e,E r, e,E ((o,+s)ij . .
w. = w, I + K + * ) » T J ' C
* = m, I+(W,+*)»T;- <X9-5«3J
Here r, is defined as in (19.31, 3); v, is the velocity of diffusion in the
direction which the electric force had one-quarter of a period earlier.
Similarly in a binary mixture the velocities of relative diffusion in the
direct and transverse directions are, to a first approximation,

*-* - »U I+(, + O,)»T" < f «.4)

respectively, cu and r having the same meaning as in (19.42, 7). No difficulties

of interpretation similar to those met in 19.31 now arise, even when (S+W)T
is large.
If in place of the circularly polarized field (19.5, 1) we have to deal with a
linearly oscillating field _ . _ .. , c\
B
' C-(o,o,£ 0 cos*<). («9-5.6)
we must represent this as the sum of the two circularly polarized oscillations
(o, - $£"„ sin st, \E0 cos st) and (o, - \E0 sin ( - st), \E9 cos ( - *»)).
Then, for example, to (19.5, 4, 5) correspond the equations
(l9 5,7)
* ' - * ' " * * % ^ ( , - H 3 ^ O M + .+(--*>•} -
and g . - ^ . ^ a a r i t f b ) . ('+">; i + ( " - ' > ? ! (,9.5,8)
382 ELECTROMAGNETIC PHENOMENA IN (19.6

the direction of the transverse velocity of diffusion being taken as that of the
first circularly polarized oscillation one-quarter of a period earlier.
These results, based on relaxation theory, agree to a first approximation
with those derived from the general theory, appropriately modified. The
modification required consists of the retention of a first approximation to
dfll)IBt in the equation for/1 1 '; if
/ I " - o(c„0cosrt + /J(c„0sinrt,
the first approximation is taken to be
dfll)
•*£- - *[-a(c„ i)sinrf+/S(c„ r) cos *<].

PHENOMENA IN STRONG ELECTRIC FIELDS

19.6. Electrons with large energies

In certain circumstances the electrons conducting electricity in gases have
energies far exceeding those appropriate to the temperature of the gas. It
was one of the simplifying features of the Lorentz approximation that in the
elastic encounter of two particles of widely different mass there is little
interchange of energy; for example, if an electron of mass m, impinges on a
molecule of mass m, at rest, and the relative velocity; is turned through an
angle X, the energy lost by the electron is
*Mg\i -cosxVfai+'a*) 1 .
which is a small fraction of its original energy, imtg*. Hence in the presence
of a strong electric field a slowly moving electron gains, on an average,
far more energy from the field during a free path than it can lose at an
encounter. In consequence, the mean energy of electrons grows until the
small fraction of it which is lost at collisions balances the energy gained
during a free path. The mean energy in the steady state is thus much larger
than the thermal energy \kT\ the velocity-distribution function of the
electrons differs widely from the MaxweUian form, and the theory of
diffusion in the electric field differs considerably from that of 10.5.
If the mean energy of electrons is sufficiently large, inelastic collisions
occur, in which part of the energy of an electron is used in exciting the
quantum states of internal motion of a molecule. In a gas like helium, which
has a high excitation potential, an inelastic collision can only occur if the
electron has an energy some hundreds of times as large as the thermal
energy, and the energy lost by the electron is large; in diatomic and poly-
atomic gases, on the other hand, a much smaller energy is able to stimulate
the rotational and vibrational motions of the molecules, and inelastic
collisions are important at much lower energies. These facts are of im-
portance in considering the mobility of electrons in gases.
19.61] IONIZED OASES 383
Another case in which the mean energy of electrons may considerably
exceed the thermal energy occurs in the study of the upper atmosphere.
The electrons are liberated from molecules by ultra-violet or corpuscular
radiation; they remain free until they combine with positive ions to form
neutral molecules, or with neutral molecules to form negative ions. The
energy required to ionize a molecule is much larger than the thermal energy
IkT, and the ionizing agent probably possesses energy exceeding this
requisite energy by an appreciable fraction; the excess energy may be used,
partly or wholly, in giving the liberated electron kinetic energy. The mean
kinetic energy of electrons when liberated is therefore likely considerably to
exceed the thermal value; and if electrons do not make too many collisions
with molecules before recombination takes place, the mean kinetic energy
of electrons present in the atmosphere is likely to be larger than the thermal.

19.61. The steady state In a strong electric field*

Suppose that electrons in a large mass of gas are subject to a strong electric
field. After some time an approximately steady state is reached, in
which the energy gained by an electron during a free path is, on an average,
balanced by the energy lost at a collision with a gas-molecule. If collisions
with the molecules are all elastic, these losses and gains of energy are
small compared with the mean energy; and since the direction of motion of
an electron undergoes a large deflection at encounter, the velocity-distribu-
tion function for electrons, though it may differ from the Maxwellian in
form, must be nearly independent of the direction of the velocity.
Suppose first that the electrons are so few in number that their mutual
encounters can be ignored. Consider the equilibrium of a mixture of
neutral molecules, of mass m„ and electrons, of mass m, and charge e„
under the influence of an electric field E. If the state is uniform as well
as steady, the velocity-distribution function / , for electrons satisfies the
equation
Ft^jjif'J't-fMg^de'dc,, (19.61,,)

where a„<fe' = bdbde and a l t is a function of; and b; also

Fj-^E/m,. (19.61,2)

* Early investigations of the steady state of electrons in a strong electic field include
those of F. B. Pidduck, Proe. Land. Math. Soc. 15, 89 (1916); M. J. Druyvesteyn,
Physica. 10, 61 (1930), and 1, 1003 (1934); E. M. Morse, W. P. Allis and E. S. Lamar,
Pkyt. Rev. 48, 41a (193s); W. P. Allia and H. W. Allen, Phyi. Rev. 5», 703 (1937);
and B. Davydov, Phys. Z. SowjUn. 8,59 (1935). Of these all save Pidduck and Davydov
ignore the motion of the molecules; Druyvesteyn in addition takes an average of the
energy-loss at collision of elastic spheres. More recent investigations have taken into
account electron-electron encounters, and have considered the approach to the steady
state, and 'runaway' effects (see 19.66).
 Copyrighted Material

384 ELECTROMAGNETIC PHENOMENA IN [19.61

The velocity-distribution function/, for the molecules can be supposed to
have the Maxwellian form for a gas at rest, i.e.

Since the distribution of electronic velocities is nearly isotropic, we put*

f*-An+**.*An+****'.<Z*/P + .... (19-61,4)

0 1 8>
where each of the functions/j ', / j ' , / j , . . . denotes a function of the scalars
ct, Ft alone, and the quantities c t , ctcit... are the vector and the tensors of
second and higher orders whose components are solid harmonics in the
components of cv i.e. functions satisfying the equation
{/*(a cPw EPw

We substitute from (19.61, 4) into (19.61,1), and equate terms involving

scalars only, terms involving the vector F„ and terms involving the tensor
e
F,Fj, etc.: then we get

1(/^+^1¾ -fjvyiF-fjnnM&dc (19.61,5)

ct dc, " * 4 &•
-ffUiA^Ft.c'J-fiA^Ft-cdgctvde'dd, (19.61,6)

and a series of similar equations. Since the second term of (19.61,4) is small
compared with the first, it is natural to expect that the third and subsequent
terms are small compared with the second; they will accordingly be ignored
in subsequent work. Thus we have to deal with (19.61, 5) and (19.61, 6)
alone, and the second term on the left of (19.61,6) can be omitted. Moreover,
since mt/m1 is very small, only the least power of mjm1 which appears on
integration need be retained. In this connection it may be observed that
mxc\ <ffjjt*,and so cjct can be regarded as at least as small as (tnjm^i.
In (19.61, 6) it is sufficient to put ct = c't = g, cY = c[, which is equivalent
to ignoring mjn^ completely. Then, transforming as in 10.5, we find

* The notation here i* rather different from that used earlier. A second approximation to
/ , was denoted earlier by/30>+/?'; it is now represented byfln+Ft.Ctfil): and so on.

Copyrighted Material
19.61] IONIZED OASES 385
As in (1945, 1), an equivalent mean free path of electrons of speed ct is
defined by /«,) = ! / « ! » , ) . (19.61,7)

Then it follows that 2 L - = -c*/,1'/^*)- (19.61,8)

dct
If the same approximations are made in (19.61, 5), the integral on the
right vanishes. It is therefore necessary to proceed to a further approxima-
tion; we retain all terms not small compared with m^nti.
Multiply (19.61, 5) by dc% and integrate over all values of c , such that
c, < v. Then we get

-*»*Mfir». (19.6.,9)
Using a transformation similar to that of 3.53,* it can be proved that
jJ!nfrg<tltde'dCldct,
integrated over all values of the variables such that c, < v, is identical with
JJJ/i/rVii*'<fci<fci
integrated over all values such that c't < v. By (341, 6, 7), since;' ~g,
cf^cl-zMfi. (gn-g'n)
= <\ - 2M,(c, + Mtgu). (gn -g'n)
and so, correct to terms of order rnjm^,
c'% = ^l-2(Ci + Mtgtl).(gn-g'n)}
= €,-(^ +MtgtJ.ign-g'nyci-Hd.ign-g'nWIci. (19.61, 10)
In this write c, = c 2 a, where a is a unit vector (so that dcx = c\dctda).
Then, by (19.61, 10), the condition c't < v is equivalent to ca < v + &v,
where Afl is a function of v, a, c, and e'; Av/v is a small quantity of order
cjc2 or (m t /w,)i, and (since £/c, is nearly unity)
Ar = C j . ( a - e ' ) +smaller terms. (19.61, 11)
Thus (19.61, 9) is equivalent to

i*F\v>tf\v) = / / / ( J J + A V i / i 0 , ^ « , ^ S ^ . ) de'd*dcv

We substitute in this the Taylor expansion

* The difficulties about convergence can be avoided if it is supposed as in 3.6 that

encounters are neglected in which the deflection of the relative velocity is less than an
assigned small quantity.
386 ELECTROMAGNETIC PHENOMENA IN [19.61
The third, fourth,... terms of the expansion, after integrating with respect
to ct, give quantities of order (A»)\ (At;)4,..., which can be neglected. Thus

SJ
+ ^\l\{\yV(e*-v)tea»w)d*'*dev (19.61,12)

In evaluating the second term on the right we can approximate by putting

g<*M 3 " <*ii(«) «*. Aw = d . (» - e').
Then, since a.e' » cosx to the same approximation, the term becomes,
using (19.61, 7),

kT.afPwct, si*, J
- 4 ^ 1 , - 0 » • ' y M l ( 1 -ctMtibdbde
kT p» 8/}w(t>)
* « ! /(») dv
In the calculation of the first term on the right of (19.61,12) it is necessary
to take into account the smaller terms on the right of (19.61,11); the inte-
gration is very involved. It is, however, possible to determine the value of the
integral indirectly.* Let it be ^(o); the value of ^ will not depend on Ft or
on/i 0) . Then (19.61,12) becomes

If F, » o, this equation is satisfied by

For this to be so, we must have

Substituting this expression and replacing the variable v by ct, we have

* A device timiUr to thii wet imd by Dtvydov {he. eU.).

19.61] IONIZED OASES 387
This equation can be given a simple interpretation. The left-hand side,
multiplied by 47rm,c$<fct, can be proved equal to the rate at which electrons
of speeds between ct and cs + dct gain energy from the electric field per unit
volume; the second term on the right, similarly multiplied, gives the loss of
energy of these electrons in encounters if the molecules mt are at rest before
encounter; and the first term represents the correction due to the relatively
slow motions of the molecules.
From (19.61, 8,13), it follows that

whence

(19.61,14)
where A is a constant. For small values of Ft, these expressions approximate
to those obtained by the Lorentz method; for large Ft, when the mean energy
of an electron is large compared with \kT, they approximate to

""-M-Jw')' ""-ST/-" '•'•"••^

The condition that (19.61, 14') should apply is that kT should be small
compared with ntiFlPjyl in the range of ct for which/J" is appreciable;
since mtc\ > jkT when this is so, the condition is that the mean value of
etEl shall be large compared with 2kT(mJm$.
If the molecules are rigid elastic spheres, so that the free path is inde-
pendent of c„ (19.61, 14') becomes

(19.61,14")
This result was originally derived by Druyvesteyn. It indicates that the
number of electrons with energies large compared with the mean energy is
much smaller than in a Maxwellian distribution with the same mean energy.
Using this result, it follows that A is connected with the number-density nt
by the relation
% .4.-4^=,,(-^,)4^
388 ELECTROMAGNETIC PHENOMENA IN [19.61
and that the mean energy of an electron is

and the mean velocity of diffusion of the electrons is of magnitude

(19.61,16)
This is small compared with the value

given by the Lorentz approximation.

