Solutions Manual for Introduction to Modern Statistical Mechanics David Wu and David Chandler New York. Oxford OXFORD UNIVERSITY PRESS 1988Oxford University Press ‘Oxford New York Toronto Delhi Bombay Calcutta Madras Karachi Petaling Jaya Singapore Hong Kong Tokyo Nairobi Dares Salaam Cape Town Melbourne Auckland and associated companies in Berlin tbadan Copyright © 1988 by Oxford University Press, Inc Published by Oxford University Press, Inc., 198 Madison Avenue, New York, New York 10016-4314 Oxford is a registered trademark of Oxford University Press Alll rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press. Library of Congress Cataloging-in-Publication Data Wu, David Solutions manual for “Introduction to modern statistical mechanics". L. Statistical mechanics —Problems, exercises. etc 2. Statistical thermodynamies—Problems, exercises, etc. 3. Chemistry, Physical and theoretical—Problems, exercises, ete. . Chandler, David, 1944 11, Chandler, David, 1944~ Introduction to modern statistical mechanics. IIL. Title QC174.844,.W82 1988 530.1°3 88-27824 ISBN 0-19-505889-5 9876 Printed in the United States of America on acid-free paperPreface The following pages describe the solutions to many of the Exercises in pavid Chandler's text, Introduction to Modern Statistical Mechanics. That text is brief but demanding with the principles illustrated only through the numerous Exercises. Especially in light of this strategy, a solution manual can serve an important pedagogical role. While we do not present herein the solutions to all the Exercises, those we have skipped require no techniques that are not already illustrated by the solutions we have included. In some cases, we have pointed out additional references. While not absolutely necessary (the text is self contained), these additions may be especially useful for the topics not covered explicitly in the main body of the text. For instance, in Chapter 6 where at least one exercise leads one naturally to perform a finite size scaling calculation, we alert the reader to some introductory literature on this important technique. In the format we have adopted, the manual should look like neatly prepared problem sets. The prose is kept to a minimum, and the figures are generally hand drawn sketches, though a few graphs were plotted with the aid of a microcomputer. Concerning the preparation of the manuscript, we owe great thanks to Mary Hammond for her expert secretarial help.Solution Manual for Introduction to Modern Statistical Mechanics Contents Page Chapter 1 1 Chapter 2 5 Chapter 3 15 Chapter 4 29 Chapter 5 43 Chapter 6 59 Chapter 7 68 Chapter 8 81Chapter 1. Thermodynamics, Fundamentals Entropy is postulated by the 2nd Law to be a monotonically increasing The first 1.2 function of E (i.e., T > 0) and an extensive function. equation of state satisfies both criteria, while the second is not so extensive since it grows exponentially with system size. . 1/2 S = LgY(8E/L) is the correct choice. Indeed, you can convince yourself that S(AE,AL, An) = AS(E,L,n) Further, +. (3) = 2 wie? tT” ene 2 implying T > 0. For the tension, we have L as — Le -~ 2 ft (Sten 7 7 bot J ue 3 or f= TY(2/2o)01 - (g/t) 1, = Ln The principles we apply are conservation of total extensive energy and 14 From the solution to equality of temperature at thermal equilibrium. Exercise 1.2, B= 177? negosu Therefore, the total initial energy is cer P9? nl) (x2)? ny Pager = BO? + 2(2) which is the same as the total final energy Pals nl) ror. Hence1 12 1 (12 001) (2))2 (2), 1/2 T= tay ray C(t Vontiie (ree yni’7 } Assuming the system is surrounded by adiabatic walls, the change in energy AE is equal to the work [pdv due to changing the pressures py and Pp - The final equilibrium state so obtained has new mole numbers nj and nz. Instead we can imagine arriving at this final equilibrium state in two steps. First we reversibly pump the particles to their final values. Then we do work by moving the pistons to their final pressures with the walls impermeable to any particles. we know the total change in energy of the system must again be AE since E is a function of state. Hence, the work done by changing n, and n> can be associated with work done by changing the pressures. ac aS op ST eth p,)T,n natural variables => as) ) dG = -SdT + Vdp + udn a » Ge Gp) tn) p,n avy) ' a | a 1 fal nie For a rubber band of length L, tension f, dE = TdS + fab + dn . E is extensive. Therefore by Euler's theorem, E = TS + fL + yn. Hence, dE = TdS + SdT + fdL + Lof + udn + ndy implying0 = SdT + Ldf + ndy - Since, E = 0s*L/n® we have ae) Le 2es*L ‘Sn’S,L 3 Substitute S = S(T,L,n) by inverting T = T(S,L,n): 2 By 29S a“ t+ Gaiw 7 ne? so S= Fe Thus, 2 n°T)2 20(ser)” Lb r wens 3 ~ ” Bate7ny * n The Gibbs-Duhem like equation is 0 = SdT + Ldf + ndu . We want to show this interdependence of these three differentials. We choose a representation, e.g., {S,L,n], though we could also use {T,f} : aT oT 48 + (Sr)gjn% * Gals 4 26L 20S 4eSL: = (2Ejas + (228)an + (- 1884)an . ne n= 3 Similarly for df and dy 3B t= (Hsn7? ar = (28)as + (- n 2 2, dy = (- es a (- SE, « OS ae n n nt1.14 Plugging into 0 = SdT + Ldf + ndy 20L 20S 4eSL Cs(=3) + oS) + al- =) as ni ni ni 2 + ts(288) + 10) + af- as) du n n 4esL 2 66S°L di 3 ) an + n(S2S nh + Cs(- +u- n’ =O. QED The Gibbs-Duhem equation, 0 = -SdT + Vdp - ndy, implies du = -sdT + vdp , with s = S/n and v = V/n . Hence, Sey = (Fe * VG, From the definition of heat capacity a SRA: = GD Dy - Further, since dE = TdS + fdl + udn, we have = Sdt + fal + ydn a(E-TS) and thus the Maxwell-type relation Hence 22.9 Chapter 2. Conditions for Equilibrium and Stability ar (a) Stability criteria are generally GA) 2 0 where X; and I, are conjugate. Therefore (i) is true, ®, = an <0, since -p and V are conjugate. Also (iii) is true since (. ap ibbs-I st < 0 by the Gibbs-Duhem equation. None of the remainder are guaranteed false by stability. (b) From Part (a), we know to examine (ii) and (iv). Note {dA = -SaT - pdv + ydn} Since (3T/9s), is positive by stability, (@ 5 > 0 implies (2), <0. But this implication contradicts (ii). Thus, (ii) and (iv) are inconsistent. To derive the analog of the Clausius-Clapeyron equation in the u-T plane, note that the y-T coexistence line requires that pi (7) =p cu,r) . But O = -Sd? + Vdp - ndp implies dp + (S/V)dT + (n/V)dy . As we creep infinitesimally along the y-T (a) ap‘) coexistence line, dp when infinitesimally close on either side, i.e., (Mar ~ &) ay = (8) Mar = 2) ay or aT 4o(T) o--e, dy 4s(T) whereny(a)_ pny(B) ae = (F) Ge. * S)(a) as = (86 ~ (8089 3S 2.19 cy = Ty - (38) > K=- Kee Assume constant n . To prove (K/Ky) - (ey/c,) we note that Vy av Key = (55) 5/ (Sp) - (2,4, 24D, 38) (aT 38) (28) _ Gy Gap = GHW GE, > c The second equality comes from realizing (. and the third uses the chain rule. Since P vi 13%) nS) pn) n and since ~ (8P/3V) > 0 by stability, we have C, > C, . Thus, Cy/Cy > 1, and since C,/Cy = Kp/K, » We have Ky > Ky This inequality says that instead of working adiabatically, it is easier to compress something if you can leak energy out of the system into a constant temperature heat bath. 2.20 (a) The rubber band heats up when stretched. Assume constant n. So, and the second derivative is positive by stability. To find the sign of (24) SPrn * We writeaL as Hen * Gein (Bern -- @), @), d(E-fL) = TdS - Ldf +... ‘ i: aL oT ae ee inpiies (3), = - 2, >0 >0 , given by stability Thus aL GPHein co. In other words, the rubber band stretches when cooled. 38) _ (a8) (at 38) (ab ©) (See = Gee Gre + Gly Gee as af) (ab L (Sp c Gre (ely) d(e - Ts) 2 as aL -(D, GP = -SdT + fd +... we Ue a6 2 oO implies by stability @), = ep Ce- 6, >0 Hence b > + In other words, the constant length rubber band Le has the larger change in temperature. 2.21(a) Since in general dE = TdS + f-dx , where f is the intensive field and x is the extensive parameter, the correct work term is HdM. (in other words, doubling the system doubles M, not H. This approximate extensivity neglects interactions such as mutual polarization which might occur if two magnetizable systems were placed adjacent to each other.) So dE = TdS + HOM - pdv + yan , and dA = -SdT + HGM - pdV - udN . Stability then requires both=), (ai Xr aH tyvin > + Using the representation {S,H,n,V} , the thermodynamic potential is = i = (2¢E-HM) E- HM. We then have M= M(S,Hsn,V) since m= (2078). | an aMly | (38 am ty * Gide + Gs)u" Gale * Gas IT ig 6 Dol), By a(E-HM) = TdS - MdH +... By the Maxwell relation 2 aT gas’ (Fe)y > 9 }y > 0 by stability. 