us ‘5. THE STATISTICAL THEORY OF IRREVERSIBLE PROCESSES 2 wo B R BLK. a 1s 6 w. ws B. 2. 2. 2. 2 ™. a. %. 7. 2 2». . aH. 2D. 2. # a. ie m 4. ». 4. 4. a 4“ “4 A. Einstein, Arm, Physik, 1, ©, Khintchine, Math, Ann, 109, 604 (1934) H. Nyquist, Piys, Roo, 32, 110 (1928), R.F. Greene, H. B. Callen, Phys, Rev, 83, 1231 (1951). ‘A, Munster, in Fluctuation Phenomena in Solids (RE. Burges, ed.) Academie, New York, 1968, p. 180, M.van Viet and J. R. Fasset, ibid, p. 268 G. Casimir, Rev. Med, Phys. 17, 343 (1948), W. Thompson, Proc. Roy. Sac, (Edinburgh), 3, 225 (1884). HL. Helmholtz, Wied. Ann., 3, 201 (1876, W. Nernst, Z, Phys, Chen 4, 129 (1863), '. R. de Groot and P. Mazur, Nom-quiirium Thermodynamics, North-Holland, Amsterdam, 1962, J. L, Doobs Stochastic Processes, Wiley, New York, 1953. Rd. Robin, J, Mash. Phys, 1,309 (1960), RJ. Rubin, J. Maik. Pays., 2,373 (1961), M. von Smotuchowski Ana. Phye., 24,756 (1906) J. Hachigian and M, Rosenblatt, nn. Mah, Sia, 2, 1112 (1988), 1N. Comngold, Ana, Pye, 6, 368 (1959). 1G Andersen, 1. Oppenheim, K. , Shuler, and G. H. Weis, J. Mark Py, §, 522 (196%), ©. D. Fokker, Ana, Physik, 43, 810 (1914). M. Planck, Sizber. Preust, Aked. Wiss, 1947, 326, N.G. van Kampen, in Flictaation Phenomena in Solids op. ct, p19, G. E. Unienbeck and L. 8, Ornstein, Phys. Rev., 36, 823 (1930) 1LN. Sneddon, Fourier Transforms, McGraw-Hil, New York, 1951. M. Abramovitz, Handbook of Maiheraricel Functions, 1. A. Stegun, ed., National ‘Bureau of Standards, Washington, 1964, p. 932. J.L. Doob, Arm Mark, 43, 351 (1942) M. Las, Rev. Mod. Phy, 32,25 (1964). M,C. Wang and G.E. Ulenbeck, Reo. Mod, Phys, 19, 328 (1948) N-G. van Kampen, Physica, 23, 707 (1987) M. 3. Lighthill, An Introduction ro Fourier Analysts and Generated Functions, Cambridge University Press, England, 1958, R. Zwantig, in Annual Review of Physical Chemistry, Vol. 16, Wiley, New York, 1965, p. 67. HB. Callen and T, A. Welton, Phys, Ret, 83,34 (195), LD, Landau and E. M. Lifshita, Statistical Physics, Pergamon Press, London, 1958, M.S. Green, J. Chem. Phys., 19, 1036 (1851). R. Kubo, J. Ps. Soe. Japan), 12, $70 (1957), 49 (1905), M. Lax, Phy, Ret, 38, 359 (1966) : RR Bellman, Introduction ro Marrtx Analysis, McGraw-Hill, New York, 1960, p. $9, CHAPTER 6 THE KINETIC THEORY OF DILUTE GASES* E.G. D. COHEN Introduction In Chapters 6 and 7 we shall atiempt to give a microscopic explanation of the macroscopic properties observed in dilute and moder. ately dense gases The macroscopic propertics discussed here can be classited into: (1) {everal properties—which ate Lo u large extent insensitive to the precise nature of the intermolecular potential field--and (2) special propertics-— which depend sensitively on the intermolecular potential field. Included in the first class are: the irreversible approach to thermal equilibrium of a 885 which initially is not in thermal equilibrium and the validity of the hydrodynamical equations (or the laws of irreversible thermodynanycs). Included in the second class are: the temperature dependence of the {thermodynamic functions and the transport coellicients of the gas, A microscopic explanation here means an explanation based on the motion of the atoms and/or molecules which make up the gas. assuming {he interatomic or intermolecular potential teld is known, The determina tion of the intermolecular potential is, in principte, a quantum mechanical Problem. We will always assume in this chapter thal the intermolecular potential is additive, depends only on the imermotecular distance, and is short ranged, (i, it falls off rapidly to zero when the intermolecular distance exceeds a few angstroms). ‘The main problem we shall he concerned with will he the statistical See references Jo4 far general background, 9 ey120 6. THE KINETIC THEORY OF DILUTE Gases problem of making the connection between the microscopic properties of the molecules, on the one hand, and the macroscopic properties of the 428 in bulk, on the other hand, The difficulty of the problem is due to the fact that matter in bulk consists‘of an enormous number of molecules. One cubie centimeter of air, for example, consists of n= 3 x 10! molecules at a temperature of °C and a pressure of 1 atm. Therefore a microscopic explanation of the behavior of matter in bulk seems to involve the solution of the equations of motion of an enormous number of particles, Since the analytic solution of the equations of motion of three particles is unknown, the solution of the equations of motion for m particles seems completely hopeless. Already in the beginning of the kinetic theory of gases in the middle of the last century, Clausius, Maxwell, and Boltzmann realized that even if one had the solution of the microscopic equations of motion for the motion of the molecules, one still would have to connect this solution with such macroscopic concepts as the local density, the local tempera ture, viscosity, ete, They realized that these macrescopic concepts were in some way related to averages over the microscopic behavior of the gas Furthermore, they realized that by considering certain well-chosen Averages, it might be possible 10 avoid the necessity of obtaining the Full solution of the microscopic equations of motion of all particles because fone could hope that for the determination of these averages different und ‘much, simpler equations might be used. Of course, these simplified equations would still comtain some features of the microscopic equations ‘of motion of the molecules but would, hopefully, be much easier to handle! In the case of a difue gas, where only binary collisions between the ‘molecules occur, Maxwell and Boltzmann assumed that it was sufficient {0 consider the average number of particles, f(rv4),which are at time ¢ at the postion + with velocity ¥, for a description of the maeroscopie Properties of the gas in bulk. In this chapter we will discuss how one hus been able to explain many of the observed macroscopic properties of dilute gases on the basts of this assumption. Also, we will discuss the equation which f satisfies, those microscopic averages with which the ‘macroscopic quantities of the gas can be identified, the question of whether or not the so-defined macroscopic quantities actually do exhibit the same Properties as the corresponding observed macroscopic quantities of the ga In the case of a moderately dense gas, Bogolubov conjectured that ‘would stil be sufficient to explain the macroscopic properties of such a 82. In Chapter 7 we will see how far this conjecture is correct This 4. The Boltzmann Equation 121 chapter is intended as an introduction to what has been accomplished so far, In fact, concrete results have been obtained only for the case where the intermolecular forces are purely repulsive and where only binary and {ternary collisions between the molecules occur, 1. The Boltzmann Equation (5) AAs we already pointed out in the introduction, Maxwell and Boltzmann realized that for an adequate description of the dilute gas one should consider (771), the average number of particles at a given position + with velocity v at time ¢, Boltzmann was the first to write down “an equation of motion” for f, that is, an equation for the rate of change of f with time. This equation, che Boltemavn equation, reads in the absence of an oulside field of fore as: 7 Z (6.1a) Rt (ula) Acer) ar where a) = J any f bb f avelfeevofienn = flew flea} 6.16) ‘The meaning of the symbols in eq. (6.13,b) will be explained below. For our purposes this equation is sufkiently general. In ease there isan outside field of force, F, present, one has an additional term on the rf side: ~(Fjm) » (@/J@v).. We can see how eg, (6.1) is arrived at by the following elementary considerations. If one multiplies both sides of eq. (6.1) by de dy, the left- hand side (Lh.s.) of eq. (6.ta) gives the change per unit lime of the number ‘of molecules with velocity between ¥ and v + dy in a volume element dt around F. In other words, the left-hand side of eq. (6.1a) then gives the rate of change of the number of molecules, dt dy, in the volume element dh dv around the point r, vin espace, the six-dimensional phase spuce of a Gingle) motecute,*-+ According 10 €q, (6.14), this rate of change is equal to the sum of a streaming term, —v. (3//8¢) dr dy, and a collision teem, yp) de dv, Of these Lo terms, the streaming term gives the rate of change of f due the molecules have a finite velocity and, consequently, 4 The volume element ded im -space should