The formula (19.61,14*), however, gives too few electrons of high energies
for actual gases. It is clear that the actual distribution of velocities is very
sensitive to the law of interaction between electrons and molecules; for
example, if the force between them varies as the inverse fifth power
of the distance, on the classical theory 4(ct) is proportional to ct, and the
distribution function is Maxwellian, though with a 'temperature' con-
siderably higher than that of the gas-molecules. It is possible to calculate
l(ct) numerically for actual gases, from the experimental results on the
angular distribution of electrons scattered by molecules, using (19.61,7) and
#¥(c«) - 2»f/(i -cosAt)<x„(cf,x)sinx</x,
where at^ct,x) is a scattering amplitude as in 17.2. From this the distri-
bution function for any Ft can be determined numerically, and the mean
energy and rate of diffusion of the electrons can be calculated. Since both
of these quantities are also measurable by experiment, a direct check on the
theory is available.
Calculations of this type have been made by Allen* for the motion of
electrons in helium, argon and neon. The calculated values of the rate of
diffusion agree fairly well with experiment for values of Ftl which are not
too large, though they are in all cases rather too low; the calculated mean
energies are much larger than the experimental, but it is suggested that this
may be due in part to an incorrect interpretation of experiment. When Ftl
is very large, fast electrons are losing energy by inelastic impacts, and the
above theory does not apply. In this case experiment shows that the mean
• H. W. Allen, Pkyt. Rn. 5». 7<>7 (>937).
19.62] IONIZED OASES 389
energy approaches a constant value as the electric field increases, and that
the velocity of diffusion increases more rapidly than the above equations
indicate.
The distribution of electronic speeds in helium is illustrated in Fig. 14.
For this gas the quantity l(ct) has a minimum corresponding to electronic
energies of about 2} electron-volts, and increases rapidly for energies in
excess of this. For comparison, the distributions corresponding to Maxwell's
and Druyvesteyn's formulae are also shown; the three curves (with the axis
of abscissae) include equal areas, and correspond to the same mean energy
(584 volts). It is clear from the figure that the calculated distribution is
intermediate between those of Maxwell and Druyvesteyn; its maximum is
sharper than that of Maxwell's curve (III), but it gives more high-energy
electrons than curve II (Druyvesteyn).

II Druyvesteyn
III Maxwell

Electronic speed
Fig. 14. The distribution function for electronic speeds (4"«J/j"): (I)calculatedforelectrona
in helium, mean energy 5 84 V., (II) given by Druyvesteyn's formula, and (III) given by
Maxwell's formula.

19.62. Inelastic collisions

The general theory of motion of electrons in a gas when inelastic collisions
are possible is very complicated, and will not be given here. However, we can
obtain some idea of the order of magnitude of the effects involved in the case
of a gas like helium, which has a large excitation potential. An electron
whose energy exceeds the excitation energy by an appreciable amount does
not undergo many collisions before it loses energy inelastically, and so
cannot obtain an energy much in excess of the excitation energy from the
electric field before this happens; its energy after the collision is therefore
small. Denote by S(ct) the number of electrons per unit volume and time
which undergo inelastic collisions and whose speeds decrease from values
above c, to values below c, because of the collisions. Then S(ct) is small
390 ELECTROMAGNETIC PHBNOMENA IN [19.62
if ct is either very large or very small; but there is an intermediate range
of values of ct which is neither losing nor gaining electrons by inelastic
collisions, and in which, therefore, S(ct) is constant. This range contains
the majority of the electrons.
The left-hand side of (19.61,9) represents the number of electrons per
unit volume and time which enter the velocity-range for which c, < v because
of elastic collisions; theright-handside gives the number leaving the same
range because of the electric field. These quantities now differ by S(v).
Hence, integrating as before, wefindthat (19.61,13) is replaced by

Here the term involving kT is unimportant for values of Ft such that

inelastic collisions are important, and so can be neglected.
It will be supposed that (19.61, 8) is unaffected by inelastic collisions; this
amounts to assuming that the resultant momentum of electrons with speeds
between ct and ct+dct is unaltered by inelastic collisions. This will be
rigorously true in the range in which electrons are neither lost nor gained by
inelastic collisions, and may be taken as a sufficient approximation else-
where. Combining (19.61,8) with (19.62,1), we have

3cJ dct /pu\ ml//

If yw> denotes the value of/} 0 ) in the absence of inelastic collisions, which is
given by (19.61,14'), the solution of this equation is

Ji J
r JOWTH'
where B is a constant. Now when c, is large /J" is small compared with
/"*; also S(ct) denotes the number of electrons with speeds greater than c,
which undergo inelastic collisions, and this, being roughly proportional to
the total number of electrons with speeds greater than ct, tends to zero more
rapidly than/"' as ct increases. For/J" to be small compared with/*9 for
targe c» we must have . . 35(Ct)<fct
JoinctlFlf»'

andso jp^g/fr (,9.62,2)

If / w is small for energies approximating to the excitation energy,/!" is

not appreciably affected by inelastic collisions. For in this case the major
part of the integral in (19.62, 2) comes from the range corresponding to
energies close to the excitation energy; thus if the energy is much smaller
19.63] IONIZED OASES 391
than the excitation energy, (19.62, 2) approximates to
/j°> = /to) x const.

As F , increases, inelastic collisions begin to affect the values o f / ^ corre-

sponding to smaller energies.
A good approximation to fi0) is obtained if we take S(ct) as constant if
E^ < $m,cf < JJlt where Et — El is the energy of excitation of a molecule, and
Ex is small compared with this, and put S(ct) = 0 elsewhere.

19.63. The steady state in a magnetic field

Consider now how the velocity-distribution of 19.61 must be modified when
a magnetic field H acts perpendicular to the electric field. In this case
(19.61,1) is replaced by

(1963,0
In this we insert
/ . = / r + ( F t . e t ) / y ' + (I*AFt).ctfl» (,9.63, 2)
where f^,/^, £jp are functions of the scalars cj, F\, H* only, and we neglect
terms involving cxct on the left. Then, equating terms involving only
scalars, and terms involving the vectors F t and If A F t , we get

S itW)(%> = JJUW-fJF)i***fdc» (19.63, 3)

(»9-63.4)
^ ( H A Ft).Ctfi» = ffviSF(HA Ft). ci - / ^ ( J f A F t ). ct}galtde'dcv
(1963. s)
The integrals are evaluated as in 19.61, giving

• w - g ^ a j / r . (.9.63,6,

We neglect the term involving kT in (19.63, 6); then elimination of/Jl) and
Ogives _n!dJPl(I+&W\ TsESfm,
l+
3cJ ecj\ m\<\) mj"'
392 ELECTROMAGNETIC PHENOMENA IN [19.64
the solution of which is

Comparing this with (19.61, 14'), we see that the effect of the magnetic field
is to reduce the mean energy of electrons, the effect being the same as if
Ft were reduced by a factor equal to a mean value of,/(1 +^ff/*/m}^).
Again, from (19.63, 6, 8),
(,963,10)
^-5¾^ ^--¾^
and so the velocity of diffusion of electrons is

Hi - jjJwJi+a'WAft)***

Thus, if //r, is constant and equal to T, (SO that T, is an equivalent mean

collision-interval), it is found that

* i+«iJ/*i/m$ • 1*9.63. »J
which is identical with the result when the electric field is small. Equation
(19.63, 11) can be taken as a first approximation when l/ct is not constant,
T, then denoting a mean value of ljcv
If / is constant and H is small the results can be expanded as series in H;
if ?M denotes the mean value of ct when H = o, it is found that the mean
energy is reduced in the ratio
1 - 0 - 6 1 8 ^ ^ ^ 5 4 , + ...,
the direct diffusion is reduced in the ratio
i-o*74«t»Wls+-.
and the ratio of transverse to direct diffusion is
1 '085«! H//w, f„ - . . . .
It is to be observed that the direct diffusion-velocity is not much less than
the ordinary velocity of diffusion in a gas subject to no magnetic force in
which the electric force is adjusted to give the electrons the same energy.

19.64. Ionization and recombination

As indicated in 19.6, a steady state is possible in which electrons with high
energies are produced by the ionization of molecules, the electrons losing
energy elastically at collisions with molecules before recombination takes
19.64] IONIZED OASES 393
place. Suppose that the gas is uniform and at rest, and that no electric or
magnetic fields act. Let <x</c, denote the rate of production of electrons
with velocities in the range ct, dct per unit time, by the ionization of
molecules, and let ftftdct be the rate of loss of electrons to this velocity-
range per unit volume by recombination; a, ft are supposed to depend only
on the magnitude of ct, not on its direction. Then the equation satisfied by
the velocity-distribution function/, for electrons is

| * = cc-fc+jfiAtt-fJlgciude'dCr. (19.64,1)

If the state is steady, dfjdt = 0. Again, if the integral on the right is

multiplied by dct, and integrated over all values of c , such that ct < v, then,
by the argument used in simplifying (19.61, 5), the resulting expression is
CqUalt
° <kTv>8ft(v) mt^f/A

Thus (19.64, 1) is equivalent to

It is, unfortunately, impossible to solve this equation completely.

Approximate solutions can, however, be given in two limiting cases, and
these serve to indicate the properties of the general solution.
Suppose first that an electron undergoes few collisions during its free
life-time. Then the mean energy of an electron is large compared with \kT,
and the term involving AT1 can be omitted from the right of (19.64, 2). The
solution of the equation can then be found in terms of integrals; for example,
if ft is a constant, and l(ct)/ct is a constant T „ the solution is

/ t = ftC»^m,)'./»-»I f " « ^ ! j c j - ( m l / m t ) r i 4 < f c 1 + i 4)

where A is a constant. For the integral

\ftdc* s 4*1 <\ft&%

to converge for large values of ct it is necessary that A = o, and so

ft = CiCn./m,)r,^-S f" a !?il« <;}-<«./»*> r , ^ ( , „ 6 ^ 3)

The number-density nt of electrons is given by

nt = 47T r%t<»i"»t>'i>-« I p a ^ ' r l - W " ^ ' • / < & , ! < / < : „

394 ELECTROMAGNETIC PHENOMENA IN [19.64
whence, by a partial integration,

-- f
». = ^{Lw»»>T»> pocJ-W')'«'<&,]" + f " ^ * , }

- jjo *4<k» (1964.4)

as can also be proved by multiplying (19.64, 2) by dct and integrating over
all values of c,. Again, the mean energy of an electron is

™*\\<\dct = » ( l + 2 ^ W i T t ^ J 0 " ^ ^ . 09*. 5)

by a further partial integration. Comparing (19.64,4 and 5), we see that the
mean energy of an electron is 1/(1 +2ml/m1rt/9) times the mean energy of
a liberated electron; and so the approximation (19.64, 3) is valid if the mean
energy of a liberated electron is large compared with
\k1\i + 2mJmlTtfi). (19-64,6)
Next suppose that an electron undergoes so large a number of collisions
during its free life-time that its mean energy is nearly equal to \kT. To a
first approximation

/, -/r - «* (^)'«p( -«vva*n

while a second approximation is given by the equation

-*-*n-£^*3?4
On integration, this equation gives

-JV«VA-'!?(g+¥/.). ,,,.64,,,
the constant of integration being chosen so as to make/, finite when c, is
small. If ct in (19.64, 7) is made to tend to infinity, it is found that
m
/;
(*-m wct-o, (19.64,8)
from which nt can be determined. This equation expresses the fact that
ionizations and recombinations balance each other.
The general solution of (19.64, 7) is

where B is a constant, whose value can be determined from the condition

19.65] IONIZED OASES 395
Equation (19.64, 9) can be shown to apply if 2mi/m1Ttfi is large, and the
mean energy of a liberated electron is small compared with (19.64, 6), r t now
denoting a mean value of l(ct)/ct.
The value of/a when electric or magnetic fields are present is very difficult
to derive. It may be observed, however, that the relaxation results derived
in 19.31 did not depend on any assumption that the distribution of velocities
is nearly Maxwellian; and these results have been shown by the exact
theory to be nearly accurate for a mixture of electrons and heavy molecules
both when the electric force is small, and, with an altered collision-interval,
when it is large. It is therefore not unreasonable to suppose that they apply
also to a gas in which ionization and recombination are taking place, using
the appropriate value of the mean collision-interval.

19.65. Strongly ionized gases

Encounters between pairs of charged particles can be neglected only in a very
slightly ionized gas. The effective radius for electrostatic interactions at
ordinary temperatures may be as much as a hundred times that for inter-
actions of charged with uncharged particles. Thus encounters between pairs
of charged particles are dominant when more than a small fraction of one
per cent of the molecules is ionized.
To illustrate the case when electrostatic interactions are dominant, we
consider the effect of a strong electric field in a fully ionized binary gas
(plasma) composed of electrons and positive ions. The electric field pre-
dominantly accelerates the electrons, giving them energy of directed motion.
Encounters with heavy ions quickly randomize the direction of this motion
but (cf. 19.6) produce only a slow interchange of energy between ions and
electrons; hence the mean energy of electrons may increase well above that
of ions. Electron-electron encounters, though not altering the mean energy
of electrons as a whole, make the electron velocity-distribution rapidly
approach the Maxwellian form corresponding to this mean energy. The
temperature slowly increases because of work done by the electric field.
The suffixes 1, 2 are taken to refer respectively to the ions and electrons.
In a uniform gas at rest, the equation (19.61, 1) for electrons is replaced by
, + /
! v£ = xff(f.fi-fMg«**'dc.; (19.65,1)

the term Bfjdt is included to take account of the slow heating of the electron
gas. As in 19.61,/] is assumed to have the Maxwellian form

/i= i
" ( 2 ^V 1 ) , '" m,c,, ' afcr,> (I9 6s 2)
- '
the mass-motion of the ions being ignored; also (cf. (19.61, 4))/1 is given by

'. =fim + Ft.cifi» + FtFt:cZctfP+..., (19.65, 3)

X4-a
396 ELECTROMAGNETIC PHENOMENA IN [19.65
where/p'./j",... are functions of cr On substituting from (19.65, 3) into
(19.65, 1), and approximating as in 19.61, we get the coupled equations

+//{/*/!*-/W>*«s*'&. («9-65.4)

+jj{rrjfr^+/rjw^-rvrtea-jrtj°^g<hMde. (19.65,5)
w, 1
In these / ,/» ' stand (or/f>{e), fjf\e)\ the term d/P/dt is omitted from
(19.65, 5) as representing only the slow change of fj?\ which is itself small.
The first term on the right of (19.65, 4) is small, since c'v c't differ only
slightly from cv ct; the second term, which is dominant, ensures that/J0) has
nearly the Maxwellian form

