SOLUTIONS TO SELECTED PROBLEMS FOR STATISTICAL MECHANICS SECOND EDITION KERSON HUANG Massachusetts Institute of Technology JOHN WILEY & SONS New York Chichester Brisbane Toronto Singapore Copyright © 1987 by John Wiley & Sons, Inc. This material may be reproduced for testing or instructional purposes by people using the text. ISBN 0 472 84827 2 Printed in the United States of America 10987654321Ll Adiabatic transformation of ideal gas: S = Const. By (1.22), which follows from the Second Law, we have y = Cp/Cy = CWP), ACV/0P), Hence av) _ 1 fav (3), * vl a The last step ts obtained by using the equation of state V=NkT/P. Thus, along an adiabatic line we have dP/P = - y dV/V. Integrating this gives the answer PV” = Const. 12 (a} In a cyclic process AU = 0. Therefore AW = AQ. which is the integral of TdS over the cycle, or the area A. The heat absorbed by the system is the area A+B. Hence the efficiency is - A. od U* AsB 7 T3B7A If the engine had operated between the highest and the lowest available temperatures, the area A would have increased, but B would have remained the same. Therefore the efficiency would increase. (b) For a arbitrary reversible cycle, the efficiency is given by the same formula as above, with A the area enclosed by the cycle in the T-S plane, and B the area beneath the cycle. If the cycle had been replaced by one that operates between the highest and lowest available temperatures, then A could only increase and B could only decrease. Therefore the efficiency could only increase. 13 In the free expansion of a real gas, no work is done, and no heat transfer takes place, by definition Therefore AU=0 by the First Law. The process cannot be represented by a path on the equation of state surface, because it is not in thermal equilibrium in the intermediate stages. The endpoints, however, are equilibrium states, and can be represented on the equation of state surface. The changes in T and V can therefore be calculated by joining the endpoints by an any reversible path with constant U. The formula AT = (6T/AV), AV merely states this fact. 1-114 The cycle is a Carnot cycle. Hence the efficiency depends only the ratio of the temperatures: y = 1-T,/T,. The work done is therefore W- nQ, = (T,/T,)Q, where Q, is the heat absorbed along the paths AB and BC. Along AB the heat absorbed is by definition L, the latent heat for 1 mole of substance, because the system transforms entirely from pure liquid to pure gas. Along BC the substance is an ideal gas al constant temperature. Hence AU=0, and the heat absorbed is equal to the work done by the system. The latter is the area under the path in the P-V diagram, or RT, log (V,/V,). Hence W = (-T,/T,) EL + RT, log (¥,/V,) ] The rest is just plugging in numbers. LS Use the Maxwell relations (A/T), = -S and (A/V), = ~ (a) Obtain ACV.T,) from P by integrating along the isotherm Ty Obtain A(V.T) for $ by integrating atong a path of constant V. We have Vv ACV,T,) = -RT, log -¥ ° 9 Vo RVTo (Te ACV,T) = casi (4) +F(V) Setting T=T, in the second equation and comparing it to the first determines F(V): FCW) = -RT, tory G may 3 [1- (b) Use P = - (@A/aT),. (c) Use W= { P dV along an isotherm.21 Water data at boiling point under 1 atm: T = 100°C = 373°K 1 = 539 cal/g = 226x 10° erg/g v, = 1671 om? v, = 1 cm? Substituting these into the Clapeyron equation gives dP/AT = 3.628 x 10° dynes/cm?/C" Hence in the neighborhood of this point AT = 276% 10° AP Using AP = -1/3.atm = -(1/3)x10° dyne/cm?, we get AT = -9.18 °C Hence the boiling point is about 90°C. 22 Let Ag denote the difference of Gibbs free energy of the two phases. The pressure at which the two phases are in equilibrium is such that dP/dT = (aAg/at)/(ddg/dp). Since this naively gives 2 » we use lHopital’s rule and differentiate numerator and denominators separately with respect to T, obtaining ap aag/at? aT ~ 9ag/aTap The denominator is Ac, the difference in the coefficient of thermal conductivity. The numerator is Ac,, the difference in specific heal al constant pressure. The latter may be shown as follows. Use the equation after (1.21): Cy = T(QV/3T}, (P/AT), Using the chain rule to rewrite (aP/3T),, we obtain Hence the analog of the Clapeyron equation reads WP/AT = dc/da Since the density is the same for the two phases, the pressure along an isotherm in the P-V diagram does not have a horizontal portion, but has a discontinuous slope at the transition volume. 2-12.3 (a) Use the Clapeyron equation to calculate P,,. Approximate Av by the specific volume of the vapor, which satisfies the ideal gas law. The Clapeyron equation then can te integrated to obtain PUT) = Ce eer where m is the mass of a water molecule. The approximation that € is constant breaksdown at very low and very high T. (b) The process is indicated by the adiabatic path AB in the accompanying P-V diagram. Since the vapor is assumed to be an ideal gas, along the path PV’ = Const. or prvr 1 = Const. Therefore PAT) = PCT C/T? But the left-hand side is P,(T) exp(a/kTr), according to (2.25), where a= 2om/p. Furthermore, we have the explicit form of PCT) from part (a). Thus, after a little algebra, we obtain ~ +) pio (4) {c) In this approximation adiabatic expansion always leads to supersaturation. Referring to the accompanying figure, we can see that V>V_, because P,>P,, by use of (2.25) and the ideal gas law. 1 t 24 Write the Van der Waals equation in the form -Rt la a. V-b y? The Helmhioltz free energy is obtained by integrating this along an isotherm: A= -RT log V+ +F(T) The entropy is therefore s = -94 - Riogv-FCT) It follows that Us G+ FC- TROD) cy = Be -TeD25 Perform the integration G = - [var graphically, as shown in the accompanying figure. The principle of minimization of G shows that the system will follow the path ABC, skipping the loop entirely. IL is obvious that the point B corresponds to a line in the P-V diagram making the shaded areas equal 26 The water will not leak out because the pressure outside balances the total pressure in the solution, and is therefore greater than the partial pressure of the water. 