be physically intinitesint large enough s9 that contains many molecules. On the other hand be small enough so that the magrosspie quanities ofthe gas do nat change appreciably acer Ae122 6. THE KINETIC THEORY OF DILUTE Gases change their positions with time. Thus molecules with velocity varound* ¢ at time ¢ will be around r + v dra little time dt later. While othet mole. ‘cules, viz., those around + — y dr, will be around r at time dt. ‘Therefore one has for the change of fdr dy by steeaming per unit time loss: all molecules in dr dy around ¢, that is, /(evt) de dv; sain: all molecules in de dv around — vi, that is, fle = vai, ¥, 0) ded; LO ae gy balance U(r — vd 0 ~ Ser, y Oded uA 0 — Fe, ¥, Ode, = (6.2) ‘The streaming term is therefore a “guin-loss™ term. This gain-ioss character is even clearer for the collision term, which sive the rate of change of fdue to the fact that the molecules change their ‘velocities with time because of the forces exerted on them during colisions, Specifically, the contribution of the last term on the right-hand side (rhs. of eg, (6.1a), —dr dv f dr, fab df dp g f(evt) fer), gives the loss Per unit time of the number of molecules fdr dv with velocity v around due to collisions with molecules with any velocity ¥, around . In fact, any such collision between two molecules with velocities ¥ and ¥, will result in two molecules with different velocities v’ and v, respectively. ‘Therefore any (v, ¥,)-collision results ina toss to / de dv because a molecule \with velocity v’ is nor part of fide dv. The number of (y, ¥,)-collisons (or so dlrect collisions) per unit time ean be computed by considering the felative motion of a molecule with velocity ¥ with respect to a molecule with velocity ¥, The relative motion is characterized by the relative velocity g = ¥ — v,, the impact parameter b and the azimuth angle y of the plane in which the relative motion takes place. The number of collisions in time di of mole- cules with velocity v with a molecule with velocity ¥,, such that impact Parameter b is between b and b + db, and the azimuth angle y is between Yand y + dy, is given by the number of molecules with velocity v in the “colision eylinder™t of volume gb db dy dr, and equals (cf. Fig.l) b db dp gf(evs) dv dt where g = el. * ity speaking, “with voces between vand y + thisimore elaborate deseipion of these fed. Siniaty, replied i the wea simply by "around 1 Sticty speaking, the igure isnot a eylinder, but common usage often refers t it 8a zallion ey For brevity, we willemvit de around #” will be 1, The Boltemann Equation 123 estiuting callision, o Fig. 1. Direct and restitutingcolisions in relative evondinate system (For Appendix B, Chapter 73r == tii Bas ed Ifthe gas is suficiently dilute so that the (collision) cylinders associated ‘with all molecules with velocity v, around r can be assumed not to overlap, then the total number of direct collisions per unit time around + and consequently also the Joss of molecules with velocity ¥ around Fis tiven by {ll naoecutes uh velocity ¥ got coals loss dey f anfaw » f dy Sten fern (63) ee {all molecules with velocity ¥ in dr around 0), ee e with velocity 4)124 6, THE KINETIC THEORY OF OILUTE GASES The contribution of the second term on the r.i.s. of eq. (6.12), adv f dy f bb f dp afl) S010, gives the gain per unit time to fdr dv duc to collisions. This gain can be computed in the same way as the loss to fdr dv was computed before, Two molecules with velocities v and ¥, will result if two molecules with, velocities v’ and ¥; (and the same impact parameter 6, but with azimuth angle y ++ =) collide (compared with Fig. 1). Therefore any such (v’,')- collision will result in a gain to fdr dv, Using then that |g'| = |v’ — vil = = g, that dv dyy = dv’ dv,, and assuming again that the collision eylind ers associated with different molecules with velocity v; around d+ do not overlap, one finds that the total number of (, vi)~-collisions (or also restcuting colisions) per unit time around r—and consequently, also the ‘gain of molecules with velocity v around r—is given by i dradvfav, | db b | dy afer) frvin) (6.30) The balance of the gain-loss contributions leads with eqs. (6.16), (6.34), and (6.3) to the following contribution to the change of f dr dv per unit time: balance: JCP) de dv (6.30) Equation (6.1) results from the eqs. (6.2) and (6.3c), after division by deay. A. Rewanxs asour J(f') (a) In J(f), 6 db dy is the classical differential scattering cross section. (b) The connection between ¥, ¥, and y’, ¥{ (and vice versa) involves implicitly the intermolecular potential field $(+) (c) The occurrence in J of two,/"s indicates that only contributions from binary collisions are considered in the calculation of 3/181. (4) The numbers of direct and restituting collisions, as given by the eqs, (63a) and (6.36), respectively, are a consequence of a certain Stoszzahl~ Ansatz.* In fact, the numbers have been computed using a statis ‘assumption; the assumption of molecular chaos, viz: itis assumed that at every time ¢ there are no velocity correlations at F so that the average * Stoszzahl-Ansatz i an assumption about the numer of colliions. ae 1, The Boltzmann Equation 125 umber of molecular pairs with velocity v and vy at r, fa(eryt), can be written as Pernt) = Men) fer) (6a) and simitarly for v’ and ¥, This statistical assumption is of a nonmechanical nature, and is the origin of the lack of time-reversal invariance of the Boltzmann equation and of the irreversible behavior shown by a gas described by the Boltzmann ‘equation (6,1) (cf, Section I1). For, changing ¢ to —r, and consequently also (= (dr/de)] to =, the Ls. of eq. (6.La) and the first term on the Lh. of eq. (6.42) change sign, while the second term on the rh. of eq, G6.ta) [ie., H(G)] does not. If, therefore, fev) is a solution of eq. (6.1) then /(¢, —¥, ~1) is nor. This is in contrast to mechanics where the basic equations (For example, the Lagrange equations) are time-reversal invariant. In that ease, corresponding to any solution 4(t) of the Lagrange equation: d (aL) aL (2b) _ a 65 aay) aq ae there is a solution q(—t) of the time-reversed (but identical) equation. aL aL (6.5b) nA —G) aq Thus 9(1) and q(—1) are both possible solutions of the mechanical equa- tions of motion. Icis clear that the statistical or “average” treatment of the mechanics of the molecules at ris the origin of this diference with mechanics. ‘The approximate nature ofthe hypothesis of molec chaos and therefore, ofthe Ssonah-Ansat 1 ested in that that te euling Bolen sgt) only desribesthe approach faite gas toward egulibfum. but not the Muay ‘roa equim shen the gas isin eulibrium, B. Sustwary ‘We conclude this section by sun he Boltzmana equation (1) A dilute gas can be adequately described by the single particle distribution function f(evt) alone. (2) Only binary collisions are considered. (3) The molecules move according to the laws of classical mechanics (4) The assumption of moleculat ehaus is used in the Stosezahl-Ansat7, narizing all the assumptions inherent126 6. THE KINETIC, THEORY OF DILUTE GASES. () A spherically symmetric intermolecular potential field (-) is assumed $0 that the initial velocities (¥’, v) of the molecules in the ges, tituting collision are equal, for all 6 and y, co the final velocities of the ‘or with velocities ¥’ and ¥; in the resi going to collide a time 1 has been computed as ifthe two colliding mole, cules were at the same position , in spite of the fact that they are outside the range of the intermolecular forces For the existence ofthe Boltzmann equation itis tol necesay to assume that intermotecular potemial Feld has a finite range, tht is. that p() vanishes for wt 1D fe One only nce o assume that 4) decease acemy Ys with needing + 40 thatthe collsion integral 4(J) exists. For the existence ef Jug), and a; hay Approach wand ¥;suBiciently fast for inreasng UL. Approach to Equlibrium I (H-Theorem and Principle of the ‘Chapman-Enskog Solution) All macroscopic properties of a dilute gas should follow from the Boltzmann equation. In this section we shall study some general questions concerning the approach to equilibrium: (A) Can one show that a gas, which is initially (ie., at time zero), not in equilibrium, will always approach equilibrium? (B) Furthermore, how does this approach to equilibrium proceed? A. HeTusonem (6) The general approach of a dilute gas to equilibrium could be proved by Boltzmann on the basis of his equation without solving it for f by the H-theorem. In fact, he proved that a dilute gas, whose time evolution is governed by the Boltzmann equation (6.1), will, for any initial state 1G), approach to thermal equilibrium which is characterized by the Maxwell distribution function f4,(v), < Lu) = nsx (. in) 66) ‘where mis the mass ofa molecule, nisthe number density, kis Boltzmann's constant, and 7's the absolute temperature. Hl. Approach to Equilibrium I ia We shall not give a general proof of this here but we shall restrict ourselves to a proof for the spatially homogeneous case where fev!) depends only on v and not on r. 1, Proof of H-thearem for fivt) H-Theorem, For any f(vt) that is a solution of the spatially homogene- ous Boltzmann equation, the function H (defined below) never increases with time. Proof. Define a function H by* He { ‘Sv In (vt) dv 7 Then using eq. (6.1), that 2/121 = 0 and that a(f,/de)/@r = 0, one has MH Vin sdv an In Inf — (6, Ten] pinser = fafa [referers m1 6a Here we have written /for f(r). for f(ev's), ete, Interchanging v and ¥, on the ths, of eq. (6.8), and adding the equation thus obtained to eq, (68), one has a - pferfenferefereingurrs -fil 69) Interchanging (v,¥,) and (¥'.