3T aM (b) Given F $9» We Want to determine the sign of Gas? at) | _ (at) (38 SHES) wm SE = Mat & Gals = - Gea Gale d(B-HM-ST) = -SdT - MdH +... _ (at am ; -- Gea GPa - By the Maxwell relation Sob eG; Gr GB >0, by <0, it arn stability given 2.23 (a) dA = -SdT + fdb + yaM x = L/M k, fh, Xq and c are independent of x, but not T. 's (Sejm 7 MKx 1 Mh (x-X 5) qe h(x (b) = Since xg is independent of L and of x, it's independent of M. 2 1 Li 2 1 2,2 weg hl- + xp] te = - Zl xO) +e Mi t 52 (e) fox => 5 kx? = ueA L 1 2. 2 rT fx = > 5 h(x x0) +e Mt eps foe fy = fp , the short phase has x = xg, and the long phase has (4) The phase transition occurs at u, = 2 a ~ x2) « Xt 3 (x Xp) re. oa implies -§ fg =f, implies kx, = h(x,- xq) + Thus 2 2 2 (kx) = nkxt - Akxy = 2ck , L or hkxyt vhk Houne + 2ce(k-h) 1 Wn-K etn Since since kyh, ©>0 , we have, Maxon ic: wink fniexo 2c(k-h) ~ kx» SS ft = . fe) x,= (x- x) implies ax =x - x Ss k L 0 Ss or / x 2 + oa - + Ax = h/k Kx 2e(k-h) hx Xo . 2.24 Near the liquid-solid phase coexistence Q tant (s) 16 n3 ty we 9ats fen 3T yewhich are functions in {n,v,T} . (2) (a) p°® and 9/5) are the densities near the coexistence line. At coexistence, the temperature is the remaining free parameter, yO yD pg GD Lg, Explicitly, on V,T on av av n,T So Boy 2 a i Bray ts)? - a (ayy? 2 Substituting 8rp(5)) = p©) into the previous equation we get 3 to) rh py . 34 Upto Therefore the densities are ole le wl whe oa Jo [Pr ol2.25 Therefore 48 - $2 fo = GF. satisfying the Clausius-Clapeyron equation. Stability requires - (3), a?o 2 ap -- (ae or Pawn a (Gan > ° From the van der Waals equation, 2). Rt at = 2ap 3e°T in (a=bp)? Instability occurs when this function is negative. The boundary surrounding the region of instability, the spinodal, is where it's 0: een, 2 RT - 2ap(1-bp)“ = 0. In the T-p plane, the spinodal is AT = 2ap(1-bp)” , i.e., the spinodal is a cubic polynomial with roots for T= 0 at p = 0, 1/b. Also note that the spinodal has extrema when at 30°p? - dbp Hi, that is when p = 1/b, 1/3b. Similarly, there is a point of inflection when a oT so = ap3p- 2-0, ap rol that is when p = 2/3b . Finally, 1 8 a Me = 35) = a7 5° These features are illustrated in the figure below. Only p 1/2b. Van der Waals Spinodal 0,30 0.20 RT/a 0.10 0.00 0.0 0.2 04 0.6 0.8 1.0 po The van der Waals equation of state for a given temperature Ty has only one interval in p, if at all, for which GB, <0, i.e., for which the fluid is unstable. This is reflected by the fact that the spinodal equation, p = ap*(1-2bp) , has only two distinct roots. In other words, the isotherm is shaped as illustrated in the figure on the next page. The Maxwell construction requires Plog) = p(p,) which can only be satisfied within the height h. Any pair of points so picked will lead to a coexistence curve enveloping the spinodal region.Van der Waals Isotherms and Spinodal 0.075 0.045, a r nn 2 a 0.015 0.015 0.0 01 02 0.3 0.4 06 pb 2.26 phase a: gp = a+ dBy 2 phase 6: sp = c¢ + d(By) At phase equilibrium, (a) (Bp) (a) (B) (a) (B) 8p = BP » By = Bu » and Bo = B =p. Thus a+ bby =c + dau) , implying Bu = The density can be obtained from the Gibbs-Duhem equation, dy = -sdT + vap , t.e., (1/V) = 9 = (2), < (3s) . So, pe) zh. 8)» 2aauand we identify the positive root as the physical root to the quadratic equation. Hence o(. (0) . Jes nacre) and + & + 2 ij ®Peransition ~ 2 * aq [b+ ¥ b’+ Ha(a-c) JChapter 3. Statistical Mechanics 3.8 The probability in a canonical ensemble is BE P(E) « Q(E)e = exp(£ng(E)-pE] . We assume that P(E) is so sharply peaked that occurs at the top (or very close to the top) of the peak. The following steepest descent calculation will validate our assumption: Expand &nP(E) = &ng(E) - BE about its maximum value &nP() which occurs at , i.e., 2 3knP(E) 1 2 9°nP(E) anP(E) = gnP() + 6E SE | + 5 (GE) a leapt eee a oo 0, because £nP(E) res has a maximum at sane) a cann-be)) 3Ee 3B a_ dena . » GE GE Vee * we aielg Ble 2 2 ES kt Hence 2 P(E) exp(- {E°SB2)") 2kyT cy For §E = 10% , N= 10!, using = 3/2 Nk.T, -109 P(E)/P() = & = 0.000 0001 iy a billion zeros That's improbable! 3.9 m/N = implies ow ol jee 2 ee, 1 + ef =m7N eee and Neo Be3.10 3.16 NI S/ky = ina(E,N) = on ater! where m = m(E) = E/e N &n N - (N-m) £n (N-m) - m gn m : 5 i 2 5 4 i e 5 o Be = S(BN)/kg Clearly, im S/k,-=0 . poe = 2B Also, N N S/kg(E,N) = N gn eae! + m(E) gn (atey - 1) E Ne = - Nan (1-55) + Efe an (Fe - 1) a(S/kg) OE = Lian ME - 1) = So 1/k aT = : The last result is in agreement with the earlier formula, Be N/m- t= e°" . Notice, however, that 1/T can be negative when Ne 7 EL sao 1 Eo 1S 1, ie, F>5 implies 7 = Ne(1 + eB)! Then S/ky = -B(A ~ ) = N an(1 + e Bey + —NBe B 1+ be” which is the same as in Exercise 3.9 from the microcanonical calculation, N= 0.01 moles is = 10°. An open, thermally equilibrated system is controlled by T, V and y. Fluctuations in the energy and the density p = N/V are given by a <(sx)?> - - 3E ¥ where & is conjugate to xror E and N varying, the grand canonical ensemble yields (22) a ae * 22) 2. (GE) > =~ V8 7 oT PV, Bu ye could calculate this last derivative precisely (see below), but for the purposes of an estimate, we only need to know that we'd only get some constant times Cy, the exact constant being irrelevant since it's still O(N). So, as an estimate 2 2 <(6E)> = KT Cy which is the canonical ensemble result. The RMS deviation =10'! ” [2 fecsey?> . kgT” 3k,N/2 1 SKRTN/2 Vi similarly, for an ideal gas, <(8N)*> = Ye(6p)*> ¥e(any?> eteio"

Wi Just for reference, let's evaluate the actual derivatives involved in <(58)?> aE: 3B + (2 GPy,6n - Gyn ( hy, Fr eS Vi Bu Boe (- 2Bu 26u Cy + (372 kgr) at uw / Gav, r 38) Gr vn - S = 1/T dE + p/T av - u/T aN O = EdB + Vd(Bp) - Nd(Bu) . wad zs = 7 at => d(Bu) = E/N dB + V/N d(Bp) and BBu E Vv (Bp), N Gey. <8 ¢ Roly e Bp = y = BP(N,V,T) =l, W So, 3 (Fey gu = Sy t 3/2 kyr) (Beyon = (3/2 + 97/0 MtIKy = 15/4 Mig «3.17 agx¢bd. Show [¥g,f| = ] <=> p(x) = 8(X-Xq) . (a) bd b b - = fo gioeGdp(xddx - [ g(xdp(xdax +f e(y)p(yay a a a bob bib =f J ecxreyptxds(y-xddxdy - [J gOct(y)pOx ply )dxay , aa aa Hence, if = , bb Jf sor pcx lety-x) - ply) axay = 0. aa This holds for all g and f if and only if p(x)[8(y-x) = ply] = 0. The trivial solution, p(x) - 0, is discounted because it is not normalized, Thus, we must have Ply) = 8(y-xXy) » » where Xp is some point in the interval [a,b] . (ob) = = f J axay roxde(yp(x,y) - f dx 2B, (x) f ay BCy BLY) « Hence, if = , J f axay eGne(y)Cpcxy) - B(x) Bp(y)] = 0 - This holds for all f and g if and only if P(xy) = B(x) By(y) - Note: The above proofs rely upon the fact that [f(x)n(x)dx = 0 for all f is equivalent to n(x) = 0 on the interval of integration. This is true because if n(x) # 0 we could pick f(x) = n(x) for which ff(x)n(x)dx > 0. For two dimensions, we can iterate theabove theorem on fox eGotfay eG) n(x,y)I calling the term in the square brackets n(x) 3.18 For N spins in a field H, E My say ee i= (a) 8,H, and N are constant in our ensemble: ny on. lt since the factor for each spin is identical, so a= (cbt, BUN © (5 cosnipun)) . aang © = 3(-B) eoBu Bul = mut £8 nyat tan (putt) eeu, but =~ NpH tanh (BuH) = kganQ + k,8 = wk, Cen(eP! + oP) - aux tanh (sult) ] (c) In the limit T>0O or Bre, tanh (Bu) + 1. So ED, = - Nu T+03.19 (a) (b) (e) i.e., the ground state has all spins + , or aligned with the fieig Lim NkgCen(eStH + @ SHH Bre ) ~ ByH tanh(pyH)] = Nk,CeyH - euH] = 0, hence, oe cL ang == hed = amy «=P = Ny tanh(pyH) <(aM)?> = = a> - ap? = 3 fa 4 sn!” 3(BH) ee = Nu? [¢N-1 )tann?( pul) +i} Hence, <(6m)?> = Nu@{1 = tann?(puH)} = Nu@sech@(auH) . Also, eee ww = pNy2{1 - tann2(eyH)} , 80 SS en (98), 5 = (eM), As p> =, tanh(puH) +1. Therefore Hy, = Nu, and <(6M)%), 0+ wuPCI=1) = 0. In other words, the ground state with all the spins aligned has no fluctuations. 