1* °n% ( ¾ ) 1 ° ^ - ^ ^ - (i9,6s 6)
*
corresponding to an electron-temperature Tt not necessarily equal to Tv
The temperatures Tu Tt are defined as such that
lmi7\-lkTv \mt7\-\kTt. (1965,7)
In (19.65, 5) we may approximate by using (19.65, 6), and also by writing
c
'i " c n c't °* ct m *he fir8t integral on the right. Then this integral becomes

independent of Tv and the equation becomes identical with that giving /J"
(and hence the electric current and conductivity) in a gas in a weak field at
temperature Tt. Thus/j" can be regarded as known from our earlier work.
In particular, by (14.1, 1) and 10.34, ^ e velocity of diffusion c, is given to
the first approximation by
ct - 2tntFtr{Tt)lmlt (19.65, 8)
where r{T^~i(2kT^mJ{i(>n\nlw\e'*e'ti\n(i+^l)}. (19.65,9)
The charges on the particles are here denoted by e'lt e't to draw attention to
their being measured in electrostatic units.
Equation (19.65,4) may be used to determine not only the deviation of
/p» from the Maxwellian form (19.65,6) but also the value of dTJdt. The
latter* is found from the equation of energy of the electron gas, obtained by
multiplying (19.65,4) by }*»,«$ and integrating with respect to c,. The
contribution to the energy equation from the second integral on the right
* See L. Landau, Phyi. Z. SoujUn. 10, 154 (1936).
19.65] IONIZED OASES 397
of (19.65, 4) is zero, by (3.54, 5), corresponding to the fact that the total
energy of the electrons is not altered by their mutual encounters. On using
the transformation of 3.53, the first integral on the right contributes

The first and second terms on the left yield respectively terms \ntkdT^dt
and (after transformations, including an integration by parts) —mtntFt.ct.
Hence finally

lntk*£ = mtn1Ft.ct-^fjnmJA-c't')galidt,dcldct.
(19.65,10)
The first term on the right of (19.65,10) represents the rate of gain of electron
energy from the electric field, the second the rate of loss of energy through
encounters with ions. In the second term, because of the smallness of
c\—c't*, it is sufficient to use the approximation (19.65, 6) for/J 0) .
In order to effect the integration in this term, we express c „ c, in terms of
r and «>,,, where — .,, .,-, . , .
Btl
' T = M ,c, + Mtcv (1965, 11)
and
M J « m , / ( m , + Mi), Aft = m^/(m1 + m't), m^^T^mJTt. (19.65,12)
Then c, =T-M'tgn, ct = T + M'lgtl, (1965,13)
and (cf. (3.41, 8) and (3.52, 5))

Also, by (3.41, 6, 7),

= 2M,{T + (M, M\ - M , M't)gn}. (gn - * ; , ) .

On substituting these values, the term in question becomes
n
^(^^T)isssiexp{~{mi+m'tn^+M'iM'tgt)i2kTi}
* ^rt <r +
<M» M* - M
» ^ t ) * * } • C*ti -gn)gbdbdedTdgn

or, substituting for A/, M'x — M , M't and integrating with respect to T and e,
/ m1mt \i m\m\{i-TJT2)
27rrii
"* \27TkT^m1 + m't)) (w, + «,)»(m, + m't)

x J J e x p { - ( « , + « i ) M i M;^'/2*71} ( ' -co*x)g*bdbdgn.

398 ELECTROMAGNETIC PHENOMENA IN (19.66
In this we now regard mJt m't as negligible compared with m,. Then the
expression is equivalent to
8 ^ , ^ , ^ ( 1 -TJTt)[r
exp I m0M.M.g*\, (I .....
—fefeU+tt./ )} I - Mr. ) —*>«**+
-1611^^-4/1-,)(7,- r j [QJWOlr.
-3-1^1-^)/^(¾).
using (19.65,9), (9.33,2) and the results of 10.34. Thus (19.65, to) becomes
tr^rinrm ffl-r.) 6 ,
{l9b5 I4)
.
it 3km, r(Tt) • '
Suppose that 7J is kept constant. Then if Ft = 0 equation (19.65,14)
indicates that any initial small difference between Tt and Tt relaxes to zero
with relaxation time T(7J). If a small Ft is suddenly switched on, Tt rises in
a time comparable with T(7,) to the steady value found by taking </7^/<ft = 0
in (19.65, 14). Because r(Ta) is (very nearly) proportional to TJ, no such
steady value can be found if Ft is large. As Tt increases above Tv the first
term on the right of (19.65, 14) (representing Joule heating) also increases,
whereas the second term (representing the loss of energy at encounters with
ions) initially increases but ultimately decreases. Thus if mtFt is sufficiently
large, the- loss of energy at encounters cannot balance the gain of energy
from the electric field for any value of 7J, and Tt steadily increases.
By considering the maximum ratio of the second term on the right of
(19.65,14) to thefirst,it can be shown that the maximum steady value of Tt
occurs when
Tt - 1 7 1 , >"tFt=[—) ^ • (19-65.15)
The corresponding diffusion velocity c, has magnitude (jkTJzm^i, com-
parable with the mean thermal speed of ions. If MjF, exceeds the value
given by (19.65, 15), in the absence of any further cooling mechanism for
electrons Tt increases without limit, and with it ct; the electrons are then
said to run away.

19.66. Runaway effects

The runaway of electrons in a strong electric field was first considered by
Giovanelli.* The runaway process can be regarded as total when the velocity
of diffusion c, given by (19.65, 8) becomes comparable with the mean
thermal velocity (8kTj7mt)l of electrons. This gives
_ p 8»,gft;» In ( i + »;;,)
m,*, -jji . (19- 0 °.'/
• R. G. Giovmnelli, Phil. Mag. (7), 40, *o6 (1949).
19.66] IONIZED OASES 399
so that the product Ft Tt is then considerably greater than its limiting value
for the runaway process to commence (cf. (19.65, 15)).
High-speed electrons are liable to run away before the electron-gas is
able to accelerate as a whole. The mean rate of loss of momentum of an
electron of velocity ct due to encounters with ions (regarded as stationary) is

"il \m^ct-c'^gbdbde

= 2ff»,lmtCtCt\(i-co$x)bdb

= nxmtetet4f^(ct)

= ™;^, n ( I + v g l )
(cf. (10.34, 4)). The corresponding retardation of the electron has magnitude
2Jr« 1 eJ 2 «2 2m ( I + vl\)lm\c\- If ct >s large compared with the mean thermal
velocity of electrons, there is a further retardation 7rn e ^ 4 ln(i+i;J,)/m|c|
resulting from encounters with electrons, since the ' field' electrons may be
regarded as at rest in comparison with the fast particle. Thus the total
retardation is
a w ^ H j e ? + K « i ' ) In (1 + O / w l c J .
If ct is so large that this is less than the acceleration F 2 due to the electric
field, encounters are unable to prevent that field from accelerating the
electron to steadily increasing velocities. The critical ct for runaway is
given by
Jw 2 f| = neffae? + i>V, 2 ) In (1 + vJO/ffitF,, (19.66, 2)
so that (cf. (19.66, 1)) \mtc\ is comparable with the value of kTt for total
runaway of the electron gas.
The rate at which high-speed electrons attain runaway energies depends
on the number of electrons in a high-velocity 'tail' of the distribution
function/,; this number is difficult to estimate, because of the failure of the
expansion method used in 19.65.* Moreover, when electron runaway once
starts, a discussion based solely on elastic encounters may be inadequate.
A sufficiently energetic electron may also produce further ionization, so
possibly increasing the number of runaways.
• See, for example, H. Dreicer, Phys. Rev. 115, 138 (1959). and 117, 339 (i960); A. V.
Gurevich, Soviet Phys. JETP, 11, 1150 (i960); M. D. Kruskal and I. B. Bernstein,
Phys. Fluids, 7, 407 (1964).
400 ELECTROMAGNETIC PHENOMENA IN (19.7

THE FOKKER-PLANCK APPROACH

19.7. The Landau and Fokker-Planck equations

As was noted in 10.34, tne dominant encounters for electrostatic interaction
are those producing small deflections. If it is assumed that in all encounters
the deflections are small, alternative expressions for d^jBt can be derived
from the Boltzmann integral.
Assume that in the encounter of two molecules m„ m, the velocities c„ c„
gtl alter by small quantities Ac,, Ac,, Af„. Then by (3.41, 6, 7)
Ae 1 --Af a Ajr u > Acj-A^A*,,. (19.7,1)
By Taylor's theorem,
/ i » / , ( c , + Ac,)-/(c,) + A c , . ^ + i ^ , A c 1 : g ^ + . . .

with a similar expression for/,. Thus if we retain only terms of first and
second order in Ac,, Ac,, we find, using (19.7,1)

-fSI{*4<,-"•£)"•
+ ibgn&gn:(Mi^-M»g^) / i / s ) * * * * ^ (»97. 2)

Now gn + Af„( =g'n) can be expressed in the form

gn + &gn = gn c o s X +g(h cos e + i sin e) sin x,
where h, i arc unit vectors perpendicular to gn and to each other; the
neglect of terms of higher order than the second in A^„ means that x* is
negligible, and that sin'x can be replaced by 2(1 -cos^)- Thus
Ifbgnbdbde - -2ngnf(t -costfMA - -*„$»>(*), (197.3)
jj^gn^gubdbde - ngHhh + u)fsm*Xl>db = #l>(*)te«U-*„*,,),
(»9-7.4)
using (1.3, 9). From these relations

2jj bgngbdbde - ^ - .Jjbgnbgngbdbde

19.7] IONIZED OASES 401
since, applied to a function of gn, -0/8c, = d/dct = d/dgn. Hence (19.7, 2)
can be written

(¥).-iJ{«-0
.jj^M^-M^/J^g^g^gbdbde^dc,.
The term involving Af, d/dct in the first of the brackets on the right vanishes
on integration, by the divergence theorem. Thus

where
Qu = WMjjj[(Ml^-Mt^fJty&gn&gntbdbtkdet (19.7,7)

Equations (19.7, 6 and 7') (in a different notation) were first derived by
Landau.* As Landau pointed out, they indicate that the changes in/, due to
encounters with molecules m, are the same as if there were a flow Q„ of
molecules « , in velocity-space.
Again using (19.7, 5) and the divergence theorem, we may further trans-
form (19.7, 7) as follows:

Qu = iM*J(M^-M*^) • {Jffifn**n**ngM*) *•

-M^fJ^gngbdbdedct

-/iSi^-Jg-.tAS,**,**,), (,9.7,8)

where tt<j> m [Uf^gbdbdedc^. (»9-7. 9)

Thus "Lt<f> denotes the sum of the values of <f> for all encounters per unit
time of a molecule m, of velocity c, with molecules m,. Combining (19.7, 6)
and (19.7, 8), we get

( ¥ ) , = -^-tA^O+I *frtfA*^ (.9.7, «•)

* L. LuuUu, PAy». Z. SowjUn. to, 154 (1936).
402 ELECTROMAGNETIC PHENOMENA IN [19.71
This is the Fokker-Planck expression for (dffjdt^, introduced into the
theory of ionized gases by Cohen, Spitzer and Routly.* A direct derivation
of this expression can readily be given; the present derivation displays the
essential identity of the Fokker-Planck approach with that based on the
usual Boltzmann integral, within the range in which both are valid.
19.71. The superpotentlals
From (19.7, 1, 3,4,9)
S , ^ = MtfftggtSXg)dct, (19.71,1)
S.AC.AC, = AfiJ/^'U-gg)g<l>^(g)det. (19.71, 2)
Some care is required in applying these expressions to electrostatic inter-
action. Difficulties regarding convergence for large values of b have already
been discussed in 10.34; m addition there are difficulties for small b, where
the deflection is not small and the assumptions of the Fokker-Planck
approach are inapplicable. Landau proposed taking a lower limit V for 6
such that the electrostatic energy |«i«J|/6' is equal to the mean kinetic
energy of a particle; a quantum lower limit has also been proposed. In what
follows we shall use the expression (10.34,4) for $i>; this is roughly equi-
valent to using a lower limit b' such that |«l<i|/&' = 4*r.t Then very nearly
0fi' - 4 ^ ^ 0 ^ 1 Mtg'Y Inv n (19.71,3)
and In vn is regarded as independent of;.
From (19.71,1,3), since %/8cx =• -gjg,
ah
Z,Act = 2 - \ (19.7:,4)

where A« = 4 ff ^i(<i'«/ w o^i^s) ,m, 'oi//i^" l « fc i- (i9-7»»5)

Similarly, since 5^/3^ Ocj = (g^-ggVg3,

Si^iAcx - g^gjf. (»9-7». 6)

where Hu = ^nM^eyjn^M,Mt)*Inunjftgdct. (197 1 . 7)
In terms of the 'superpotentials' hlt, Hlt,

^r E J T 1 acJ+2 dddd-y^&J- {l971 8)

'
The introduction of A,f, / / „ is due to Rosenbluth, MacDonald and Judd. J
The notation hlv Hlt, like the £,Acj, EJACJACI in the Fokker-Planck
* R. S. Cohen, L. Spitzer ind P. McR. Routly, Phyi. Rn. So, 330 (1950).
t Note th»t in deriving (19.71,2) the approximation $!> «• 2 $ , ' ha* been uaed; cf.
(«0.34. 4. 5)-
X M. N. Roaenbluth, W. M. MacDonald and D. L. Judd, Phyi. Rto. wj, 1 (1957).
19.8] IONIZED OASES 403
expression, conceals integrations involving / „ so that the Boltzmann
equation remains an integro-difTerential equation. Rosenbluth and his
co-workers gave equations for calculating hlt, / / „ when / , is expanded in a
series involving spherical harmonics in the components of c,; but often
a relatively crude approximation is made by taking the velocity-distribution
function/j of the 'field' particles in (19.71, 5 and 7) to be Maxwellian.
However, as Spitzer and Harm* showed, the values of the transport co-
efficients calculated using this approximation are subject to fairly large errors.
In many problems the Fokker-Planck equation is at least as difficult to
apply as that based on the usual Boltzmann integral. However, it sometimes
has advantages when detailed information about the velocity-distribution
function is required, e.g. in the region of high velocities.