¥{) on the rhs. of eq. (6.9), using the fact that dv dv, = dv’ dy, and that g'= g’, and adding the equation so obtained to eq, (6.9), one has a = ferfonfarofavetn (4 as the rh.s, of eq. (6.10) is always <0, From eq. (6.10) it follows that dHH/dr < 0 for all t, thus H can never ease. Jun <0 6.10 * The symbol H does not stand for capital but for capital (grck) a, thusemphasze Ing the connection with the entropy 1 & S, Infact, in equilibrium, when f(t) fa, I ~ Sik + constant,128 6, THE KINETIC THEORY OF DILUTE GASES 2, Proof of Approach of {to Equilibrium H will continue to decrease* until finally, as > co, dH/dé = 0 for all ¥5%iy by y (that is, for all possible binary coltisiors), and - LOS) = SOF) (6.11) Taking the logarithm of both sides of eq. (6.11), we have Info) + Inf) = In fw) + Io foo) 6.12) This equation shows that In /(vt) is an additive collisional invariant, hat is, the total of in summed over the two colliding molecules before. and after-a collision is the same. Since the oniy independent additive collisional invariants are the number (1), the momentum (mv), and the (Kinetic) energy (Jim) of a molecule, In f must be. linear combination of the five invariants mentioned, that is, Inf) = a+ bev + feme? or also SW) = aexp [—Ielv ~ bje)') (6.13a) where a, , and c are constants, By using the defitions: " =fro av (6.14ay a= [fio dv = 0 (6.140) InkT = froin = wid (6.140) one can relate the five constants a, b, and ¢ to the usual constants density (n), average velocity (u), and temperature (7) thet characterize thermal equilibrium, One obtains £0) = fut) = (sti) o0( a) (6.136) where we have set u = 0, * That H cannot decrease indefinitely is due to the fact that H must have a miimam yalue, For if this were not so, and could be —, then this would meen that S/'nf ds diverges. This could ony be 0 if, fr v= « andj ~-0, Inf would go faster to —o than f+ 0, This ean only be 50 if In/ goos 10 infinity faster than et 36 Sf mo do—the total kinetic energy of the gas—must certainly exist. However, if Info", with n > 2, then fneexp(—o%), and H will certainly ext? Tr i 11. Approach to Equilibrium £ 19 Therefore, f approaches, through binary collisions, monotonically to the equilibrium distribution function given by eg, (6.13), Strictly speaking, only obtains fy afer an infinite time. For all practical purposes, however, J will equal fy to a very good approximation after very many collisions have occurred, that is, when £>> fyyy» the meanefree time, ‘The proof of the approach to equilibrium in the spatially inhomogene- ‘ous case proceeds in a similar fashion. One considers, instead of the H- function defined by eq, (6.7), the function H' = f H dr (f depends on Fy, and (now). (1) First one proves an H-theorem for H’. dH'/dt contains now contributions from the streaming—as well as from the collision--term in the Boltzmann equation. Since the contribution from the streaming term can be shown to vanish, H’ changes only through collisions. One can show, by using the preceding arguments, that H’ never increases (H-theorem) (2) Tite proof of the approach of fto equilibrium now proceeds in two steps. (a) First, one has from the H-theorem that /, in the course of time, will approach monotonically a focal Maxwelledistribution function fle), which is given by an expression of the same form as that on the hs. of eq, (6.13), except that now n, u, and 7 will depend on r and ¢ (the average velocity w isin general not zero), that is, m_ P47 mvten) 1) = ae — ave 6.134 fie = m0 res) °° (—Femen) 6H Here, the peculiar velocity, V(rr), defined by V(r) = = w(ee) (6.156) is the velocity of a molecute at ¢ relative to the average velocity w at r, Ineq, (6.15) the functions n(r,), u(, 1), and 7(e, )are the local number density, the local velocity, and the local temperature of the gas, respec- lively. These are defined by [compare eq. (6.14)] rer = [rer an 6.163) nrtvtrt) freer dy (6.160) incokT(ep = [ ey ae (6.160) (b) Then, to determine the r and 1 dependence of the five functions130 6. THE KINETIC THEORY OF DILUTE Gases and T, one substitutes fy into the Boltzmann equation. Since JC) vanishes, the change of fy with time is given by the streaming term only, One can then demonstrate that the only solution of this equation for f, is the (total) equilibrium distribution Function fy given by eq. (6.6)* so that n ad Tare independent of r and s and u = 0. In reality, the approach to total equilibrium in a spatially inhomo- geneous system will also go in two steps, as is suggested by the proof Just given. First, in a time of the order of the meanefree time fry, the system approaches monotonically a state of local equilibrium in wl the velocities are already distributed according to a Gaussian (Maxwellian) distribution, except that the five constants which characterize this (local) velocity equilibrium vary still with position and time. Then, in a time of the order of the traversal time of the container imac [see (B)}, it approaches (not necessarily monotonically) the state of tatal equilibrium where also all spatial variations have disappeared. For the first step, one should ‘emphasize the word “approach”: The state of local equilibrium is never reached, but only attained to a first approximation, This follows from the fact that fis not a solution of the full Boltzmann equation but only of the equation J(ff) = 0. The function fy, however, isa solution of the full Boltzmann equation! B, THE CHAPMAN-ENSKOG SOLUTION (7,8) Although the H-theorem establishes once and for all the fact that a gas will approach equilibrium, it does not provide information about how this approach to equilibrium proceeds, other than that this approach proceeds in two steps, In fact, it seems that in order to obtain more information, fone would have to obtain a solution of the Boltzmann equation itself; a formidable undertaking in view of the fact that the equation is a non= linear integrodifferential equation in six variables, in addition to the time. However, for the case that the gas is near equilibrium, Chapman and Enskog have been able to avoid this formidable problem. They have been able, by obtaining a special solution of the Boltzmann equation, to gain mote information about the approach to equilibrium, while at the same ‘In the (core realistic) case that an outside force Feld F(e is present, un additional term appears inthe equation fr f. But for some very special academic) cases, the solution of this equation for fy is the Maxuell-Bollzmann distribution Tune Sux = fu exp—ViCVkT) where H(A) is the potenlat of the outside force field Fl): Fe) = ~aM(n/er, I, Approach to Equilibrium 131 time establishing the connection with the equations of hydrodynamics. The details of their work will be given in the next section. In this section wwe will sketch the general ideas behind the Chapmen-Enskog solution of the Boltzmann equation. The Boltzmann equation describes mathematically the time evolution ofa dilute gas. However, physically, it wll be clear that the time evolution of the gas, or of any system for that matter, will be dominated by the characteristic relaxation times of the system, In our case of a dilute gas there are three characteristic relaxation times which are related to three basic lengths in the gas: the (effective) range of the intermolecular forces ‘op.the mean-free path J, and a macroscopic length for example, the length L of the vessel, which contains the gas.