Tensor renormalization group approach to critical phenomena via symmetry-twisted partition functions

This paper demonstrates that the tensor renormalization group (TRG) method, when applied to symmetry-twisted partition functions, provides an efficient framework for detecting spontaneous symmetry breaking and accurately determining critical temperatures and exponents in the 2D Ising, 3D $O(2)$, and 2D $O(2)$ (BKT) models.

The Gist

Imagine you are trying to understand the behavior of a massive crowd of people at a concert. Sometimes, they move in perfect unison (a "symmetric" state), and sometimes they break into chaotic, independent groups (a "broken symmetry" state). In physics, these crowds are made of atoms or spins, and figuring out exactly when they switch from one state to another is a huge challenge.

This paper introduces a new, clever way to spot these switches using a mathematical tool called the Tensor Renormalization Group (TRG). Here is the breakdown of their discovery using simple analogies.

1. The Problem: How to See the Invisible Switch

In the past, physicists tried to find these "phase transitions" by looking at individual people in the crowd (local order parameters). It's like trying to guess the mood of a stadium by staring at one person's face. It's hard, slow, and often misleading.

The authors propose a different approach: The "Twisted" Boundary.
Imagine the concert hall is a giant loop. Usually, the people at the left edge are connected to the people at the right edge in a normal way. But, what if you "twist" the connection? You tell the person at the far right to act as if they are the opposite of the person on the far left.

• If the crowd is chaotic (Symmetric): They don't care about the twist. They just keep doing their own thing. The "twisted" crowd looks exactly like the normal crowd.
• If the crowd is unified (Broken Symmetry): The twist breaks their unity. They can't maintain their synchronized dance anymore. The "twisted" crowd behaves very differently from the normal one.

By comparing the "Normal Crowd" to the "Twisted Crowd," you get a perfect signal (an Order Parameter) that tells you exactly when the crowd is about to switch states.

2. The Tool: The "Smart Squeezer" (TRG)

Calculating the behavior of billions of atoms is impossible for a standard computer. It's like trying to count every grain of sand on a beach one by one.

The authors use Tensor Renormalization Group (TRG). Think of TRG as a smart photo compressor.

• Instead of keeping every single grain of sand (every atom), it looks at the patterns and compresses the information, keeping only the most important details.
• It does this repeatedly, zooming out from the microscopic level to the macroscopic level, until it can calculate the total "energy" of the system without getting bogged down in the details.
• The paper shows that TRG is actually better at handling these "Twisted" crowds than traditional methods because it can easily apply the twist rules to the compressed data.

3. The Experiments: Three Different Crowds

The team tested their method on three famous physics models, acting like different types of crowds:

A. The 2D Ising Model (The "Yes/No" Crowd)

• The Crowd: A grid of people who can only face "Up" or "Down."
• The Twist: Flip everyone's direction at the edge of the room.
• The Result: They found the exact temperature where the crowd stops flipping randomly and locks into a single direction. Their calculation matched the known perfect answer, proving their "Twisted" method works.

B. The 3D O(2) Model (The "Spinning Top" Crowd)

• The Crowd: People in a 3D room, each holding a spinning top that can point in any direction (like a compass needle).
• The Twist: Rotate the tops at the edge of the room by 180 degrees.
• The Result: This is a very hard problem. Previous methods struggled to find the exact "critical temperature" where the tops start spinning in unison. Using their twisted method, the authors pinpointed the temperature with high precision ($T_c \approx 2.2017$). This is a major breakthrough because it's the first time this specific method has nailed this difficult 3D problem.

C. The 2D O(2) Model (The "Superfluid" Crowd)

• The Crowd: A 2D layer of superfluid (like liquid helium).
• The Twist: This is special because the crowd doesn't just "break"; it undergoes a BKT Transition. Imagine the crowd forming tiny whirlpools (vortices). At high heat, the whirlpools are everywhere and chaotic. At low heat, they pair up and cancel each other out, creating a smooth, flowing state.
• The Result: By twisting the boundary, they could measure the "stiffness" of the flow (helicity modulus). They found the exact temperature where the whirlpools pair up ($T_{BKT} \approx 0.8928$). This confirmed that their method can detect even the most subtle, exotic types of phase changes.

4. Why This Matters

Think of this paper as inventing a new kind of thermometer.

• Old thermometers (Monte Carlo simulations) are like trying to measure the temperature of a soup by tasting a single spoonful. It works, but it's noisy and slow.
• This new method (TRG with Twisted Partition Functions) is like having a sensor that measures the entire soup's structure at once. It's faster, cleaner, and gives a much clearer picture of exactly when the soup boils or freezes.

In a nutshell: The authors showed that by mathematically "twisting" the rules of a system and using a smart compression algorithm (TRG), we can detect exactly when matter changes its fundamental nature (like water turning to ice, or magnets turning on) with incredible accuracy. This opens the door to solving even harder physics problems that were previously too difficult to crack.

Technical Summary

1. Problem Statement

The paper addresses the challenge of efficiently detecting phase transitions and characterizing critical phenomena in statistical mechanical systems, particularly those involving Spontaneous Symmetry Breaking (SSB) of both discrete and continuous global symmetries.

