Part IV: Advanced Topics | Chapter 2

Universality

Upper critical dimension, epsilon expansion, Wilson-Fisher fixed point, scaling functions, and the O(n) model

Historical Context

Kenneth Wilson and Michael Fisher developed the epsilon expansion in 1972, providing the first systematic method to compute critical exponents beyond mean-field theory. Their key insight was to treat the spatial dimension \(d\) as a continuous variable and expand around the upper critical dimension \(d_c = 4\) in powers of \(\epsilon = 4 - d\). The resulting Wilson-Fisher fixed point gave exponents in remarkable agreement with experiments and numerical simulations.

The epsilon expansion unified decades of experimental observations and proved that universality was not merely empirical but followed from the structure of the renormalization group. The O(n) model framework showed how different universality classes emerge from the symmetry of the order parameter, and the concept of the upper critical dimension clarified when and why mean-field theory succeeds.

1. The Ginzburg-Landau-Wilson Functional

Derivation 1: From Lattice to Continuum

The partition function for the \(O(n)\) model in the continuum limit is:

\[Z = \int \mathcal{D}\boldsymbol{\phi}\,\exp\left(-\frac{\mathcal{H}[\boldsymbol{\phi}]}{k_BT}\right)\]

where \(\boldsymbol{\phi}(\mathbf{x})\) is an \(n\)-component field and:

\[\mathcal{H}[\boldsymbol{\phi}] = \int d^d x\left[\frac{1}{2}(\nabla\boldsymbol{\phi})^2 + \frac{r_0}{2}\boldsymbol{\phi}^2 + \frac{u_0}{4!}(\boldsymbol{\phi}^2)^2 - \mathbf{h}\cdot\boldsymbol{\phi}\right]\]

Here \(r_0 \propto T - T_c^0\) (bare mass), \(u_0 > 0\) (bare coupling), and \(\mathbf{h}\) is the external field. The cases \(n = 1\) (Ising),\(n = 2\) (XY), \(n = 3\) (Heisenberg) correspond to different universality classes.

The gradient term \((\nabla\phi)^2\) penalizes spatial variations -- it encodes the rigidity (stiffness) of the order parameter field. Higher-order gradient terms (\(\nabla^4\phi\), etc.) are irrelevant near \(d = 4\).

2. Upper Critical Dimension

Derivation 2: Dimensional Analysis of the Coupling

Under a rescaling \(\mathbf{x} \to \mathbf{x}/b\), the field scales as\(\phi \to b^{(d-2+\eta)/2}\phi\) (from the gradient term). At the Gaussian (free) fixed point (\(u_0 = 0\), \(\eta = 0\)):

\[[\phi] = \frac{d - 2}{2}, \qquad [r_0] = 2, \qquad [u_0] = 4 - d\]

The scaling dimension of \(u_0\) is \(\epsilon = 4 - d\):

\(d > 4\): \(u_0\) is irrelevant, Gaussian fixed point is stable, mean-field exponents are exact
\(d = 4\): \(u_0\) is marginal (logarithmic corrections)
\(d < 4\): \(u_0\) is relevant, Gaussian fixed point is unstable, need new fixed point

\[\boxed{d_c = 4 \quad \text{(upper critical dimension for short-range } O(n) \text{ models)}}\]

3. The Epsilon Expansion

Derivation 3: Wilson-Fisher Fixed Point

Set \(d = 4 - \epsilon\). The dimensionless coupling \(u = u_0\mu^{-\epsilon}\)(where \(\mu\) is a momentum scale) satisfies the beta function:

\[\beta(u) = \mu\frac{du}{d\mu} = -\epsilon u + \frac{(n+8)}{6}u^2 + O(u^3)\]

The fixed points are \(u^* = 0\) (Gaussian) and the Wilson-Fisher fixed point:

\[\boxed{u^* = \frac{6\epsilon}{n + 8} + O(\epsilon^2)}\]

The critical exponents to first order in \(\epsilon\) are:

\[\eta = \frac{(n+2)}{2(n+8)^2}\epsilon^2 + O(\epsilon^3)\]

\[\nu = \frac{1}{2} + \frac{(n+2)}{4(n+8)}\epsilon + O(\epsilon^2)\]

For the 3D Ising model (\(n = 1\), \(\epsilon = 1\)):\(\nu \approx 1/2 + 3/36 = 0.583\) (exact: 0.630). Higher-order epsilon expansions with Borel resummation give extremely accurate results.