COLLISIONLESS PLASMAS
19.8. The usual Enskog expansion of/,,
/==/^+/^+/^+... (19-8.0
is effectively an expansion in powers of the collision-interval T,. In 19.31 it
was pointed out that in rarefied plasmas (strongly ionized gases) in the
presence of a magnetic field the time &>,"' might be comparable with T,, and
expressions attaching equal importance to w,-1 and r, were derived. For
sufficiently rarefied plasmas in a magnetic field, an expansion of/, in powers
of w,_l has been suggested, neglecting encounters completely.! This pro-
cedure regards the spiral-radius in the magnetic field as analogous to the
free path in an ordinary gas, the magnetic deviations corresponding to
collisions.
The analogy is not exact, since the magnetic 'collisions' affect only the
motions perpendicular to H. If the electric field E has components E„ E±
respectively parallel and perpendicular to H, the component E, accelerates
charged particles freely; thus a solution for/, resembling the earlier normal
solutions can in general be obtained only if E, is small. Again, in the earlier
theory it was possible to express a normal solution in terms of a finite
number of parameters—partial densities and the mass-velocity and tem-
perature. Because magnetic 'collisions' lead to no interchange of energy
between particles, h e r e / , depends on an infinite number of parameters.
The equation satisfied b y / , is now the collisionless Boltzmann, or Vlasov,
equation; assuming only electro-magnetic forces, this is

In (19.8, 2) we substitute an expansion of form (19.8, 1), in which the

• L. Spitzer and R. HSrm, Phys. Rev. 89, 977 (1953).
t G. F. Chew, M. L. Goldberger and F. E. Low, Proc. R. Soc. A, 236, 112 (1956).
404 ELECTROMAGNETIC PHENOMENA IN {19.8
1
successive terms are supposed to be of order i, w? , &>,"',...; also d^f/dt is
supposed divided into parts of similar orders, and e,EJm„ e,E,/m, are
arbitrarily regarded as of order w, and i respectively. Then on equating
terms of successive orders, we get

i(E±+c.AJI).|£-o, (I9-8, 3)

~ir+e'^+m,\E'-w.+^+c'AH)-w.i = 0, (19.8,4)

where dof^/dt and the other parts of dfjdt are so far undetermined.
Let c0 here denote a velocity such that
£ 1 + c 0 Aff = o, (198,5)
and write C, — c, — c0. Then the general solution of (19.8, 3) is
f!»-Fj(C*,C..H.,r,t), (19.8,6)
where F, is arbitrary. This implies that in velocity-space /J*> is symmetric
about an axis through c, « c0 parallel to H. Thus to the first approximation
the mean velocity c, has the same component perpendicular to H as c„; since
the component of c0 parallel to W is not determined by (19.8, 5), c0 can be
chosen to equal the mass velocity of the gas, to the first approximation. The
function F, may be chosen to equal, to a first approximation, the actual
velocity-distribution function /„ averaged over all directions of the com-
ponent of C, perpendicular to H.
The first approximation to the pressure tensor is

P« = 2:J«.c.c./.«»<fc.. (19.8,7)
It does not represent a hydrostatic pressure, but is symmetric about the
direction o(H. Up, zndp± denote the pressures parallel and perpendicular
* * iF>-p±U+(p,-p±)HHIH*. (19.8,8)
No simple general rule for assigning the value of dg/f'/d/ in (19.8,4) can
be given. If (19.8, 4) is multiplied by an arbitrary function ^ C j , C,.H, r, t)
and integrated over all c„ the 'collision term', involving EL + c,f\H, can
readily be seen to vanish; thus we find an infinite set of equations for the
quantities djS»/8t, where ^ _,„ _ ^ , , ^ ^ ^

This whole infinite set of equations has to be taken into account in defining
dofv/St, in place of the equations for dtn/8t, B0ejdt and dtT/dt in the
discussion of 7.14. Thus a general method of solving (19.8, 4) cannot be
given; we must either assign d^f^jdt arbitrarily and test what are the
19.8] IONIZED OASES 405
consequences, or else content ourselves with such information as can be
derived without explicitly solving (19.8, 4).
As an example of such information, if (19.8, 4) is multiplied by tn,c, and
integrated with respect to c,, we obtain on summation over all s the equation
of motion n
P - $ = - V. p«» + / » A H, (19.8,10)

where/" is the current density given by

y» = SV.Q" = X'.jflr)c.<ic„ (19.8,11)
and the volume charge J K / , is regarded as negligible. The equation
a
(19.8, 10) is the only one now available to determine the component of/0*
perpendicular to H\* for its effective use, we need to know the values of p,
and pL.
Again, sum over all t the equation obtained by multiplying (19.8,4) by
m,C,C, and integrating with respect to c,. From the double products of the
resulting equation into HH and into the unit tensor U, it is found that
%4 = -p,v.c0-2P,™^-(v.Qy.!jj!+*£,.?», (1941,)

^ = - 2 / > x V . c 0 + ^ ^ P - i ( V . Q ) : ( U - 5 J ) , (19.8, .3)

where Q is a third-order tensor measuring the flux of mCC, i.e. (in an
obvious notation) f
Q = SJ/.l0,*.C,C.C.<fc.; (19.8,14)
/ ° ' is the electric current corresponding to thefirstapproximation/}°\
When E, = o, equations (19.8, 12, 13) can be simplified by using the
equation of continuity D0p/Dt = —pV.c0 and Maxwell's equation
8H/dt - -curlfi;
since now E= —C0AH, this gives

^-«.(VcD)-»(V.c0).

The simplified forms of (19.8, 12,13) then are

S(&) - - ^ <»<"«•-»»>. C9-8..6)

* E. N. Parker, Phyi. Rev. 107, 914 (1957). Parker gave an explictt expreaaion for thia
component of /"' in terms of p, and p±.
406 IONIZED OASES [19.8
To use these equations we need to know Q. If it is assumed that Q is
negligible, we obtain the 'double-adiabatic' equations

In general Q is unknown. Its rate of variation can be obtained by multi-

plying H by m,C,C,C„ integrating and summing; the resulting equation
involves the mean value of a function of fourth degree in the components of
C,. Similarly the time-derivative of this mean value is given by an equation
involving the mean value of a function of fifth degree; and so on. This type
of difficulty is one often met; it can be overcome only by arbitrarily assuming
a relation between moments of sufficiently high order.
Because of this difficulty a number of discussions of collisionless plasmas
have been based on particle orbits instead of the Boltzmann equation.* The
two approaches are fundamentally the same; a theorem due to Jeansf states
that the most general solution of the collisionless Boltzmann equation is
/„ = ${<xt, a,,..., a,), where a„ at,..., a, are functions of e„ r, t such that the
equations a, = const, give six independent integrals of the equations of
motion of a molecule m,. For example, if <?o/»<0,/<" -= o and c0 is negligibly
small, the equations for the d0J^ldt of (19.8,9) are satisfied by assuming that
along any line of magnetic force/}0) is a unique function of a„ /?„ where

and V, is the electric potential. This agrees with the particle approach, since
a, is the total energy and fi, is proportional to the magnetic moment of a
particle m„ which is an adiabatic invariant of the motion. However, the
development of the theory in terms of the subdivision indicated in (19.8, 1,
3, 4), is parallel to, rather than identical with, the theory based on particle
orbits; and its connection with the theory given earlier in this book is not
close.
• See, e.g., H. Alfvta, Cotmieal Electrodynamics chapter a (Oxford, 1947, 1963); M. N.
Roienbluth tnd C. L. Longmire, Attn. Phyt. 1,120(1957). The relation of the Boltzmann
and particle approachea ia diacuaaed, e.g., by K. M. Wataon, Phyt. Rev. 10a, 1a (1956);
S. Chandraaekhar, A. N. Kaufman and K. M. Wataon, Am. Phyt. a, 435 (1957), and
5, 1 (1958); E. N. Parker, Phyt. Rev. 107.924 (1957); S. Chandraaekhar, Platma Phytict
(Cambridge, i960).
f J. H. Jeana, Month. Not. R. attr. Soe. 76, 70 (1915).
 HISTORICAL SUMMARY
Daniel Bernoulli (1700-82) in 1738, Hcrapath (1790-1868) in the years
1816-21, Waterston (1811-83) in 1845 (though his paper was not published
until 1893) and Joule (1819-89) in 1848, made early significant contribu-
tions to the kinetic theory of gases. A new phase began in 1858, when
Clausius (1822-88) introduced the concept of the mean free path; his work
extended over nearly twenty years, from 1857. Maxwell (1831-79) in 1859
obtained formulae for the transport coefficients of a gas (viscosity, con-
duction and diffusion), using the mean free path and his own new concept
of the velocity distribution function/. He found the equilibrium form/** of
/ , though his proof (both then and at a second attempt in 1866) was
imperfect. In later years Meyer (1834-1909), Stefan (1835-93), Tail
(1831-1901), Jeans (1877-1946) and others devoted much effort to im-
proving the free-path theory; but already in 1866 Maxwell turned from this
to base a new theory on equations of transfer that he formulated. From
them, however, he could obtain formulae for the transport coefficients only
for the special ' Maxwellian' gas, in which the molecules interact with an
r - 8 force; for such a gas he did not need to determine/.
Maxwell's new method inspired Boltzmann (1844-1906), who in 1872, by
his //-theorem, much improved the proof for Maxwell's function /<0).
In the same paper he gave his integro-differential equation for/, and in-
ferred the specially simple form of/for a Maxwellian gas. In 1879 Maxwell
used this form to show that in a very rare gas at non-uniform temperature,
at rest, there are stresses not included in the Navier-Stokes equations.
Again he discussed only a Maxwellian gas, 'for the sake of being able to
effect the integrations'. In 1880/1 Boltzmann published three long memoirs
in which he reviewed Maxwell's 1866 and 1879 papers. He said that the
solution of his own equation for/is easy only for a Maxwellian gas; 'in all
other cases, and especially for...elastic spheres, the solution.. .meets
with great difficulties'. He added that one must almost despair of the
general solution of his equation.
In 1887 Lorentz (1853-1928) further improved Boltzmann's discussion
of fm. He also, in 1905, achieved the first success, after Maxwell, in
obtaining accurate formulae for the transport coefficients for another
special type of gas. Hilbert (1862-1943) m I 0 1 2 discussed the theory of
Boltzmann's equation, but without arriving at a satisfactory solution.
Pidduck (1885-1952) in 1915, by numerical treatment based on Hilbert's
discussion, obtained an accurate value for one special gas coefficient, the
self-diffusivity, in a gas composed of elastic spheres. However, neither his
work nor that of Lorentz gave any hint as to how to obtain general results.
Enskog (1887-1947) in 1911/12, and Chapman in 1912, wrote on the
kinetic theory, but without success in solving the general problem. Not
[407]
408 HISTORICAL SUMMARY
until 1916 could Chapman, using Maxwell's equations of transfer in an
extended way, publish accurate general formulae for the transport co-
efficients (without needing to determine f). In 1917 Enskog's Uppsala
inaugural dissertation gave the general solution for/, yielding transport
coefficients identical with Chapman's. This identity of results independ-
ently obtained by different routes led to their acceptance by many physicists
who might not wish to check the complicated mathematics.
The half-century between Maxwell's 1866 and Chapman's 1916 kinetic
theory communications to the Royal Society, during which exact theory was
limited almost entirely to a Maxwellian gas, left to Chapman and Enskog
the discovery of the complete equations of diffusion and heat conduction in
a mixed gas, revealing the phenomena of thermal diffusion and the diffusion
thermoeffect. Thermal diffusion in liquids had been discovered experi-
mentally by Ludwig in 1856 and studied by Soret in 1879-81. However, it
vanishes for a Maxwellian gas and so, despite its appearance in Enskog's
first 1911 paper, for a Lorentzian gas, its general occurrence in mixed gases
remained unsuspected until in the summer of 1916 it was confirmed by
Dootson in conjunction with Chapman. Even then, for another quarter-
century it seemed to be little more than a scientific curiosity, until in the
thermal diffusion column invented by Clusius it was applied on a large
scale for the separation of uranium and other isotopes for a variety of
medical and technological purposes.*
Since 1920 the ideas of Chapman and Enskog have been considerably
developed and generalized. A variety of molecular models have been
studied, including those with internal energy and those in which quantum
effects are important. The theory has been extended to ionized gases and to
very rare and very dense gases, each of which presents its own methodological
difficulties. One further gas coefficient—the volume viscosity—has been
isolated. These later developments are described in the body of this book;
the present Historical Summary therefore deals mainly with the history of
our subject prior to 1920.
Earlier editions of this book included a more extended Historical
Summary, partly reproduced above.