2-3PROBLEM 2,3PROBLEM 2.5 2531 Under standard conditons the density of a gas is given by Loschmidt’s number n= 2.69x 10"? molecules/cm? The temperature is 300°K, corresponding to kT = 1/40 eV = 4x10" erg One gets for H, ans 3 “Ene 3.4.x 10 For other gases the above ratio will decrease as the root of their masses. 3.2 All molecular inteactions are electromagnetic in origin because they arise from the overlap of the electronic clouds of atoms. The atomic nuclei are kept sufficient far apart so that the strong but short-ranged nuclear forces do not play a role. The weak interactions are weaker than electromagnetic forces by the order of tol, while gravitational forces are weaker by the order of 107%, 3.3 The scattering of two hard spheres each of diameter a can be reduced tu the problem of the scattering of a point particle by a fixed hard-sphere potential of radius a (a) Refer to the accompanying figure. We have 26 +6: 7. Hence the impact parameter 6 is related to scattering angle 9 by b=asinfi=asin(6/2)= a (-ctos 6/2 The classical differential cross section is a() = orgie gs = a/4 The total cross section is ma’. (ob) The quantum mechanical total cross section in the low-energy limit is ta?, and in the high-energy limit it is 2na% It comes as no surprise that the low-energy cross section differs from the classical vatue, for the de Broglie wavelength is much greater than the diameter of the hard sphere. Bul it seem strange that the high-energy cross section is twice the classical value. 3-1The physical explanation is that half corresponds to the geometric cross section (the classical value), and the other half represents the amount of scattered wave necessary to cancel the incident wave behind the sphere, thus producing a shadow behind the sphere. For the scattering of two identical particles, however, the total cross section is only half that for the scattering from a fixed potential, because the phase space in the final state is only haif as large, owing to the fact that interchanging the two particles does not lead to a new state. In the case, the quantum mechanical total cross section in the high-energy limit agrees with the classical value, But the differential cross section still disagrees. This example shows thal we should not expect the classical description of scattering to be quantitatively correct in the subatomic world. 34 The Boltzmann transport equation is a set of coupled equations for the distribution functions f and g: If ager (2+¥,9,+ Fv, } ty = dat des a > (Ar F9,+ 07,,) g > Jeet Tig where J,, and J,, are the usual collisions integrals as given on the right-hand side of (3.36), each with their own T- matrices. Jig is the collision integral describing collisions between different types of molecules: TCP) = |e apy dps SP PIT? Pa es - 0, 8) where T is the appropriate T-matrix for the scattering. 3.5 (a) The accessible phase space consists of two horizontal strips in the p-q plane, of width 1/2, and located at the top and at the bottom of the unit square. (b) After not too long a time, the distribution looks like that in sketch (a), which can be obtained by "folding (b). After a long time, the distribution looks like that in sketch (c), which can be obtained by “folding” (d).Incident PROBLEM 3,3 33(@) (b) P P oT . ero ' Pobal* Eft p ME iy 4 Hit uy a i EL Ey cH () (d) PROBLEM 3,541 A neutral molecular beam is passed through bwtween the cogs of a rotating wheel. Only molecules of a particular velocity can go through. The velocity can be selected by adjusting the speed of rotation of the wheel. The flux of the fillered beam as a function of velocity is then compared with the Maxwell-Boltzmann distribution. 42 In the rotating frame of reference, the gas sees a potential or) = 4 mur? where w is the angular velocity of rotation, and r is the distance from the axis of rotation. The Maxwell-Boltzmann distribution is 2 aa . - + o(r)| Br) = ce i reo) where ( is a constant. 43 (a) Escape velocity: vg = (2gR = 1.56 x 10° cm/s, where g is the acceleration of gravity, and R the earth's radius. The kinetic energy of an H, molecule moving at the escape velocity is J anv,? = $¢mc%)(vy/c2) = (2 GeV) (2.x 10°) = 2.7 eV At a temperature of 300°K, kT = 1/40 eV. The ratio of escape kinetic energy to thermal energy is X) = 2.7/(1/40) = 108. The fraction of H, gas that can escape is F = i Pv ICV) =e = e108 vey We need not do the integral in greater detail. The above gives a good approximation to log F. (0) The high-velocity tail of the Maxwell-Boltzmann distribution does get depleted, but the rate is so slow that other things can happen long before its effect is perceptible. For example, the sun may die. 4-14.4 The Maxwell-Boltzmann distribution for a gas hot enough for the average velocity to be relativistic is -B,p?eur (CB) = Ce where RP is the momentum and m the rest mass. We use units in which the velocity of light is unity (c=1). The constants € and B are determined by n= jes (P) . €= b fe pm? ICP) where n and ¢ are respectively the density and average energy (including rest energy), The pressure is 2 Ps | dp apy, FCP) = § jis —faes (PD >a * 3 ape The last form is obtained by using v,= p,/\(p"m? and noting that in the integrand we can replace p,? by dp? To do the integral above, make the substitution p=m sinh @. Define TOO = | d6 cosh 6 sinh’a eAcoshe (r= Bm) QO Then n= 4n€m? 10) o fea | 46 cosh’e sinh’e eh a » pv em sinh*9@ e@7acoshe N 7 310) I do sinh’e ¢**° The second equation above determines 4 as a function of € alone. The last equation says PV/N=F(e). The temperature T is defined by F(€)= kT. To find the relation between T and 4, do the following integrals: 4-2100 = - 2 ABA Kan | da sinh*e e*oste 2 42 (5/2 KO) 5 ait where K, and K. are standard Bessel functions. Using the Tecursion relation KOd=-K,00+4K,00 and the fact 15/2}=31(/2) we get X = me?