Home / Courses / GATE Physics / Practice / The Ising model and Landau theory (NET)

The Ising model and Landau theory (NET)

Statistical physics · 50 questions and templates for GATE Physics, in order from warm-up to top rank. Check each answer, then read the full solution.

Loading questions… (they need JavaScript; nothing you type is sent anywhere)

Written questions with solutions

The same questions as a list, so you can print them or read without the practice box. Templates (fresh numbers) appear only in the practice box above.

  1. NATWarm-up

    A ferromagnetic Ising model H=−J∑⟨ij⟩sisj (J>0) lives on a square lattice of N sites with periodic boundaries. Its ground-state energy per spin is ______ J.
    Answer and solution

    Answer: -2

    Each site has z=4 neighbours, so there are zN/2=2N bonds. All spins aligned make every bond contribute −J: E0=−2NJ, i.e. −2J per spin. The ground state is twofold degenerate (all up or all down).

    Why people get this wrong: Counting z=4 bonds per spin double counts each bond.

  2. MCQWarm-up

    In zero field, which Ising systems with short-range ferromagnetic coupling show a spontaneous magnetisation at some T>0?
    1. Only the one-dimensional chain
    2. The square lattice (2D) and the cubic lattice (3D), but not the 1D chain
    3. All of them, in every dimension
    4. None: spontaneous magnetisation needs a field
    Answer and solution

    Answer: B

    In 1D a single domain wall costs only 2J but gains entropy kln⁡N, so order is destroyed at any T>0. In 2D a wall around a domain costs energy proportional to its length, which beats the entropy at low T (Peierls): there is a transition, at kTc=2.269J on the square lattice (Onsager). The lower critical dimension of the Ising model is 1.
  3. NATWarm-up

    Mean-field theory predicts kTc=zJ for the Ising model. For a face-centred cubic lattice, kTc/J is ______.
    Answer and solution

    Answer: 12

    Each spin feels the average field zJ⟨s⟩ of its z neighbours; m=tanh⁡(zJm/kT) has a nonzero solution for kT<zJ. The fcc lattice has z=12, so kTc=12J in mean field (the true value is about 9.8J).
  4. MCQWarm-up

    Landau's free energy for an Ising-like transition, F=F0+a(T−Tc)m2+bm4−hm, contains no m3 term because
    1. the free energy must be symmetric under m→−m when h=0
    2. cubic terms are always small
    3. it would make F unbounded
    4. Landau theory only works in 4 dimensions
    Answer and solution

    Answer: A

    The Ising Hamiltonian is invariant under flipping all spins when h=0, so F(m)=F(−m) and only even powers appear. Where symmetry allows a cubic term (e.g. the nematic-isotropic or three-state Potts transition), the transition becomes first order.
  5. MCQWarm-up

    With ni=(1+si)/2 (ni=0 or 1), the Ising model maps onto a lattice gas. The ferromagnetic coupling and the magnetic field correspond to
    1. repulsion between atoms; temperature
    2. attraction between neighbouring atoms; chemical potential
    3. hard-core exclusion only; pressure
    4. attraction; volume
    Answer and solution

    Answer: B

    Substituting s=2n−1 into −J∑sisj−h∑si gives −4J∑ninj−(2h−2zJ)∑ni+ const. A ferromagnet (J>0) is a gas with nearest-neighbour attraction 4J, and the field plays the role of the chemical potential. The Ising critical point is a liquid-gas critical point, which is why they share exponents.
  6. NATWarm-up

    How many microstates does an Ising model of 3×3 spins have?
    Answer and solution

    Answer: 512

    Each of the 9 spins is ±1 independently: 29=512 configurations. For 10×10 there are 2100≈1030, which is why Monte Carlo sampling is needed instead of summing every state.
  7. MCQWarm-up

    An Ising ferromagnet H=−J∑⟨ij⟩sisj (J>0, each pair of neighbours counted once) has N spins on a square lattice with periodic boundaries. Its ground-state energy is
    1. −NJ
    2. −2NJ
    3. −4NJ
    4. −NJ/2
    Answer and solution

    Answer: B

    In the ground state all spins are parallel, so every bond contributes −J. Each spin has z=4 neighbours, and each bond is shared by two spins, so there are Nz/2=2N bonds. E0=−2NJ.

