Tensor Renormalization Group Calculations of Partition-Function Ratios

This paper employs bond-weighted tensor renormalization group calculations to demonstrate that dimensionless partition-function ratios for the Ising and Potts models obey finite-size scaling and yield critical values consistent with conformal field theory predictions, while also revealing logarithmic corrections in the four-state Potts model.

The Gist

Imagine you are trying to understand how a crowd of people behaves. Do they move randomly like a chaotic party, or do they all march in perfect lockstep like a military parade? In physics, this is called a phase transition. It's the moment a material changes from one state to another, like ice melting into water or a magnet losing its magnetism when heated.

The scientists in this paper, Satoshi Morita and Naoki Kawashima, are trying to find the exact "temperature" where this change happens and understand the rules that govern it. To do this, they use a clever mathematical tool called Tensor Renormalization Group (TRG).

Here is a simple breakdown of what they did and why it matters, using some everyday analogies.

1. The Problem: Counting the Impossible

To predict how a material behaves, physicists need to calculate something called the Partition Function. Think of this as a giant "scorecard" that counts every single possible way the particles in a system can arrange themselves.

• The Challenge: For a small system, you can count the arrangements. But for a real-world material with billions of atoms, the number of arrangements is so huge it's like trying to count every grain of sand on all the beaches on Earth. Traditional computer methods (like Monte Carlo simulations) often get stuck or take too long, especially when the system is very complex.

2. The Solution: The "Zoom-Out" Camera (TRG)

The authors use a method called Tensor Renormalization Group. Imagine you have a high-resolution photo of a forest.

• The Old Way: You try to count every single leaf. Impossible.
• The TRG Way: You take a step back. You group leaves into branches, branches into trees, and trees into a forest. You keep "zooming out" and simplifying the picture, but you keep the most important information (the "shape" of the forest) while throwing away the tiny, unimportant details (the exact color of one specific leaf).
• Bond-Weighted TRG (BWTRG): The authors used a special, upgraded version of this camera that is even better at keeping the important details while zooming out. This allows them to simulate much larger systems than before.

3. The Secret Sauce: Ratios of Scores

Instead of trying to calculate the total "score" (the Partition Function) which is huge and messy, they decided to look at Ratios.

• The Analogy: Imagine you want to know if a cake is perfectly baked. Instead of weighing the whole cake (which might be heavy and hard to measure), you compare the weight of the cake to the weight of a smaller piece of the same cake.
• The Magic: They defined a specific ratio (let's call it X1 and X2) that compares the "score" of a square system to the "score" of a larger system made of two squares.
• Why it's cool: This ratio is dimensionless. It doesn't matter if you measure in inches or centimeters; the ratio stays the same. It acts like a universal fingerprint for the material.

4. The Crystal Ball: Conformal Field Theory (CFT)

The paper connects their computer calculations to a beautiful theory called Conformal Field Theory (CFT).

• The Analogy: Think of CFT as a "Crystal Ball" or a "Recipe Book" written by mathematicians. It predicts what the ratio should be if the material is at the exact moment of changing phases (criticality).
• The Prediction: The theory says that at the critical point, this ratio should be a specific, universal number (like 1.76 for the Ising model). It doesn't depend on the specific material, only on the "class" of the phase transition.

5. The Experiment: Testing the Crystal Ball

The authors ran their "zoom-out camera" (BWTRG) on three famous models:

1. The Ising Model: The simplest magnet.
2. The 3-State Potts Model: A slightly more complex version.
3. The 4-State Potts Model: A very complex version.

The Results:

• For the simple models (Ising and 3-State): Their calculated ratios matched the "Crystal Ball" predictions almost perfectly. It was like looking at a map and seeing the mountain exactly where the map said it would be. This confirmed that their method works beautifully.
• For the complex model (4-State): They found something interesting. The ratio didn't settle on a single number immediately. Instead, it crept toward the target value very slowly, like a snail. This "slow creep" is called a logarithmic correction. It's a subtle signal that the 4-State model is behaving in a unique, tricky way that only advanced methods can spot.