3.20 In an ensemble with constant M, all members have the same number ofspins up. For a given H, this ensemble is the microcanonical ensemble since E= - HM is also fixed. Note we can't generalize this ensemble to one with fluctuating extensive variables because there is only one: M, which we fix (assuming N is constant), i.e., we can't have E flowing independently of M from the system to an energy bath. So we want to use &, H, and N as our natural variables. In this ensemble, N Me won = u(n,- n_) where n, = number of spins up n_ = number of spins down = ul2n-m) , andn,+n_=N and E-- Hy(an,-N) since a 1 “ bq, (nln, 1N-n, 13) ‘oH: So pt S28 ont t(nen, 17 Qu ony + ee” For N >> 0, mel 2 - o a in, [nin n, + (N-n,)&n(N-n,) ~ NJ = ECan n, = anion.) = % tn 2 + Ml” By Thus, Ae 7 * 1 +e Bull hence Me u(2n, - N) = Ny tanh(BuH)which agrees with Exercise 3.19 and (after a change in notation) with 3.9 as well. Substituting n_ = M/2y + N/2 into the above equation for BH gives Mu ¥ My 1 BH oy tn (SH 3.21 In the {A,B} basis, Hy = ( In Pauli spin matrices notation, Hy =~ Ao, and m= wo, Then H = Hy - ME = Ao, - wEo, « 01 pena A] [lat]B>] , ice. | +> eS (1) and |-> + gj So vicosoua el cea ff cet 1o ap->+ wa (ty) [\]--2 jeer: B=). (b) Hin the {+,-] basis is -if i z Ba Since H is diagonal, e 6H - |® an » and e . BA Q = Tre BH = Tr 8 a = ef et +e PS as expected. Doing the trace with | a> and | B> gives fA 0 We Qe ca | | a> + a> = Cl+> + [142 e 0 BA |B = Ll - [>Icosh Bd sinh BA sinh BA cosh BA These matrix elements are the same as those arrived at by rewriting 8h ( *,] in the {A,B} basis. Now performing the trace, oe te BH - 2 coshga = ef + 284. In the end, the route to the partition function is unimportant, since the trace is invariant to a change of basis. (ec) Using the {+, -} basis, (2) cm = $ te(me™PH) <+|m|+>e7P® + <=|m]->e7F4] Since ce|mje> = 5 () = 0 and similarly <-|m|-> = 0 , therefore cm = 0. BA BA (4i) Similarly <+| |m| [+> =e" and <-| |m[ |-> =e" .So <|m|> = [ul (411) <(4m)®> = - emp? = y? -uE -4 (4) For E#0, H=~ do, - uBo, = (OY he (i) The energies then satisfy the equation det(EI-H) = 0. so > - (ue)? - 9! =O , or B=+t and q = eb Mub)?+ a? BM ue)? 0° Then A(B) = -8 ' &nf2 cosh(a/(uE)* + a*)) and the free energy of solvation is ACB) - ACO) = -B7'{ anl2 cosn( B/C uk)*+ a°)] - anl2 cosh(ga)] } . 2 tea, + do, ai where c = BuE and d= pa . (iy eo BH n=0 To evaluate the nth term, note 2. 62 2. Ceo,+ do]° = cfr + ar + ed(o,o, + o,0,) - 2 = (c+ d°)I since {o,,0,} = 0. 80 [eo,* doy] = (e% a?y"/? x for n even «tate gre (co,+ do,] for n odd and ett. FL n=0 even a fe, en ef G voy" as, * d ,) n=0 (2, ge KE, ge n odd= cosh(¥e"+ d@) » r+ sinn(ve** at) (—S— g + 4.) . sO x ons a rs a? Since TrI = 2, and Tro, = Tro, = 0, 2, ae . Q = 2 cosn(avy"B + Av) in agreement with (i). (e) To evaluate . solving for |+> in the {a,B} basis, bra) a ii aaa implies -s ue) ly | ° fore 4 7 o° + (B-uB)* having energy - V(uB)* + 0? = - & E+ E having energy Aye)? + ae =e. tyE = ae, 2 Therefore = ——HE__{ exptg/u2e?+ 0? J - expl-a/urE™* 0°) J7a 2 - HE tanntai2e? + 07) [zee + 12 2 2 2 Similarly < 2 |m| +> = |u| 4+ ao + (EB is true for any & since <|m|> = Te(|ult e°*)/tr eo uE) uE) » hence <|m|> = | This ‘BH In other words, increases as the field increases. There is a competition between the field E which would make |A> and |B> the eigenstates, and the tunneling 4 which has |+> as eigenstates. Only as IE] + ©, does mp, 7 ¥ and Cm, | = - uy where all spins are3.22 aligned with (or against) the field. = B&ng As a final note, although , is {n general no longer equal to - ae unless m commutes with H. Systematic analysis of the fluctuation in that case is in fact equivalent to analyzing the quantum dynamics of the system, which is not the focus of this text. ap = p/(1-bp) - Bap , pe < ‘ vee, (a) For constant v= 13 <(5p)%>'/2 7p = <(an)®>!/2sem> 2, _ (aa = ye ey (8 scam” = tare, * “Gea a.v ~ YoGpp ev = ay (282)71 2 Py = = ayy Lie) “ 1 - 2Bap(1-bp) or ) At the critical point opp 1 (288), =, - aban - op *B (i-bp)* Do some algebra! 1 3b 27 and 8, = ————5 = 2ap,(1-be,) 2a(2/3) — °c” (c) Use the result of Part (a) with V = 100b, p = > and 1 3d 1 270 . 8 = 5 “gy + Where x will be close to 1: 2 !” 1 7 5V301 ~ 17x)Thus we have the table x «5p)°>!/2/p 1.1 0.38 1.001 3.7 1.00001 37 (d) We're looking for "significant" density fluctuations in a volume ~ (10004)3. Typical molecular volumes are ~ (5A)3 = b Ve V = (200) = 8 x 10° * <(op)291/2 To estimate significant ———— » we checked the CRC for the index of refraction of distinguishable (by signt) fluids: HO = 1.33 n-Hexane = 1.37 or a =~. For small differences in p, we can consider n« p . So take (4p)? significant “802. 9,03. Solving for the difference in 1 B= B/X at p =o, = Fp » We Bet: -3 1 (<5p) 291/27! w- ix = to? — (Se), ve e 21/2 -1 Then x= 14107 (se —) ~ 14109 uneeeeen) = 10" In other words, critical fluctuations will be easily observed optically if you are within 0.1% of the critical temperature. 323 APB , NO+N,=N. We can make an analogy with the ideal gas analysis of Sec. 3.6. @, 2. 2 <(N 7 )> = SND = <> N - b Cen ings? - enn ;>] + y= 1, if molecule i is in state A 0 , ctherwiseiMag? ~ engy>cngy?d Se ey } . : i = J (sng,> - (1 > ) , since each molecule is equivalent Weng >< - A> = XyKAN 1 A'B where Xqs G/N, and xp = 1 - xy.Chapter 4. Non-Interacting (Ideal) Systems 4.5 Begin with the equation at the top of page 94 BA = -tnQ= J dw g(w) tn [ePMo/? _ 9 BMw/2, 0 =O) Law wt an potit/2 — 9 BHa/2y °o As long as Bhwy/2 >> 1 , we can approximate this integral in the limit of B= as ® 2 0 aim ea = (2) f aww! shure D To 0 (B=) 7 ND“ uot -s-"o 28 zoriy f° Hence 2 cep = HM ND: gf a(-8) 'n,v 7 20DeT) * Similarly, 2B) ~ ge? 3, (219) vt Gry = > Kee” og (Gear) Bha/2 -Bhw/2 = ke 3S ee a = > kgB” 3g C r de (0) “aera Brave Me’? 1 2 “0 2 (ND D-1 2 2 2. ~~ ke (A) J du! (tw/2) C-(4a73—— ere Tpiare) y. 4% 0 e -e Change variables to x = Bhw and let k,O) = Huy : Be gto D 0 (e%-1)* In the limit T + 0 , we encounter the integral = Dt x 8 I c » + (D+) dx 0 0 e e= 0 + (D+1) r(D+I) o(De1) = (D+1)t g (Det) . —— Gamma Reimann Function Zeta Function For D = 3, we note (4) = 1'/90 , and at say Ey? Jj = amp cy> ij, since ny and ny are uncorrelated = ini cn’s = : i=454 since cny> (1-6) 4,11 The expression for the electronic partition function of an H atom, Bey Bep/l Begin? (T) = Be * Be te + Be + ... is only exact in the Sint case of an isolated atom. When the atom is put in a box or if there are other atoms in the universe, the convergence of energy levels *£-0asn-+*- is lifted. The reason for this is that as n increases, so does the energy and spatial extent of the atom, The expectation value for the repulsive energy of two atoms then has no upper bound, and will increase as the spatial overlap increases. 4,15 For an ideal mixture x Ny A q- A BO Ny Np! Further annd aenQ -ay, = (B28) + Buy = (2259) . A ON, BL V.N5 B ON, BA VINyUsing %nN! = NinN - N we get ~inQ= - Ny in q, + in Ny! - Ns Bn gy + Sn Nel, so that ua tg implies -fn q, + in N, = -inq,+ nN, , or 1 4.16 (a) Q= apa = HP ( * OR 1 x @ Ny OPN) “NI Wasnt fa SB wyco Nah OPN, N,N, A Ne = J oes a, a = J exp(-sa(n,, N,)) NoGN, N,IN,! A B P A B ane Ae such that Ny ¢ Ng = N 3a 3A (>) ug = ig implies = = SB qa) A B # = N= const implies = = pk (2) A B a Equations (1) and (2) imply Therefore, quite generally, A(N,,Ng), 15 minimized when Ng + Ng = N are partitioned to satisfy chemical equilibrium. More specifically, with the ideal expression in Part (a), we have 3 o-~ aN, {= tng H(N-NAD1T + Ny an gy + (N-NA)EN gy] = - en(N,/(N-N,)] + &n(q./q.) , applying Stirling's A A a’ 9g ‘ approximationor associating the solution to this equation with = N- , we obtain 6N,2/ = 44/03 Ny NB Sy 417 Q@ = | — as shown above . ee NW! ars (iy + Ng= ND Ny Me Ny Ge 5 Sq i i 98 WNOo=- No -_— —=_ A ANT Wt Q ‘a Not Nass A B Ng B (Ny + Ng = ND (Ny + Ny = ND #9, QgN ~ “A Cay * a5) ¥: N d Gey Since A and B play symmetrical roles, ag @, * ap) Hence, Mp>/ = 44/4 + Similarly, N,N, a? 2 A (8Q_y2 <(N,- )°> = /N4.18 For an ideal gas of structureless particles 2,2 nek’ Totengend ®t ke ax ny n 2) (a) Then assuming a macroscopic volume, z 2,2 p= Lf mkPantr + BCH KAM Gy aus (4,33? {2 and + 227 2 = et! n° -2eH 41 ff 2 - 2nen®)1/2 so pp> th fv? ants + ae * ) ax a= (28) ven 0 ° kK ° 2 feo et J ene E fx ax fa kel ° - Fenty 2 k=1 To calculate the density, 22, . =? am . . 1 . 2 ze ae i a Cr (an)3 S ‘rk i k Ke Kad 1+ ze o ox? - 4 oh fg? unk dx. aw vr 0(bd) = b ey = fu dk « ao to obttP72m =u) « e 2,2 ma f B(y - Wek’ /2m) Saas unk? gn (1 + Jak 8 om = may o 3 Vv = B 3m 3 S572") - 5 on 33 fsve 3 "2B (BpV) = = pv (a) We want to invert y = pra = 5 (-1)**! 2hy_3/? fet Using z= J cy) in the method of undetermined coefficients, y= J GOET CE oyykyhy 93/2 a1 k=0 +t at ey 2 = (1) a1 = 1 Sa 9% TT pe Reg aly hel 2 Bel ae . gel iD) 2-1 (-) 2] 2-2 62 2: +L Re 2 | Ob [i G19? & s “1 Be 0 %2* t, V37e— [2] Po Equating terms, 0 = f5,2(¢g) which implies ¢, = 0 and then - Tee, and 2 O= 0, - oF. 2 Therefore r+ (p03) savB +... 