/kT This result could have been derived much more directly by a partial integration at some earlier stage. But we do all the integrals for completeness. It is clear that the temperature reduces to the usual one in the non-relativistic limit: . 2 2) 2) mkt m/KT) 41 +Cp/m0) =(m/eT) [1+ a c CareT) d oCp/my’ Co MAT) HeGpamy72) Lr gr P*! 45 Consider an ensemble of stamps. Each member of the ensemble is a stamp on the table top, constantly bombarded by air molecules, so that its height fluctuates. The temperature is determined by the air. The height distribution is Maxwell-Bolzmann: f= Cemenar where m is the mass, g the acceleration of gravity, and h the height. We have neglected the kinetic energy of the stamp. Putting in m=.1, g=980, h-10$84-8, kT=1/40 eV=4x10"'*, all in cgs units, we get the ridiculously small number of logf = -2x10° But the above will become the order of 1, if the mass is decreased to 10°. that of a speck of dust. 46 The probability of finding one molicule in volume v is v/V. The probability of not finding a particular molecule in v is thus given by !-v/V. The probability of not finding any molecule in v is, in the limit No, V-o, with N/V finite, P = a-wvy = Mog’ -wv) 3 ew 4-3(a) Under standard conditions, there are about 3x 19’? molecules/cm?. Hence N=2.7x 10", w/V = 107, and -1919 logP = - 3x10". P » 10° 23 (b) If v=1A, then we must multiply the above by 19%, with the result P=1-3x10° 47 Under standard conditions, the characteristic time for a gas to randomize is of the order of the collision time 1, which is of order 10° 5, During a time interval 1, the probability that an initial state will recur is the ratio No. of systems in ensemble with vy empty _ eon No. of systems in emsemble where we have used the result of the last problem. To see a recurrence, we should be prepared to wait for a time t= Ten 2 ioe? This number is so targe that the units make no difference. 48 (a) If no constraint is placed on the average velocity, then the average velocity will be zero, because all values of the velocity will be equally weighted, and any particular value is counterbalanced by its opposite. aa wme)= cer ™ 49 Use the method of Lagrange multipliers: SH+ 2d, dnd, fe = 0 dp (log f+1 +2, +X, p2/2m) of - 0 « logf = -1-,- 2, p2/2m5.1 System ¥ (cm/s) » Com) tT (s) H, at stp Lsxi0° Bxl0 ® 6x10" p in hot plasma 4xi0° 3 10° p in very hot plasma 3x10" 100 4xia® p in sun’s corona 8x10° 10°° 10° . 5.2 Let A = Area of the hole T = Time during which hole is open E = Total amount of energy escaped N = Total number of molecules escaped (a) The flux for molecules in d°v is d°v v,{, where the x- axis is normal to the hole, and f depends only on the magnitude of V. Write d°v = 2nv dv d€cosa). Integrating 8 from 0 to 7/2, one obtains the answer mv°fdv. (6) The particle flux is v,f, and the energy flux is p°/2m, times the particle flux. Hence E = | dp (p?/2m) v, £ = tACHC/2m?) C2mkT,)° v,20 =z 0 raf @py, £ = TACIC/2m) (2mkT, 3° 20 Alter the hole is sealed off, the totat energy E in compartment 2 remains unchanged when the gas comes to equilibrium, and of course N is also constant. Therefore € = E/N = 2kT,. (c) By definition € = Sut. Hence : 2 T= $7, $.3 (a) If the atoms in a gas do not interact, then compressing the system will not produce density waves, for the atoms merely bounce off walls, but not off one another. (o) An ideal gas is not a gas without interactions, but one in which the density is small enough for PV=NkT to hold, but large enough that the mean free path is much smaller than relevant wavelengths of density variations. 5-15.4 (a) Let the temperature variation over a wavelength \ of the sound wave be denoted by Ad. Then Heat flux * K A6/h Q, = Heat transferred over distance \ during a period 1 KAgX 1 c= K 40/c = Q, = Heat required Lo equalize A@ = Apcy AB The process is adiabatic if Q,/Q, << 1, or K << Acpcy (b) For air at 300° at 1 atm we have p = 0.08 tb-ft? c¢ = 1088 ft/s Cp = 0.240 Btu/Ib--F K = 0.0157 Btu/hr-ft--F Taking a sound wave of wavelenty d=10 ft, we have depey * 3x10” Btu/hr-ft--F >> K 5.5 (a) Moluecules emerging from the disk make collisions, and rebound back and forth, within a mean free path of the surface of the disk If the mean free path is small compared with the dimensions of the disk all this happens within a thin boundary layer, and we may replace the disk by disk plus boundary layer, with some appropropriate boundary condition. We are in the hydrodynamic regime. (b) Put T= 0 im (5.100)-(5.102): a a * 9 gP =0 a. Kye a” Bey’ e The second of these equations shows there is no force on the disk. The problems becomes one of heat conduction through the gas. (c) Since the mean free path is much greater than the dimensions of the disk, molecules hitting the disk had its last collision very far away, with Maxwellian distribution corresponding to temperature T. Molecules that emerge from the disk go off to "infinity", for there are no collision to send them back Let the 5-2density of the gas be nm. Let the x axis be normal to A, pointing in the direction A to B. Calculate the forces F, and F, on sides A and B, using the method of (4.20) Remembering that the disk is of unit area, we have ad a 3. 