HISTORICAL REFERENCES
Where page numbers are given in parenthesis, they refer to pages of this
book where the reference is to be found.
L. Boltzmann, 187a (p. 73); Wien Ber. 8x, 117 (1880), 84, 40 and 1230 (1881);
ColltcUd Works, 2, 388, 431, 523.
S. Chapman, Phil. Trans. R. Soc. A 213, 433 (1912); 2x6, 279 (1916); Proe. R.
Soe. A 98, 1 (Dec. 1916); S. Chapman and F. W. Dootson, 1917 (p. 268).
* Cf. L. B. Loeb, Tht Khutic Theory of Gaits (Dover edition, 1961), pp. xiii, xiv.
 HISTORICAL SUMMARY 409
R. Clausius, Annln Phys. 105, 239 (1858).
D. Enskog, Phys. Zeit. 12, 56 and 533 (1911); Annln Phys. 38, 731 (1912);
Inaugural Dissertation, Uppsala, 1917.
J. H. Jeans, Phil. Mag. 8, 692-9, 700-3 (1904).
H. A. Lorentz, 1887 (p. 73); 1905 (p. 188).
J. C. Maxwell, i860 (p. 72); 1867 (p. 68); 1879 (p. 281).
O. E. Meyer, Annln Phys. 125, 177, 401, 564 (1865).
F. B. Pidduck, 1915 (p. 195).
J. Stefan, Wien Ber. 65, 323 (1872).
See also: S. G. Brush, Kinetic Theory, 2 vols. (Pergamon, 1965, 1966); (1957/8)
Ann. Set. ix, 188 on Herapath, 13, 273 on Waterston, 14, 185 on
Clausius, and 14, 243 on Maxwell; also Am. J. Phys. 29, 593 (1961),
30, 269 (1962) on the later developments.
S. Chapman, in Lectures in Theoretical Physics vol. 9C (ed. by W. E.
Brittin), pp. 1-13 (Gordon & Breach, 1967).
 NAME INDEX
Note: where an author's name is used adjectivally (Boyle's law, Boltxtnanrit equation, etc.),
only the first appearance is listed here. For later appearances, see the subject index
Aichi, K., 172 Chahine, M . T., 29<
Alder, B. J., 292 Chandrasekhar, S., 406
Alfven, H., 406 Chaos, F., 3J8
Allen, H. W., 383, 388 Chapman, S., ix, 103, 131. 169. 172. 178.
Allis, W. P., 379, 383 I 9 i 221, 268, 2J2, 290, 3J2i 366, 407j
Altermann, Z., 169 408. 409
Amdur, I., 279 Charles, J. A. C , 38
Arthurs, A. M., 342 Chew, G. F., 403
Avogadro, A., 38, 33 Choh, S. T., 318
Chu, C. K., 206
Barua, A. K., 279 Clauser, F . H., 352
Baumann, P. B., 246 Clausius, R., 299, 4Q7. 409
Becker, E. W., 270, 235, 279, 331, 332 Cleave, A. B. van, 246
Beenakker, J. J. M., 246, 261. 278, 3 j 6 Clusius, K., 268, 270, 234, 279, 354, 408
Bendt, P. J., 278, 332 Cohen, E. G. D., J18, 327, 330. 33a
Bennett, C. O., 312 Cohen, R. S„ 178. 402
Bernoulli, D., 407 CondifT, D. \V., 220
Bernstein, I. B., 399 Cook, W. R., 172, 183
Bernstein, R. B., 142 Coremans, J. M . J., 246, 336
Beyrich, W., 27s. 279 Cottrell, T . L., 246
Bhatnagar, P. L., 104. 108. 295 Courant, R., 114. 128
Binkele, H. E., 246 Cowling, T . G., 244, 2<6. aoo, 3 j i . 362.
Bird, R. B., xi, 162. U k . ' I S . 329 364. 166, 368, 376
Boardman, L. E., 268 Cozens, J. R., 279
Boer, J. de, 184, 313, 327, 323, 330, 332, Crookes, VV., 10a
Curtiss,C. F., xi, 167.197,221,318.319.343
Bogoliubov, N. N., 313, 3 ' 6
Bohr, N., 362 Dahler, J. S., 137, 220, 222. 27s. 318
Bottzmann, L., ix, 37, 46, 67, 73, 77, 79, Dalgarno, A., 342
2QS• 299. 4Q7. 408 Dalton, J., 4 j
Born, M . , 3 j j Dave, S. M . , 271
Boscovich, R. G., : 6 3 Davies, A. R., 332
Botzen, A., 310 Davydov, B., 383, 386
Boyd, C. A., 279 Dawson, J., 42
Boyle, R„ 38, 227 Debye, P., 128
Braune, H., 246. 279 Deming, W. E., 3_io
Brittin, \V. E., 452 Dickel, G., 2 M
Brokaw, R. S., 244, 2so. 252, 256 Diller, D . E., 333
Bromley, L. A., 2^6 Din, F., 42
Brush, S. G., 227. 409 Dognin, A., 237
Bryan, G. H., 197. 2<7 Dootson, F . W., 168, 408
Buckingham, R. A., 186. 330. 332 Dorfman, J. R., 318
Buddenberg, J. W., 244 Dorsman, C, 246
Bunde, R. E., 240. 279 Dreicer, H., 399
Burbury, S. H., 29. Drickamer, H . G., 312
Burgers, J. M., 352 Druyvesteyn, M . J., 383, 387
Burnett, D., ix, xi, 127, 128. 169. 244, 281. Durbin, L., 312
284. 289
Enskog, D . , ix, 42, n o , i n , 114, l i s . 1 » .
Carnot, S., 41 118, I29j i l l , 143i Uli i l i ? i ! i ?97i
Cercignani, C , 296 308,4021408, 402
[ 4 1 1]
412 NAME INDEX
Eucken, A., aaa, a j i 250, 356 Holford, J. O., 346
Houen, J. N , 379
Favin, S., 379 Holtamark, J., i n
Faxen, H., 333 Holway, L. H., ao4
Feberwee, A., 378 Hoover, W. G., aoo
Flynn, G. P., 310 Huber, M., ajo, 574, 879
Flubacher, P., 379 HUckel, E., 178
Fofckena, K., a;6 Hulbert, E. 0., 100
Fowler, R. H., 35, 68, 74 Hutchinson, F., 879
Francis, W. E., 344
Furry, W. H., 14a, a6o Ibbs, T. L., a68, 369, 877, a79
Irvine, J. W., 379
GUM, R., 36a, 379 Itean, E. C , 184
Garda-Colin, L. S., 318 Itterbeek, A. van, 236. »46. 333, 336
Gibert, R„ a37 Iwaaaki, H., 246
Gibcon, R. O., 309
Giddings, J. C , 378 Jackson, E. A., ifis
Gille, A., a*6 James, C. G. F., i£a
Gillee, D. C , 33a Jeana, J. H., jf, 73, 80, 93, 103, 192, 897,
GiovaneUi, R. G., 308 406. 407. 400
Glueck, A. R., 184 Jeffrey*, H., 328
Goldberger, M. L., 403 Johnson, M. H., 109
Gora, E. K., 3X3 Johnston, H. L., 829, 831, 836, 846
Graaff, W. dc, 41 Jones, J. E., ttt Lennard-Jones, J. E.
Grad. H..xi. 58.73. iai. in. 176, a9a, a93. Jones, R. C , 142, 18a, 183, 106, 869, 333
804 Joule, J. P., a, 32, 407
Graham, T., 141 Judd, D. L., 40a
Gray, P., 344, aj6,151
Green, H. S., 313 Kaufman, A. N., 406
Green, M., 318 Keesom, W. H., 4a
Greentpan, M., aoa Keller, W. E., 330, 33J
Grew, K. E., a68, 369, 334, 377, 379 Kelley, J. G., 879
Grilly, E. R., aa9, a3i, 346 Kelvin, Lord, 266
Groech, C. E., am Kestin, J., 229, ajj, 846
Groea, E. P., 104, 106, .108. aon Keyes, F. G., a to. 256
Gurevich, A. V., 399 Kihara,T.,xi, 160.161.166.173.184,187,193
Gwathmey, E., ja6 Kirkwood, J. G., 311
Klemm, A., 879, 33a
Haar, D. ter, aoj Klestov, R. E., 840. 879
Haat, N. de, 331, 354 Knaap, H. P. F., 846
Hainiworth, W., aai Knudsen, M. H. C , i s s
Halpern, O., 336, 330 Kobayashi, R., 31a
Hamilton, J., 330 Kohler, M., 131. aao
Hanka, R. V., 310 Kramers, H. A., 337
Haring, A., 379 Kranendonk, J. van, 184
Hlrm, R., 178, 403 Krook, M., 104, 108, 893
Harpe, J. P., 378 Kniakal, M. D., 390
Hasse, H. R., 12¾ 183 Kurs, F., 846
Heberling, R., 146
Heijningen, R. J. ). van, a6i. 378 Lamsr, F. S., 383
Heinxinger, K., 33a Lammeren, J. A. van, 43
Hellund, E. J., m , 343, 348 Landau, L., 396. 401
Herapath, J., 407, 409 Landshoff, R., 179
Herxfeld, K. R., 345, 246 Langenberg, D. N., 42
Hilbert, D., 46« 114, UJU US, 407 Laranjeira, M. F., 274
Hilaenrath, J., 4a Leeuwen, J. M. J. van, 33a
Hinchfelder, J. O., xi, 167, 184, 187, a j i Leidenfrost, W., 346
»54. 343 Lemaire, N. A., 310
Hoffmann, D. K., 31a Lennard-Jones, J. E., 17a, 180.18a.a81.a80
 NAME INDEX 413
Liboff, R. L., 178 OfTerhsus, M. J., 1 6 ^ 33*
Liepmann, H. W., 293. 296 O'Neal, C , 25¾ 232. 256
Liley, P. E., 184 Onnes, H. K., »46, 331
Lindsay, A. L., 256 Onsager, L., 142. 144. »60
Linke, R., 246 Oosten, A. van, 278
Liouvilte, J., £, 198 O'Toole, J. T . , 318
Lipsicas, M., 312 Ouboter, R. de Bruyn, 256
Litovitz, T. A., 24;, 246
Loeb, L. B., 408 Paemel, O. van, 246
Longmire, C. L-, 406 Parker, E. N., 405, 406
Lonius, A., 261, 278 Parker, J. G., 252
Lorentz, H. A., 2 1 188,104, 407. 400 Parker, W. H., 42
Lo»chmidt, J., 32, 278 Partington, J. R., 4»
Love, A. E. H., 76 Paul, R., 279, 354
Low, F. E., 403 Pedersen, P. O., 362, 366
Lu, W. K., 220 Pekeria, C. L., 160. 19¾
Ludwig, C , 408 Percival, I. C , 34a
Pereira, A. N . G., »53
Maau, O., 246 Persico, E., 178
MacDonald, W. M„ 402 Piddington, J. H., 368
Madigosky, W. M . , 310 Pidduck, F. B., 125,127. »12.383.407,409
Margenau, H., 186
Maaon, E. A., xi, 161, >66, 184, 186, 187, Rankine, A. O., 246
193. 108. 224. 225, 217. 2 l 8 . 240, 2j2. Rayleigh, Lord, 230
»53. »54. »55. »59. »64, 271, 275. 276, Ree, F. H., »95
»77. 278, 279. 3»2i 333. 3 4 i 354 Rice, W. E., 187, 237, 238, 323
Maaaey, H. S. W., 323, 327, 3*8, 32¾ 336, Rietfeld, A. O., 333
34* Rocco, A. G. de, 246
Maxwell, J. C , ix, 5J 46, 4¾ S i 63. Z*J I t Roos, B. W., 33a
72, 78, u i j 172, 173, 222, 281, 452, Roaenbluth, A. W., 209
408. 409 Rosenbluth, M. N., 299. 40». 406
Mayer, J. G., 316 Rosa, J., 279, 310
Mayer, M. C , 316 Routly, P. McR., 178, 40»
McCloskey, K. E., 246 Rumpel, W. F., 279
McCoubrey, J. C , 246
McCourt, F. R., 312 Salwen, H., 29s
Meyer, E., 232 Sandler, S. I., 127, 2»». »73, 334
Meyer, O. E., 2¾ '03. 407. 400 Sather, N. F., 197. 222
Michels, A., 4¾ 309, J I O Saxena, S. C , 166, »40. 254, 264, »71. »73.
Mifflin, T . R., 312 »78. »79
Milne, E. A., ix Schamp, H. W., 187
Mirsky, L„ 120 Schatzki, T. F., 279
Misenta, R., 331 Schirdewahn, J., 279
Mohr, C. B. O., 323, 327, 328, 322, 330 Schleiermacher, A., 230, 256
Monchick, L., 184,187. 108. 224. 225, 238. SchlUter, A., 102, 368
»3». »35. *2Z. »79. 34» Schmeisaner, F., 331
Moran, T. 1., 279 SchrMinger, E., 127
Morse, E. M., 383 Schumacher, E., 354
Mott. N. F., 1 8 1 328 Seager, S. L., 278
Mott-Smith, H. M., 293 Sengera, J. V., 308, J i 8
Muckenfust, C , 127, 221 Semagiotto, F., 296
MUller, Fr., 229, 232. 246 Sessler, G., »92
Mundy, J. N., 274. »79 Shchegolev, G. T., 240. »79
Munn, R. J., 187, »77 Shilling, W. G., 42
Shuler, L. M., 279
Nagashima, A., »35 Shupe, L. E., 310
Naraaimha, R., 29¾• 296 Simons, S., 206
Narath, A., 246 Slots, W., 279
Ney, E. P., 279 Smith, C. J., 246,168
414 NAME INDEX
Smith, F. J., I|7J 276 Velda, C. A., 333
Sneddon, I. N., 183 Vermeer, W„ 2j6
Snider, R. F., 318, i n Vlaeov, A. A., 403
Sommerfeld, A., i n . 336
Sonine, N. J., 127 Wainwright, T. E., 299
Soret, C , 408 Waldmann, L., xi, 60, i 7 i 220, 260. 270.
Sorg, K., 246 2 7 i 2J4,279, 328, 332, 336, 34¾ 343,
Spitzer, L., 178. 402. 403 345.313
Spotx, E. L., 16¾ {£4, 235 Walker, R. E., 240, 264, 278, 279, 354
Srivaatava, B. N., 279, 354 Wang-Chang, C. S., iTi 208, 292, 294.
Srivastava, I. B., 279 336. 342
Srivastava, K. P., 279 Waaaenaar, T., 42
Stefan, J., 452, 409 Waaailjewa, A., 100
Stehl, O., 332 Wateraton, J. J., 407, 400
Stein, N., 279 Wataon, G. N., 9 i 325
Steingrimsaon, V., 279 Wataon, K. M., 406
Stockmayer, W. H., 187 Wataon, W. W., 274, 279
Strehlow, R., 240. 279 Weber, S., 246, 331
Strunk, M. R., 279 Weiaemann, S., 279
Suetin, P. E., 240. 279 Wendt, R. P., 279
Sutherland, W., 108, 180, ifia Weatenberg, A. A., 240. 264. 278, 279,
Svehla, R. A.. 184 354
Whitelaw, J. H., 246
Taconia, K. W., 256 Whittaker, B. T., 91
Tait, P. G., 88, 92, 402 Wild, N. E., 268
Takayanagi, K., 342 Wilke, C. R., 244
Tanakadate, A., 172 Winn, E. B., 279
Taxman, N., 208 Winter, E. R. S., 279
Taylor, B. N., 42 Wobaer, R., 229, 212. 246
Taylor, M. H., 187 Woemley, D.. 274. 279
Thomaa, L. H., 202 Wolkera, G. J., 42
Thome, H. H., t n Wright, P. G., 244, 256, 35J
Timmerhaut, K. D., 312
Tolman, B. C , 201 Yoa, J. M., 3SJ
Tonka, L., 379 Yun, K. S., 25s
Townaend, J. S., 362 Yvon, J., u s
Trautz, M., 236, 246
Zaitaeva, L. S., ano. 256
Ubbink, J. B., aso. 251, 236, 111 Zandbergen, P., 246. 336
Ubiech, H. von, 253, 256 Zehle, F., 279
Uehling, E. A., ijjj 323, 33¾ 336 Ziering, S., 295
Uhlenbeck, G. E„ i7i. 208. 113, 118, 323, Zink, R., 246
 SUBJECT I N D E X