* For a dilute gas that obeys the Boltzmann equation, one has KIEL The characteristic times f, ary ANG fyyurs COFFeSponding tor, f, and L, respectively, can be obtained by dividing ry, /, and by a characteristic velocity. It is convenient to choose the velocity of sound for this: v, I0* cm sec *. Then one obtains the following table which is characteristic for a dilute gas at O°C and 1 atm pressure TABLE I [) |e we [we |e we" | or | oe For our ease, therefore, the three characteristic times are far apart v 1, & tote Haar The first inequality: 1, & tgp. exprenses the fat that the gus is dilure and that each molecule spends most of its time in a free Might which is ‘ccasionally interrupted by w (binary) ‘or, replacing 1, & by the original inequality’ ry < F and using that ~ Hang one sees this, ws 4 ly tra ithe intermatecula Force ae short ranged. Hor Coworub forces for example, whore #ir)~ Hr, soca distinction not132 6. THE KINETIC THEORY OF DILUTE Gases implies that 73 « 1. The second inequality, taty < face permits—for a large class of initial states—to distinguish in the time evolution of the pas towards thermal equilibrium two steps or stages: 2) For times 0<1< tarp @ kinetic stage, where for an adequate description in its approach to equilibrium, the gas has to be described by J(evt) which has to be obtained from a general solution of the Boltzmann ‘equation; (b) For times £>> tmeps & Ayidrodynamical stage, where the gas, for an adequate description in its approach to equilitxium, can be described by the five hydrodynamical quantities n(et) (et) and’ T(t), instead of by J(evt). The five bydrodynamical quantities are defined by eq. (6.16) and ate, therefore, related to the first five moments a} f with respect to v only, ‘Therefore, itis clear that the hydrodynamical description of the gas by ‘means of 2, u, and Tin the hydrodynamical stage is much simpler than the description by means of the full distribution Function f(evt) in the kinetic stage. The approach of the gas to equilibrium is thus accompanied by a simplification of the description of the gas. ‘The existence of these two stages, in particular the hydrodynamical stage, can be established on physical grounds by the following argument, For almost any initial state at 1 = 0, every molecule in the volume element dr around r will have collided several times after a time ¢ 3> ues During each collision the velocities of the colliding molecules change ‘enormously due to the strong intermolecular fores, AS a consequence, all functions depending on the velocities of the molecules at r, in particular J(es), will change very rapidly with time, Hovever, five functions, viz., net), w(e), and T(t), will not change at all because of collisions. This is so because the number, momentum, and energy of the molecules are conserved during collisions. Therefore, as far as collisions are con cerned, n, u, and Tat x will not change af all éuring the time ¢ and will remain constant during all the collisional turmoil that is going on at during the time r. As a consequence, all the rapidly changing quantities at r (and in particular /) will adapt themselves to the “prevailing’” n, 4u, and Tat r during the time ¢, and will have become dependent on n, uy and 7 after a time £9 fyqy. The preceding argument only considered the change of n, u, and T due to collisions at r, but neglected the change of 1, u, and T due to spatial inhomogeneities in the gas at r, that is, due to the gradients of 1, 4, and Tat r. Now, ifthe variations of 1, u, and T over a mean-free path are small—as is almost always the case—then one can in first approxima- tion neglect the change of 2, u, and T due to spatial inhomogeneities at, Il, Approach to Equilibrium 1 133 and the preceding argument still applies. More precisely: if one calls any of the macroscopic quantities (x, w, or 7) M, then the argument applies if the relative change of M over a mean-free path is small; in other words, if Tgrad MK 1 where AM is the difference in Mf over L.* Therefore, ote would indeed expect on physical grounds that after a time ¢ >> fap, the t'me evolution of the gas can be adequately described by the time evolution of the five hydrodynamical quantities alone, and that the full (ev) is not needed anymore, Continuing this physical discussion ofthe time evolution of the gas, one would expect after atime ¢ 9 fmace, When every molecule has traversed the vessel many times, that all spatial inhomogeneities have been equalized and that the hydrodynamical state of focal equilibrium of the gas which is characterized by the five functions n(r1), u(r’), 7(et) will have gone over into the final state of fora! (thermal) equilibrium of the gas, characterized by two parameters a and Talone, We remark that the approach of the gas to equilibrium is then again accompanied by an increasing simplification of the description: sage: Kit pteatystat sent descriptions flo) eh. Te) eT A different way of expressing the existence of two stages in the approach of a dilute gas to equilibrium is to say that one can distinguish in the approach of such a gas to equilibrium two processes: (a) A fast process on the time scale of /yy, Which establishes (in first approximation at least), after a time >> nny & cal (velocity) equilib- (b) A slow process on the time scale Of fyinee Which teads, after a time 1) fyqes 10 total equilibrium, The uniformity parameter jis propor tional to the ratio of these two characteristic times, vi: H~tastolovece Af he system is uniform and AF = 0, then x = 0. Thus 1 measures the deviations ‘oun spaial Hemogenely, and Is consequently éaled tne uniformly parameter,134 6. THE KINETIC THEORY OF DILUTE GASES ML. Approach fo Equilibrium It—The Chapman-Enskog Solution: ‘Connection with Hydrodynamics (9) In ths section we shall make the physical results ofthe previous section ‘mathematically more precise. We shall study the approach to equilibrium of the gasin the hydrodynamical stage, thats, forties ¢ > fy when the 408s already near equilibrium. As argued in the previous section, we can {ook then for special solutions of the Boltemann equation which are such that /, as far asits time dependence is concerned, is completely determined by its fist five moments: 1», u, and 7 alone. Thus Kew) > flew | nu) (6.17) so that Yee, after | nut) _ afer | nat) an a a ana Alev|nut) Bu, ater} nut) ar ate a 18) Here /(ev| nu) indicates that the dependence of fis through n(r1), (rt) and Ter), Assuming eq. (6.17) one can actually find a special solution of the Boltzmann equation for f of this form in successive approximations by expanding /(rv| nu7) in powers of the uniformity parameter j (1 < 1) {lev | muT) = filer | maT) + wher | nut) + piper |auT) +--+ (6.19) In this expansion the parameter is meant as a parameter indicating the order of magnitude of the terms. It is used like the parameter 4 in the Rayleigh-Schridinger perturbation theory in quantum mechanics to classify terms according to their magnitude. Like 4, xis set equal to one at the end of the calculation,» Substituting this expansion in the rh.s. of the Boltzmann equation and using that d/dr ~ 4, one obtains the following expansion for the r.h.s, of the Boltzmann equation (6.1)*: ve Ihde + wf Se As i has not been assumed that the + dependence of f should be only th 1, and T, it has been tacitly assumed in obtaining (6.20) cha the variation Off with ¢ ‘heough any explicit dependence of fon r (chat i, not via mw, of T) does nol eweed the variation of fwith# through m, 1, and 7: (HUN JUHI] Peo 620) Ht. Approach to Equilibrium 1 135 To obtain a z-expansion of the L.h.s. of the Boltzmann equation one needs to know, in view of eq, (6.18), the z-expansion of an/ar, u/dt, and 07701. Now equations can be obiained for these quantities by multi- plying both sides of the Boltzmann equation with I, v, and 4m, respee- tively, and integrating over ¥, Using then that the collision term, J(f'), does not contribute due to the conservation of number, momentum and energy in a collision, one obtains the following five consereation equations which expe the conten of number, momentum, and eneEy ively, during the time-evolution of the gas: "pation of Contry (Conseration of Newer) da nM nd, (6.21) a ar, Equation of Motion (Conservation of Momentum): 7 nm Bt =~ G=xye) (6.21) Equation of Energy (Conservation of Energy) 3,d0_ _ ak 3,4tt PED, (6.216) de ore Here