7 a F,=8 av Qmv, v, f,€¥,T) = ¥nkT ¥e0 F,- | dv my, ¥, in-s| dv my, v, fyC¥,7,) = - 4 ak(T-1,) ¥y60 v0 Hence the net force of the disk is > 3S Fos Fy+F, = -2%cr-7 Since T,>T, the net force is directed normally from face B towards face A. 5.6 Assume that the white side is perfectly reflecting to incident tadiation, while the black side is perfectly absorbing. Sunlight has two effects on the vane: (1) Radiation pressure: On the white side, a reflected light team transfers twice its momentum to the vane. On the black side the mamentum transfer is just the momentum of the light beam absorbed, Thus there is an excess pressure on the white side. (2) Heating effect: Since the black side absorbs energy from radiation, while the white side does not, the temperature of the black will get higher than that of the white side, creating a situation similar to that described in Prob.s.5. (a) At high densities the mean free path is small, and we are in the hydrodynamic regime, similar to that discussed in part (b) of Prob.5.5. As we found there, we have heat conduction in the gas, but no net force acting on the vane due to thermal effects. The force from radiation pressure is small, and easily overcome by viscosity. (b) When the gas is sufficiently dilute, we can treat the vane as if it were in a vacuum. The radiation pressure causes the vane to rotate, black side jeading. (c) When the gas density is ~- dilute encugh so that there is excess pressure acting on the black side, as discussed in 5.S(c), and -- dense enough so that there is viscosity to counter the radiation pressure, 5-3then the vane will rotate with white side leading. A qualitative estimate of the density n may be obtained by assuming that the mean free path is of the order of the size of the vane: a on d Wao tem Take the molecular diameter to be 2A. Then n= 10 cm? Assuming the ideal gas law, we obtain at T=300'K a pressure Pp — nk = 40 dyne/em? = 4x10 atm = 03 mmHg 57 To use the Boltzmann equation (3.8) to calculate the distribution function of the charged molecules, we must take the force F to be where E is the external electric field, and F is the force due to the lattice of fixed ions. We make the approximation an . Fea, i. te Hence we have to solve the problem "from scratch”, and cannot use the hydrodynamic equations derived in the text. The equation to he solved is (f + Vv, + eE9,] { + z) = § We can neglect g on the left-hand side, and drop the first two derivative terms on the left-hand side. (a) g = TeE-Y, = at Ep > Lz 1 3 0) > 20 pay m 960) (b) = XP? = a (é pO +e) B = hae fe p(F-p) pe FE} 3. 0 _ 2ek Pp? - —fh 2 = sek gh > 3am XT [eee Po > Smet Gi? = We have use the fact that the average of p?/2m is 3kT/2. The electrical conductivity is ¢ = ne?/m. 5-46.1 From (6.12) we have log! = Ing [8CE+0) - EF = tog L(E+A) + tog [1 +H] The second term is negligible when N-o. Similarly, from (6.13), we conclude jog! = log w, because as No, log! is of order “ N, whereas A is independent of N. 6.2 Although the uniform ensemble includes all states with energy below E, the states near the maximum energy E are overwhelmingly numerous, because the number of states increases with a rate of order e%, (which is of order e!°") The situation is typical of exponential growth, familiar from the common observation that there are more physicists living at present than all physicists that ever lived. 6.3 Use E as the unit of energy, ie, set Eel. {a) With U fixed we have n, = U, ny-N-U The number of ways to choose n, paricles out of N is the binomial coefficient (A) whose logarithm gives $/k. Using the Stirling approximation, we obtain S/k = N log N ~ U log U-(N-U) log (N-UD (b} The average occupation numbers are n, = U, ny = N-U. The fluctuations are of the order of the uncertainty in U, which has been neglected. (c) Using 1/T = dS/dU, we find im) The temperature is negative for 5| dCcos oy) : “4 -ae(eo Qe (oy -B H(pq) + 6 2, cos a where Q.°" is the partition function in the absence of magnetic field. a 1 (@) M= kT A tog Q, = Nu(cosne - 4) LM wf 2 bt) x= EM - E+ -csch’o (b) x (4 esel ) (c) At high temperatures 6-0. Thus wld dy ley). ° itl aft $0°)| = $e Hence the Curie constant is C = p?/3k. 7.6 Solution given with the problem. 7-277 (a w- bop s a/v) = it ‘i , yy _ a T p= KE -av 20 )-%- 4. Foie) (b) From the result of Prob. 7.6, az (kT b’ - a’) , K = T' 2m BCHy-E)T where s=T or t, E is the incident energy, Hy the unperturbed Hamillonian, T the reaction matrix, and n labels a complete set of unperturbed states. The cross section for a partially polarized beam is Oo = fop+Cl-Do) = Tr(pK) where the density matrix p is given by p= fit tog {1-zexp|[-8 +0, (0e4)]} k-0 Boltzmann: log @ = iogz -4pe Wg - log {1 - eo) 8.6 Enough hints have been given with the problem One might note the following points: (1) The energy of an H, molecule is the sum of its translational energy and its internal excitation energy. Even though the temperture is high enough for us to treat the translational part of the energy classically, we must treat the internal excitation part quantum mechanically. (2) The two electron in H, are in a triplet (symmetric) spin state in ortho-hydrogen, and in a singlet (antisymmetric) spin state in para-hydrogen. The Pauli principles requires the two- electron wave function to be antisymmetric under the interchange of alt coordinates. Thus the spatial part of the wave