A, 163; numerical values, for r~" model, Buckingham molecular model, 186 n, 330;
172: for 12,6 model, i 8 s , 186 (Fig. 2I; modified, i£6_
for exp;6 model 18& (Fig. 9)
*i("), A / " ) , 171: numerical values, 172 c, 163: values for r~' model, 171; for 12.6
Activation energy, 85 and exp; 6 models, 18s. 186 (Fig. 2)
Alternating electric fields, 379; with mag- Carnot cycle, 4J
netic field, 380 Centres of force, 8, ja, 160. 230, 264. 3»3
Ambipolar diffusion, 368 Centrifugal force and potential, 77
Angular momentum of molecules, 201 • 201. Chaos, molecular, s L 23. 22. 83, 297, 298:
2 i i , 342 see under Molecular chaos
Angular velocity of molecules, 206. 208, Charles's law, 38, 40. 42. 4S
Coefficient of conduction, diffusion, etc.;
Approximate formulae: simple gas, thermal see under Conduction, Diffusion, etc.
conductivity, 100. 10;; and viscosity, Collision cross-section, I55n, 168; quan-
08.104: mixed gas, diffusion, 102.106; tum, 327, 337
heat conductivity, 100. 255; and vis- Collision frequency, 88; velocity-dependent,
cosity, 108, 243 22
Apse line, 55, £2 Collision interval, 88, 2¾ 104, 120, 365,
Atmosphere, solar, 273, 362; Earth's upper, 376; relation to relaxation time, 105,
362 n, 383 245
Attraction, molecular, 35, 52, 172, a32. Collision-interval theories; set under Rela-
235. 266, 308: van der Waals, 183. 308 xation theories of transport phenomena
Avogadro's hypothesis, 38, 42J number, 32, Collisional transfer, 2971 3°ii .107. .lis
88 Collisionless plasmas, 403
Collisions, molecular, 3, 52, 52. 86, 297;
B, 163: values, for r~* model, 171; for 12,6 between charged molecules, 176, 395;
and exp;6 models, i 8 s . i8fi (Fig. g} ternary, multiple, 177. 297. 317. 3 ' 8 ;
Balancing, detailed, 68, 204 inelastic, 382, 389; see also Encounters
BBGKY equations, 312 Column, thermal diffusion, a&S
Best possible approximation, 131. 348 Compressibility, 308
BGK (Bhatnagar, Gross, Krook) approxi- Conditions of solution of equations, n j j
mation, 104, 108, 20$; see also Relaxa- I I I , 123, L32, u s , « s
tion theories of transport phenomena Conduction of electricity, 358; slightly
Boltzmann distribution, io<. jog. 339 ionized gas, 360; multiple mixture,
Boltzmann's collision chains, 205 361; Pedersen, Hall and Cowling
Boltzmann's constant, 37 conductivities in magnetic field, 366;
Boltzmann's equation for the velocity- see also Alternating electric fields
distribution function/: simple gas, 46, Conduction of heat; general expression,
" o : mixed gas, 47» 132, 343; gas with 43; elementary theories, 100. IOS. 109;
internal energy, 200. 338: 'rough exact theory for simple gas, 125; mixed
sphere' gas, 218; dense gas, 300; gas, 142. 346; gas with internal energy,
ionized gas in magnetic field, 361; a i t , 213; 'rough sphere' gas, 219;
collisionless, 403 Eucken's approximation and modifica-
Boltzmann's H-theorem; simple gas, 67J tions, 222. 224; third approximation
mixed gas, 8OJ for gas with internal (for simple gas in non-uniform motion),
energy, 205; connection with entropy, 285. 287. 293; rare gas, I S I ; dense gas,
78; with reversibility, 7.9 307: ionized gas in magnetic field,
Bose-Einstein particles, 326, 330, 33s direct and transverse conduction, 368,
Boundary pressure, 32 376
Boyle's law, j8, 4$, s i 184 Coefficient of conduction, A, for
Bracket integrals, 8¾ 212, 345; inequalities simple gas; approximate formulae, 100.
for, 84 105; exact general formula, ta<: nota-
Bryan-Pidduck 'rough sphere"' molecular tion for approximations, M9: first and
model, 107. 217 second approximations, 160; rigid
[4 5]
416 SUBJECT INDEX
Conduction of heat (coot.) Detailed balancing, 68, 304
elastic tpherei, 168; point-centre* of Diameters, molecular, 52, 86, ai8, 298;
force, 17a: inverse square force, 170; numerical estimates from viscosity,
Lennard-Jones model, 180. 183; 228, and from diffusion, a6j; quantum
Sutherland's model, i8a;exp;6model. theory, 32a, 329
186: gaa with internal energy, a n . Diatomic molecules, 4JL ao8, 340
aaa. 224: 'rough sphere' gas, 119: Diffusion: elementary theories and their
dense gas, a08: experimental values, defects, 101-103. 106-8. 100: exact
theory, 140: mechanically similar
Coefficient of conduction. A, /or gat- molecules, 194: isotopic, 19¾. 267:
mixture; approximate formulae, 109. gas with internal energy, 216; 'rough
*55J general formula, 14a;firstappro- sphere' gaa, ato: dense gas, 31a;
ximation (general expression), 164. multiple mixtures, 109. 344; in
*53; Lorentzian gas, 189. 191; con- ionized gas in magnetic field, direct
vergence of successive approximations and transverse diffusion, 364, 373:
for Lorentzian gas, 19J; variation with ambipolar, 368; thermo-effect, 103.
concentration-ratio, comparison with 143. 270: resistances opposing diffu-
experiment, 253; gas with internal sion, 106, »09; third approximation
energy, 254; in multiple mixture*, 347. to diffusion velocity, 390; pressure-
349: in mixture including electrons, diffusion, a<7. 291; forced diffusion,
154 as8; thermal diffusion, 103, 141,
268-78. 351; »ee alto Mobility
Conduction-viscosity ratio f; free Coefficient of mutual diffusion Dlt;
path formula, toj; notation for appro- approximate formulae, 10a. 106;
ximation*, 160; first and second exact general formula, 14JJ for
approximations, 160. 169, 173. 247. Lorentzian and quasi-Lorentzian gas,
S48; 'rough sphere' gas, aao; gas 189. loi; notation for approximations,
with internal energy, a»3; comparison n o : first and second approximations
with experiment, monatomic gases, (general expressions), 163, 164: first
ano; polyatomic gases, 351; polar approximations for rigid spheres,
gases, *s» point-centres of r-' force, attracting
Conservation of number, mass, momen- spheres, 12,6 and exp;s models, an8.
tum and energy, 45, S5i Sli IS2, 121 and rough elastic spheres, 359; for
U4. »03, 3»3 highly ionized gas, 170: for dense gas,
Continuity, equation of, 50, 133, ao3 31a: second approximation, 359; later
Convection, of heat, 248,391; of electricity, approximations for Lorentzian gas,
35l 19a: variation with concentration-
Convergence of formulae, tai. ia8. aoa: ratio, 260. afii; dependence on density,
tee alto Lorentzian gas 147. and temperature, 264: units, a6a:
Corresponding states, law of, for A, /t and experimental values, 263; quantum
D,„ 180, 1A1 corrections, 332; multiple mixture,
Critical temperature, 273 345. 346, 348: in magnetic field, 364.
Currents, electric, tee under Conduction of 373; measurement by diffusion thermo-
electricity effect, ajo
Coefficient of aetf-diffisslon D„, 103.
Dalton's law, 45 a6<: mean free-path formula, 103:
Damping by viscosity, 227. US; by volume exact formula for rigid elastic spheres,
viscosity, t 9 i a i i Z4.S > 93. >ndfirstapproximation for general
Debye shielding length, »78 smooth molecules, 194: first approxi-
Degeneracy: in gases, 333. 333. 1 3 i 336; mation for molecules with internal
rotational, 33¾ 34a energy, 217: Kelvin's calculations, a66;
Dense gases, 297-321: Boltzmann's equa- numerical values from isotopes, a67.
and from similar molecules, 268; dense
tion, 300; equation of state, 303, 307.
316: Enskog's theory of viscosity, 306,
and of thermal conduction, 302; com- gs». 3 «
parison of Enskog's theory with ex- Coefficient of diffusion of internal energy,
periment, 308: gas-mixtures, 311: DM, 224: comparison with Du, aa*.
diffusion, 31a:BBGKYtheory, 312-19
Density: uniformity of, 4_i mats-, number-, Coefficient of thermal diffusion Dr, HL,
»7. 44i partial, 44 189, 345
 SUBJECT INDEX 417
Diffusion (cont.) 62. 64. 2Q3. 304; molecules with
See alio Thermal diffusion ratio, thermal internal energy, 200; quantum theory
diffusion factor of, 1*3; iee alto Multiple encounters
Dissipation of energy: tee under Damping; Energy: heat-, 1, 2. 361 203. 314; equation
electric, 266 of conservation for simple gas, 51, 203,
Distribution of velocities, 27. ioq. 31» 31s; »nd for mixed gas, 134. 203; tee
Distribution f u n c t i o n / , 28, 44J Maxwell's, oboEquipartition, Internal, Rotational,
Si 68-73; in • conservative field of Vibrational
force, 25;in a rotating vessel, 78 j for Enskog's equation of change, 49; his method
gas with internal energy, 100: many- of solution of Boltzmann's equation,
particle function, 312; see alto Boltz- simple gas, lie—19. mixed gas, 134-
mann's equation 2; theory of dense gases, 297-308
Divergence, of a vector, 13^ 25; of a tensor, Entropy, 4JJ connection with Boltzmann's
15 //-function, 78
Dyadic, 14J differential, i £ Equation of change (or transfer or conser-
D y n a m i c s of encounters, 4: for' smooth' vation) of molecular properties, es-
molecules, 53, S i i6_2i 317: for pecially number, mass, momentum and
molecules with internal energy, 200; energy: simple gas, 47. 40. w . «,1;
for rough spheres, 218 mixture, 132. 133: molecules with
Dynamical systems; see Hamiltonian func- internal energy, 203; dense gases,
tion, Liouville's Theorem 3<>3, 3>5i ionized gas, 363
Equation of state, 35 n. 38, 52, i S s n; for
E, 163; rigid sphere model, 169; r* model, rare gas, 3 j j for dense gas, 303, 307
171: connection with diffusion co- Equipartition of energy, 4_i, 82, 208, 340
efficient, 163 Erf *, ox, 2 j
Eigenvalues and eigenfunctions, 105, 175. Ergodic theory, 205
206 Eucken's theory of heat conduction, 222.
Elastic molecules: rigid smooth spheres, 2. 247. 252; modified theory, 223, 224.
<:2. 86, 226, 229, 2<i8. 262: rough 248, 252, 342; extension to mixed gas,
spheres, 107. 217. 226, 259 »54
Electric conductivity, 359, 360. 361. 375: Excitation energy, 337, 389, 391
Pedersen, Hall and Cowling conducti- Exp;6 molecular model, 186; viscosity,
vities, in magnetic field, 366 237. 240: diffusion, 264, 265; thermal
Electric fields, 176; in magnetic field, diffusion, 271. 275. 277; Lorentzian
366. 391: alternating, 379; strong, gas, 191
382-99
Electrona, 193. 333, 354. 367; electron-gas, Factor, thennal diffusion ( a » ) ; see Thermal
323. 336; degeneracy, m , 336; dif- diffusion factor
fusion, thermal diffusion and electric Fermi-Dirac particles, 326, 330. 332, 333
conduction by, 3551 361, 367, 37°; Fifth-power l a w of intermolecular action
thermal conduction by, 3HS. 368, 376; (Maxwell), 52, 227. 247. 265: diagram
with large energies, velocity distribu- of orbits, 52J theory for, 173: eigen-
tion, 383-95; electron temperature, value theory, 125
388. 3o6: inelastic collisions, 389; Flow, of molecular properties, 2 9 - 3 1 :
recombination. 392; runaway electrons, number, 31; momentum, 32J heat, 43
32? Flux-vector, 31, 305; thermal, 43
Element: of volume, rjj of spherical Fokker-Planck equation, 400
surface or solid angle, I4j of general Force, external: influence on density dis-
surface, 29 tribution, 7.S-7: occurring in Boltz-