is the substantial time derivative, a ayy ent y yp wre a wa a ae, ‘The summation convention is used so that if two identical greek indices occur, itis understood that a summation over the x-,-,and z-components has to be carried out. The pressure tensor PK is given by Ph = PR f= [rmee, Ww (jen (6.22) while the heat Mux vector JB is given by ix (6.23) Bee Lf) = [siya Giaen The superscript K indicates that in the computation of the pressure tensor, and the heat flux vector J,. only kinetic contributions due 10136 6, THE KINETIC THEORY OF DILUTE Gases the transitional motion of the molecules have been taken into account. ‘The rate of strain tensor D,, is defined by 1/au , Ou 7 Dy = A(e 4 eee tae) om. We obtain from eqs. (6.21) dn/at, Ou/, and 97/21 in a power series in by expanding Af and J® in powers of su, by substituting the p-expansion 4, (6.19) for / into the expressions (6.22) and (6.23), respectively, and collecting all terms of the same order in j. One obiains then the fotiowing, geexpansion for dn[@t, Guj@r, and 87/2: (6.25) (6.250) SaDay) (6.25e Here one has used that : FE = Phy + uPhy bo (6.26) with ok = f fymvV,V, lv (6.260) f fam he (6.26b) and that JE ek + gd 6.2 with ee a f fobmvey, dv (6.274) JK= fi fp ymveV, dv (6.27) (We have omitted the appendage (i = x, y, 2) or (i,j = x.y.) from the preceding equations, From now on it is always assumed that the summa tion of f and f runs over the axes x, y, and z.) Using now eqs. (6.18) and (6.24)-(6.27), one obtaine the Following 111, Approach to Equilibrium Ht 137 srenpansion for the 1... of the Boltzmann equation (6.1): nf 20(2) 4 Aho, (24) 4. Aa /2T) 2 ARGS BBG) te to 20 Equating then the coeficients of equal powers of on the Ls. and the ris. of the Boltzmann equation, one obtains, with eqs, (6.20) and (6.28), the following set of equations for fy fi IUofo) = 0 (6.29a) OW: Jef) + MLS = ue) + x. Gi, + ae) +r & (6.290) “The eqs. (6.28) can be solved successively for oy |A, SOLUTION OF THE fy EQUATION ‘The equation for fy is identical with that obtained from the equation dtifdi = 0 in the spatially inhomogeneous case. Consequently, the equation has the local Maxwell distribution function fa. given by the 44, (6.188) a5 solution Equation (92%) is suisied by any Function of the (orm (6.15), with mu, and 7 arbrary functions of rand t, andeot necessarily the local density velocity and emer sure, respectively. We choove however—in accordance with the considerations ofthe previews tection-—n, u, and 7 as the focal hydrodynamical quantities, and this Enable us to abtaina well-defined yolaton seheme for the eqs. (6.2). Because the n, u, and T for the gas also follow from the eqs. (6.16), these same equations must be satisfied iff, instead of fis used, Thus, one ‘can require from all higher approximations, f, with [> 1, that fiir (6.30 fon veo (6.306) fran 00 Therefore, since J() is the collision term, one sees that (as was antic ipated in the previous section on physical grounds) f does approach ‘rough collisions the local eyuilibriuin distibution function fo138 6. THE KINETIC THEORY OF DILUTE Gases With f= fa, one can now evaluate the rhs. of the conservation ‘equations; in particular, one can compute 4 and Jf. In doing 80 one ‘obtains the Euler equations of hydrodynamics for an ideal gas: (31a) (6.310) — Py 316) Here one has used that, with f= fo, the pressure tensor is of the form: Pia = PR bys (6.32) Here p* is the hydrostatic pressure given by the ideal gas faw: pi m= nk (6.32) while the heat fux vector JE is given by J&=0 (6.3) and ek, the local (kinetic) energy density, is given by the ideal gas expression: ak (6.34) ‘One might wonder why the ideal gas laws ae obtained forthe pressure and the energy ery inspite ofthe fact thatthe Boltzmann equation contain eonteibutons fon binary collisions through the colsion term JC). The reason s thatthe colision term in the form of eg. (618) does not contribute to the enservation equations. In ordet to obtain contributions to the pressure and the energy density from binary collisions, the dferene in poriion of the two colliding molecules mst be taken ita account in the clision term, tha is, in J() the wo °s should wo be taken both a the same position We see that in local equilibrium (f= J), the thermodynamic functions such as the pressure and the energy arc the same functions of n(rt) and T(t), 8 they are in (total) equilibrium of w and 7:* B. SOLUTION OF THE fi EQUATION (LINEARIZED BOLTZMANN EQUATION Taeory) Using eqs. (6.25) and (6.31)-(6.33), the r.s. of eq, (6.296) can be determined, ‘This leads to the Following linear inhomogeneous integral * Combining eqs. (6.312) and (6.316), one obiains the adiabatic equation of state: dnt *Mah 0. IM. Approach to Equilibrium It 139 equation for fy: Joh) + IS 3 Inducing the Function &,, through the equation: faker | na) = flee | nuT}D (ev | au) (6.36) eq, (6.35) can be transformed into a tion for: aa) Pal (6.35) ear inhomogencous integral equa 1 (SPE (May Here the (symmetric) linear binary collision operator / is defined by 10) = far fon ft 6f aytor) + (99 ~ 8) ~ B49] 638) where, on the ¢.h.s, we have only indicated the velocity dependence of &,, ‘According to the theory of integral equations, the inhomogeneous integral equiation [eq. (6.37)| for ®, is soluble if ts r-hs. is orthogonal (© the solutions of the associnted homogeneous integral equation: 1) =0 (6.39) Since this integral equation has as independent solutions only the five collisional invariants 1, mV, and JmV*,* one easily convinces onesell that the solubility conditions for the integral equation (6.37) are Fulfilled.t ‘The general solution to eq. (6.37) can now be written as: general solution = particular solution plus 5 constants times 5 solutions of homogeneous equation. ‘To make the general solution of eq. (6.37) definite, one fixes the live constants by the five conditions for f= 1 feq. (6.30)}: [rived =0 (6.80) “the tise collisional invariants 1, m,n ate linear combinations of the Five ints F, my, at mentioned before. ian evalvation of lat + ¥ = 2f/8e to Oy, theorthoge onaity conditions ate nothing dhe than I ule’ equations of hydrodynamics?140 6. THE KINETIC THEORY OF DILUTE Gasts We have now established the existence of a unique solution of the integral equation (6.37) for ®. One can find the form of ©, by using that the integral operator 7 is a linear operator, that the r.h.s, of eq. (6.37) is linear in the gradients, and thit the gradients of T and w have a different tensorial character, i can then be shown that ® is of the following general form: ar ar, and, by substituting (6.41) for ‘, into the integral equation (6.37), that AAV) and B,,(V) satisfy the integral equatio Mev | nu) = —A4v) — Bag V)Dap (641) (mv? tay= -(Ge = ) (6.424) (mV,V, mv? 16, = — (mv KB) = — ("Ee ea (6426) Furthermore, it follows from the eqs. (6.42) that 4,(¥) is a vector, while BAW) isa traceess tensor because the tensor on the rhs. of eq. (6.42) is traccless. As V is the only vector available, the functions ,(V) and B,(V) must be of the form*: AM) = A), (6.43a) BAY) = (6.436) Because of eq. (6.40), the (scalar) function A is subject to the condition: Jrraany dro (6.44) While the function B(V) is subject to no restrictions. The functions 4(V) and (V)can be found from the integral equations (6.42a) and (6.426), respectively. Before we do so, we first detive the form of the conservation equations with f= f, + f,. We remark that as Li = J, we have from eqs. (6.41), (6.37), and (6.30): CW A~ aTf6r, and Aur, that is, ~ grad T and grad w. ‘+ For a more elaborated discussion of this point, see relerence 10. 1 The first wo conditions in eq (6.40) are automatically fulfilled for 4,(V), while ail three condivons in eq. (640) are automatically flied for B,(V>, This follows from the genera form of A,(¥) and B,V), respectively [se 6g. (6:3) IIL. Approach to Equilibrium 1 141 (2) fr is independent of n ‘The last result follows from the fact that the .h.s, of the integeal equation (6.37) must be independent of n as is the r-h.s. With f= fe + fi, one can evaluate the r,s. of the conservation equa tions. In particular, one can compute # and JX. In s» doing one obtains the Navier-Stokes equation of hydrodynamics: (6.44a) (6.440) nse a3 )Py (6.44) dt Here we have used that, with f= fy + fi, the pressure tensor 9, is of the form PE = Phot Pham PX, + Phy (6.4Sa) Here the off-diagonal part :*K, of # obeys Newton's law of viscous low: Py = —2S(Dyy = 4Dau5,.) (6.456) while the heat flux vector Ji obeys Futrier's law of heat conduction: or Oo The coofficionts n¥ in eq. (6.4Sa) and 2¥ in eq, (6.46) are the transport coefficients: the shear viscosity and the heat conductivity, respectively, of the dilute gas.