function is antisymmetric for ortho-hydrogen, (hence odd relative orbital angular momentum,} and symmetric for para-hydrogen, (hence even relative orbital angular momentum.) The triplet spin states are degenerate, giving the factor 3 in the answer.oD 91 a= > XQ, NeO Qi = 3k |e Oz) = 4 fe exp [-¥+19 tog z+ log Oz) log Q, rae ~N log Z + tog fz) 00 where Z is determined by the saddle-point condition, which is identical to the second equation in (8.71). The explicit form of ihe, grand partiton function 9(z) is given in (8.63) and 9.2 (a) From Prob. 7.6, oo BR = i= dr4nr?ftr), = fr) = eM -4 9 Using (9.57) for v(r), one finds fr)=2 e200? y= fame 7inkT where the upper sign is for Bose statistic, and the lower sign for Fermi statsitics. The integral to be done is elementary. One obtains as final answer Py _ 4255/2, 43 i = 1 *2 nd where n=1/v is the density. (b) The order of magnitude of the temperature below which quantum effects are important may be estimated by putting nn 1, or kT ~ Qa? n?/8/m As a numerical example put n=10°'° cm. One then finds T~ 0.1 °K for H,. For the other gases, the temperature varies inversely as the mass. Thus it is half this value for He, and 1/14 for N,. 9.3 Solution given with the problem 9-194 (a) A line of charge of density ¢ per unit length sets up an electrostatic potential ® = -0 log a@ at distance a perpendicular to the line Consider such a line to be a generator of the cylinder, with surface charge density g(q) per unit area, g(6)=g(-6). At a point at normal distance z from the axis of the cylinder the potentiat is 1 o(z) = -2} d@ (0) log (I - 22 cos 8 + 22) Q Hence the answer follows. (b) By Gauss’ theorem the discontinuity in electric field is 4mg(0). Hence the discontinuity in 1/v is 2ng(0). 9.5 (a) The equation state is a = log(i+z)+ t log(] + 2*¥) av Zz, az “Tez "T+ n where n is the density. Take the limit Vo. The second term of the first equation is 0 for zl. The second term of the second equation is 0 for z<], ans o for z>l. Thus there is a first order phase transition at z=1, at which both P and n are discontinuous. The discontnuity in n is a. The density of the two phases are 1/2, and w+ 1/2. (b) There are three roots in the complex a plane, at z=-1, and z=exp(tin/aV), AS Vo, the latter pair of roots pinch the real axis at z=1. (c) The gas phase is the low-density phase, corresponding to z<1. The equation of state is obtain by ignoring the second terms in the parametric equations, and eliminating 2: P/kT = - log (1-n) As n increases from 0, the first singularity encountered is at n=l. But a phase transition actually occurs at n=1/2, as we have shown earlier. 9-210.1 (a) Do the integrals in (10.16) and (10.17) with f(@r)=1 for ra The second cluster integral is straightforward: a = deri = 2 (ay 2 n 3 AX. 7 The third cluster integral consists of four terms, the three terms in which only two of the particle interact are identical: [er ary dry flr,2) f(r) = V [a Wry5 fry_) f(r,5) = V (4na?/3)° The term with all three particle interacting can be calculated as follows. Use particle 1 as origin, and let T and p be respectively the vectors from 1 to 2 and from | to 3. Let the angle between these vectors by 8. Let s denote the side opposite the angle 6 in the triangle 123. The integral we have to do is Is [es Wry Bry Mr, 9) Mr) M9) = V | er fe f(r) £9) f(s) s? = 12+p?-2rp cose If r+p/5 One can get 0° from (8.68) for the Fermi case, and (8.72) for the Bose case. The answers follow by doing the elementary integrals. 10.4 For the complete set of wave functions ¥,, we can use the free-particle. wave functions ®, as defined in (A.14) of the Appendix. This problem is an exercise in counting, as explained in the Appendix. The main point is that, when one takes the expectation value of an operator symmetric in all the particles, one needs to symmetrize (or antisymmetrize) only either the initial state or the final state. 10.5 According to the variational principle (9.82), > ce my 2 Fest n n Take 3 to be the Hamiltonian of the Heisenberg model. Take the state M> to be a direct product of individual spin states, 10-2which are eigenstates of o,. The diagonal matrix elements of # give the Hamiltonian of the Ising model, because the diagonal elements of o, and o, vanish, and those of 0, are +1. 10.6 N @ trp = [Ze] = ce = 1 C = (P+e8y! (vo) om = (e®- ee? + e*) = tanh (c) First consider the term Tr(p#) = - 2,(Mets,] iG x SS, + wHDs)} (s}i [e(s,e(s,)-e(sy)] S\S; = [Zets,)s,| [Dats,) s, = m {s} 8, Sp where we have used the fact that Trg = 1. Thus Tr(p#) = - €(Ny/2) - pHmN where Ny/2 is the total number of nearest-neighbor pairs. Next calculate Tr plogp = Pes -a6so]2 oe g(s) = Y Lets) log g(s,) s i i 5 = NSig(s) log g(s) = log C +B tanhB Ss The answer follows straightforwardly. (d) Note B= tant'm, c= {i-m?2. WAN = ~-Seym? ~ Him + kTm tanh 'm + kT[ Hog (1-m?)- tog 2 The rest is straightforward. (e) The free energy is w(m). Get result by straightforward substitution. 10.7. The solution is fully discussed in the reference given. 10-3Ih oa «58 (a) Electrons: $-2 for 2 spin states. n= 6.6 x 10°" cm? (for Pt} € = 3.6 eV (b) Nucleons: $=4 for proton, neutron, 2 spin states each. n= 1,72x 10°? em? & - 38 Mev (c) He®: S=2 for 2 spin states. n= 216x107 cm" 3 € = 10° eV = 12°K m2 The answer can be obtained by integrating the pressure (11.31) along an isotherm Write the Fermi energy as €,=c v4, where c is a constant. Then Integrating this immediately gives the answer. The constant of integration is determined by the fact that at T=0 the result must reduce to the energy (11.28). 