6. The Twist: Anisotropy (The Tilted Room)

They also tested what happens if the material behaves differently in different directions (like a room that is long and narrow instead of square).

• The Finding: They discovered that the "universal number" changes depending on the shape of the room. If the room is tilted, the ratio changes in a predictable way. This is like saying the "perfect cake" ratio changes if you bake it in a rectangular pan instead of a round one. Their method successfully predicted these changes.

Summary: Why Should You Care?

This paper is a victory for computational physics.

1. New Tool: They proved that this "zoom-out camera" (Tensor Renormalization) is a powerful tool for solving problems that were previously too hard for computers.
2. Universal Truths: They confirmed that deep mathematical theories (CFT) accurately predict the behavior of real-world materials, even in complex scenarios.
3. Detecting the Subtle: They showed that this method is sensitive enough to detect tiny, weird behaviors (like the logarithmic corrections in the 4-State model) that other methods might miss.

In short, they built a better microscope to look at the "phase transition" of matter, confirmed that the universe follows elegant mathematical rules, and found a few new wrinkles in those rules that we didn't see clearly before.

Technical Summary

Here is a detailed technical summary of the paper "Tensor Renormalization Group Calculations of Partition-Function Ratios" by Satoshi Morita and Naoki Kawashima.

1. Problem Statement

The paper addresses the challenge of accurately determining critical temperatures and critical exponents in statistical physics models, particularly those where traditional Monte Carlo methods face difficulties (e.g., sign problems or large system sizes). While dimensionless quantities like the Binder parameter are standard tools for finite-size scaling (FSS) analysis, they require calculating higher-order moments of the order parameter.

The authors investigate an alternative class of dimensionless quantities: ratios of partition functions ($X_1, X_2$, etc.). These ratios, originally proposed by Gu and Wen, are defined based on the geometry of the system (e.g., comparing a square system to a rectangular or skewed system). The core problem is to:

1. Verify if these ratios obey the same finite-size scaling laws as the Binder parameter.
2. Determine if their universal values at criticality can be precisely predicted using Conformal Field Theory (CFT) and modular invariance.
3. Analyze their behavior across different universality classes, including cases with logarithmic corrections (e.g., the 4-state Potts model).

2. Methodology

The authors employ the Bond-Weighted Tensor Renormalization Group (BWTRG) method, a variant of the Tensor Renormalization Group (TRG) that improves accuracy by weighting bonds based on their singular values.

• Models Studied: Three 2D lattice models belonging to different universality classes:
  • Ising Model: $c = 1/2$ (Unitary minimal model).
  • 3-State Potts Model: $c = 4/5$ (Minimal model).
  • 4-State Potts Model: $c = 1$ (Compact boson CFT / $Z_2$ orbifold).
• Definitions of Quantities:
  • $X_1(T, L) = [Z_{L \times L}(T)]^2 / Z_{2L \times L}(T)$: Ratio comparing a square to a $2:1$ rectangle.
  • $X_2(T, L)$: Ratio involving a skewed boundary condition (equivalent to a $45^\circ$ tilted square).
  • Generalizations ($X_{m \times n}$): Ratios defined for larger cluster sizes ($3 \times 1$,$4 \times 1$, etc.).
• Theoretical Framework:
  • The authors derive the universal values of these ratios at criticality using CFT.
  • They utilize the modular invariant partition function on a torus, $Z_{PP}(\tau)$, where the modular parameter $\tau$ corresponds to the aspect ratio of the system.
  • For example, $X_1 = Z_{PP}(i)^2 / Z_{PP}(2i)$, where $\tau=i$ represents a square and $\tau=2i$ represents a $2:1$ rectangle.
• Anisotropy: The study extends to anisotropic systems where correlation lengths $\xi_x \neq \xi_y$, adjusting the modular parameter to $\tau = i(\xi_x/\xi_y)$.