2 -Bp*/2m apy? (0) n> = 2 ~ pide? 720 for pre ca 1 + ze BP /2m vy 3343 g7Bp /am (eo) <[ppp=—4y a i np? par e 8P d (2n)3 Vp 2 ae im. anew? 8For a single particle Vp = 1, so <|p|> = ) n/N, or = 2) h/<|p|> (a) BP/p = f5,2'%) p ay Mea? © tonty?s29/8y = ga8yPy 27? & 22.) pa 1+ para? so, The finite pi? correction is due to the negative correlation between fermions (i.e., the Pauli exclusion principle implies that no two identical particles can simultaneously exist at the same point in space). However, as pi?» 0 , the effect of the exclusion principle diminishes as the particles become farther apart. 2 ay -172 °° i 120 (a) 4 .(2) = on ax xz le ean? faye Ca teY + 1! 0 Since the derivative of the distribution (z 'eY + 1]°! is sharply peaked at y = {nz , we integrate by parts to get f5j9(2) == fay y3/? Arey ¥n2 4 yy? wr =O Expanding y3/2 around the peak at y = Qnz gives = t Byte) = fae EB, Caney + ayatanny'/7e + 3/8(4ney”/?e? « 3¥a “nz (ert) - t = © (anz)3/? foa a avn sanz (e°+ 1) o t ong. t es ee un = feet * (ee + 1) The first integral is 1 and the second is of the order z™ So pad = (anz)3/? uy3¥eand z=e © where eg = (H/am)(6x°p)°/> . 4 2 F (bd) Bpa ten(z)=t f ax x* intise ) ae vio , a" fel B £3/2(2) To get the correction term in (fe,)"*, we calculate f(z) to order t?: 2 t fat 42. +0 since the integrand is odd, and -2 (e+ 1) ° tet e ae ce | -» (e411) 3 4 3/2, 1 “Ve So spi= ec, f dpi CBE) + F (8"ep) a see] 0 30a 4 3/2 2 0° “1 = we C5 (sep) + “G (Bey) tore] 2 7 Hence p= Zp cpt + 3 (Beg) + oJ The pressure is not zero at T=0 because the exclusion principle requires all but one fermion to have non-zero momentum. (a) Simply plug in the Fermi distribution given on page 96. ~B(e-e,) (b) Since e ex1, 1 then py = 17x so, “Ble-e,) a(e-ep) P= 2p, - ati +e ) = 2,01 - (l-e~B(e-ep) = 2p.e Similarly, =pfe, +e.) n=, faxtire . z Jake ee (2m) (2a) -Be =e Furs ax wexp(-eh-k?/2m,) (29) 0 Be we Pant « Therefore, . 3 ye -BE pn = p./ne PF - -B(-e-y) BR Be n22 Pad I ap = pv e = ep Ve adsor bec degeneracy classical limit Wad _ bu, oft Pad Vv . Also Bu = inp a? for the classical gas, and y has the same value as in the above equation for p.4- Hence, = ofty3 og hae, which implies 3, ad/g 7 Pgh 4.23 (a) B+ E+ + Ey then one a i EG ) “BE Bl *s -(We ") Qe *. ~BEp - )Qe Y=a a a> wae” A B c apc= oft) 4 lB). fF) Let B= Ey + Eo, Ey = zero point energy . “B(E,+ Ey) -BEy _ -BE, thn g@-fe 4 % we Fe A, A A ~BE, -BE, ang 3 ‘A A

Hye oR 7a), = iegtd , = - kge og Ee which is independent of Eo. In other words, fluctuations produce Cy, and Eg has nothing to do with fluctuations. () Sree * where energies ©, and e, are relative toe, , our choice for the 1 2 O° zero of energy. (elec) _ cy * ~ yep a1, 2, = kyr { ty (g,e7e Be58 -Be Be 2 “5 (g,e,e 1s gene ¥ } e1ec (0) @* rans Verec’Snuc ‘ot “Wvin/ Tan * 8° — pttrans ) + clelec) (nuc) , Q(rot) , (vib) cy = Cy Cy rey Cy cy .Now (elec) BB. ase cy = kB Lee - = ainq/3(-B) or A= ~gna/B. On the other hand, gg will affect S because even though the average energy doesn't care about the Be, Be, - 0. 0. degeneracy, = g e,2 /a,e €q > the free energy is affected. 4erec Serec = kpingy .4.2 N Q= drans Sint” Thus the degeneracy accounts for an additive term to the entropy, Nt If we write Sint ~ “Bey = Be (21+ 1)(2Tg+ 1) Woden Sot * Trot 4 . if . Mt’? vib 2 we are assuming: a) (2) (3) cay The volume is large enough that qi, is independent of volume. This 1s important when distinguishing C, from Cy, Since the difference will be that from Qtnangs 1-@+) the ideal gas value. Excited electronic (or nuclear) states are sufficiently high in energy to be ignored in Q. Internal nuclear motion is well approximated by a rigid rotor~ harmonic oscillator, which neglects: - anharmonicity in the vibrations - centrifugal distortion - rotation-vibrational coupling. The high temperature formula for go, assumes that ®t CT (i.e., 12.1 K << 300 K) From the partition functions, we obtain = -a8nQ/ap = Ne Hw, + NKBT + N coth(Blw,/2) + 3/2 NkyT . ° 2 Therefore,= 5/2 Nky * wo)? eseh® (Blu /2)/ kt? and C, = Cy + Nk, M9 2 a, 2 = 1/2 Wk, + era esch"(Bhw/2) - Furthermore, Bia, = Nky{ 5/2 + = coth(Bhuy/2) + ng, + &n(T/0,04) Bhw)/2 Bw, /2 -anle 9 =e 9 p+ ent 3/2 no - ann +1 The necessary numerical input is: . _ 3700 . . Tong . T= 298K Bhuy = 355 op 7 121K b= 6.63 x 10 feng «5 kT w@he 3 “ Mose Bs wa 2 on nue 7 (2 5 1) (25+ 8 rT ? 4.06 x 10 “em ~ ~ u! pro! pp ?? yielding = 51.7 cal/mol K , (Note: CRC value of 47.5 neglects the tn g,, contribution of 4.14 cal/mol K) and C, = 6.97 cal/mol K .BM opm, 3/2 neh GF) h (bo) z= e8PY 5 ppv av p= z/p aunz ap + FEE ve implies z= Hence, p= p/B , the ideal gas law. ¥<(8p)"> . ¥<(6n)?> e N (e) > 2,8 yi ye <(8W) = SE = VE = Ve = a> Ye(Sey> on = Ska = 2.01% 101°.

) + $ (sup? a°anP(N) =) al N= i.e., assuming a Gaussian distribution around with variance <(an)>® , we get “19712. et ss w el D2 10 uniich as ‘very unlikely .5.1 5.3 Chapter 5. Statistical Mechanical Theory of Phase Transitions Since the number of nearest neighbors in a D-dimensional cubic lattice is 2D, 1 Ey = 7 wN (3+ 20) = - ow, where the % corrects for double counting pairs. Note that for a two- dimensional triangular lattice, each spin has 6 nearest neighbors, so that in that case where b, = 5,5 Sis; 7 £1 18 the bond between spin i and itl, The b's completely determine the s,'s up to the value of s,. This degeneracy is related to the fact that only N-1 of the bj's are independent: N N ee tert iat However, as N + © , the contribution in energy due to the NEM bond Net N-1 byt b, is small compared to the other bonds yy So int N N 8d 2 s155,, BJ J by 7 i= - I @ ist Neo [bo = 41] N w mw (CY FD). (eP4a eo BY)N . Cacosn (pays . Note: With some patience, the reader can compute Q for finite N without neglecting edge effects. Also of interest is the transfer matrix method illustrated in the solution to Exercise 5.21.5.4 Note that the Ising model energy can be written as N ' - mH J (anj-1)- 3 LY (an,-1)(2n,-1) at? ay J N ' ' = -2mi } ony + mH Wy J onyny + 2s T (np +n) - a fot = 4 ag ff id j 1d N same as z J ny N2/2 ist As a result, the answer is Q sinc (B,N,H;J,m) = exp(@N(mH - Jz/2)) = LATTICE (B,N,use) cas with w= 2mH - 2Jz, e = 4J and z is the number of nearest neighbor sites to any given lattice site. e2(BuH + Bzdm) _ | 2(BuH + Bzdm) +4 5.6 m= tanh (ByH + 62Jm) = or 22(BuH + BzJm) 2(ByH + Bzdm) _ m +m=e or 28(uH + zdm) = in(22) which implies 8 am an (e) ,atH=0. Now, an(2) = an(tem) = an(-m)= m+ 2m/3+. 1 (am +2 m/3+. So 8 * 23am or fo? = Jz(8-B,) Thus, for T > Ty 1/2 me (T, - T) or p= 1/2. In reading this solution, try not to confuse the reciprocal temperature g with the critical exponent 8. Finally, as T+ 0, 8+, it's easiest to analyze efzdm _ ,-B2dm m= tanh(pzgm) = © _ efzJm , .-Bzdm which clearly tends to -1 as 8+@ ifm<0, and+las Bre ifm>oO. The complete mean field solution is plotted in the figure. Mean Field Magnetization 1.0 06+ m Magnetization 0.2 + 0.0 1 L L 0.00 0.25 0.50 0.75 1.00 k,1/2DJ TemperatureIn this figure, only positive m are plotted since the curve is symmetric about the temperature axis. Note that k,T/2DJ is v1"), where rie) fs the mean field theory prediction of Tg. 5.8 In the mean field theory, different spins are uncorrelated. Thus (Bhp = ~#B SO lsdap — 5D Flsss hep ii = HEY sue - 3D Jis(3i) yp (8) ae » The desired result follows from the association (sj)yp =m. Al- ternatively, you may begin from the perspective of Sec.(5.5) where the partition function in this approximation is shown to be InQ = In Qur — B(AE) up where —B(AE) ap is given by Eq.(d) of p. 138, and Qup = [2cosh(6uH + BzJm))" . The desired result for the internal energy is obtained from differ- entiating this approximation for InQ with respect to —6. Here, in the differentiation, note that m is a function of §; in fact, from Eq,(b) on p. 196 with AH = zJm/y, we find a = (uH + 2JM)(1 - m?)/[1 - BzJ(1 - m2) for T < T,. For T > T, and H = 0, mean field theory gives m = 0, and thus predicts zero internal energy in that case. This prediction is incorrect since (although the average coupling to the field is zero — i.e., there is no broken symmetry and long ranged order) there are still correlations between nearest neighbor spins.5.18 g(K') = 2g(K) - A(K) where h(X) is analytic at K = K' = K'(K) = 3/8 &n cosh(4K) KK) = Ky and W(Ke) = g(Kg) . Assume Ca |K - K,[% near K = K that is, C= { a less singular part } + alK- K|°. The second term with a > 0 dominates as K+K, . 2 Then since c = Ss e(K) , dk we can write g(K) - { a less singular part } + b|x - K,]?