Elementary theories: of viscosity, 32» 104; mann's equation, (i) independent of
of thermal conduction, 100. 105. 222: velocity 46, (ii) ionized gas in magnetic
of diffusion, 101. 106. 109; defects, field, 362: influence on diffusion, 107.
103; transport in magnetic field, 141. 257; electric, and resulting
364-70 electric currents, 359, 360. 361: in
Encounter, molecular, J, 2i 46^ £2j dyna- magnetic field, 366, 375; alternating
mics of binary encounter for ' smooth' electric fields, 379; strong electric
molecules, £3j 53, 169; and for rough fields, 382, 383
spheres, 2• 8; geometry of, 55, statistics Force, intermolecular, 1, jSj 38, S i 65;
of, 58, 323; direct, inverse and reverse, point-centres of, 52, 169; Maxwell's
418 SUBJECT INDEX
Force, iatermoleeolar (font.) H-theorem, Boltzmann's, 67. 71. 77. 78,
fifth-power law, sit m i inverse- 29, 8oj gas with internal energy, 305
aquare law, 176: Sutherland model, History of Chapman-Enskog methods, 131.
i8on, 18a; Lennard-Jonea' model, 170. 407
183; Maton'a exp; 6. model, 186; other Hydrodynamics, 48, 51
model*, 187; influence on preaaure, 34, Hydrogen, ortho, para, a68, 33a
308, 316; influence on coefficients of Hydrostatic pressure, 34, 45, M A aoa.
viscosity ft, aao-38, diffusion D„,
i;8, »64. and thermal diffusion, a7i.
Impurity, effect of, on viscosity, »30, on
»73, »7.1. a76: numerical values, for like
molecules, 13a. ail. »37. »38. for un- thermal conductivity, a$3; diffusion of
like molecules, »63. a6«, a trace gas, 260, 334
Free path, molecular, 3, 86j mean, j , 82, Index of intermolecular fore* v, 170.
8¾. pa. 07; Tail's mean path, 88; de- 100: » » Si '73; » ™ a, 176; connec-
pendence on apeed, ooj frequency- tion with temperature-variation of
function for free paths of assigned viscosity n, »30, and diffusion co-
length, 93j free path theories of vis- efficient D„, 164; numerical values
cosity, 92, conduction, 100. and dif- from /t (simple gas), aia. and from
fusion, IOI. and their defects, 103; Dn (mixed gas, »-„), 265
Induced diffusion, 345, 146. 356
equivalent, for Lorentzian gas, too, for
general gas, a8i, »91; numerical esti- Inelastic collisions, a n . 38a, 389
mates, 88| quantum theory of, Integrals: involving exponentials, 10:
336 vectors and tensors, at, 319; multiple,
Frequency of colUalooa, 86, 88. 00: see ao, transformation of, ao, 64; theorems
alto Relaxation time on transformations, aa. 64. 8a
Function: Bessel, 112, iao: Erf, 91, o<; Interaction, molecular: >ee Force, molecular
gamma, 19, 171: Gaussian, 29J Le- Interchange of internal and translatory
gendre, 334; Sonine, 117: spherical energy, 5a, aia, ajA 337
harmonic, 176, 284; symmetric, 114 Internal energy of molecules, 17.41.4a. <i.
101. io7-aa<. as 1—3; quantum theory,
Gaa, 1, n simple, a6j mixed, 44, ue alto 336-4Z
Mixtures; multiple, 343; Lorentzian, Interpretation of kinetic-theory results, 6
188. 378; Maxwellian, 171; quasi- Interval, collision, 88, oa, iao: use in
Lorentzian, ioi: with internal energy, approximate theories, 104. 364; set
'97, 336; dense, ao7: degenerate, m . also Relaxation time, Relaxation
333, 31S_i ionized, 156, 358; tee alto theories
Force, intermolecular Invariants, summational, 42,11a. 110. aoa:
Gaa constant per mole, 39J numerical for molecules with internal energy,
value, 4a aoa. aoi. aot. aoo; for dense gases,
Gaa reactions, activation energy, 85 303
Geometry of molecular encounter, 55 Inverse encounter, 6a, 64, aoi
Gradient vector, ia_; velocity-gradient Inverse square law, 176
tensor, 18 Ionization, 383, 300; and recombination,
Grad'a ii-moment approximation, 393 191
Graham phenomenon, 141; change with Ionized gas, 102,109,176,334,338-406; in
temperature, »41 magnetic field, 361-70. 39'. 403-6;
Gram-molecule; it* Mole in magnetic and alternating electric
field, 379; in strong electric field,
Hall current, 166.17<; conductivity, 366 381-99; Fokker-Planck equation for,
Hamiltonian function, 198,199, afti n 400: coUisionlesa, 403
Hamiltonian equations, 108 Irreversible processes, thermodynamics of,
Heat, specific; e„ 39, loo, 107, aaa, a47: !44>2iZ
e„ 40; per mole, 40; numerical values, Isotopes, 195, 262, »68, a j i 330, i«a:
43 mixtures, ton. » 1 : self-diffusion, a67.
Heat energy, 1, 2i 3¾ zoj, 314; -density, a68; separation, 268, H4: thermal dif-
37; mechanical equivalent of, a, 32 fusion, SJJj 27¾ a j i 353, 354
H-function, Boltzmann's, 67. 71. 78. 70:
connection with entropy, 78J history Jaeobians, jo, S i 6¾ 63, 150, 36a n
of notation, 29 n Joule's equivalent, a, 37
 SUBJECT INDEX 419
Kihara's approximations: for simple gas, Mixtures, 4Ai non-uniform state, general
160; for mixed gas, 166: for diffusion, theory, 132-48 (binary), and 343-57
*59> 260. 261; for self-diffusion, IQS; (multiple); in magnetic field, 364. 370;
for thermal diffusion, isotopic mix- tee also Conduction,Diffusion, Thermal
tures, 196, »71, »77. 353; Lorentzian diffusion. Viscosity
and quasi-Lorentzian gases, 221 Mobile operator, 48
Mobility of charged particles, 360
Landau equation, 400, 401 Models, see under Molecules
Lennard-Jonea m o l e c u l a r m o d e l , i7g; Mole, 32, 40, 42
11,6 model, 182; other cases, 183. 187: Molecular chaos, J8J 21, 29, ?& *97. aog;
Lorentzian gas, 101; comparison with diameters, tee under D ; encounters,
experiment for viscosity, n s - 7 : see under E ; force or interaction, tee
diffusion, 264, 265, and thermal dif- under F; hypothesis, 1 , 7 : mass, J, 26,
fusion, 276, 222 2 2 , 4 2 , 4 4 ; models, tee under Molecules;
Liouville's theorem, 1, too. 2ai momentum, 32, 44J motions, 24:
Loaded spheres, 107. 222. 27< property ¢, 22, 42, 132: range of in-
Lorentzian gas, »88: diffusion, 180, 260: fluence, 6jj 177; speed, 25J streams,
heat conduction 180: viscosity, 100: interaction of, 323; structure, 6, 52.
convergence of approximations for, 197; variables, 25, 132, I29j 217:
1Q2, 103: in magnetic field, 378: in velocities, 25, 3_6j weight, 3.9.
strong electric field, 382 Molecules, 1; attracting 2, J J , 52, 179.
Loschmidt's number, 3J2, SS 182. 308; charged, 176. 358; diatomic,
43. 52, 208, 326, 340; exp-six (modi-
Mach number, 204. 296 fied Buckingham) model, 186: general,
Macroscopic properties, 2, 26, so, L2i; scale 52. loo. 336: identical or unlike, 326;
of variation, 26, 121. 2&1 ionized, 176, 358; Lennsrd-Jones type,
Magnetic fields, phenomena in presence 179. 183.187: Maxwellian, 173:mech-
of: Boltzmann's equation, 361, solu- anically similar, 194: model, 6,52,197:
tion for ionized gas, 370, Lorentzian monatomic, 42, J2j nearly smooth,
case, 376, 378; motion of charged 222: point-centres of force, $2, 160;
particle, 363, 364; conduction and dif- polar, 187: polyatomic, 4Jj 52, 251,
fusion, approximate theory, 364. 368, 326; rigid, 1, 6, 26J rigid elastic
general theory, 373, 376, alternating spheres, smooth, 26,52152,86, loaded,
electric fields, 100. 379; strong electric 107. 222. and rough, 26, 107, 217: ro-
fields, 301; collisionless plasmas, 403 tating, 6, 26, 102, »o6. 208, 217;
Magnetism in sun and aunspots, 362 smooth, 26, 52, 247; spherically sym-
Mason and the exp; 6 model, 186, 237, 27s metric, I, 6, 26, J2j 197; sphero-
Mason-Monchick formula for A, 224, 248, cylinders, 107.22j; Sutherland, 180 n,
? 5 i 255, 342 1S2; see alto Internal energy, Quantum
Mass: molecular, 2, 26, 3¾ 42, 44, 88j theory
-density, 26j -velocity, 44, 112. Moment-integrals (Grad), 294
-angular velocity, 226. Moment of inertia, 206. 208, 218, 340
Mazwelltan distributions o f molecular Momentum, molecular, 32,44J equation of,
velocities, 5, 68, 6¾ 72, 104. 120: in 5'. 133, 303;angular,203;generalized,
smooth vessel, 23; in presence of 108
external forces, 25: m rotating gas, 77. Multiple encounters, 177. 297. 317. 318
206: in gas mixtures, 82,108. n < : in a Multiple mixtures, 108. 343-57, 361; in
dense gas, 301, 315; in an ionized gas magnetic field, 365, 368
in a magnetic field, 363; divergence
from, in an ionized gas in a strong Nernst effect, 368
electric field, 382, 387 Normal states and solutions, 120. 203
Maxwellian molecules: tee Fifth power law Nuclear spin, 326, 330. 332
Mean free path; see Free path Number-density, 22, 44, 112: -flow, 3 j
Mean velocity, 22, 14J. 345 n Numerical values; tee under quantity re-
Mechanical equivalent of heat, J, 2, J2 quired
Mechanics: classical, oj generalized, 108:
quantum, & 322-342: statistical, j , 6, Operator: differential (vector), 12. 18. n .
37. 42. 80, 82, 158, 333 n, 334 200: mobile, 48
Metals, electron-gas in, 336, 366, 378 Orthogonality conditions, M4
420 SUBJECT INDEX
Paradoxes of kinetic theory, 29, 80, 3*4 n Reactions: tee under Gaa
Parameter! of Enakog'a solution, i n . 118. Recombination in an ionized gaa, 302
« 9 , 403 Relaxation time, 105. lao. i n , 244, 248,
Partition function, 207, 339 355, 204: connection with collision-
Path, free or mean free; see under F interval, iojj 102, 245. 2$ 1, and with
Peculiar velocity, 22, 4Ai apeed, 271 d»* eigenvalue theory, lot: equivalent
tributkm of, 70, 71 (Fig. 2) collision-number, 245
Pedersen electrical conductivity, 366 Relaxation (or collision-interval) theories of
Pendulum oscillationa, aa7 transport phenomena: simple gas, ther-
Persistence of velocities, 91,98,103,100: mal conduction, 105. 224. and vis-
quantum, 337 cosity, 104: mixed gaa, diffusion, 106.
Peraittence-ratio, 94J mean, 91 100. thermal conduction, 109. 354,
Phase apace, 108: internal, 100: quantum and viscosity, 108: ionized gas in
equivalent, 337 magnetic field, 364, 368, .381
Phase-angles (quantum), 338 Relaxation of internal energy, a n . 224.
Planck's constant, ] » , 333 245. 248; rotational, 24s. 251; vibra-
Plasmas, tee under Ionized gaaea tional, »45, 251
Point-centres of force, 52, 169, 190, 330, Reverse collisions or encounters, 203
264. a7* Reversibility, 79, 8a
Polar gaaea, 187, aj8, 349, 35¾ 356 Rigid molecules, 3, 6,J6J elastic spheres,
Polynomials: Hermite, 137 n; Laguerre, smooth, 2i S i SI 86, 92, 168, 339,
ia7 n; Legendre, 334; Sonine, 137; 363. attracting, 183. 33a. rough, 36.
expansion in, LlS n 197. 317. and loaded, 107. 333; non-
Position-vector, u apherical, 306; spherocylinders, 107.
Prandtl number, 350 221
Pressure, 8, l i IS, 45J on boundary, 3a; Righi-Leduc effect, 369
hydrostatic, 34, 45, 303; internal, 34, Rotating gas, 22, 206
2SJ partial, 45, 106, 138, j n , 343; Rotating molecules, 3, 35, 301 • 206. 308,
-tensor, 34, 4J; successive approxima- SIT, a46, 351
tions, i n . first, i n . IAA.second, 116. Rotation-vector, 14, ait, 173
144. third, 386; low pressures, ojj Rotational energy of molecules, 35, 32,
ionized gaa in magnetic field, 370, 378 206. 308; lag in transfer, 251; and vol-
Probability, 4, S, 38,58,2¾ 29,199.298,333 ume viscosity, 216. 245
