Fisher Curvature Scaling at Critical Points: An Exact Information-Geometric Exponent from Periodic Boundary Conditions

This paper establishes an exact information-geometric exponent describing the scaling of Fisher information scalar curvature at critical points for lattice spin models under periodic boundary conditions, deriving a universal formula in terms of critical exponents that is validated through exact transfer-matrix and Monte Carlo simulations across 2D and 3D Ising and Potts models.

The Gist

Imagine you are trying to understand a massive, complex crowd of people (like a stadium full of fans) by looking at how they interact with their immediate neighbors.

In physics, this "crowd" is a material (like a magnet or a gas), and the "interactions" are the tiny forces between atoms. Usually, scientists study these materials by looking at big, average properties like temperature or pressure. But this paper takes a different approach: it looks at the microscopic geometry of the connections between every single atom.

Here is the story of the paper, broken down into simple concepts:

1. The Map of Interactions (The Fisher Manifold)

Imagine every bond between two atoms in a grid is a road on a map. If you have a 2D grid of $10 \times 10$ atoms, you have hundreds of roads.

• The Old Way: Scientists used to look at the "weather" of the whole city (Temperature, Pressure). They found that near a critical point (like water turning to steam), the "shape" of the weather map gets infinitely sharp.
• The New Way: This paper builds a map of every single road between neighbors. Because the number of roads grows with the size of the system, this map is huge and changes shape as you add more people to the stadium.

2. The "Curvature" of the Crowd

In geometry, "curvature" tells you if a surface is flat like a table, curved like a sphere, or saddle-shaped like a Pringles chip.

• The authors calculated the curvature of this massive "interaction map."
• The Discovery: As the system gets closer to a critical point (where a phase transition happens, like a magnet losing its magnetism), this curvature doesn't just get big; it explodes in a very specific, predictable way.

3. The Magic Formula (The Scaling Exponent)

The paper found a "secret code" that predicts exactly how fast this curvature explodes. It depends on two famous numbers in physics:

• $\nu$ (Nu): How far the "influence" of one atom reaches (the correlation length).
• $\eta$ (Eta): How weird the interactions get at the very smallest scales (anomalous dimension).

The formula they found is:
$$\text{Curvature Growth} \approx \frac{d\nu + 2\eta}{d\nu + \eta}$$
(Where $d$ is the number of dimensions, like 2D or 3D).

The Analogy: Think of it like a recipe. If you know how far a rumor spreads ($\nu$) and how distorted the rumor gets as it travels ($\eta$), you can predict exactly how chaotic the "rumor map" will become at the moment the whole city goes crazy.

4. The Proof: Testing the Recipe

The authors didn't just guess; they ran massive computer simulations to test this formula on different "universes" (models):

• The 2D Ising Model (The Gold Standard): This is the simplest magnet model. The formula predicted the curvature should grow by a factor of 1.111... (which is exactly $10/9$).

  • The Result: Their computer simulations matched this number perfectly. It's like predicting a coin flip will land heads 50% of the time, and after a million flips, it lands heads 50.0001% of the time. Confirmed.

• The 3D Ising Model (Real-world magnets): Here, the prediction was almost 1.02.

  • The Result: The simulations were very close, showing the formula works even in 3D, though it takes longer to settle down.

• The Potts Models (More complex crowds): These models have more than just "up/down" states (like a crowd with 3 or 4 different team jerseys).

  • The Result: The data was "wobbly" (oscillating). Why? Because at these specific points, the math gets messy with "logarithmic corrections" (think of it as a slight delay in the system reacting). The data wasn't perfect yet, but it was wobbling around the predicted number, suggesting the formula is right but needs bigger systems to settle down.

5. The "Check Engine" Light (Ricci Identity)

To make sure they didn't make a calculation error, they used a mathematical "checksum."

• Imagine you have a complex machine with four gears. The authors proved that if Gear 3 turns, Gear 1 must turn in the opposite direction at exactly half the speed.
• They checked this rule for every single simulation. It held true to 5 or 6 decimal places. This gave them confidence that their numbers were real and not just computer noise.