-% where b is a constant. Plugging this into the first equations above, we get {a less singular part} + b|x'= Ky|2* = 2{{a less singular part} + b]k = K|P 9) = nc As K*K,, the dominant contributions are |= jee, b|K'= Ky |°% = 2b|K - Ky Taylor expanding Kk'(K) about K = K, gives 1 aK) K = Kt Ogle * (K - K,) ce Hence, aK" . 2a _ jee Lg las (KD LEO = ale - I, ‘c or = 2, which implies (2-a) tn = a2, aK") 2-0 | Il c aK) aK Kk,or ax! a-2 sn 27 an ala Note that the actual value of h(k) is irrelevant. N (35 + 85,,)n/2 + 9555, ,K 5.21 (a) trig) = J Te yan B pile BRAT ot i ; - expl J (hs, + Ks,8,,,)] = @ Sy sSpy eee Syntt ja Sse N N A, 0 0 (>) tr(q’) « ofp | jJ=tr { * th oat Qh I oO () @- E | [at = q] = Qe-ayQ-b) > = 12 (asda + ad - 0% 0 implies Thus, a, + eeosn(n) + /e™sinn?(n) + 6 = eteoshih) + /sinn’(n) + ey. Further, asNee , Hence, £89 2 an(a,) = K + anfeosh(n) + Yeinn®(h) + e“*)sinh(h) + —Sinh(h)eosh(h) 3 (Ang (4) = 3 GR) = 5 Bs cosh(h) + ¥sinh“{h) + e inne (n) + ok Lim = lim sinh(h) + ( L* sosh@n)/ve inh (i) +2 a he neo cosh(n) + Ysinn@(n) + eK =0 5.22 (a) Q(K,h,N) = , (hea De Fag dinents (23) . ftv?) bef ytnen te #934 woe {odd s,| . 9h 2)leyat syleneK lay. 9¢8y] olay. pty Emr Cay. 2794), = Ceckiny 37? cx" jn! na) Thus, _ 2) (0, #95) #neK(9, +85) . 2318485) > [h+K(s,+85)] (h 72)(s\+83) +K'sis = £(K,n) le 3] For all the possible states of Sy and 83: 8, * 8371 e(Kshy eft K's g2tneK) , 72K a 81 = 7-33 ftc,nye"® eh a oh (2) sp sgectt e(Kynje tt KR" . go 2K, geht 2K @) By multiplying Eqs (1),(3) with Eq(2) twice quickly yields £ (Kh) = 2feosh@(h) cosh (2K+h)eosh(-2ken)]'/"Division of Eq(1) by Eq(3) quickly gives nal cosh(2k+h) phen + 5 tn C cosn(=aKeny J Division of Eq(1) times Eq(3) by Bq(2) twice easily leads to K' = gnfcosh(2k+h )eosh(-2k+h)/oosh*(n)] 74. (b) For Kh >O: [2K +h] >|-2K +Al, so that AGRE). > 1, implying 4! > A. Further, Ak. cosh(2K +h) cosh(—2K +h) cosh" (h) = ake + e7tkK— Ay(entK th 4 em *) < (eh Fem) + e-F) “ Hence, K' < K. So, the flows as written cause h to increase and K to decrease as illustrated in the picture,(eo) AnQ zZl= (nh, K) = Fancet + eo ®)" g(h,K=0) &n(2cosh(h)) As written in part (a) and (b), flow from g(h,K = 0.01) will be towards the h-axis, away from (h,k) = (1,1). We would like then to invert the parameter flow, i.e., find nh = h(h',k') ; K = K(n',K") . Using these inverses, our errors will diminish with each iteration, since then g(h,K) = 1/2 g(h',K') + 1/2 f(h',K') One method to get reversed flows (from low K to high K) is to first flow forward f, om (1,1) (to lower K values), and then retrace the steps taken back up to (1,1):On the reverse trip, d-e-f, we can use our low K "accurate" approximation for g, the error halving with each iteration. aE. a 2.00285 1.00 1.00 2.41 1.94 0.469 2.84 2.81 2.23 x 10% 2.86 2.86 7.09 x 1076 2.86 2.86 7.5 x 10712 Generate starting at (h,K) = (1,1) and moving down according to flows. Start at bottom with -12 g(h = 2.86, K= 7.5 x10 “) = (2.86, 0) = &n[2cosh(2.86)) , and iterate upwards using gine) = Seth", K+ dan rene). The convergence to h = 2.86, K = 0 exhibits a stable fixed point, and in fact any h = hg, K = 0 is a fixed point in flow. The exact answer is (from Exercise 5.22) gna(Ket, hel) 2 x + gn(eosh(h) + Ysinh@(h) + eK ) = 2.00285 N which matches the RG value of 2.00285 above. 2h An analytic inverse in closed form can be found by using e@" and Wk e"K as variables, which reduces the problem to the finding of the roots of a 4th degree polynomial.5.24 Parts (a) and (b): The rules for construction of legal configurations admit only three energy levels: (4) == © if any partition has other than two atoms. Now, given that there are two atoms per partition, we can have % oO oO oO oO oO oO oO aay BL Of gf tere 3 ig | 1B sane) Having two atoms on the same side of a partition means that the two atoms in the next cell must belong to the next partition. The degeneracy is 2, and the energy is E- N-O-0. (iii) The only other noninfinite energy states are those where each partition has an atom on each side: means that the next partition is either lo or ol , and the choice between these two possibilities for each partition is unrelated to other partitions, e.g., 0 0 0 slo lo o owt These states have energy E = Ne, and the degeneracy is 2, (ec) AS a result of the preceding analysis, Q=24 28 evONe (a) Ae 209, an(2 + (207BE)N Ni” =pn =BN Therefore, via’ x { vk,T an2 + €, 2eF®> 1, Nee N 0 » 20 Fe 4Hence ~Be <1 ang >4 <> a> 38) 0, 208 4 (e) A phase transformation is associated with 8 = i an 2, or € kgin2 This problem illustrates how phase transitions can arise due to competition between entropic and energetic effects. P ° Pp 5.26 (a) Q= oe cea)P/? 5 expt J x u,u,,,Jf dB expl-se’/20 + (ey Su,de) . 7 \ oye Matar | a The integral can be evaluated to be e S Pp a -8/20LE-o/B(eu ) ap? exp{e/2al$ (en } uy r1°} [ae © et ist eo 2 P Pp = expt 88 ¢ Jue Lupa Ae: va : 2P’ i det So, ge Sees lim ((eayP/2 5 gtel B Poo (b) From the text, we know ¥ Bik uu + hu] P/2 gi-1 lim y fel ate) = oe (ea) ut i = 2 cosntp/a* + y7E*) Hence,ae erBEr/20 (BE) = f dE e ote) / 2 uf ae 8B /20 2sinn(avAé + 2B?) weyva* + yee? f ae BE /20 2eosn(va? + u?E*) AS B+, coshax, sinhax > e’* , so uf dB expl-eB*/20 + ava’ + yPEI(wE//a> + y°B*) {de expt-p8°/20 + ava? + ye?) = uf ot co) (eke yee) fa oe BW(E) where the last equality defines W(E) and the weighting function ec PN(E) as gre , the weighting function becomes a delta function « 6(E~ B,,) where Bai, is the E that minimizes W(E). Exploiting this idea ls a method of "steepest descent.” Note aW_eE 2g Set - ——b4— nich implies E, = 0 or op aE ‘min’ 2 pe + SE In the latter case, Be = ou - A°/u? which has the real solutions B= tule 25) uw when gy°/n > 1. When this condition is met, we have broken symmetry. In other words, E, starts contributing to W(E) at op’/o = 1. For gu@/A 61, 1, as5.27 (a) {s,= +1) Now, 80 B= , #0. This emergence of broken symmetry will cause <(4m)*> to diverge at o = = af? . More is said about this Sorit broken symmetry in the isomorphic problem, Exercise 5.27. Doing the sum over spins with the transfer matrix method (Exercise 5.21): N Bh+aJ .-Bd N e e exe D Bhs, + BJs,s;,,] = Tr(g') , where g = or gbneay) the eigenvalues of M = ( : ares, = (232) + tq’) = tr = oN a where A, > A_ are the eigenvalues of g , and Q(h5B,N) Thus, 2 -aNne/20 (48, Ny ~ 178 &£nQ(h36,N) aL = = 178 EW ana, + an(1 + GOI") - ann? /20) For large N, therefore, a = N Nh nyp,n) = - 8 aun A(h;B,N) BM, + 35 2 ———t =n & - a= dan (cosh(gh) + Ysinn? (gh) + e “89)3 (b) At this point, notice that A(h;B,N) is even inh, t.e., if h minimizes A , so does -n . Further, 2 lim 7 n a NC Z5 > d - [ld - |n|o=Thus, we're looking for bistability in A like this: A A B< Berit B > B crit Since { is even and A +, A will have a bistability in h when hee 3 ni 69 for h'> 0 near zero; i.e., A is concave downward at h = O when aA/an' < 0 for h' near zero, Hence, 3A/ah = 0 can be used to determine the critical temperature. Alternatively, note that taking the g>= limit gives aim cosn(gh) = efl"1 | iim sinh(gh) = sign(h)-e®!9l, and hence Bee peo 2 ith, pee ney 1 anceb lhl A(h, Be, N) = NU Z5 7d - g inte 1 2 nl =n C3579 - [ali and does show bistability. However, we don't get an equation for 8 from this analysis. crit (c) Near h = 0 for h > 0, we find 3A) = we B= sinn(an) 278%) n=0 # e284 = Nn i - ls since sinh(gh) = ghThis derivative changes in sign as a function of B when 2B e crit? crit 1 5 0 which is an invertible function in @ since it's montonically increasing. Another way to view the analysis of this Exercise is to solve for # = 0 to find the stable minima in A(h) . This equation can ve written A'(hmin,B) = 0 , an implicit function for B = B(h,,,) Bont tS then the limiting value of @ = B(hy.t). ashy +O, aA = and so we look for hh += 0 and solve for Berit: hoChapter 6. Monte Carlo Methods in Statistical Mechanics To perform Exercises 6.10, 6.11 and 6.14-6.16, you will collect statistics from long Monte Carlo runs. These exercises help illustrate what must be done to obtain reliable information from such calculations. In this regard, efficient and fast computer codes are important, and as emphasized in the text, much can be done to improve the programs listed in the text. At the very least, one should use a compiled version of the codes, As a standard of excellence, however, you might want to view Jeffrey Fox's Monte Carlo of the Ising magnet." It utilizes binary arithmetic only and machine language, and it runs from 102 to 103 times faster on a PC