Skip to content
Tutorials Center
My learning

Topics

Lesson 29 of 39 · Phase Transitions and the Ising Model

Mean-Field Theory of the Ising Model

Derive the mean-field equation m = tanh(β(zJm + h)) for the Ising model, find the critical temperature and exponents, and compare with exact results in different dimensions.

Exact solutions of the Ising model are rare. Mean-field theory is the first approximation almost every statistical physicist reaches for: replace the messy interaction of a spin with its neighbors by an interaction with the average spin. It is easy, it predicts a phase transition, and it gets the qualitative picture right. Its errors also teach you when fluctuations matter.

The mean-field idea

Consider one spin s_i in a lattice where every site has z nearest neighbors (z = 2 for a chain, 4 for a square lattice, 6 for a simple cubic lattice). Its energy contribution is

Text
-s_i ( J Σ_{j neighbor} s_j + h )

Replace each neighbor s_j by its average value m = ⟨s⟩. The spin then feels an effective field h_eff = h + zJ m, and its own thermal average is that of a free spin in a field:

Text
m = tanh( β (h + zJ m) )

This is the self-consistency equation: m appears on both sides. The picture behind it is that each spin is surrounded by a "mean field" produced by all the others, and the ordering is a collective effect that sustains itself.

Spontaneous magnetization and the critical temperature

Set h = 0. Then m = tanh(βzJ m) always has the solution m = 0. A nonzero solution exists only if the slope of the right side at m = 0 exceeds 1, i.e. βzJ > 1. This defines the mean-field critical temperature

Text
k T_c = z J

For T > T_c only m = 0 exists (paramagnet). For T < T_c two symmetric solutions ±m₀(T) appear and are stable: spontaneous magnetization. Near T_c, expanding tanh x ≈ x - x³/3 gives

Text
m ≈ √3 · (1 - T/T_c)^(1/2)

The critical exponent β = 1/2 is the power with which m vanishes: m ∝ (T_c - T)^β. Above T_c, a small field produces m = h / (k (T - T_c)), so the susceptibility diverges as (T - T_c)^(-γ) with γ = 1.

Do not confuse the critical exponent β with the inverse temperature β = 1/kT. Both symbols are standard, so context tells you which is meant.

Solving the equation numerically

The code solves m = tanh(m T_c/T) by fixed-point iteration (starting from m = 1, which converges to the positive root), plots m(T), and tests the exponent by fitting a power law close to T_c. It also checks the susceptibility above T_c with a finite difference. Units: J = 1, k = 1, z = 4.

Python Mean-field magnetization, exponent and susceptibility

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The fitted exponent is very close to 0.5, and the susceptibility matches 1/(T - T_c).

How good is mean-field?

Mean-field theory ignores correlations: it assumes each spin sees the average of its neighbors, not their fluctuating values. This is more accurate when each spin has many neighbors (large z, or high spatial dimension), and worse in low dimensions. Compare its prediction kT_c = zJ with exact or numerical results:

Lattice z Mean-field kT_c/J Accurate kT_c/J
1D chain 2 2 0 (no transition)
2D honeycomb 3 3 1.5187
2D square 4 4 2.2692
2D triangular 6 6 3.6410
3D simple cubic 6 6 4.5115 (numerical)

Mean-field theory overestimates T_c, always, because ignoring fluctuations makes order too easy. In dimension d ≥ 4 its exponents become exact (the "upper critical dimension" for the Ising universality class is 4). Below, true exponents differ: for the 2D Ising model β = 1/8 and for the 3D Ising model β ≈ 0.326, compared with mean-field 1/2.

Python Mean-field versus accurate transition temperatures

Press Run (or Ctrl + Enter) to see the result here.

Program input · what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The ratio shrinks toward 1 as the number of neighbors grows. In 2D, mean-field overestimates T_c by roughly 65 to 100 percent; in 3D, by about 33 percent.

Why bother?

Mean-field theory is not just historical. The same logic (a self-consistent field) underlies the Weiss theory of magnetism, the Hartree-Fock approximation, the BCS theory of superconductivity and the van der Waals equation of state for fluids. It also motivates the more general Landau theory that closes this chapter.

Check your understanding

Quick check · 4 questions

of 4 answered You scored / 4 · perfect!

  1. 1. In mean-field theory for the Ising model, a spin interacts with

    Each neighbor spin is replaced by m, giving the effective field h + zJm.
  2. 2. The mean-field critical temperature for coordination number z is

    A nonzero solution of m = tanh(βzJm) requires βzJ > 1.
  3. 3. The mean-field exponent β describing m ∝ (T_c - T)^β is

    Expanding tanh near T_c gives m ∝ (1 - T/T_c)^(1/2).
  4. 4. Mean-field theory is most accurate when

    More neighbors average out fluctuations; in d ≥ 4 the Ising exponents are exactly mean-field.

Summary

  • Mean-field: m = tanh(β(zJm + h)).
  • kT_c = zJ; spontaneous magnetization with exponent β = 1/2; susceptibility 1/(k(T - T_c)) above T_c.
  • It overestimates T_c, predicts an order in 1D that does not exist, and gives wrong exponents below 4 dimensions.

Next: The 2D Ising Model: Onsager's Exact Results