11.3 Put z=1 because there is no particle conservation The energy density u and particle density n are given by _4 _ 4493, me? + (p2/2m) w= ve em, = sla pomp Pom) 4 4 - 4/,—1 ae ven 4 le Ppenc™p2am) , | The factor 4 comes from the fact that there are 4 kinds of nucleons: spin up or spin down, proton or neutron. The rest energy mc? corresponds to a temperature of 10$82713 -K Thus the factor exp(-me/kT) is ridiculously small for ordinary temperatures. There is no need to do the integrals with any greater accuracy than Tt-tuo n ~ exp(-me7/kT} For such small numbers, wits are irrelevant. Nucleons don’t like to be around (unless required by law,) because they can make so much entropy by decaying. This shows that, but for the law of baryon-number conservation, all would be gamma rays. 14 . (a) Assume classical statistical mechanics applies to the normal modes of the latlice at room temperature. By equipartition of energy, each harmonic degree of freedom contributes kT/2 to the internal energy. Hence Cirtice = SNK {b) The Fermi energy of the electrons is typically 4eV, which gives a Fermi temperature of Sxl0° ‘K. Hence room temperature (300 “K) is very low temperature for the electrons. Using the low-temperature expansion (11.28), one obtains C, = WkT/3€,. Hence electron -2 Carectton/Crattice = TKT/9€p ~ 10 ILS The pressure P must be the same in both compartments for equilibrium. (a) At T=0, P=-dE/dV, and E=(3/5)Ne,. The Fermi energy for spin s fermions is & = eln/(ase1]? where n=N/V, amd c depends on the mass, and is the same for the two gases. Thus P = Const. n/3/(2841)"? Equating the pressures of the two gases leads to ny/ng = ((2s,+1)/(2s,+2))"° = 27° (b) As To the gases approach classical ideal gases, with equation of state P=nkT. Hence n,/n2~l. 11.6 The diamagnetic contribution to the magnetization is given by (11.89) as ~ p/v, where = et/2mc. The paramagnetic contribution is given by (11.112) as ur/v, where r is the fraction of spins aligned with the field minus the fraction 11-2aligned against the field) When ali particles can be accomodated in the lowest Landau level we have r=1. Hence the total magenetization is 0. 11.7 Find the internal energy u per particle by differentiating (11.128) with respect to -8. Find the specific heat by differentiating u with respect to T. The calculations are straightforward. The point of this problem is to serve as an exercise in calculation in a more complicated case than Pauli paramagnetism.12.1 (a) In d dimensions the Helmholtz free energy per unit volume of a photon gas is | ak k*! log 1 - e FF) 0 a. ee ~ §(2n)? where C, is the surface area of a d-sphere. The number of photons per unit volume is wo 2c. da 1 Zfahty (any), ehh These integrals can be put in the forms oO NATCa-19 Xo ol s # where N, is a constant. The entropy per unit volume is obtained Uhrough o = da/dT: o a id o = N,1 kava dx gy = NTN Cae) Dv! 4 n=l The result s=o/n follows immediately. A table of numerical values follows: ad s d s 1 0 8 8.981 2 2.218 9 9.990 3 3.602 10 10.995 4 4.790 il 11.997 5 5.887 12 12.998 6 6.938 13 13,999 7 7,966 (ob) For a Boltzmann gas, one replaces the factor (e-1)! by e* in the the integrals for o and n Since dxxte™ = dtThis shows thal the density of states goes like e"'. In a fanciful interpretation, one might say that the d in the exponent comes from the volume of a d-sphere, while the term ] comes come the permutation of the distinguishable particles. 12.2 By plotting the data on a log-log plot, one can show that the specific heat is proportional to T° near T-0. This is characteristic of excitations with an energy-momentum relation of the form €,= ¢ p. Jn this case, (hese are phonons, and c is the velocity of suind The heat capacity per unit volume is half of (12.16), ( because the latter was derived for photons with two posstble polarizations, whereas there is only one kind of phonons in liquid hetium--the longitudinal kind.) One can extract c from the data. The answer is 226 m/s. Examining the plot of the data will reveal a deviation from the T° behavior for T>0.8°K. An exponential function would give a better fit there. The exponential behavior is characteristic of excitations separated from the ground state by an energy gap. These are the "rotons" discussed in Chapter 13. (See Fig. 13.4.) 12.3 In (12.64) G is expressed as a function of T and V, instead of its natural variables T and P. Cailing the function in (12.64) f(T,V), one has GCTP) = FCT.VCTP)) from which one would get (aG/aT}, = (af/aT), + CAf/3V), (AV/9P), For TT,, start with U/N in (12.62) and take two derivatives with respect to T, discarding terms that vanish at T=T.. One gets Cy _ 1585 (2) 72 3 Ba) 75? 01 a7 Nk ~ 4 (2.612) 7? * 2.612 ml Zit) Now take one more derivative with respect to T, and let T approach T, from above. From the first term one gets the result of approaching T, from below. From the last term only the term containing the derivative of 2°! 