than the Ising model program listed in the text. As well as providing rapid access to good statistics, the greater speed permits one to view rather large systems in real time. The yiews are both instructive and pleasing to the eye. Some new books on Monte Carlo methods in particular and computer simula- tions in general are worthy additions to the bibliography of this chapter: M. H. Kalos and P. A. Whitlock, Monte Carlo Methods Vol. Basics, (Wiley Interscience, New York, 1986); H. Gould and J. Tobochnik, An Introduction to Computer Simulation Methods, Parts 1 and 2, (Addison-Wesley, Reading, 1988); M. P. Allen and D. J. Tildesley, Computer Simulation of Liquids, (Oxford U. Press, Oxford, 1987). The last of these is especially pertinent to the material covered in Chapters T and 8. a * + Fox, "Fox's Ising Model Simulation for PC Compatibles," placed in the Public domain on 3/25/88.6.10 Above Tg, we find cj; decays more or less exponentially as 1 gets farther from j, and random thermal fluctuations destroy long-ranged correlations. In fact Sy + 0 for i far from j until T+ aS from above, at which point o,) +0 . Below T,, we have broken symmetry, and as T> 0, we find || is significantly greater than zero. This is an indication that we are frozen into either the up state or the down state. One result of this broken symmetry is that as our runs apparently show ? a5 and hence + 1 a8 well, we will observe o,, + 0asT+0. 1.00 = oes Spontaneous magnetization 0.60 \| 0.40 0.20 0.00 kaT/J 0.60 Correlations in the 0.48 Ising Magnet 036 0.24 <8; Sj > - org 0.00 00 08 1.6 24 3.2 4.0 kRI/d Fig. Representative results for exercise 6.10. The lattice was equilibrated for 100 passes, and then statistics were compiled for 5,000 passes. Spin-spin correlations were measured up to the fifth nearest neighbor.6.11 Fig. The effect of the interface is to introduce long-ranged correlations between spins along the interface, even while T << Tg and the correlations in the bulk are only short-ranged. However, as defined in the code given in the book, the ground state of the 20 x 20 lattice, even with the biasing field on, has all the spins aligned the same way (with the exception of the biased spins). Simply setting the initial configuration to be an interface would sample a metastable well. To observe the interface fluctuations we can increase the bulk of the lattice relative to the interface by changing the dimensions to be 40 x 20. Note that in implementing the calculation, there are four columns of spins each that can be considered bulk or interface. In our calculations the dependence of and <8;s;> on the distance from the biasing field is averaged over. Correlations shown in the graph are $9842 - S0g80g» . 0.8 Correlations with spins in the interface 06 0.4 02 0.0 1 2 3 4 5 Neighbor distance These representative results were obtained on a 40 x 20 lattice at k,T/J = 1.2 using 100 passes to equilibrate and 10,000 passes to mBasure correlations. The biasing field keeps spins 1-20 up and 21-40 down in the top row. Spins in columns 1, 20, 21 and 40 are considered to be in the interface, and columns 10, 11, 30 and 31 were taken for bulk measurements. The upper line shows correlations along the interface, and the lower line shows the correlation of a spin at thg intgrface into the bulk. Correlations between bulk spins are - 10°” 10~" and are not discernible on this graph. r6.12 A straightforward algorithm is to choose a large enough, or equivalently, to scale the range of the uniform random numbers [0,1] to (-Xnax+ XmaxJ SO that most of the Gaussian will fit in this range. Then trial x's will be accepted with a probability = p(x) , normalized so that for all x, prob(x) ¢ 1. The trouble is that the fraction of accepted trial x's tends to be small. This fraction is just that of the area under the curve to the rectangle: 1.0 0.0 -xX x max max In other words, if we pick x uniformly from the interval, we will then reject most choices. This inefficient method does not exploit any form of importance sampling. There are a couple of possibilities for improvement. ax? One is to imagine that the distribution p(x) = va/a e' is due to an energy B(x), where pE(x) = ax? , and to employ the usual Metropolis algorithm with diffusive motion. Then trial samples will tend to come from the regions of lower E(x) and higher probability. The drawback is that one random number x;,, is correlated with the previous one x, from which it made its move to x,,, = x, + 6x , with 6x small. Nonetheless, the calculated moments should be correct. Another method, equivalent to the "force bias" Monte Carlo of problem 6.13, is to change + , to reflect the more likely regions of vw p(x), thereby increasing the A ,'s . vw 62The first algorithm described above corresponds to m , = constant xx 12 and A ,«e-aX " , as we said, that scheme is not very efficient. If xx instead we make a change of variables, such as x = f(y), with y picked uniformly from [0,1] then the probability of picking x in [xgsxq + dx] is df” vax « KeKy ay, ax ' where y = f-'(x), So if we want to achieve 100% acceptance A, = 1, then our trials must be chosen witha , = Yasn e xx xx and thus x 2 : 0 _ 2 - Jae ™* which implies f Neg) = Vat fo dx = ert (xy) , oO i.e., f is the inverse of the error function, Note this scheme only generates positive x; the sign should be assigned randomly. In some problems, such a perfect solution is not possible. But as long as A, increases and begins to approach unity, we have made good progress. In the present context, any change of variables favoring sampling smaller x is better. For example, let x= e where y € [0,1] is uniformly picked, This implies6.13 So the x are picked with probability Ty Sst and so we must accept with probability * + ex) 2 2 (_+ ex) ~ax Ay « 7 ‘t/a @ to preserve w, . The proportionalities are chosen so that m, <1 and a, $1. Again only positive x are generated and the signs should be x randomly chosen. Note only one index is retained because any given choice is completely independent of previous choices. (a) The condition of detailed balance, w ,/w, = exp(-BsE ,), w' “vly vy with a6 ,=E ,~E , gives wi VN mA, ~ my Ay {S = exp(-BsE ,), or A ,=—%*™ exp(-pae , = 3 ' w vw iy vw viveviy vw Since A , is a probability, A ,< 1 vw vw * "BE, Tsliy “BOE , ip e WwW ¢1,weletA = e WW and ' vw Fe vw A , =1 3 otherwise, we can switch indices, v#v', L.e., viv “BAE , e vv and A jad vy let A ,=aA, =1 So then w viv Py voo= vel woyem ye vw w' I-p, v' = v-1 This form, which is independent of v, is possible because of the symmetry of the problem — each pair of adjacent levels isequivalent to any other pair. With detailed balance, designating >E , v' = vtl , we therefore have, ~Bhw po. 2 = > aL Note: in the limits Ww>0, p+1/2, and fw>e, po ° BCH” ME + BL )) enya 6.14 (a) Q=f agltre pple ° o-BE 2 BH {age tr me aang 1 39 = are eo oCBE. DD) Q (SEL) Q Evaluating the trace gives oo 2 Q= f age FF 0 2 cosntala® + ute + B)*)!/71 BE aL i = 2s) J dBe cosh(BE) , and so 3g op g BE /20 ap El elog.e 3G) - 2f ae sinh(BE) = 5 2B + Bo) (u/a) = a: = aru) J dbo PE 24 sinnGe) y (E+ app /E from which the desired formula follows directly. 6.16 The modification to include windows is simple. Since we are basically adding a potential that is infinite outside the window (i.e., not under the umbrella) and zero otherwise, we need only check if we've steppedoutside the window, This is done at the time we also check 2 BAE 5 pnd . Moves outside the window are rejected. we can scan through each window for the minimum in A(M). Once one is found for a given T, we can assume for T' near T, that Moi, is near Mains 80 We need only scan a few neighboring windows. However, as T+ T, , it will be hard to find the minimum since a wide range of M are equally probable. We find that the barrier drops faster than the minimum moves in as illustrated in the schematic figure. Notice that the criteria ofa vanishing bistability can be used unambiguously to define a critical temperature for finite systems, of course, such a definition leads to system size dependent T, which becomes the true critical temperature only as N= Lx L> =. So, our Main 2 several values of L. The L + @ behavior of Mgin/L° is the exact M(T), results for the "phase diagram," (S") vs T is shown in the figure for and we've plotted M(T) on the same graph for comparison. 