* They depend on the temperature 7 but not on the density n because f; is independent of n. Using eqs. (6.26b), (6.36), (6.41), and (6.43b)—and the fact that averages of odd functions of ¥ vanish when evaluated with fg(¥) (which is IN dh (6.46) symmetric in V)—one obtains the expression for 1/8: kit) = 2 [pews av (647) =e | Similarly, using eqs. (6.27h), (6.36), (6.41), (6.436), and (6.44) one obtains * For the bulk viseosity see Chapter 7.la 6. THe KINETIC THEORY OF DILUTE Gases the expression for 2* aN) = 5 fm [GG - Janne To obtain the explicit temperature dependence of 7 and AK one must know A(V) and B(Y), and therefore solve the integral equations, (6.42), for A(V) and B(V). A solution of these integral equations can be obtained by expanding the functions (V) and B(V) in terms of a complete set of ‘orthogonal functions of ¥. A convenient set is formed by the Sonine polynomia, which are eigenfunctions of the linear integral operator 1 for the case of Maxwell molecules, that is, for the case of the inter- molecular potential (+) ~ I/r4. The convenience ofthe Sonine polyno expansion, even for realistic (nonMarwell) molecules, is due to the fact that the r-h.s, of each of the eqs. (6.42) can be directly related ta a Sonine polynomial in ¥. For Maxwell molecuies this implies that only one Sonine polynomial is needed for the computation of A(V) and 8(V), that is, for the computation of 7! and 2X, In principle, for a general intermolecular ‘potential, all polynomia of a complete (infinite) set of Sonine polynomia are needed to compute A(V) and B(V). To determine A(¥) or B(V) then, fone has to solve an infinite set of linear equations for the coefficients of the Sonine polynomia, This set of equations can be solved successively by {aking into account more and more coefficients. In practice, the expansion seems to converge rapidly, and when one restricts oneself to one coeff cient—that is, to one Sonine polynomial—in the expansion of A(V) or BUV), we already get results for 7! and 2%, which are correct to a few per cent! In fact, in so doing, the following expressions are obtained for ‘if and 2, valid for a general (short-ranged) intermolecular potential (6.48) (6.49) (6.50) Here 3D is a weighted (with g*) temperature average—with 4 Maxwell distribution function—over the transport cross section ('(g), characteristic tor IM. Approach to Equilibrium 1 143 viscosity and heat conductivity. The cross section Q(g) is defined by os) = [ab b fave — cost (6.52) and is a weighted average [with (1 — cost z)] of the differential cross section b db dp. The angle z = x(b, g) isthe scattering angle in the collision, characterized by the impact parameter 6 and the relative velocity 4. This angle depends, apart from on 6 and g, also on the intermolecular potential field g(e), and itis in chis manner thatthe transport coefficients, i and 28 depend on the intermolecular potential field. The term cy = ‘Uelm, isthe specific heat of & monatomic ges per unit mass If evaluated numerically for a realistic intermolecular potential $(+) as, for example, for the 12-6 Lennard-Jones potential, the values of iit and AX given by eq, (6.49) and eq, (6.50), respectively, agree very well ‘vith experiment over a wide range of temperatures (compare Section 1V, Subsection 3, and Chapter 9). ‘Asa function of bg. and g(r), z follows from the meshanies of wo particles in infinite space, and ie given by the expression is the smallest raot of the equation obiuined when the expression under the root in the denominator is set equal lo ze. If the molecules are point centers tf fore and if the (repulsive intermotecular potential #(r) =r then x reduces 10 eal positive rot of the equation: where 245 ae tty ra aa seein soa aeenery = Tint144 6. THE KINETIC THEORY OF DILUTE GASES and aky~ ime ‘Te emperare dependence of rect herr dey he eng of he molecularrepulsion n! (12), ete s TV. Discussion In this section we sball discuss a number of poicts connected with the results obtained in the previous section, (1) We first want (0 verify that the expansion parameter meter 4—used to indicate the order of magnitude of the terms in the Chapman-Enskog. ‘expansion--is indeed the parameter defined in Section II (3). Using eq. (6.19) {see remark about 1 following eq, (6.19)] and (6.36), we have Ly Lh Now in case there is, say, a temperature gr n= aint = -AQN)- es where 4,(V) satisfies (compare eq. (6.42a)] the integral equation: 1Avyva = -( = ay Because the operator /~ 70, where Q is the total scattering cross section, fone must have that A~ I/nQ~ /. Therefore, indeed Jar Tox TL , if the temperature gradient is in the x-direction. A similar discussion can be given in case there is a velocity gradient (2) In the preceding section, we have derived for limes £3 ftp that the approach of dilute gas to equilibrium is governed by the hydrodynamical ‘equations. We remark that the hydrodynamical equations were obtained from @ molecular point of view and not on the basis of continuam IV. Discussion M45 mechanics, as was the case in Chapter 2. For that reason, we were able to Sbtain the explicit n and T' dependence of the thermodynamic (pX, eX), fs well as of the transport properties (9, 2%) of the gas in terms of the intermolecular forces. (G) As far as experimental verification of the results obtained in the previous section is concerned: (a) the density independence of » and A Myer a wide range of densities—corresponding roughly to pressures of 6.01 atm to about 5 alm ({his depends also on the temperature) —has been very well confirmed for a larger number of gases; (b) the Eucken relation Wes = 5/2 agrees very well with experiment for monatomic gases (that jg, the noble gases); and (¢) the temperature dependence of the transport coeflcients * and A® for a realistic intermolecular potential presents @ sensitive test of the theory of the (linearized) Boltzmann equation. “All these points are further discussed in Chapter 9. Here we would like to say a few words about point (c). A comparison between theory and tuperiment can be made by using the law of corresponding states. This law holds for all substances that ean be described by an intermolecular potential which can be characterized by wo parameters (for example, by the 12-6 Lennard-Jones potential characterized by « and a).* All physical quantities can then be reduced, hat is, be made dimensionless, with the help of the three molecular parameters, a a, and € The law of corresponding states says that if the molecules of these substances move according to the laws of classical mechanics, any func~ tional relationship between physical quantities that holds for any one of these substances can be transformed into a relationship valid for all these substances, provided that the physical quantities occurring in that relationship are replaced by the corresponding reduced quantities, ‘Allthe noble gases can be very well described by a 12-6 Lennard-Jones potential. It follows then, for dilute gases, that the properly reduced Viscosity coefficient should be the same function of the reduced tempera~ ture T* for all noble guses.t Tndced, using the expression (6.49) with eq, (6.51) for 1, one obtains the relation: ‘ 5 wary ed 6.5) aT) =a ‘ wst be the same forall substances, as for example, the form of the potential the « and # values will be diferent, of course, fo 124 Lennard-Jones potential Substances which are nol isotopes it Fr the light noble guses, however, se point (4) in this section.46 6. THE KINETIC THEORY oF DILUTE GASES where (63% . 