    Why people get this wrong: −4NJ counts each bond twice, once from each end.

  8. NATStandard

    Two Ising spins interact through H=−Js1s2. At kT=J, the probability that they are parallel is ______ (to three decimal places).
    Answer and solution

    Answer: 0.881 (±0.001)

    Four states: two parallel (energy −J) and two antiparallel (+J). Z=2eK+2e−K with K=J/kT=1. P∥=2e2e+2e−1=11+e−2=0.881. Equivalently ⟨s1s2⟩=tanh⁡K=0.762=2P∥−1.
  9. NATStandard

    In the zero-field Ising chain, ⟨sisi+r⟩=(tanh⁡K)r with K=J/kT. At K=0.50, ⟨s0s3⟩ is ______ (to three decimal places).
    Answer and solution

    Answer: 0.099 (±0.001)

    With bond variables τi=sisi+1 (independent in zero field, each with mean tanh⁡K), s0s3=τ0τ1τ2, so the correlation is tanh3⁡K=0.46213=0.0987. It decays exponentially at any T>0: no long-range order.
  10. MCQStandard

    Three Ising spins on a triangle have antiferromagnetic coupling H=J(s1s2+s2s3+s3s1), J>0. The ground-state energy and its degeneracy are
    1. −3J, 2 states
    2. −J, 2 states
    3. −3J, 6 states
    4. −J, 6 states
    Answer and solution

    Answer: D

    Not all three bonds can be antiparallel (frustration). The best is two antiparallel and one parallel: E=−J−J+J=−J. Only all-equal states (2 of 8) have E=+3J, so the other 6 states all have E=−J. Frustration leaves a large ground-state degeneracy; on the triangular lattice it gives residual entropy 0.323k per spin.
  11. NATStandard

    In a Metropolis simulation of the square-lattice Ising model at kT=2.0J, a proposed flip raises the energy by 8J. The probability that it is accepted is ______ (to four decimal places).
    Answer and solution

    Answer: 0.0183 (±0.0002)

    Metropolis accepts with probability min⁡(1,e−ΔE/kT). Here e−8J/2J=e−4=0.0183. Flipping a spin si changes the energy by ΔE=2Jsi∑jsj∈{−8J,−4J,0,4J,8J}; 8J is a spin aligned with all four neighbours.
  12. MCQStandard

    Why does the Metropolis rule Pacc=min⁡(1,e−ΔE/kT) sample the Boltzmann distribution?
    1. It satisfies detailed balance with respect to e−E/kT, and the moves are ergodic
    2. It always lowers the energy
    3. It visits every state equally often
    4. It is exact only at T=0
    Answer and solution

    Answer: A

    For two states with E2>E1: P(1→2)/P(2→1)=e−(E2−E1)/kT/1, which equals p2/p1 for Boltzmann weights. Detailed balance makes the Boltzmann distribution stationary, and ergodicity (any state reachable) makes it the unique limit.
  13. NATStandard

    The Bethe (cluster) approximation gives Kc=12ln⁡zz−2. For the square lattice (z=4), the predicted kTc/J is ______ (to three decimal places).
    Answer and solution

    Answer: 2.885 (±0.002)

    Kc=12ln⁡2=0.3466, so kTc/J=1/Kc=2.885. Compared with mean field (4) and the exact Onsager value (2.269), Bethe is better because it treats one spin and its neighbours exactly. It still gives mean-field exponents.
  14. MSQStandard

    Which statements about mean-field theory of the Ising model are correct?
    1. It always overestimates Tc, because it neglects fluctuations
    2. It wrongly predicts a transition for the 1D chain (at kTc=2J)
    3. It becomes exact for long-range interactions or above four dimensions
    4. It gives the exact 2D exponent β=1/8
    Answer and solution