1.0 N =) & 08 Magnetization for — 2 Different Lattice Sizes 06 kg Fig. These results were obtained with 5000 passes/window, and each window had the width L®/20 except for the L = 40 case where we used L°/40 window widths.As described above, we determined T, for each value of L by extrapolating to Maj, = 0. The results of those extrapolations are given in the table. We then extrapolated T, to the L + @ by using a procedure known as finite size scaling. In particular, we assumed b (tT, - TCL)" «@ 1/L where T, = T,(*) . We determined the exponent b from the slope of [T, - Ty(L)] vs. nL. This procedure gave b= 1,0 +0. Then a plot of T,(L) vs. 1/L gave Ty = 2.26 + 0.02 as the 1/L + 0 intercept. This estimate can be compared with Onsager's Ty = 2.269. L Tg(L) 10 2.60 20 2.43 30 2.38 4o 2.34 A discussion of the finite size scaling method is given in Sec. 1.2 of Binder and Stauffer's Chapter in Applications of the Monte Carlo Method in Statistical Physics, ed. K. Binder (Springer-Verlag, N.Y., 1984). Also see Part 2 of Gould and Tobochnik's Computer Simulation Methods referred to at the beginning of this manual chapter.Chapter 7. Classical Fluids 7.10 The kinetic energy and pair potential parts of will be the same as discussed in the text. For the three-body term, we write < } Bon B) he z N(N=1) (N-2) a ( f0)5 i tsjone “12°F 23 i>j>het N -avir’) _ NET (we2) Protas? * 6 -su(r’) N 1 Lee Sty fons a (rio, yg) MENTV(W-2) fdr“ 3e AU) N if ar BUCr) 1 (3/N) =gfor, far, f ar, o y+ For a uniform (homogeneous) system, Fn, Ros f3) . pe (ry X23) » and we can integrate out one of the degrees of freedom to get the volume: ens 3,03) (3) = EV Say f dtgg oe leygs tog) wl? Ueyas Nos) 1 2 (3) (3) “N° Ge if Wyo f Gry, 8 (hy o» T3) ue (rye 2o3) . N TW ap = ar Je er a an [ drN Q BU’) , v 1/3 We change coordinates to x, = nr, + thereby accounting for each of the three cartesian components, dx, = Vdr, . Hence,Tele N a N N sp = inv I dx owls) uw? x. where the integration is now over the unit volume, not the system volume. Differentiation is now easy, directly affecting the pair potential. Accounting for the sum over N(N-1)/2 equivalent pairs, differentiation yields N J a nov-texpt-8 5 av! 3x, jy iCaucv'/3x,,)2a0v"/3x, 53 x,.V'3 B idje1 aa ov N foaxNexpt-e ) ucv!/3x, 93 idjet N N & farNncn-tyexpt-@ J utr, ,)Idur,,)/dr, o3r,, /f arexpt-B Y wlr,,)) ov gat td 12/8 y2l 2 ea td u(r.) 1 ay Jon, Sor, oe N N. N-2 ,-BU(r™) No-BU(r") Py NNT) f de® e BOE Dy JF anbe BO'F du(r 15) ayo 12 for, for, on, 4) qe Therefore, for a uniform system 2 = 1 - (89/6) f de g(ryr du(r)/dr Q.E.D. fo apte"B™)~ 4) ene far rte BU). 4] 0 3 . » 3 - - d z a pee. 4g I, «tg [Eg BDL g =e o 3 As r+, u(r) > 0, and 3 3 Te pe OU) ys EE > z le V+ Z Ebulr)] + 0 for quickly enough decaying u(r), i.e., faster than r~3, Therefore, theTAZ boundary term in the integration by parts gives zero. As a result, 1 Bu (r 1 . 2 .-Bu(r) d =p Jaete MMP 15H) we tee oTBar) ge =~ 8/6 f op oo BUT) pw Q.E.D. For hard spheres 3 B,(T) = - 2nf- J, dr r?] = For a square well 9 a By(t) = - 2n {| - J, arr? + far r@te® 1] } At the Boyle temperature, 1 Boe F o's - eo B(o'3- = 0, therefore, 3 e/kgt, = in 2 a - a or kan (0" ("3 99) The figure shows a graph of the square well By with o' adjusted such that ° 2 o J derluyse) = ef den? which implies o’ = 9!/3, . With this choice of a’, the square well By1.28 (Ld) Bt. should be a reasonable estimate of However, unlike the square well By, Bit”) vanishes at very large temperatures since oF) +49 as peo. 3 2 2m0 3 is Square well 1/7 0 The velocity or momentum averages are isotropic and factorable from the coordinates. Therefore, only one average is necessary to answer all parts to this Exercise. 2 2 2. 2 -pp°/2m -Bp°/2n (a) <> = fapve sf ape © 2 ~8pe/2m J ap, (p,/m)* e = 2 J ap, 7 Bp,/2m since the p, degree of freedom is statistically uncorrelated from the others. Performing the integral gives 24 Da Alternatively, 221 2.2 1 8 wot R BB rg SP 2m = (2/30) 3 kt = 1/80 n7.29 (b) (ec) (d) (e) (a) (b) 2 2 2 25 ey2 1 ev, Yy> = = vet v2 ov 2 30v% 2 2 x y z x 8m x " ° «vy + adevpev.> + d° x xy’ y 2 <(yt bv) > 2) ey2 Ceo") Classically, these averages won't be affected by isothermal volume/pressure changes (i.e. through the potential) since the momentum distribution is dependent only on the temperature. When the classical assumption™breaks down at extremely high pressures, one would then need to consider the quantum dispersion of particles since the dispersion (i.e., the uncertainty principle) leads to statistical correlations between positions and momenta. Since the Hamiltonian is separable, one can factor the partition function as discussed in Chapter 4. This gives —Bhw/2 3N 3N = (eMw/2 _ .~Bhu/2y where we have used the fact that & +n) Mw with w = vk/m is the nth energy level for a one-dimensional harmonic oscillator. In the classical world, the partition function still factors, Noon 1 N N .-BH(r”, Pp) a+ Je fap’ e 727-31 2 2 ° =Bk(x - x,)°/2 ° -Bp°/2m ‘0 x - oy (J axe Jf an, e 1 fe. (ae 3N ni BK B = (Bitw) 3% +6 /2 (ec) Since e ~ 1 BMu/2 for small B , th + Bey ~~ Bey ~3N ~ (pho) "= Garassical * 3" (a) For p+ Pop (but p < Pop) we approach the g(x) for a close packed system, 1.e., delta functions spaced { apart, The lower the density, p, the broader the peaks, and the more rapid the decay to unity at large x. g(x) at high Pp x(b) (a) (e) Since u(x) is @ for [x] < % and zero otherwise, g(x) is therefore a step function at low p. as p20, glx) =e Bulx) g(x) at low p 2 g 1 0 0 1 2 3 4 5 ~ x/i Notice for a hard potential system, gu(x) is either infinite or zero, So Bu(x) is independent of temperature. Thus g(x) is independent of temperature. The configurational properties of hard core potential systems have no temperature. Since the integral of pg(x) determines the number of neighbors, and since there will be one neighbor between 0 and (3/2)%, we have (3/2)2 mq 4X Bop (X) = Gp oO =k Note that the momenta are uncoupled from the positions, and that the momentum distribution is proportional to exp(-ap*/2m) . Thus, Gi) eof) asfe ° a 2 2 e BP /2m Bp /2m a Bhoand re tua. 4 (ii) <3 avo = Finally, (iv) = + = 0 since the "potential energy" is zero for all acceptable configurations. = (f) Since gp = p + bp® + , we can identify bas the second virial coefficient. Hence ° 2 pbe-l yf axte MUO 7a S fo ax(-1y =e. 1 Note that b is independent of temperature. (g) The partition function for the system can be calculated exactly. The easiest way is to think of the excluded volume due to the rods (NR) as being subtracted out of the possible position space of each rod (L). Secondly, configurations which are identical except for the switching of rods are counted as the same state, i.e., we need to divide by a factor of NI. So, L-NR L-NE L-NB 14 a-F Lf af ax... Joa wow Jy Tg 2 eo Lt N - Wow (L~ NR). a This value for Q can also be arrived at from scaling and changes of7.32 coordinates of the configuration integral: First, adopting one particular ordering of the rods, L L L 1 -aU a= Jf ax, f ax, ... J dye ge tg ti tg oN L=(N=1)2 L-(N=2)2 L 4 -4A4 ax, [ ax, ... f ow? wg fee 14 which becomes after a change in variables to xj - xj - ik, L-na. L-Ne L-NR 1 1 ' ' Q- 7 y ax) f diye I axy * N-1 14 N = yp opr (bh - ue) yh NE Then an ep = nay z . Thus, b = 2, as already noted, (a) Neglecting the internal structure in the low density limit, the gaseous Ar is an ideal structureless gas. Thus, for M atoms (id) (id) _ _ agng nr 2 Bh™ )3/2 = an ( San ) + gnp = £(B) + kop , with p= MV. (>) Let u,(R) stand for the total potential energy for all N water molecules with RN denoting all the coordinates necessary to describe the configurations of all N water molecules. In the low density limit, different argon molecules do not interact with eachQ = expl-NE,,(8)-Mf(8)] other, but they do interact with the water molecules. Therefore, quite generally, we can view the total potential energy of the solution as M N u(r) + Fou ¢ mt ist where r, is the position of the i*? argon atom and there are M such i atoms. The partition function is then 1 M air Jan" far, +-dzy exp {-ecuy") + 3 vies y3} wT ist N u N -MF(B) -BU (R) MM -8U(r,3R>) ~BU (CR ) 2 forse “ Cfar,.+.faryy Be iN “yfar eo Mo -BU(r; RY) wefan <1 e@ be ist where Q, and <...>, denote the partition function and ensemble average for the pure liquid water solvent. Now, since the argon is dilute, different argon atoms are uncorrelated. Thus, the average of the product can be replaced with the product of the average. Hence, due to indistinguishability, ees) oy Bu Crys) e-g, Sp We dy yo Mle(a) + Bou] pan, 7BUE ys BD =ayve 27 ,_ where Me ce ys Note that since the solvent is isotropic, the last average is independent. of The desired formula for y is obtained on differentiation of 4nQ with respect to M with Q given by the last equation. Due to , = 1, it is obvious that Ay = 0 when U(ry; RN) is zero.