6.55 and cee aeaCr) me aaierio £(8) (6.56) Using for each noble gas valucs of e and o determined (rom the tempera ture dependence of the second virial coefficient the experimental values of 1, when properly reduced according to eqs. (6.54) and (6.55), indeed yield a universal curve for n* as a function 7 (see Fig, 2) (4). At high temperatures, where the repulsive part ~I/r? of the intermolecular potential dominates the scattering of the noble gas atoms, 7* should behave ~T**9, Or, the slope of the plotied curve for large values of T* should be ding (TM _ 1 dint 6 el hich is to a good approximation, so. The good agreement over a wide range of temperatures for a variely of substances (including even some with only approximately spherically symmetric intermolecular potentials like Os, Na, Ha) is not only a check on the validity of the Boltzmann equation and the Chapman-Enskog solution, but also, of course, on the 12-6 Lennard-Jones potential as an effective intermolecular potential. (4 Curves for some light gases (D,, Hy, He, *He) are also drawn in Fig, 2. These show deviations from the law of corresponding states, ‘especially at low temperatures (15). This is due to the fact that, for light molecules, quantum mechanical deviations from the classical motion become important at low temperatures. The reason for this is that, at low temperatures, the de Broglie wavelength associated with a light molecule, Jats becomes appreciable compared to the size of the molecule (that to ) so that quantum-mechanical diffraction effecis become important and deviations from the classical behavior occur. The larger Ayg/@ is the larger the deviations will be. Now: Tear (6.58) 1 For our purposes this procedure ie sufsient. Sse, however, Chaptet 9, Section I IV. Discussion 147 where fy i8 an average momentum of a molecule, Therefore, the quan- tum mechanical deviations from the classical behavior due to diffraction cffects are determined by dat = Ag AT (6.59) o ov mkT iT. where the quantum mechanical parameter A* is given by h Ate (6.0) ovme ‘This, at a given reduced temperature 7*, the quantum deviations from the classical (corresponding states) behavior will be larger as the value of TABLE 2 the | ote | ty | De ro |__|] ao | 268 | 173 | oes Atis larger, that is, the lighter and smaller the molecule is and the weaker the intermolecular potential, Table 2 lists the value of A¥ for various light gases, and itis seen in Fig, 2 that the deviations from the classical behavior do indeed increase with the value of A* at a given T*. In addition to diffraction effects, quantum-mechanical statistics effects will also occur if 2yyy is of the order of the average distance between the molecules, AU the densities and temperatures in which we are interested, these statistics effects are of importance only during the collision of molecules.* Both hinds of quantum mechanical effects require a modification of the Stoszzahl -Ansatz. In our ease, the only change one has to apply is that the classical differential cross section is replaced by the quantum-mechanical differential eross section: bab dy» Mg. ghsin xd (661) Here Mle, ) ean be found by computing the quantum-mechanical phase shifty 9,(k) characteristic oF the relative motion of te molecules with So sutclly how Pssst, Aya wil ssome of the onder of the average intermolecular distance, ws that statics elects oscur due to a quantum degeneracy fshy gas ava whole eee roerence Ph148 6. THE KINETIC THEORY OF DILUTE GAStS m4 -16 =o =05 08 or is nt Fig, 2, Reduced viscosity coeficients from eqs. (6.49), 6.53), and (6,56) plotted against rece temperate’ i yh/\/T® ves In TCs the lea eae Caper Points: A *He; @ He; O Hy; C1 Dy: x Ney + A. e relative energy /°k!/2m = 4mg? and angular momentum (I(/ + 1)h]!* = Amgb. As the m(k) depend implicitly on A*, the transport coefficients n* and A* will also depend on A*. “The statics eects dating the collision make neces the use of wavefunctions, which ae peopel synod with respon tothe eaoiiats of the two cling parses, Thus fr Head H, which ones ofan even numbe of elerentary partes {eae roto, and eons) an law Bow ih se} way wae. tanto synthe coordinates ofthe two calling particles shouldbe wed. This implies that only n(& with even canbe wed Inthe compaton of Tg. 3)- For He wbich const of an odd number of elementary partes, haa spin of fe and follows Fermi-Dine sats! wavefunaons bolt srmtie and anttyometicn the coordinates ofthe cling parties an seat. This pes that 4), wth both ven and odd, are used inthe computation of 2). Tait leads 1o large eens in fore and ‘He, al the sana Thee dferences cto quantom sates come ination o those due to thediereorin N= which area comyuenc fhe ference inmass (cand o ate ese for He and He snc the ae topes) 1 This is based on a theorem of Ehresfest and Oppenheimer (16). This theorem Sates that composite particles like atoms or molecules consisting of a number of slementary particles con be treated as peint pais fallowing w veflnie quantum IV, Discussion 149 (5) As we pointed out before, the Chapman-Enskog solution of the Boltzmann equation isa special solution to which, presumably, the generat solution of the Boltzmann equation approaches for a large class of initial conditions. One can ask: What is the nature of this approach? It seems certain that the approach is exponential with a characteristic time of the order of finups that is, ~ exp (times) This is made plausible by Grad’s thirteen moment expansion of / (17). Grad considered special solutions of the Boltzmann equation where it is assumed that / depends ‘on thirteen—instead of five—-moments of f; that is, f depends, in addition to its five moments with 1, ¥, and v?, also on those with uF, 08, 2, 0.0 124, and vty, The special Solution of the Boltzmann equation, fa, which Grad obtained on the basis of this assumption seems to be more accurate than the Chapman-Enskog solution fey» =/(tv | a7), discussed in Section IIT. However, Grad was able to show that the additional eight ‘moments, when computed with the (more acturate) q decay exponentially ina time of the order of fap to the values they would have had if they hhad been computed directly on the basis of the Chapman-EnskoB fo. using five moments only! This illustrates again the possible simplification ‘of the description of the gas after a time 1 > fry because the “extra information™ which these eight additional moments contain, in addition to, u, and T, has disappeared after #9 foey (6) The Euler and the Navier-Stokes equations are the hydrodynamical equations which correspond to taking / = f, ot f = fo + fu respectively, im the general conservation equations. They can be trusted to describe adequately the approach of the gas to equilibrium, only as long as ye & 1, that is, as long as the (relative) variation of the macroscopic variables over 1 is small. If one has, however, a rarefied gas where / is not /, the dispersion and the absorption of sound as a function of 4,, obtained from the (linearized) Navier-Stokes equations, agree well with experiment. With decreasing 4,, however, deviations are ‘observed which can be accounted for by using the (linearized) Burnett ‘equations instead of the (linearized) Navier-Stokes equations (/8). Thus in this case the Burnett equations are able to describe the behavior of the 8 under conditions in which «is not very much less than one. V. The Connection with Irreversible Thermodynamics One of the basic assumptions of irreversible thermodynamics is that the same thermodynamic laws and relations hold for the local thermodynamic ‘quantities in a system which is (in first approximation) in local equilib- rium, as hold for the ordinary thermodynamic quantities in a system which is in total equilibrium. tn particular, itis assumed that the so-called Gibbs relation (/9) Tds = de + pd(\[p), (6.62a) valid in (total) equilibrium, is still valid in the form: piille (6.620) at in focal equilibrium, Most of the results of irreversible thermodynamics are obtained from this equation by using it to define appropriate fuxes J, (i= 1, 2,..+.1) and associated forces X,(7=1,2,....1), which ate such’that li relationships exist between them, vie., daSux, 6 16.63) ¥. The Connection with Irreversible Thermodynamics st for which the matrix of the coefficients L, is symmetric, that is, (ja bdo (6.64) tn (6.622), 5, e, and I/p are the specific entropy, energy, and volume, respectively, while in (6.626). 