    Answer: A, B, C

    Fluctuations disorder the system, so ignoring them raises Tc; mean field even predicts order in 1D. For infinite-range coupling or d>4 (the upper critical dimension) mean-field exponents are exact. Mean field gives β=1/2; 1/8 is Onsager-Yang.
  15. NATStandard

    In a 1D Ising chain of N spins, one domain wall costs 2J and can sit in about N places. With J/kT=5, the free energy of creating a wall becomes negative for N larger than about ______ (to the nearest whole number).
    Answer and solution

    Answer: 22026 (±2)

    ΔF=2J−kTln⁡N<0 when N>e2J/kT=e10=22026. A long enough chain always breaks into domains: no order at T>0. In 2D a domain wall of length L costs 2JL while its entropy also grows only ∝L, so order survives at low T.
  16. MCQStandard

    The critical exponents of the 2D Ising model are
    1. β=1/2, γ=1, ν=1/2, α=0 (jump)
    2. β=0.326, γ=1.237, ν=0.630
    3. β=1/8, γ=7/4, ν=1, α=0 (logarithmic)
    4. β=1/4, γ=1, ν=1/2
    Answer and solution

    Answer: C

    Onsager (1944) and Yang (1952): C∼ln⁡|t|, m∼(−t)1/8, χ∼|t|−7/4, ξ∼|t|−1, η=1/4, δ=15. The first option is mean field, the second 3D Ising, the last the tricritical mean-field values.
  17. NATStandard

    In Landau theory F=a(T−Tc)m2+bm4−hm. The ratio of the susceptibility above Tc to that below Tc, at equal |T−Tc|, is ______.
    Answer and solution

    Answer: 2

    Above: m0=0 and 2a(T−Tc)m=h, so χ+=12a(T−Tc). Below: m02=a(Tc−T)2b and F′′=−2a(Tc−T)+12bm02=4a(Tc−T), so χ−=14a(Tc−T). Ratio χ+/χ−=2, a universal mean-field amplitude ratio (about 4.4 in 3D Ising, 37.7 in 2D).
  18. MCQStandard

    For the liquid-gas transition, viewed as a lattice gas (Ising) transition, the natural order parameter is
    1. the pressure
    2. the temperature
    3. the density difference ρliquid−ρgas
    4. the latent heat
    Answer and solution

    Answer: C

    In lattice-gas language the magnetisation m=2n−1 maps to the density. Along the coexistence curve ρl−ρg∝(Tc−T)β with β≈0.326, the 3D Ising value.
  19. NATStandard

    In the fully aligned ground state of the simple cubic Ising ferromagnet (z=6, coupling J, no field), flipping one spin raises the energy by ______ J.
    Answer and solution

    Answer: 12

    The flipped spin now disagrees with all 6 neighbours. Each of those bonds goes from −J to +J, a change of 2J, so the cost is 6×2J=12J. At low temperature such flips are suppressed by e−12J/kT, which is why the magnetisation stays very close to saturation there.

    Why people get this wrong: Each broken bond costs 2J (from −J to +J), not J.

  20. MCQStandard

    A 10×10 square lattice with periodic boundaries has how many nearest-neighbour bonds?
    1. 100
    2. 400
    3. 50
    4. 200
    Answer and solution

    Answer: D

    Each of the 100 sites has 4 neighbours, and each bond joins two sites: 100×4/2=200. Equivalently, one horizontal and one vertical bond per site.

    Why people get this wrong: 400 counts every bond twice.

  21. NATExam level

    Two Ising spins have H=−Js1s2−h(s1+s2) with s=±1. At kT=J and h=J, the mean value ⟨s1⟩ is ______ (three decimal places).
    Answer and solution

    Answer: 0.931 (±0.002)

    Energies: (+,+): −J−2h=−3J; (−,−): −J+2h=+J; (+,−) and (−,+): +J. Weights at kT=J: e3=20.086 and e−1=0.368 (three states). Z=20.086+1.104=21.189. ⟨s1⟩=[e3−e−1+e−1−e−1]/Z=(20.086−0.368)/21.189=0.931.