(ce) (d) At phase equilibrium, the chemical potentials are equal. Hence, . . = 5 bh fp) + dnp, = £(B) + &np, + BAu , implying pp = pe and since for an ideal gas, ap = pg Bp = 264 . Q.E.D. We can simply plug into the result for du derived in Part b Alternatively, we can follow the suggestion noting Au = change in Helmholtz free energy due to introducing one argon atom into pure liquid water. =~ kT tn[Q(1 argon, N waters)/Q,] from which the desired result for exp(-BAy) follows immediately Next, define dy, according to -BAU -Bau, ({r - ry |) e *-cne 5 ; w i Then ~Blu,y(e-e, D aban g, g(r) = y(r) so y(o) = lim, g(r) . Hence, ro¢ BP/p = 1 + (po"n/2) go")7.35 (a) Let U) = 2 Cup (rj) + day (ry 5] . Then from 1 dung a da/da a = = 1 a aay mforee one finds using arguments similar to those in Exercise 7.32 that z aa/dr = (p°v/2) f de gy(rduy(r) « Integration over 4 from 0 to 1 yields the desired result. ~B(UG* UL) BU, ) aay = farNe 9 Ty farNe 9 Further, W> =< JY ur). = i rT TS = (p°v/2) f ar gyirduy(r) « Therefore, since A - Ay = - kT 2n(Q/Qo) y 5p far gy(ryu,r)8.7 Chapter 8, Statistical Mechanics of Non-Equilibrium Systems Since Hl2]= 1 2 > q* 0 z2 e70t et where @ is the Heaviside step function. We can then write -1 gt ky 0) = oO >. BA(O) = Xq <6[q(0) - Q*J> <6Lqto) - a*I> , and since the distribution for v(Q) is even and the time origin is 81arbitrary, . gt Kpqg6O) = <|v]> <6¢a-q 72x, Recognizing ATO race] = ofv(0)] gives the transition state theory expression for Kgs?) t t “+! Sanity) = 2 dt' = 2f at' et /* 0 an 0 t/t = 2cv?> tr - oT), Integration and enforcing the initial condition of ar?(o) = 0 gives ar?(t) = 2cvont + 2cvoe?te t/t = 1] For t << t, to and order int, 2, 2. 2528 : t, 2 aR“(t) = 2tt + 2evr[1 - = + —5- 1) Tog? = event? es K(ygt)> i.e., inertial (non-diffusive) behavior. For t >> 1, 2,2 an?(t) * 2¢v2ort = 2¢v2 >a? = 2ev2oa (ter) which represents the long time diffusive behavior, The function is plotted for all time in the figure. aR) 828.20 Since 1 1a 2 1 ey2 ps 5 J at = & gE aR ore Zevon then v= BAe. ow. & mv > For ar, Md = 40 g/mole, so at T = 80K t = 60 fsec . The primary variable x(t) drives the bath, and the driven bath affects the primary variable through the force f(t) = f(t) + Ja xp(tt) x(t") t ' da ' ' - - at’ ¢ —t - f(t) - 8 J bh ey! Spite) ait) we fy tt) + BC, (0) x(t) = BC, (t) x(0) £ ' by: Beet -8 J dt’ cy(e-t") kt") (a) where the second equality uses the connection between y,(t-t') and Cy(t-t') , and the third arises from integration by parts and the fact (essentially a notational convention) that (0) = f= J e,y, = £,(0) f£,(t) differs from f(t) only as time evolves from the favevan phase space point at t = 0. Now consider the average of f(t), averaged with the initial conditions x and x held fixed: l = 6 tyltixes by oP x,x fixed at t-0 This equation follows from the fact that the distribution of initial bath variables differs by the factor (gxf) from that when the bath isuncoupled from the primary variable, Let us employ the yy's that are the normal modes of the bath, where mode y; has the frequency wj. Then we have . BxOs¥, 1 = Te, i . fo eee ty: zt x,x fixed i axe, y 3 i = i Cy coswyt ggre~ kn Argon Gas t The very slow decay has a time scale of t , the mean collision time in the gas,(b) (e) Argon Solid The decay in the oscillations is due to the processes described above. Note that even if the lattice were one-dimensional, the minima of would not drop to zero because the ensemble average is over various initial conditions, so the turning point times are averaged over. In three dimensions, we have orbits withno turning points. This is reflected in the three separate phases that are averaged over: ae = ely2 205) + v2 Ben) veley aye = <(vE(O) + vyto v3(0)) (vy) vy 6b) + vitt))> = 3 in the case of an x x isotropic lattice. (a) CO in Solid The small oscillations are due to the vibrations in the lattice. The long time decay is due to the infrequent process of a CO molecule flipping over completely in the lattice, 8.23 For a small perturbation from the equilibrium zero-field distribution, the fluctuation-dissipation theorem tells us the rate of relaxation to equilibrium is the same as that of a spontaneous fluctuation to that non-equilibrium distribution, At time t = 0, the field is shut off but the flow ¥(0) = 0 . Since the flow will dissipate with time, Vit) = AV(E) « <6y(0)-8y(t)> = 8.26 or <¥(O)ev(t)> The relaxation time is then = or t = pmD . Substituting ¢,(t) = ¢- c,(t) - ¢,(t) into the rate equations gives e,(t) =- (k3,* ky3) e,(t) - ky gep(t) + ky ge ett) =_- ko3 e, (t) - (kot ka3) on(t) i k23¢ or, in matrix notation, a fo} fk 30 3 Mt oy 13] % at fey kag Kaa* Kos! |@2 Ko3 “4 The homogeneous part can be solved by expressing | ,'{ in terms of the 2 eigenvectors of the above matrix, while the inhomogeneous part just adds a constant to d, and dj: written in the basis that diagonalizes the above matrix, mAye Fy 4. @ a d,(t) = fye + const oo) a | 1 > ae fa. ~Agt 7 Ody fa a,(t) = tye ? + const! where f, and f are constants determined by the initial conditions. Thus the concentrations c, and cy can be written as the sum of two decaying exponentials, mast wAgt y(t) = Ae + Be + en(t) = Aoe + Boe + £65: where , | ms are eigenvectors of the above matrix. (a) S0, do, (t) = te? + ae? seytt) = Age By oe where 27 Ck, + ky) * (ka 5% k,3)1/72 / z (CK, + k (k 39% Ky3))/2 }e + kok 13 13°23, -BQ (b) For e <1, then k, k, >> k k 137 "23 31” 32° We can write ai Cikg,+ k 30) + (3 + ky) 1/2 2. + a7 {C0k3,+ Ky g) + (got k53)]/2} (k,ka5 + kK, 4Ka5) 31% 32 * Kayko3 * Ky 3K3a = [kay kyo )/2 + (Kyat ky3)/2] - + + 2 wD 1 = egy kegy + kgyhag + Ky gkga)/Elk a, Kyg * Kgot Kpq)/2I A, 2 18 now conveniently written in terms of the small parameters 1 kay +kg9, So that we can expand the root Ye = 1-7 : = Lok, + kgo)/2 + (ky + k53)/2] 1 2 «frei -5 kk (kay Ky3 + Kgot k33)/2) IE 4 31% 30" Ka)Kag * Ky 3kgz 13° M3So the + root is om (ky 3+ k3)/2 + Of) [2 - OC€)] = kyg * kg = Thhansient and the - root is W(kgy+ Ky gt Kot og) (34k 13° "32% Ko3 32 * K31Ko3 * Ky 3kge Wig" Ky3) * thy (3ykog * Ky3kaa 23) * Tex * From the above, k, k. “1 23 13 -1 we =k. (<3 —) + k (2 ) ct i , ram 31 Kg * 3 transient and so the relaxation is dominated by t + When ko =k. , rx 23 13 “1 Than 7 (Kgy* Kgg)/2 (c) As shown above, the faster transient decay occurs on a time scale of 1 ict (ky 3 + ky3) Koz if 1g 2 k3- (d) The two decay rates are analogous to the two rates in the reactive flux description: 1,01 = Tkransient “ Tram * The connection can be made by imagining preparing the system at the transition state, i.e., in state 3. Then the decay into states 1 and 2, c,(t) and e(t), follow the two decay rates, one much faster than the other. But c (t) « <én, (0) 6n4 (t )> by the regression hypothesis, and the time derivative of <én,(0)én,(t)> is just the flux in the reactive flux picture. In particular, Toy is on the order of ka, or k3p, and is the plateau value for the reaction rate.8.28 (e) This is similar to the transition state theory idea where e 8° is the probability of getting to the transition state 3, and D = 1/n is the rate to cross the barrier (3) once there. As we showed previously TST) cly|> <é(a-a%)> « 4 5 m2 T = 300K n= 0,01 poise Dap = 1X 10Pem?/see My, = 40 amu Myo 7 18 amu (a) and (b): The velocity distribution is just the Maxwell-Boltzmann distribution and is independent of the potential. So wi : a <3 my ,¥e2 - ee 3 kgT = 1.87 x 10° cm/sec” or evel. 4.3 x 10" om/sec for Ar atoms in the vapour and in solution. (ce) (a) Assuming inertial motion in the vapour, the time to move 10A is 2 51/2 t= 10A/'/* = 2.3 psec Since AR@(t) = 6Dt , the time to diffuse 10A in solution is t = 170 psec. We are justified in assuming diffusive motion since 170 psec is long compared to the velocity relaxation time calculated in part (e).(e) (ft) The relaxation of the non-equilibrium distribution will follow the same rate as that of spontaneous fluctuations, such as in , The decay time for is given by Tre1ax 7 Bm ~ 16 fsec . 1 14 From 8.24 D= 5-3 aap a ant) So if n doubles, D halves, and the diffusion time t = 47 doubles to 340 psec. In contrast, since the temperature is still 300K , !/? = 4.3 x 104 om/seo remains the same. 2