4. Universal Scaling Functions

Derivation 4: Equation of State

The scaling hypothesis gives the equation of state in a universal form:

\[h = m^{\delta}\,f\left(\frac{t}{m^{1/\beta}}\right)\]

where \(f(x)\) is a universal scaling function. All systems in the same universality class share the same \(f(x)\) (after non-universal metric factors are fixed). This predicts data collapse: plotting\(h/m^{\delta}\) vs \(t/m^{1/\beta}\) should yield a single curve for all temperatures and fields near the critical point.

The correlation function also takes a universal scaling form:

\[G(r, t) = \frac{A}{r^{d-2+\eta}}\,\tilde{G}(r/\xi)\]

where \(\tilde{G}(x)\) is universal with \(\tilde{G}(0) = 1\) and\(\tilde{G}(x) \sim e^{-x}\) for \(x \to \infty\). Universality extends beyond exponents to entire functional forms.

5. Lower Critical Dimension and the Mermin-Wagner Theorem

Derivation 5: Goldstone Mode Fluctuations

For continuous symmetry (\(n \ge 2\)), the ordered phase has Goldstone modes (massless excitations corresponding to rotations of the order parameter). Their fluctuations can destroy long-range order. The mean-square fluctuation of the order parameter direction is:

\[\langle(\delta\theta)^2\rangle \sim \int \frac{d^dk}{k^2} \sim \begin{cases} \text{finite} & d > 2 \\ \ln L & d = 2 \\ L & d = 1 \end{cases}\]

For \(d \le 2\), fluctuations diverge with system size, destroying long-range order. This is the Mermin-Wagner theorem:

\[\boxed{d_l = \begin{cases} 1 & n = 1 \text{ (Ising, discrete symmetry)} \\ 2 & n \ge 2 \text{ (continuous symmetry)} \end{cases}}\]

The 2D XY model (\(n = 2\)) is marginal: it has no true long-range order but exhibits a Berezinskii-Kosterlitz-Thouless (BKT) transition with quasi-long-range order (algebraic correlations below \(T_{BKT}\)).

The Large-n Limit

For \(n \to \infty\), the O(n) model can be solved exactly in any dimension (the spherical model). The critical exponents are: \(\eta = 0\),\(\nu = 1/(d-2)\) for \(2 < d < 4\). This provides a useful check on the epsilon expansion and serves as a starting point for the \(1/n\)expansion.

6. Applications

Key Applications

Superfluid helium: The lambda transition belongs to the 3D XY universality class (\(n = 2\)). The specific heat exponent\(\alpha = -0.0127\) was measured in microgravity (Space Shuttle) to extraordinary precision, confirming the epsilon expansion predictions.

Conformal field theory: In \(d = 2\), critical points are described by conformal field theories. The conformal bootstrap provides exact results for the 3D Ising model exponents to unprecedented precision.

Quantum phase transitions: A quantum phase transition in \(d\)spatial dimensions maps to a classical transition in \(d + z\) dimensions (where \(z\) is the dynamical critical exponent).

Random systems: The Harris criterion states that quenched disorder is relevant if \(\alpha > 0\), changing the universality class.

7. Computational Exploration

This simulation explores epsilon expansion predictions, the Wilson-Fisher fixed point, universality across O(n) models, and data collapse.

Universality: Beta Function, Epsilon Expansion, O(n) Exponents, and Data Collapse

Python

script.py176 lines

import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt

# ============================================================
# Universality: Epsilon Expansion and Wilson-Fisher
# ============================================================

fig, axes = plt.subplots(2, 2, figsize=(12, 10))
fig.suptitle('Universality and the Epsilon Expansion', fontsize=16, fontweight='bold', color='white')
fig.patch.set_facecolor('#0a0a0a')

for ax in axes.flat:
    ax.set_facecolor('#1a0a0a')
    ax.tick_params(colors='white')
    ax.xaxis.label.set_color('white')
    ax.yaxis.label.set_color('white')
    ax.title.set_color('#fca5a5')
    for spine in ax.spines.values():
        spine.set_color('#333')