    Why people get this wrong: The mixed states contribute +1 and −1 to s1 and cancel; only the two aligned states matter for ⟨s1⟩.

  22. NATExam level

    Three Ising spins on a ring have H=−J(s1s2+s2s3+s3s1) with J>0. At kT=J, the mean energy is ______ J (two decimal places).
    Answer and solution

    Answer: -2.79 (±0.01)

    All aligned (2 states): every bond satisfied, E=−3J. Otherwise (6 states) exactly one bond is satisfied and two are broken: E=−J+2J=+J. Z=2e3+6e−1=40.17+2.21=42.38. ⟨E⟩=[2(−3)e3+6(+1)e−1]/Z=(−120.51+2.21)/42.38=−2.79J.

    Why people get this wrong: On a ring of three, a single flipped spin breaks two bonds, not one.

  23. MCQExam level

    In the Bragg-Williams (mean-field) treatment of the Ising model, the free energy per spin is f(m)=−12zJm2−kTs(m) with s(m)=ln⁡2−12[(1+m)ln⁡(1+m)+(1−m)ln⁡(1−m)]. Expanded in powers of m, the coefficient of m4 is
    1. −kT/12
    2. zJ/12
    3. kT/12
    4. kT/4
    Answer and solution

    Answer: C

    Expand: (1+m)ln⁡(1+m)+(1−m)ln⁡(1−m)=m2+m46+…, so s(m)=ln⁡2−m22−m412. Then f=−kTln⁡2+12(kT−zJ)m2+kT12m4+… The m2 coefficient changes sign at kTc=zJ, and the m4 coefficient is positive, so the transition is continuous: a Landau free energy derived from the lattice model.

    Why people get this wrong: The quartic term comes from the entropy, so it carries kT, with a positive sign since −kTs is added.

  24. NATExam level

    The exact critical temperature of the honeycomb-lattice Ising ferromagnet (z=3) is kTc/J=2/ln⁡(2+3). Mean-field theory gives kTc=zJ. The mean-field value is ______ times the exact one (three decimal places).
    Answer and solution

    Answer: 1.975 (±0.003)

    Exact: 2/ln⁡3.732=2/1.317=1.519. Mean field: 3. Ratio =3/1.519=1.975. The error is largest for lattices with few neighbours, where the local field fluctuates most; for the face-centred cubic lattice (z=12) mean field is only about 23% too high.

    Why people get this wrong: Using log10⁡ in the exact formula gives a ratio near 0.86, less than 1, which mean-field theory never gives.

  25. MSQExam level

    Below Tc, Landau-Ginzburg theory F=∫[c(∇m)2+a(T−Tc)m2+bm4]dx allows a flat domain wall between regions with m=−m0 and m=+m0. Which statements are correct?
    1. The wall's thickness is of the order of the correlation length ξ.
    2. The wall grows thicker as T→Tc from below.
    3. The wall's free energy per unit area falls to zero as T→Tc.
    4. At the centre of the wall, m=m0/2.
    Answer and solution

    Answer: A, B, C

    Minimising gives m(x)=m0tanh⁡(x/w) with w=2c/a(Tc−T), twice the correlation length below Tc, so the wall widens as ξ∝(Tc−T)−1/2 diverges. Its energy per area scales as cm02/ξ∝(Tc−T)3/2, vanishing at Tc, which is why domains form and vanish freely near the transition. By symmetry m=0 at the centre.

    Why people get this wrong: The profile is odd about the centre: tanh⁡(0)=0, not 12.

  26. MSQExam level

    For a ferromagnetic Ising system in a uniform field h, which statements about its phase diagram in the (h,T) plane are correct?
    1. Below Tc, sweeping h through zero causes a first-order transition.
    2. On that line the magnetisation jumps from −m0 to +m0.
    3. The line of first-order transitions ends at the critical point h=0, T=Tc.
    4. Above Tc, the magnetisation also jumps when h passes through zero.
    Answer and solution

    Answer: A, B, C

    Below Tc the two ordered states are degenerate at h=0; any field picks one, so m jumps from −m0 to +m0: a first-order line along h=0. As T→Tc the jump 2m0 shrinks to zero, ending the line at the critical point. Above Tc, m=χh varies smoothly through zero. This is the same picture as the liquid-gas coexistence line ending at its critical point.