5, ¢, and 1/p ate the specific focal entropy, Energy, and volume, respectively, which depend on r as well as on ‘The term didr is the substantial time derivative defined before, The total energy has been defined by eq. (2:30), only here we have set the Kin cnery from the centrokeravty motion and the external potential nergy equal 1o zero, therefore e = eK /m, ger asthe ft to prove tat Muxes and fores cou be defined for which the Onsager reciprocal relations hold {see eq. (6.64) and Chapter 5]. ‘The proof was based on the hypothesis that the decay of microscopic fluctuations of a system in equilibrium follows the macroscopic linear laws {eg. (6.63)] and, in addition, only used microscopic reversibility, that is, the time-reversal invariance of the microscopic equations of motion (ef. eqs. (6.5a,b)]. For a single gas of the kind considered here, no ‘Onsager relations eq, (6.64)] exist, Therefore, in this section we sl only prove that eqs, (6,62) and (6.63) are valid for a dilute gas, on the basis of the Chapman-Enskog solution of the Boltzmann equations, A proof of this was given for the ist time by Prigogine (20), For a mixture Of two (dilute) gases there is a relation of the form eq, (6.64) which ex- presses the equality of the thermal dilfsion eoefcieat and the Dufour Coefficient, A proof of this relation, as well as the eqs. (6.626) and (6.63) forthe eine of binary motu, s presented in Aprende & at Chap ter 7. (A proof of the Onsager reciprocal relation between the dilfusion Reticions in multicomponent mature hus been given by de Groot and Marat (21)) A, Paoor oF 1H: Gris RELATION the Chapman timber (a) To prose the Gib estaton om the ass of the Ca solute athe Bevmann equation, and to define Boxes and res J enSunclyc is onvenent to tanaorm he eg. (662) sn Jr, but equivalent, equation (eq. (6,698,b)}. This is done with the192 6. THE KINETIC THEORY OF DILUTE GASES (p = nm) momentum, and energy*: (6.65a) (6.65b) (6.65¢) 1 poe (6.66) where ®,, is defined by Pua = Pu Py 6.67 Introducing the entropy flux vector /,(3) by the relation ds dis) = (6.68) €q, (6.66) can be written in the form of a balance equation: ds __ asa Pa On where the entropy production o(3) is given by + o(s) (6.694) : 7 (6.696) In the following we shall prove eqs. (6.68) and (6.69a,b). rather than eas Prove eqs. (6.68) and (6.68,b), rather th (0) In order to prove a relation concerning the rate of change of the local entropy of a dilute gas we must first have a definition of the entropy of such a gas in terms of the distribution function fjust as, say, the (local) energy density « was defined by eq. (6.16¢). ‘The Euler~ and Navier-Stokes equations of hydcod ; tions of hydrodynamics obtained before follow ‘rom the general conservation laws (Eqs. (6:6)] under the special conditions that the a ror ane ht x eto eyes (622 an 639). 0 (45) V. The Connection with Ireversible Thermodynamics 133 One defines the entropy density SE by Sf = ku (670) where HE = Heal) = ~ferseeou = In f(evt)] (6.7) ‘This definition of SK is such that in local equilibrium, when f= fe, the Joual Maxwell distribution function, SE is the same function of » and 7 vie in equilibrium, and reduces to the known expression for an idcal fs SUE = -kHK(e |) = —3n nA = Gn = inn where A= (2nmk TY, As before, the superscript K reminds us of the fact that for the dilute gas only kinetic contributions to S and # are considered. We now calculate dHX/dr from eq. (6.71). Using the Boltzmann equation (6.1), one obtains due ~ favor ze + fariasn (62) at “The two terms in the rhs. are treated 28 follows: (1) The first term on the rhs. of eq. (6.72) can be rewritten with eq, (6.150) in the form = Janine 2ffoer = Jaryinso.. + [oer (6.73) Using ¢q. (6.19), expanding the logarithm, and keeping only terms i this capansion up to O(j), one has, with eq, (6.27), a a (ak) ~ 20 favsiagy) = 2 (2 (624 all Stas] a } Using eqs. (6.71) and (6.16a) one also has that = fern PP me - (HE + yD, (6,75) while with eqs, (6.156) and (6.163. forge aad. (676) Pn154 6. THE KINETIC THEORY OF DILUTE Gases Adding eqs. (6.74), (6.75), and (6.76) yields for eq. (6.73) xy 9 (58 ~ faving, LZ = 2 (2) _ x Jormmeg males) ~ #4 6m (2) The second term on the r.h.s. of eq. (6.72) can be reduced with eq. (6.19); expanding the logarithm, keeping only terms of O(a) in this, expansion, one obtains ferman =faling + Hoc + stor (68) (One easly verifies, using, for example, eq. (6.15a), that the terms on the chs. of eq. (6.78) containing In fy vanish, Then, by using eqs. (6.36) and (6.41), the integral equations (6.42), and eqs. (6.26) and (6.27), one finds for the r.h.. of eq, (6.78), the following result Saint | P&, dv in fJ( Sat ferinsep = RIOT 4 2s 0, Using eqs. (6.77) and (6.79) and the fact that for any thermodynamic quantity its specific value 4, (that is, its value per unit mass) and its density A, (that is, its value per unit volume) are related by (6.79) dA, _ day the the a a, ote AD Equation (6.72) can be writen in the form: aus 7 (680) where (6.806) and (6.02) (6.x0ay In the transition from eq, (6.80c) to e4, (6.804), one has used the Fact that, as aceess tensor, the ner product of the tensor D,,and the traceless tensor Py. isequal to the inner product ofthe traceless tensor (D,,"~ $D,,,) and the traceleos tensor By, 22) References 135 Using eq, (6.70), one sees that eq, (6.80) is identical with (6.69). There fore, with (6.80), (6.68), and ultimately (6.62), have been proven for a dilute gas. ‘According to the principles of irreversible thermodynamics, the expres sion (6.696) is of the form: t on byan, 7 ‘This defines the fuxes and forces X; for whieh the linear relations (6.63) should hold. By comparing with the eqs. (6.690), (6.70), and (6.804), and by using eqs. (645b) and (6.46) for the fluxes J and 2 , respectively, one sees that the linear relations (6.63) are indeed satisfied and, in fact, are nothing else but Newton's law of viscous flow and Fourier's law ofheat conduction. REFERENCES 1. 8, Chapman, and T. G. Couling, The Mathematical Theory of Non- Uniform Gases, Cambridge Univesity Press, England, 1953 2, L. Waldmann, Eneyclopedia of Physles (S. Fligge, ed.) Vol. XI, Springer-Verlag, Berlin, 1958, p. 298 2. 1.0, Hirschfelder, C.F, Curtis, and RM Bied, The Molecular Theory of Gases land Liquids, Wiley, 1968, Chapters 7 and 8 4. GE, Uhlenbeck and G. W. Ford, “Lect in Applied Mathematics, Nol, \. Amesican Ril, 1963, Chapters 4 and 6 Reference 1. Chapler 3: ref. 2, p. 35; ret 4, Chapter 4 Reference 1 Chapier 4; ref. 2, p. 338% ref. 3. p. 464% Reference 4, p. 10% G1, Unlenbock, *The Boltzmann Equation.” in Probability and Related Topics in Pheu! Setencess M. Kas; Lectures on Applied Mathomaties, Wol. Interscience New York, 1959, p, 182; E.G. D. Cohen, “The Boltzmann Equation and is Generalization to Higher Densities," in. Fundunemal Problems in Statistical Meclunies (EG, D. Cohen, ed), Novth-Htelland, Amsterdam, 1962, p. 110. 0. Reference 1, Chapter 7; rf 2, p- 384: ref 3, p. 466: tf. 4, Chapter 6 10, 8.T-Choh and G. t Ublenbock, The Kivene Theory of Dense Geter; Ph.D. They, University of Michigan, 1998, pp. 13 and 93. Reference 1 Chaplet 9, p. 102; ret, 2, p92 ref 3. p 485 and p. 5275 ef. 4. p 107 114 12 Weterence 1B, Reference 1 14 Reference 1, p. 3921 te. 2. p97. 1S, Meference 4, Chapter 1720 2p. $82: el 3, p. 397 1, PU venfes and J. R. Oppenbermer, Phys, Rer., 37,333 (931) 4, Chapter 4. 1 Chapin 105 ref 2, p98 ef 3, p46, 25 tf.156 6, THE KINETIC THSDRY OF DILUTE GASES 1”. Ba iw, 2. 2 2. Reference'3, p. 492; H. Grad, Encyclopedia of Physies (S. Fligge, ed.), Vol, XI SpringerVerag, Bevin, 1958, 2260. ee : utes Chapter 3:3, D: Foy On the Theory of Sound Propagation Mone stont Gsetond Binary Ga Maas: PhD. Tas, The Receller Unie 1967. * E S°R. De Groot and P. Mazur, N-gram Thernodanics, NothMotn Amsterdam, 1962. os oe : sad I Piggies Ply, 18, 271 149) felette 1p. 18 Reference Ip - Chapter 9, CHAPTER 7 THE KINETIC THEORY OF MODERATELY DENSE GASES," E.G. D. COHEN I, Introduction and Approach to Equilibelum 1 (6) In order to explain the experimental facts observed in moderately dense gases on the basis of the dynamics of the motecules in y-space, one needs a generalization of the Boltzmann equation to higher densities, To this day no one has been able to generalize Boltzmann’s procedure, that is, to look at the motion of the molecules in j-space and compute the change of f with time by generalizing the Stoszzaht-Ansatz to include the effect of triple and higher order collisions. Therefore, one has had to derive the generalized Boltzmann equation ina systematic way, that is, one has had to start from the basic equation ‘of statistical mechanics, the Liouville equation, expand in some way in the density, obtain in first approximation the Boltzmann equation, and then, in the next approximation, a correction term incorporating the effect of triple collisions, and so on. Bogolubov was the first one to carry out this program. His work has rcatly influenced the development of kinetic theory in the last ten years. Very recently it has been found that his ideas cannot be entirely correct and only lead (o a correct derivation of the Boltzmann equation and to the first correction (due to triple collisions) to this equation. Since we restrict ourselves here to this case and since, moreover, we believe that the + Vor general hackyround, se ceferenees (1. 137