32/8T survives. Thus a) fay) _ 35 aflaz) _ _ 3.66 (+ Whar (i rot; 7 2 leaT 7H) =F 12.5 One has to eliminate z from the paramtric equations To do this systematically, write 12-3£. T Mig a (y Sublitute for °/v its power series in z, and reexpand in z. Determine the coefficients a, by comparing this to the original expansion. * n=O 12.6 (a) From (8.62), 72 jog 9 = - © log (1-20 *) = | dp p log (i-ze™?) Pp n a Expanding in powers of z, one obtains o 7? Pp _ log 2 | 1 ~kBp272m kr YA ae Eth dp pe = 52 lr) © ik g,(2) = > Fe (dilogarithm) k=1 (b) N ,ale2@ 1 Sz te a eee} =z log (1-2) from which one can explicitly solve for the fugacity: z=i-eMr geNA? (c) The integral for N diverges when zl, in contradistinction to the 3-dimensional case. Hence any number of particles can be accommodated in the excited states. The zero-momentum state is not forced to accept particles, and consequently will not be macroscopically occupied Physically the reason is that, whenever possoble, the particles like to be spread oul to gain entropy. The explicit equation of state is a regular function:13.1 The energy in the volume element AV=(dz)d, (Limes unit jJength in the third direction.) is composed of a potential energy and a kinetic energy: 2 PE = mgzav. KE.» “4 pray where m is the mass of a He atom, n the density of the liquid, ¢ the acceleration of gravity, f the superfluid fraction, and h is Planck's constant. The momentum of a suerftuid particie is taken to be that of a free particte with de Broglie wavelength equal to the thickness of the film Thus the energy of the volume element is AE = mn Az (gcd + (h°/2m2)fd"1 Minimizing with respect to d, one finds d= c/iz, c = (wm) {f/g Numerically one has d = (3x10 cm) ff7z (z in cm) 13.2 (a) From hace, - nh/m , we obtain by symmetry 2nrv, = nh/m , hence v, = ms/(mr) where n=1,2.... (b) The kinetic energy per umit volume of the superfluid flow is 0s, 272, where p, is the mass density of the superfluid. Hence the energy per unit length of the cylinder is an a E= | do | drcv,? = (r'ps?n?/m?} tog (a/b) 0 bd (c) The angular momentum per unit length of the cylinder is J = 2np.} drr(rv,) = Gantp_/m) (a? - 67) s a 3 > 2 where b° may be neglected compared to a’, because v is of atomic size, whereas a is macroscopic. A vortex line carries a lot of angular momentum (~a"} for the relatively small excitation energy (~log a). If the fluid in the cylinder rotated rigidly, it would have required much more energy for the same angular momentum. Hence rotating a bucket of superfluid helium will create vortex lines, 13-113.3 Refer to the sketch accompanying the problem Draw any imagnary path P connecting point B on one side of the channel to point A on the other side of the channel. Along P we have A | ds-y. = wm | ds-V@ = (/m) do e 8 We may use this to calculate Ag. Now suppose one vortex line of wit quantum of clock-wise circulation is created on the lower wall. Its total circulation is tos Woe h/m Suppose this vortex line moves across the channel and annihilates itself on the opposite wail. As it moves across, we may distort our path P to make way for il, but eventually it must cross P, because it has to disappear into the opposite wall. Thus, we may deform our path P to avoid the vorlex, but eventually a part of our original path must break away as a closed path about the vortex, (like in complex integration when a pole crosses a contour of integration.) After the vortex crosses the path P, Ao is given by (7m) Ao + | ds-y, + (2m/m) P Thus, So gains 21 each time a unit vortex of clock-wise circulation move across the channel. A chemical potenial difference du between A and B can therefore be maintained by a rate of vortex migration aN/dt = Am/h 13-213.4 We wish to calculate the diagonal matrix clement M, of 3 X= Der -1) i In the above sum only the following types of terms have diagonal matrix elements: pa eo Dalal, ~ Qonyé -N a0: tat =N- 2 péq, k=O: a! 92 paqa, = N zo pé#q, k=q-p: Same as above The sum of these gives mv 1S9) 13-314.1 For the new pair-wise interaction, every pair in the one- dimensional chain interacts with each other. The energy cosl to create a domain wali may be estimate by treating the system as if it were a continuous distribution of spins: L AE = | are = N(L'*-1) 1 where L is the length of the chain, and N the total number of spins. The entropy remains k fog N. Thus the change in free energy is (putting LaN. and ignoring ail proportionality constants,) of the order of AA = N°"-T log N For T>0, this is positive when N-o for o<2. In that case domain wall formation is discouraged, and spontaneous magnetizaton will occur at sufficiently tow T. 14.2 Continue to use the boundary conditin that all spins at the boundary are +, The domain walls are closed two-dimensional surfaces of area b. The maximum number of spins enclosed by a domain wall of area ® is (b/6)"*. To bound the number mb) of domain walls with area b, try to construct a continuous surface with b plaquettes. The first one can be located N ways. Each succisive one has less than 12 choices. Thus m(b) In the limit Noo do the integrals by the method of saddie-point integration. The saddie point is al @=4: 9/(20) = tanh +h) Thus the magnetization is 6/20. It satisfies a mean-field condition of the same form as (1441). Q 17.3 w(m) = bty m* +s, m3 +a, m4 The conditions for a minimum are W = rym+3s,m?+4u,m? = 0 Ww" + rg + 68pm 12um’ < 0 From y'-0 we have m=0, or m=-m,, with m, = 4 -Gsy/4u,) + x 3s,/4u,) - (ry/t)] Denote the lowest minimum by m. Let 1, decrease from a large initial vatue: To/tty > (389/4ug)”: 1-0. (0-0. Tg/Uy = (38q/4up Ys m==38_/(B8up) is a minimum; but o(m)-(3/8)°/8. Hence ti-0. When r, continues to decrease, the secondary minimum gets lower, and equais zero when =O, ¥’=0, p20, at (A) Ty = Sy/(2uy) . Mh = ~85/(2uy) When rp continues to decrease from this value, the the lowest minimum of ym) occurs at m<0. Take r, to be the temperature. Then a first-order phase transition occurs at the value of r, in (A) At that temperature the order parameter has a discontinuity given by the value of m in (A). The behaviors of y{m) and that of m are summarized: in the accompanying sketches. If one add an external field through a term -Hm, the qualitative behavior is the same, with the transition temperature becoming dependent on H. 17-217.4 We wish to do the integral Then the k-integral becomes a