PHYSICS 140B : STATISTICAL PHYSICS HW ASSIGNMENT #5 SOLUTIONS (1) Consider the S = 1 Ising Hamiltonian Ĥ = −J X Siz Sjz − H X Siz + ∆ i hiji X Siz i 2 , where Siz ∈ {−1, 0, +1}. (a) Making the mean field approximation in the first term, but treating the other terms exactly, find the corresponding mean field Hamiltonian ĤMF . (b) Defining θ = kB T /zJ, h = H/zJ, and δ = ∆/zJ, find the free energy f = F/N zJ as a function of θ, δ, h, and m. (c) Write the mean field equation for this model. Solution : (a) The mean field Hamiltonian is ĤMF = 21 N zJm2 − (H + zJm) X Siz + ∆ i X Siz i 2 . The mean field free energy is then " FMF = 21 N zJm2 − N kB T ln 1 + 2 e−∆/kB T H + zJm cosh kB T # . (b) The scaled free energy is " m+h f = 21 m2 − θ ln 1 + 2 e−δ/θ cosh θ # . (c) The mean field equation is ∂f =m− ∂m sinh 1 2 m+h θ exp(δ/θ) + cosh m+h θ =0. Sjz , where the individual spin polarizations take values Siz ∈ {−S, . . . , +S}. Find an expression for the mean field value of Tc . Hint: Expand the free energy for small m = hSiz i, and find what value of T makes the coefficient of the quadratic term vanish. (2) Consider a spin-S Ising model. The Hamiltonian is Ĥ = −J 1 P z hiji Si Solution : The mean field Hamiltonian is ĤMF = −zJm FMF = 2 1 2 N zJm P i Siz , hence the mean field free energy is − N kB T ln S X ejzJm/kB T . j=−S Note that S X ju e Su =e −ju e Su =e j=0 j=−S Thus, 2S X 1 − e−(2S+1)u 1 − e−u sinh (S + 12 )u . = sinh( 21 u) f = 12 m2 − θ ln sinh (2S + 1) m/2θ + θ ln sinh(m/2θ) . The mean field equation is ∂f = m − (S + 12 ) ctnh (S + 21 ) m/θ + ∂m Now use the Laurent expansion for ctnh (u) = m = 31 S(S + 1) 1 u + u 3 − 1 2 u3 45 ctnh (m/2θ) . + . . . to obtain m + O(m3 ) , θ from which we identify θc = 31 S(S + 1), or TcMF = S(S + 1)zJ/3kB . (3) Consider the O(2) model, Ĥ = − 12 X Jij n̂i · n̂j − H · X n̂i , i i,j where n̂i = cos φi x̂ + sin φi ŷ. Consider the case of infinite range interactions, where Jij = J/N for all i, j, where N is the total number of sites. (a) Show that " # Z P βJ X N βJ 2 exp n̂i · n̂j = d2 m e−N βJm /2 eβJm· i n̂i . 2N 2π i,j (b) Using the definition of the modified Bessel function I0 (z), Z2π dφ z cos φ e , I0 (z) = 2π 0 show that −β Ĥ Z = Tr e = Z d2 m e−N A(m,h)/θ , where θ = kB T /J and h = H/J. Find an expression for A(m, h). 2 (c) Find the equation which extremizes A(m, h) as a function of m. (d) Look up the properties of I0 (z) and write down the first few terms in the Taylor expansion of A(m, h) for small m and h. Solve for θc . Solution : (a) We have J Ĥ =− kB T 2N kB T Therefore −Ĥ/kB T e X i n̂i 2 − H X · n̂i . kB T i " # X 2 1 h X = exp n̂i + · n̂i 2N θ θ i i " X # Z N m+h N m2 2 = + d m exp − · n̂i . 2πθ 2θ θ i (b) Integrating the previous expression, we have Y Z dn̂ −Ĥ/kB T i −Ĥ[{n̂i }]/kB T Z = Tr e = e 2π i Z h iN N 2 = d2 m e−N m /2θ I0 |m + h|/θ . 2πθ Thus, we identify θ ln(N/2πθ) . A(m, h) = 12 m2 − θ ln I0 |m + h|/θ − N (c) Extremizing with respect to the vector m, we have ∂A m + h I1 |m + h|/θ =0, =m− · ∂m |m + h| I0 |m + h|/θ where I1 (z) = I0′ (z). Clearly any solution requires that m and h be colinear, hence I1 (m + h)/θ . m= I0 (m + h)/θ (d) To find θc , we first set h = 0. We then must solve m= I1 (m/θ) . I0 (m/θ) 3 The modified Bessel function Iν (z) has the expansion Iν (z) = 1 ν 2z ∞ X k=0 1 2 k 4z k! Γ(k + ν + 1) . Thus, I0 (z) = 1 + 41 z 2 + . . . I1 (z) = 21 z + and therefore I1 (z)/I0 (z) = 21 z − 1 3 16 z 1 3 16 z + ... , + O(z 5 ), and we read off θc = 12 . 4