6. Why This Matters

• It's a New Lens: Previous studies looked at the "thermodynamic" map (low-dimensional). This paper looks at the "microscopic" map (high-dimensional). It's like looking at a forest from a satellite vs. looking at the roots of every single tree.
• It's Universal: The formula works for magnets, different types of phase transitions, and even continuous spin models.
• It's a Falsifiable Prediction: The authors say, "If you run a simulation on a 5-state Potts model (which shouldn't have this behavior), the formula should fail." And indeed, when they checked a model that doesn't have a smooth phase transition, the formula didn't work. This proves the formula is specific to the right kind of physics.

The Bottom Line

This paper discovered a new "law of geometry" for critical systems. It tells us that when a material is on the verge of changing state, the shape of the network of its interactions becomes hyperbolic (saddle-shaped) and explodes in size according to a precise mathematical recipe involving the correlation length and anomalous dimensions.

It's like finding that the way a crowd cheers at a stadium follows a specific geometric rule that depends only on how fast the sound travels and how loud the noise gets, regardless of whether it's a football game or a rock concert.

Technical Summary

1. Problem Statement

The paper addresses the geometric structure of statistical manifolds at critical points in lattice spin models. While the Ruppeiner metric (defined on the low-dimensional thermodynamic state space of temperature $T$ and magnetic field $h$) is well-known to diverge as the correlation volume ($|R_{Rup}| \sim \xi^d$) near second-order phase transitions, the behavior of the Fisher information metric on the microscopic coupling-parameter manifold remains less understood.

Specifically, the author investigates the scalar curvature $R$ of the Fisher manifold defined on the space of all bond couplings $\{J_e\}$ for a lattice with $n=L^d$ sites and $m=d \cdot n$ bonds. The central question is: How does the scalar curvature $|R(J_c)|$ of this high-dimensional ($m \sim L^d$) manifold scale with system size $n$ at the critical coupling $J_c$?

2. Methodology

The study employs a combination of exact analytical methods and large-scale numerical simulations across various universality classes.

• Model Setup:

  • Models: 2D and 3D Ising models, 2D Potts models ($q=3, 4$), 2D Clock model ($q=12$, BKT transition), Ashkin-Teller model, 3D XY, and 3D Heisenberg models.
  • Boundary Conditions: Periodic Boundary Conditions (PBC) on tori ($L \times L$ or $L^3$) are used to preserve translational invariance and enable Fourier decomposition. Open Boundary Conditions (OBC) were tested but found to suffer from slow convergence due to broken symmetry.
  • Manifold Dimension: The Fisher metric is an $m \times m$ matrix where $m = d \cdot L^d$ (one parameter per bond).

• Computational Techniques:

  • Exact Transfer Matrix (TM): Used for small systems ($L \le 9$ in 2D, $L \le 3$ in 3D) to provide exact benchmarks.
  • Monte Carlo (MCMC):
    • 2D: Multi-seed Wolff-cluster sampling with FFT-based cumulant accumulation ($L=10–24$).
    • 3D: GPU-accelerated Wolff-cluster sampling with translational symmetry averaging (symavg), which reduces variance by a factor of $\sim L^3$ by averaging the Fisher matrix over all lattice translates.
    • Potts Models: Two-phase pipeline (chunked MCMC followed by GPU/CPU curvature assembly) with Jackknife error estimation.
  • Curvature Calculation:
    • The scalar curvature $R$ is computed using the Riemann tensor derived from Christoffel symbols.
    • Christoffel symbols are calculated via the third cumulant $\kappa_{abc} = \langle (\sigma_a - \mu_a)(\sigma_b - \mu_b)(\sigma_c - \mu_c) \rangle$.
    • Ricci Decomposition: The authors utilize a structural identity ($R_3 = -R_1/2, R_4 = -R_2/2$) derived in a companion paper [7]. This allows the total curvature to be computed from only two independent Ricci scalars ($R_1, R_2$), significantly reducing computational cost and serving as a rigorous consistency check (verified to 5–6 digits).

3. Key Contributions & Theoretical Formula

The paper proposes and validates an exact scaling exponent $d_R$ for the divergence of the Fisher scalar curvature:
$$|R(J_c)| \sim n^{d_R}$$
where $n = L^d$ is the number of sites. The exponent is given by:
$$d_R = \frac{d\nu + 2\eta}{d\nu + \eta}$$
Here, $\nu$ is the correlation-length exponent and $\eta$ is the anomalous dimension.