    Why people get this wrong: Above Tc there is no spontaneous magnetisation, so nothing jumps.

  27. MCQHard

    The square-lattice Ising antiferromagnet H=+J∑⟨ij⟩sisj (J>0) in zero field orders below a Néel temperature of
    1. zero: it never orders
    2. kTN=1.134J
    3. kTN=2.269J
    4. kTN=4J
    Answer and solution

    Answer: C

    The square lattice is bipartite: flipping every spin on one sublattice (si→−si) turns each bond term +Jsisj into −Jsisj and maps the antiferromagnet onto the ferromagnet with the same partition function in zero field. So kTN=2.269J, Onsager's value, with staggered magnetisation as the order parameter. On the non-bipartite triangular lattice the antiferromagnet is frustrated and does not order.

    Why people get this wrong: The mapping needs zero field and a bipartite lattice; a uniform field becomes a staggered one under the flip.

  28. MSQHard

    Which statements about Monte Carlo simulation of the Ising model are true?
    1. The Metropolis rule satisfies detailed balance with respect to the Boltzmann distribution
    2. Near Tc, single-spin-flip updates suffer critical slowing down: correlation times grow as a power of the system size
    3. Cluster algorithms (Swendsen-Wang, Wolff) greatly reduce critical slowing down
    4. On a finite lattice the simulated heat capacity truly diverges at Tc
    Answer and solution

    Answer: A, B, C

    Metropolis: P(a→b)/P(b→a)=e−(Eb−Ea)/kT, detailed balance. Near Tc large correlated regions must be turned spin by spin, so τ∼Lz with z≈2. Flipping whole clusters at once brings z well below 2. In a finite system Z is a finite sum of exponentials, analytic in T: the heat capacity has a rounded peak whose height grows with L, but no true divergence.

    Why people get this wrong: True singularities need the thermodynamic limit; finite-size data show peaks that sharpen with size.

  29. NATHard

    A system has Landau free energy density F=a(T−Tc)m2+bm4 with a=0.50 J m−3 K−1, b=0.25 J m−3 and Tc=100 K. The jump in its heat capacity per unit volume on cooling through Tc is ______ J m−3 K−1.
    Answer and solution

    Answer: 50 J m⁻³ K⁻¹ (±0.1)

    Below Tc: m2=a(Tc−T)2b and F=−a2(Tc−T)24b. Then C=−T∂2F∂T2=a2T2b, which at Tc is 0.25×1000.5=50 J m−3 K−1. Above Tc, m=0 and this part of C is zero: a finite jump, α=0 in mean-field theory.

    Why people get this wrong: Forgetting the factor T in C=−T∂2F/∂T2 gives 0.5 instead of 50.

  30. MSQHard

    Which statements about Landau theory are true?
    1. It assumes the free energy can be expanded as a power series in the order parameter near the transition
    2. If the system is symmetric under m→−m, only even powers appear
    3. An allowed cubic term generally makes the transition first order
    4. It treats fluctuations correctly in three dimensions, so its exponents are exact there
    Answer and solution

    Answer: A, B, C

    Landau theory expands F(m) analytically and keeps the terms the symmetry allows. With a cubic term, F develops a second minimum away from m=0 before the quadratic coefficient vanishes, so m jumps (as in liquid-crystal and some structural transitions). It is a mean-field theory: it neglects fluctuations, which dominate near Tc below four dimensions, so in 3D its exponents (β=12, γ=1) are not exact.

    Why people get this wrong: Landau theory is qualitatively right in 3D but quantitatively wrong near Tc; it is exact only above the upper critical dimension.

← All topics · Timed mock test →