3. Key Contributions

• Validation of Scaling Laws: The paper confirms that partition-function ratios obey the same finite-size scaling form as the Binder parameter: $X(T, L) = f(L/\xi) \approx f(L^{1/\nu}t)$.
• CFT Prediction Verification: It demonstrates that the universal values of these ratios at the critical point match the theoretical predictions derived from modular invariant partition functions with high precision.
• Observation of Logarithmic Corrections: In the 4-state Potts model, the authors successfully observe the expected logarithmic corrections to finite-size scaling ($X \sim g(1/\log L)$), a phenomenon often difficult to detect numerically.
• Anisotropy Analysis: The study establishes that the universal value of the ratio depends on the ratio of correlation lengths ($\xi_y/\xi_x$) in anisotropic systems, providing a method to estimate this ratio from numerical data.
• Extension to Non-Square Lattices: The methodology is successfully applied to the honeycomb lattice, deriving new universal values based on the lattice's rhombic unit cell geometry.

4. Key Results

• Ising and 3-State Potts Models:
  • The numerical values of $X_1$ and $X_2$ near the critical temperature converge to the CFT-predicted universal values over a wide range of system sizes.
  • Finite-size scaling plots collapse onto a single curve when scaled by the correlation length, confirming the validity of the scaling hypothesis.
  • The critical exponents ($\nu$) estimated from the data match exact solutions (e.g., $\nu=1$ for Ising, $\nu=5/6$ for 3-state Potts).
• 4-State Potts Model:
  • Unlike the other models, $X_1$ does not plateau at a constant value for finite sizes.
  • Instead, the deviation from the universal value scales inversely with $\log L$, confirming the presence of logarithmic corrections characteristic of the $c=1$ theory.
  • The bond dimension ($\chi$) significantly affects the convergence; larger $\chi$ is required to see the asymptotic behavior due to the infinite number of primary fields in this model.
• Cluster Size and Skew Effects:
  • Quantities defined from larger clusters ($X_{3 \times 1}, X_{4 \times 1}$) also follow the scaling laws but show larger deviations from universal values due to the difficulty of representing long-range correlations with finite bond dimensions.
  • "Skewed" ratios (like $X_2$) generally exhibit larger deviations than non-skewed ones, attributed to the difficulty of square-lattice tensor networks in representing diagonal correlations.
• Anisotropic Ising Model:
  • The universal values of $X_1$ shift systematically as the coupling ratio $J_y/J_x$ changes.
  • The numerical results align perfectly with the CFT prediction that the modular parameter scales with the correlation length ratio.

5. Significance

• Alternative to Binder Parameter: The partition-function ratios offer a robust alternative to the Binder parameter for locating critical points, especially in systems where calculating high-order moments of the order parameter is computationally expensive or prone to noise.
• Direct Access to Universal Data: These ratios provide a direct numerical route to extracting universal CFT data (scaling dimensions, central charge) without needing to diagonalize transfer matrices for the entire spectrum.
• Handling Logarithmic Corrections: The ability of Tensor Network methods (specifically BWTRG) to resolve logarithmic corrections in the 4-state Potts model highlights their superiority over standard Monte Carlo methods in critical regimes with marginal operators.
• Anisotropy Estimation: The method provides a practical tool for estimating the anisotropy of correlation lengths in complex systems where exact solutions are unavailable, potentially improving the efficiency of finite-size scaling analyses by optimizing system aspect ratios.
• Lattice Independence: The successful application to honeycomb lattices suggests the method is generalizable to various lattice geometries, provided the modular parameters are adjusted for the unit cell shape.

In conclusion, the paper establishes partition-function ratios as a powerful, universal, and computationally efficient tool for analyzing phase transitions and critical phenomena using tensor network renormalization group methods.