# --- Panel 1: Beta function and Wilson-Fisher fixed point ---
ax1 = axes[0, 0]
eps = 1.0  # d = 3

for n, color, label in [(1, '#f87171', 'n=1 (Ising)'),
                          (2, '#60a5fa', 'n=2 (XY)'),
                          (3, '#34d399', 'n=3 (Heisenberg)'),
                          (0, '#a78bfa', 'n=0 (SAW)')]:
    u_range = np.linspace(0, 2.0, 200)
    beta_u = -eps * u_range + (n + 8) / 6.0 * u_range**2
    u_star = 6 * eps / (n + 8)
    ax1.plot(u_range, beta_u, color=color, linewidth=2, label=label)
    ax1.plot(u_star, 0, 'o', color=color, markersize=8)

ax1.axhline(y=0, color='white', linestyle=':', alpha=0.3)
ax1.plot(0, 0, 'o', color='#fbbf24', markersize=8, label='Gaussian FP')
ax1.set_xlabel('Coupling u')
ax1.set_ylabel('beta(u)')
ax1.set_title('Beta Function (eps = 1, d = 3)')
ax1.legend(fontsize=7, facecolor='#1a0a0a', edgecolor='#333', labelcolor='white')

# --- Panel 2: nu(epsilon) for different n ---
ax2 = axes[0, 1]
eps_range = np.linspace(0, 2, 200)

# nu = 1/2 + (n+2)/(4(n+8)) * epsilon (first order)
exact_vals = {
    1: {"d3": 0.630, "d2": 1.0},
    2: {"d3": 0.672, "d2": None},  # BKT in 2D
    3: {"d3": 0.711, "d2": None},
}

for n, color, label in [(1, '#f87171', 'n=1 (Ising)'),
                          (2, '#60a5fa', 'n=2 (XY)'),
                          (3, '#34d399', 'n=3 (Heisenberg)')]:
    nu_eps = 0.5 + (n + 2) / (4.0 * (n + 8)) * eps_range
    ax2.plot(eps_range, nu_eps, color=color, linewidth=2, label=label + ' (O(eps))')

# Mark exact d=3 value
    if n in exact_vals and exact_vals[n]["d3"]:
        ax2.plot(1.0, exact_vals[n]["d3"], '*', color=color, markersize=12)

ax2.axvline(x=1.0, color='#fbbf24', linestyle=':', alpha=0.5, label='d = 3 (eps = 1)')
ax2.axhline(y=0.5, color='white', linestyle=':', alpha=0.3, label='Mean-field (nu = 1/2)')
ax2.set_xlabel('epsilon = 4 - d')
ax2.set_ylabel('nu')
ax2.set_title('nu(epsilon): Lines = O(eps), Stars = Exact')
ax2.legend(fontsize=7, facecolor='#1a0a0a', edgecolor='#333', labelcolor='white')

# --- Panel 3: Exponents vs n (at d=3) ---
ax3 = axes[1, 0]

n_vals = [0, 1, 2, 3, 4, 10, 100]
# Best known values (from conformal bootstrap, MC, epsilon expansion)
nu_best = [0.588, 0.630, 0.672, 0.711, 0.749, 0.888, 0.990]
beta_best = [0.302, 0.326, 0.348, 0.366, 0.383, 0.440, 0.495]
gamma_best = [1.157, 1.237, 1.317, 1.396, 1.448, 1.644, 1.960]

ax3.plot(n_vals, nu_best, 'o-', color='#f87171', linewidth=2, markersize=6, label='nu')
ax3.plot(n_vals, beta_best, 's-', color='#60a5fa', linewidth=2, markersize=6, label='beta')
ax3.plot(n_vals, gamma_best, '^-', color='#34d399', linewidth=2, markersize=6, label='gamma')

# Large-n limits
ax3.axhline(y=0.5, color='#f87171', linestyle=':', alpha=0.3)  # nu MF
ax3.axhline(y=0.5, color='#60a5fa', linestyle=':', alpha=0.3)  # beta MF
ax3.axhline(y=1.0, color='#34d399', linestyle=':', alpha=0.3)  # gamma MF

ax3.set_xlabel('n (order parameter components)')
ax3.set_ylabel('Critical exponent')
ax3.set_title('Exponents vs n (d = 3)')
ax3.legend(fontsize=8, facecolor='#1a0a0a', edgecolor='#333', labelcolor='white')