Products: of vectors, scalar and vector, 11; Rough spherical molecules, 36, 107. 317;
of vectors with tensors, I6J of tensors conductivity, 319. 348; viscosity, 110.
with tensors, 16 336; volume viscosity, 210. 220; diffu-
Properties of molecules, soj flow of, soj sion, 216. 239; disadvantages as a
we also Macroscopic properties molecular model, 220
Runaway electrons, 398. 300
Quantum theory, 8, 9, 183. 208, a n . 333-
43: collisiona in,] 333: equations of Scalar product of vectors, 11; of tensors, i6_
steady state, 330: phaae angles, 338; Screw motion, 76, aoo
comparison with experiment, 330-33 Self-diffusion: mean free path theory, 103:
Quasi-Lorentzian gas, 193. 371 accurate theory, 194; second approxi-
mation, 105: Kelvin'a calculations,
R, gas constant, 34; numerical value, 43 266; diffusion of isotopes, 267. and
Radio propagation, geomagnetic influence similar molecules, 268; quantum cross-
on, 363 n, 379 sections, 328
Range of molecular influence, 6$, 178,21a n Separation of gases, 268. 369, 353
Rarefied gas, MS, a8i. aos. 306: viscosity, Shock waves, 381. aoa. 304. 395. 306: pro-
08: stress system, 38;. 288, 291, 294: files by B.G.K. method, aftj, and by
heat flow, 284. 286, 291, 204 methods of Grad, 394, Wang-Chang,
Rate of strain tensor, 15, 380 304. and Mott-Smith, aoj
Ratio: of specific heats, 41, 4 1 347: Slip at wall, 52, 228
thermal conductivity to viscosity, f — Smoothed distributions, 2fil
A//K,, see under Conduction; persis- Solar atmosphere, 273, 36*
tence-ratio of velocities after collisions, Solid angle, U
04. 95; thermal diffusion ratio kr, see Solubility of integral equations for/1", 1 IJ,
underT 133. 137. *«»
 SUBJECT INDEX 421
Sonine polynomials, J2J T h e r m a l diffusion ratio k r ( - DrIDit)
Sound waves, see Damping and factor <x„ ( - k r /x,x,), I 4 i 14¾
Specific heats, 2¾ 4¾ 222, 247; monato- 268; general expressions, I 4 i 147,165;
mic gas, 42, 240; diatomic and poly- for Lorentzian gas, 18¾. rot, and con-
atomic gas, 4_li 208; rough sphere gas, vergence, 102: first approximations to
218; numerical values, 42 k r and a „ , 165; measurement by the
Speed, molecular, 25; peculiar, 27J numeri- diffusion thermo-efTect 270: isotopic,
cal values, j6^ of effusion, 36^ of 27i. 275; sign of k r and a „ , 232, 274:
sound, 36 variation with concentration ratio, 273,
Spherical molecules: see under Molecules 277 (change of sign, 274). and with
Spherocylinders, 197. 221 temperature, 276, 277 (change of sign,
Statistical mechanics, 3, 6, 37. 42. 80. 173, 276); numerical values, 276, 278;
8¾ 108^ 333 n, 334 multiple mixtures, 345, 3<i.3S3. 3_S5
Statistical weight, 333, 340 T h e r m a l flux, 4 i 45, 25¾ 314; approxi-
Statistics: of encounters, 58; Bose-Einstein, mations to, first, m : second. 12;. 142.
3*6, 1 » : Fermi-Dirac, 326, 333 2«3, 107; third, 285
Steady state: uniform, 67J in smooth Thermal stresses, 291
vessel, 7 j ; in presence of external Thermo-efTect of diffusion, 103, 143, 270;
forces, 7_5J 258; mixed gas, 80, 82; gas application to the measurement of D„,
with internal energy, 203. 20s. 206. kr and a „ , 270
322; dense gas, 301, 315; ionized gas in Third approximation to velocity distribu-
magnetic field, 3 7 ' : quasi-steady, in tion function, chapter Li, 280-06
rotating gas, 22, 206; in strong electric Thirteen-moment approximation, 294
field, 383 Time of relaxation: see under R
S.T.P., 26 n, 36, J3, 8¾ 228, 263. 281. 297 Trace gas, and viscosity, 239, thermal con-
Streams, quantum interaction of mole- duction, 253, and diffusion, 261. 254
cular, 323 Transfer, equation of: see under Equation of
Stress-system: see under Pressure change
Summational invariants: see under I Transport phenomena, 97, 322. 336. 364
Sun, atmosphere, 273, 362; sunspots,
magnetism, 362 Ultrasonic waves, 24s. 292. 3 to
Superpotentials, 402 Uniformity of density, 4
Surface element, M, 30. 32 Unit vector, 10, i_4j tensor, i j
Sutherland's constants: S^ (diffusion),
182, 258: £ (viscosity and conduction), Van der Waals' attractions, 308
182. 232; numerical values of, 221 Variables, molecular, 25, 1Q7. 109. 217:
Sutherland's molecular model, 180 n, 182, of encounter, 54
232. 256; Vector, joj divergence of, i_3J flux-vector,
Sutherland's formula for viscosity of mix- 31; gradient vector, 12, 25, 200:
tures, 108. 243 integrals, 2j_; position-vector, I2_;
Swing separator, 270 products with vectors, 11, and with
Symmetric; tensor, i_5j function, 114 tensors, ^6j rotation vector, 12. 211.
372: unit vector, 10, 14J of tensor, 24
Temperature, 8, 3 i ali 4Si 5fii *22, l i t : Velocity, molecular, j , 25_, 36J distribution,
absolute, 27J internal, 215; electron. 27; distribution function, see under D ;
.106 -domain, 25J -gradient tensor, 18:
Temperature drop at wall, 101. 248 mass-, 27j 44, I4J, 3$8; mean, 27, 44,
Tensor, 14^ conjugate, L5J divergence, 1%; »43, 3<8; peculiar, 27J -space, 25_j
non-divergent, L5J products with angular velocity, of molecules, 26, 206.
vectors or tensors, 16; rate of strain and 208, 217. of gas, 76, 206; see also
rate of shear tensors, 15, 280;' sym- Persistence
metric, I J ; antisymmetric or skew, 23; Vibrational energy of molecules, 25, 37.
unit tensor, 1 s ;vector of tensor,24 ;velo- 201. 208, 245. 251
city-gradient tensor, I 8 J integrals, 2J Virial coefficients, 237, 309, 316
Ternary mixture, 343, 3*3. 354 Viscosity, equations for stress components,
T h e r m a l conduction: see under C 126
T h e r m a l diffusion, IOJL, 140, 14s: w also Viscosity coefficient, fi, for simple gas (see
under D ; column, 268. swing separator, below for mixed gas): elementary
270: in multiple mixtures, 351 theories, 92, 104, and effect of low
422 SUBJECT INDEX
Viacoaity coefficient (COM.) proximateformulae,108.243; accurate
pretsurei, 58; accunte general for- general formula, 144. 148: fir*t ap-
mula, 1*6. 131; notation for approxi- proximation (general expreaaion), i6s.
mation*, n o ; approximation* for rigid 238: Lorentxian gaa, too. 191. 103;
elastic apherea, 168, aao, point-centre* variation with concentration-ratio, a30-
of force, 17a. t73, aa6, inverae- 42. and with temperature, 242: mul-
aquaie law, 179., Sutherland molecule*, tiple mixture*, 248. 352; ioniaed ga* in
18a. aa6, Lennard-Jonea ia, 6 mole- magnetic field, .169. 378
cule*, 181. aa6, exp;6 molecule*, 186. Vlasov equation, 403
226, molecule* with internal energy, Volume-element dt, 13
ai4. 34». and rough apherea, aio. aao; Volume vitcotity: in denae gaa, 307.110; in
Kihara'a approximation, 161; Enekog'a gas with internal energy, 197. ai4. a n ,
denae-ga* expreaaion, .107; numerical 244; rough apherea, 219: relation to
value* of /t at S.T.P., 228; dependence relaxation of internal energy, a n , 245
on denaity, 227. and on temperature,
229; value*, inferred from p, of the Wall, presaure on. H i alip at, 99, 328;
molecular diameter, 228, 230, force- temperature-drop at, 101. 348
index v, 232. Sutherland conitant Waaailjewa formula for conductivity of gas-
S, 233. and the ia, 6 and exp; 6 mixture*. 100. a<6
conitant* «/*, a and %,, 237. 238; tee Wave field* of molecules, 322. 333
alto Volume vitcoaity Weight, molecular, 39J atatietical, 333,
Viacoaity coefficient, /L, for mixed gat: ap- 34?
 REFERENCES TO EXPERIMENTAL DATA
FOR PARTICULAR GASES
K E Y . Simple gas: £ = specific heat, V = viscosity, C = thermal conductivity,
SD =• self-diffusion, ITD « isotopic thermal diffusion.
Gas mixtures: MV = mixture viscosity, MC = mixture conductivity, D = diffusion,
TD = thermal diffusion. In each case the table number of the second gas is given in
parentheses following the page number.
Quantum results and those at high pressures are distinguished respectively by the
addition of Q and P.
1 Acetylene, C,H, V, 228, 212. a n ; C, 242
2 Air S, 43J V, 228, 232, 233, 217: C, 242
3 Ammonia, N H , S, 43^ V, 228, 212. 233, 234: C, 240: D, 263 (4, 11, 12, t£, 18, 24^
265 (4. LSll ITD, 234, 276; TD, 234 (18)
4 Argon, A 5 , 43J V, 228, 232, 233, 2 3 1 238; MK, 241 (11); C 240: MC, 254 ( n ) ;
D, 261 (11), 263 (3, 6, 11, 12, 13, 18, L9, 22, 24k 265 (3, II, ifQj SD, 267; / T D ,
276; TO, 225 (13). »78 (6, 11, 12. a , 18, ifi, 22, 2 4 ^ VP, 310; C/>, 310
5 Carbon monoxide, CO S, 43J V, 228, 232, 233. 237, 238; C, 240: / ) , 263 (6, io, 12,
15, 22X 265 (22); S D , 266, 262, 268; ITD, 221; 7"D, 278 (12)
6 Carbon dioxide, CO, S, 4 3 ; V, 228, 232, 233, 232, 238: C, 240: D, 263 (4. S. 8 . 1 1 . 1 2 .
16, 18, 12, 21, 22L 265 (11, 12, 18, 12, ?2h S D , 266, 262, 268J ITD 27s: TD, 278
(4. " . 12, ta, 222
2 Chlorine, CI, S, 43J V, 228, 232, 233. 237: C, 243
8 Deuterium, D , S, 431 P, 228, 232, 237: C, 249J D, 203 (6, 12I; TD, 278 (19)
2 Ethane, C,H, S, 43J V, 228. 233, 238J C, 249J D, 263 (12, 10)
10 Ethylene, C,H, S, 43J V, 228, 231; C, 249J D, 263 (5, 12, i_92i TD, 278 (12)
U Helium, He S, 43J V, 228, 230, 231, 23¾ 233, 237. 238; MV, 241 (12, 42; C, 249;
M C , 254 (4); D, 2DJ (4), 263 (3, 4, 6, 12, 15, I8J 15, 22, 24), 263 (4, 6, 18, ifili
SD, 267J / T D , 276; TD, 278 (4, 6, la, L£, 18, 13, 242J KP, 310; VQ, 331
l i Hydrogen, H, S, 431 P, 228, 23¾ 233, 2 3 ^ 238: MV, 241 (11), 243 (13); C, 240:
D, 2J6J (19), 263 (3, 4, SJ 6. ?, 9, 10, 11, 16, 1 3 , 2 1 , 2 2 , 23), 263 (6, 22I; SD, 266,
267: ITD, 2 7 i 276; TD, 278 (4, J, 6, 10, 11, 16, 18, ifi, 22ij VP, 310, PO, 332
13 Hydrogen chloride, HCI V, 228, 232, 233^ 237: MV, 243 (12); C, 249; S D , 267; TD,
225(4)
14 Hydrogen sulphide, H,S S, 43J V, 228, 233: C, 249
15 Krypton, Kr V, 228, 233, 237, 238J C, 249J D, 263 (3, 4, 11, 18, 24), 265 (3); SD,
267: / T D , 276; TD, 278 (4, 11, 18, 24J
ifi Methane, CH, S, 431 P, 228, 232. 233, 237. 238: C, 240: D, 263 (6, 12, 22), 26$
(22); SD, 267J / T D , 276; TD, 278 (12)
12 Methyl chloride, CH.C1 V, 228, 233J C, 240
18 Neon, N e S, 4 j j P, 228, 232, 233, 232, 238J C, 249: D, 263 (3, 4, 6, 11, 13, 24),
265 (4, 6, 11); SD, 262J ITD, 276; TD, 224 (3), 278 (4, 11, u , 15, 242; VP, 310;
CP, aio
12 Nitrogen, N , S, 43J V, 228, 232, 233, 234, 232, 238J C, 249J D, 261 (19), 263 (4, j ^
9, 10, i i , 12, 22,24), 263 (6, 11); SD, 262, 268: ITD, 276; TD, 278(4, 6, 8, 11, 12,
2 i 2221 PP, 309; CP, 310
ao_ Nitric Oxide, N O S, 43J V, 228, 232. 233. 237; C, 242
21 Nitrous Oxide, N , 0 S, 431 V, 228, 23¾ 233^ 237: C, 249; D , 263 (6, 122; SD, 268:
TD, 278 (19)
22 Oxygen, O , S, 43J V, 228, 23¾ 233, 232J C, 249J D. 263 (4. 5. 6. n . 12. 16. 19. 24).
26n (5, 6, 12, I6}J S D , 266, 262i / T D , 276; TD. 278 (4, 6, 12, 19J
23 Sulphur dioxide, SO, S, 43J V, 228, 232, 233, 237; C, 249; D , 263 (12)
24 Xenon, Xe S, 43J V, 228. 233, 232, 238; C, 249J D, 263 (3, 4, 12, 15, 18, 15, 221;
SD, 2621 ITD, »76; TD, 278 (4, 11, ! i i J l i VP, 310
[423]