Gaussian integral, and can be done by completing the square. One obtains © 2 l= | ae’ ett Oe 0 where x is the magnitude of the d-dimension vector x. Making the substitution te she’ 2a one obtains the finat integral representation: J = (n/2y? pls oma ytd» | dv TO* * coshi(1-a/20v] o 17.5 The equation to be solved is (17.45); (¥? - rg) m(x) ~ 4u, mE) = 800 For r,>0 this has been done in the text, with solution given in (17.46). We only have to solve it for ry<0. In this case there is spontaneous symmetry breaking. When X=0, the solution is fh = Fro/4y For weak source (A\0) we write m-m+im, and keep only terms to first order in 6m The equation becomes (¥" + 2rq) bm(x) = 800 Thus, the solution is of the form as for r,>0, except that Ty is replaces by -2r,, or rol 17.6 Consider a Landau free energy of the form v(m) = dom? +r) m? tu, m? +s, m> 17-3Near the tricritical point, both Ty) and uy) approach zero, and the m® term is dominant. The question is whether we can ignore spatial fluctuations and take m to be uniform (mean-field theory). For mean-field theory to be good in the neighborhood of the tricritical point, the characteristic length of the m® term must be much shorter than that of the gradient term, which we shall take to be the Gaussian correfation iength Sq = tot”? The characteristic length of the m® term is determined by the coefficient sp, which, by an analysis similar to (17.95), has dimensionality 2d6 {s,] = L Thus, for mean-field theory to te good we require pre << | s/G, which means So rot3 << | Since ry~0 al the tricritical point, this can be satisfied only if d>3. Thus the critical dimension associated with a tricritical point is d.-3. 17-4Pod Nt wle s|y ° ad -So/Zu, > *T | 31 |x. i oy PROBLEM 17.3 17-518.1 In the RG transformation (18.15) the constant Nu is deternimed by the condition Des) = 1 {s) which leads to the explicit solution 2 log DPfs's) e®° = F(K,) (s} {s} The point is that Ny depends only on the parameters K, and not on N. Making an RG transformation according to as: 25) give the dlock-spin energy, which we can write in the form E{s'} = G{s'} + F{K,} where Gf{s'} = - log >Pis's) exo(-Sk,5,) (s) o to extract K's, we use (18.18): Ka = 2" S SaGls'} + F(K,) (s') The first term does not afepend on N’, the number of block spins, because of the factor 2. That makes it an average over the biocks spin configurations. Therefore K's is a function of the K.’s only, and itt does not matter how many RG transformations had been carried out before this one. 18.2 (a) In the lattice of sketch (b), let the spins on circled sites be denoted by o, and the others by s. Each o has two and only two nearest neighbors of type s. The sum over the o’s can be carrried out one-by-one independently. For a given o, call its nearest neighbors s, and s,. Then we try to define a renormalized interaction between s, and s, through ese des? = 2coshl2J(s, +53} 0 By making a table of both sides for all possible values of (s,s) and matching them, we find ev = jee +e) The spins s, and s, become nearest neighbors on the lattice in sketch Cc). Thus the derivation above hoids for every nearest neighbor on the new lattice. Hence the stated result follows. 18-1(b) Let £02) = bx), Repeated iteration of this function, ie, f(fCx)), FCICFGX))), etc. can be done geometrically as shown in the accompanying sketch. Consider starting from the three different values of x labeled abc on the x axis. The value b leads to the new fixed point x’, determined by fc) = x" - with solution x" 1.95. It is clearly unstable, as can be seen graaphically. Since x’ = f(x), we have wx =f) - 16°) > Gx) £OD A short calculation shows f(x") = xt-G")°, (c) In term of the temperature, x = exp(1/T). (We ignore all constnats.) The fixed point x" corresponds to a critical temperature T°. Near the fixed point, write T = T'+6T. We have x = el/T » el/T(1-8T/T). Hence oT = fOryeT = ZT v= 1/D, = log [@x"V/log 2 18.3 The model is symmetric under interchange of m, and m,. When one of the m, vanishes, say m,-0, the model reduces to the Landau-Wilson model. When g. =28 there is rotational symmetry in m,-m, space, and the model reduces to the X-Y model. Put m, = mcos 8, m,=msin®, y=m,/m,. Then Elm] = |cen{}one + ine? + 3 m+ g,m'[1+(y-2Xsin20)"/4}} The symmetry between m, and m, becomes the invariant under rotation of @ through #/2, The system is unstable for y<-2, because E{m] is not bounded form below. (a) For mean-field theory assume m and 6 are both independent of x. Then the Landau free energy is & = Arm? + g,an*Lly-2Xsin20)"/4] The relevant derivatives are ay/am = rom+ 4g,m?[1+(y-2 sin26)°/4) a’y/am? = x + 12g,mft«y-2Xsin20)'/4] aysae = 4 m*ty-2) Sintae) a°p/ae" = 2g,m*Cy-2) cos(49) 18-2For ry the only minimum of y ts m=0. For r,<0 a minimum occurs at m>0, and we have spontaneous symmetry breaking. In this Jf y>2, then we have either 6=0 or G=7/2. This means either mem, m,=0, or m=O, m=m If -2 dx _ 2€(1-x) (6+y) dt > FG y? dy €yly-2) 9-6) G7 36+y? Near a fixed point, put xe et Ox, ys yt ay Then a ox . ax _ e rt] dr (3) (y TT Lboc where = of - ev A= afi Gbry"*) Bp - —2.f;- 2y"(6+y") 364y"? 364y"? C= er (Cy'-2XLy"-6) + yCy"-6) + y'(y-2)] We only need to look into the cases y*,6, because the other cases are familiar: y" A B C Eigenvalues 2 2-2e/5 e€/100 9 -€/5 2-€/5, €/5 6 2-2€/3 -€/360 -€/3 1.55, 0.45 Thus. the fixed-point at y*-6 is unstable in both directions, while that at y’ is unstable in one direction and stable in the other. The crilical index v for y"=2 is __t in, e Y= 3475 = 2 I? 18-4o cb a t x* PROBLEM 18.2 185m=0 Line of first-order trams ition transition lL SYSTEM -2 UNSTABLE PROBLEM 18.3