# --- Panel 4: Data collapse demonstration ---
ax4 = axes[1, 1]

# Generate "data" for a 3D Ising-like system
beta_exp = 0.326
gamma_exp = 1.237
delta_exp = (beta_exp + gamma_exp) / beta_exp  # from scaling

np.random.seed(42)
t_vals_data = np.array([-0.3, -0.2, -0.1, -0.05, 0.05, 0.1, 0.2, 0.3])
colors_data = plt.cm.coolwarm(np.linspace(0, 1, len(t_vals_data)))

for i, t in enumerate(t_vals_data):
    h_range = np.linspace(0.01, 2.0, 50)
    # m(h, t) from scaling: h = m^delta * f(t/m^{1/beta})
    # Simple model: m ~ (-t)^beta + chi*h where chi ~ |t|^{-gamma}
    if t < 0:
        m0 = (-t)**beta_exp
        chi = abs(t)**(-gamma_exp) * 0.1
    else:
        m0 = 0
        chi = t**(-gamma_exp) * 0.1
    m_vals = m0 + chi * h_range
    m_vals += 0.02 * np.random.randn(len(h_range))  # noise

# Scaled variables
    x_scaled = t / np.abs(m_vals)**(1.0/beta_exp)
    y_scaled = h_range / np.abs(m_vals)**delta_exp

mask = m_vals > 0.05  # avoid division issues
    ax4.plot(x_scaled[mask], y_scaled[mask], '.', color=colors_data[i],
             markersize=4, alpha=0.7, label='t=' + str(t) if i % 2 == 0 else None)

ax4.set_xlabel('t / |m|^(1/beta)')
ax4.set_ylabel('h / |m|^delta')
ax4.set_title('Data Collapse (Scaling Function)')
ax4.set_xlim(-30, 30)
ax4.set_ylim(0, 50)
ax4.legend(fontsize=7, facecolor='#1a0a0a', edgecolor='#333', labelcolor='white')

plt.tight_layout()
plt.savefig('output.png', dpi=150, bbox_inches='tight', facecolor='#0a0a0a')
plt.close()
print("Plot saved to output.png")

# Numerical results
print("\n" + "=" * 60)
print("UNIVERSALITY: NUMERICAL RESULTS")
print("=" * 60)

print("\n--- Wilson-Fisher Fixed Point u* = 6*eps/(n+8) ---")
for n in [0, 1, 2, 3]:
    u_star = 6.0 / (n + 8)
    print("  n = " + str(n) + ": u* = " + str(round(u_star, 4)) + " (at eps = 1)")

print("\n--- Epsilon Expansion: nu to O(epsilon) ---")
for n in [0, 1, 2, 3]:
    nu_1 = 0.5 + (n + 2) / (4.0 * (n + 8))
    print("  n = " + str(n) + ": nu(eps=1) = " + str(round(nu_1, 4)))

print("\n--- Best Known Exponents (d = 3) ---")
header = "  " + "n".ljust(6) + "nu".ljust(10) + "beta".ljust(10) + "gamma".ljust(10) + "eta"
print(header)
eta_best = [0.032, 0.036, 0.038, 0.037, 0.036, 0.025, 0.005]
for n, nu, beta, gamma, eta in zip(n_vals, nu_best, beta_best, gamma_best, eta_best):
    row = "  " + str(n).ljust(6) + str(nu).ljust(10) + str(beta).ljust(10) + str(gamma).ljust(10) + str(eta)
    print(row)

print("\n--- Critical Dimensions ---")
print("  Upper critical dimension: d_c = 4 (short-range O(n))")
print("  Lower critical dimension:")
print("    n = 1 (Ising): d_l = 1")
print("    n >= 2 (continuous): d_l = 2 (Mermin-Wagner)")
print("  Marginal case: d = 2, n = 2 -> BKT transition")

print("\n--- Large-n Limit (spherical model, d = 3) ---")
print("  nu = 1/(d-2) = 1.0")
print("  eta = 0")
print("  gamma = 2/(d-2) = 2.0")
print("  beta = 1/2")
print("\n" + "=" * 60)

Click Run to execute the Python code

Code will be executed with Python 3 on the server

8. Summary and Key Results

Core Formulas

Wilson-Fisher: \(u^* = 6\epsilon/(n+8)\)
\(\nu = 1/2 + (n+2)\epsilon/[4(n+8)]\)
Upper critical dim: \(d_c = 4\)
Scaling function: \(h = m^{\delta}f(t/m^{1/\beta})\)

Physical Insights

Universality class = (d, n, range of interactions)
Epsilon expansion: perturbative access to non-MF exponents
Mermin-Wagner: no continuous symmetry breaking in \(d \le 2\)
Data collapse tests the full scaling function

← Renormalization Group Non-Equilibrium Stat Mech →