Physical Mechanism:
The derivation relies on four key ingredients:

1. $Z_2$ Selection Rule: Suppression of the energy-channel three-point function forces curvature anomalies into the $\sigma$-channel.
2. Fourier Decomposition: On a torus, the Fisher metric decomposes into blocks. The dominant contribution comes from the antiperiodic sector (half-integer momenta) where the minimal eigenvalue scales as $\lambda_\sigma \sim L^{-(2-\eta)}$.
3. UV-IR Interference: The curvature arises from a convolution of Christoffel symbols involving both soft modes (IR) and UV modes, creating a compound ratio of exponents.
4. Diagonal Block Cancellation: In the bulk of the Brillouin zone, the Fisher block is approximately diagonal, leading to zero curvature per block. The non-zero curvature is a residual effect concentrated at the softest mode $k_{min}$, where eigenvalue anisotropy breaks the diagonal approximation.

4. Results

A. 2D Ising Model (Definitive Confirmation)

• Prediction: With $\nu=1, \eta=1/4$, $d_R = 10/9 \approx 1.111$.
• Findings:
  • Exact TM ($L=3–9$) yields an effective exponent $d_{eff} = 1.1115 \pm 0.0002$ (deviation of $1.8\sigma$ from theory).
  • MCMC ($L=10–24$) confirms all effective exponents are within $2.2\sigma$ of $10/9$.
  • The Fisher vertex correction exponent $\alpha_f = 2(d_R - 1)$ is confirmed as $2/9$, ruling out alternative hypotheses (e.g., $1/4$).

B. 3D Ising Model

• Prediction: With $\nu \approx 0.630, \eta \approx 0.0363$, $d_R \approx 1.0188$.
• Findings: MCMC data up to $L=10$ shows oscillating convergence around the prediction. A power-law fit ($L=5–10$) yields $d_R = 1.040$, consistent with the theoretical value given slow convergence and finite-size effects.

C. 2D Potts Models ($q=3, 4$)

• Predictions: $q=3 \to d_R = 33/29 \approx 1.138$; $q=4 \to d_R = 22/19 \approx 1.158$.
• Findings:
  • Effective exponents oscillate non-monotonically (e.g., $q=3$ oscillates around $\sim 1.20$ before dipping).
  • This behavior is attributed to logarithmic corrections of order $O(1/(\ln L)^2)$ arising from marginal operators (specifically the $c=1$ operator for $q=4$).
  • The data is consistent with the theory but requires much larger systems ($L \gg 100$) for definitive confirmation.

D. Other Models

• BKT Transition ($q=12$ Clock): As $\nu \to \infty$ and $\eta=1/4$, the formula predicts $d_R \to 1$. Results show a slow directional decrease toward 1, consistent with the limit.
• Ashkin-Teller Model: A continuous family test along the self-dual line shows $d_R$ varying smoothly from $10/9$ to $22/19$ as the coupling parameter changes, confirming the formula's validity across a continuum of universality classes.
• 3D XY and Heisenberg: Results for $L \le 10$ show convergence toward predictions ($d_R \approx 1.019$ and $1.017$ respectively), with 3D models converging faster than 2D.

5. Significance and Implications

• New Geometric Scaling Law: The paper establishes that the Fisher curvature on the microscopic coupling manifold diverges with a distinct exponent $d_R$, which is fundamentally different from the Ruppeiner curvature exponent ($d\nu$).
• Dimensionality Dependence: Unlike the Ruppeiner metric (fixed 2D manifold), the Fisher manifold dimension $m$ grows with system size ($m \sim L^d$). The exponent $d_R$ reflects the collective geometry of this growing high-dimensional space.
• Structural Consistency: The verification of the Ricci decomposition identity ($R_3 = -R_1/2$) to high precision across all models validates the numerical pipeline and the underlying geometric structure of discrete exponential families.
• Falsifiability: The paper provides a clear protocol for falsification. If effective exponents deviate by $>3\sigma$ from the formula across consecutive system sizes, the theory is rejected. Currently, all tested second-order transitions are consistent with the formula.
• Limitations: The formula applies strictly to second-order transitions with scale invariance. First-order transitions (e.g., 2D Potts $q=5$) do not converge to a rational $d_R$.

Conclusion

This work provides a rigorous, information-geometric characterization of criticality. It demonstrates that the scalar curvature of the microscopic Fisher manifold follows a universal power law determined by standard critical exponents $\nu$ and $\eta$. The results bridge the gap between information geometry and statistical mechanics, offering a new tool for probing the collective geometry of phase transitions in high-dimensional parameter spaces.