• Limitations of Traditional Methods:
  • Monte Carlo (MC): While effective for local order parameters, MC simulations struggle with symmetry-twisted partition functions. Calculating the ratio $Z[A]/Z[0]$ (where $A$ is a background gauge field) suffers from severe overlap problems because the twist is imposed on a codimension-1 surface wrapping the spacetime, making reweighting inefficient.
  • Tensor Renormalization Group (TRG): Standard TRG methods excel at computing partition functions but have historically focused on local correlators. Extracting long-range behaviors or specific topological/order parameters like the helicity modulus or twisted partition functions directly has been less straightforward compared to MC.
• Theoretical Gap: While the Gu–Wen ratio (a ratio of untwisted partition functions with different volumes) is known to detect discrete symmetry breaking, its utility for continuous symmetries (like $U(1)$) is limited due to the presence of massless Nambu-Goldstone modes and the difficulty in distinguishing trivial gapped phases from conformal phases in low dimensions.

2. Methodology

The authors propose a framework where symmetry-twisted partition functions serve as order parameters, computed efficiently using the Tensor Renormalization Group (TRG) method.

• Symmetry-Twisted Partition Functions:
  • The authors utilize the fact that locality in field theory constrains the behavior of partition functions under symmetry twists.
  • For a system with global symmetry $G$, a background gauge field $A$ (flat connection) is introduced. The twisted partition function $Z[A]$ is defined by imposing twisted boundary conditions (holonomies) along non-trivial cycles of the torus.
  • Order Parameter: The ratio $R = Z[A]/Z[0]$ acts as an order parameter:
    • In the symmetric phase (high $T$), $R \to 1$ as volume $L \to \infty$.
    • In the SSB phase (low $T$), $R \to 0$ (for discrete symmetries or $d \ge 3$ continuous symmetries) due to domain wall formation or vacuum mismatch.
• TRG Implementation:
  • Tensor Network Construction: The partition function is expressed as a contraction of local tensors $T_x$.
  • Symmetry Blocking: For systems with on-site global symmetries (e.g., $Z_2$ or $U(1)$), the local tensors are decomposed into irreducible representations. The bond indices are blocked by symmetry sectors.
  • Twist Implementation: Instead of modifying the bulk tensors, the twist is implemented in the final tensor trace step. For a twist $g_0$, the trace is weighted by the character of the representation: $Z_{g_0} = \text{tTr}_{g_0} [\prod T_x]$. This allows the computation of twisted partition functions with the same computational cost as the untwisted case.
  • Algorithms Used:
    • BTRG (Bond-weighted TRG): Used for 2D models.
    • ATRG (Anisotropic TRG): Used for 3D models.

3. Key Contributions

1. Efficient TRG Framework for Twisted Partition Functions: Demonstrated that TRG can compute symmetry-twisted partition functions naturally by exploiting the symmetry structure of tensors, avoiding the overlap problems inherent in Monte Carlo methods.
2. Universal Order Parameters: Established that the ratio of twisted to untwisted partition functions is a robust order parameter for both discrete ($Z_2$) and continuous ($U(1)$) symmetry breaking, capable of identifying critical points solely from the partition function data.
3. Direct Helicity Modulus Calculation: Showed that for 2D $O(2)$ models, the twisted partition function allows for the direct extraction of the helicity modulus (superfluid stiffness), a key quantity for identifying Berezinskii–Kosterlitz–Thouless (BKT) transitions.
4. Finite-Size Scaling Analysis: Applied finite-size scaling to the twisted partition function ratios to precisely determine critical temperatures ($T_c$) and critical exponents ($\nu$).

4. Key Results

The authors applied their method to three benchmark models:

A. 2D Classical Ising Model ($Z_2$ Symmetry)

• System: 2D Ising model with $Z_2$ spin flip symmetry.
• Result: The ratio $Z_{-1}/Z_1$ (twist by $\pi$) acts as a Binder cumulant.
  • $Z_{-1}/Z_1 \to 1$ in the symmetric phase ($T > T_c$).
  • $Z_{-1}/Z_1 \to 0$ in the broken phase ($T < T_c$).
• Validation: The curves for different system sizes cross at the exact critical point $T_c = 2/\ln(1+\sqrt{2})$. The value at $T_c$ matches the theoretical prediction from 2D Ising CFT.
• Scaling: Finite-size scaling of the ratio yields the critical exponent $\nu = 1$, consistent with exact solutions.

B. 3D Classical $O(2)$ Model ($U(1)$ Symmetry)

• System: 3D classical $O(2)$ model (XY model) where $U(1)$ symmetry breaks to trivial.
• Challenge: The Gu-Wen ratio is not applicable here; the twisted partition function is required.
• Result: The ratio $Z_{\pi}/Z_0$ (twist by $\pi$) clearly distinguishes phases.
  • Determined Critical Temperature: $T_c = 2.2017(2)$.
  • Determined Critical Exponent: $\nu = 0.663(33)$.
• Significance: This is the first precise estimate of the critical exponent $\nu$ for the 3D $O(2)$ model using the TRG approach. The results are consistent with high-precision Monte Carlo ($\nu \approx 0.6718$) and Conformal Bootstrap ($\nu \approx 0.6717$) results, though TRG errors are currently larger.
• Note: The authors discuss finite bond-dimension effects in the thermodynamic limit, showing that smaller volumes are necessary to extract $\nu$ via scaling, while large volumes require careful extrapolation.

C. 2D Classical $O(2)$ Model (BKT Transition)

• System: 2D $O(2)$ model, which does not exhibit standard SSB due to the Mermin-Wagner theorem but undergoes a BKT transition.
• Method: Extracted the Helicity Modulus $\Upsilon_\alpha$ from the twisted partition function ratio $Z_\alpha/Z_0$.
  • $\Upsilon_\alpha \propto -\frac{T^2}{\alpha^2} \frac{L_2}{L_1} \ln(Z_\alpha/Z_0)$.
• Result:
  • Observed the "universal jump" of the helicity modulus at the transition.
  • Determined BKT Transition Temperature: $T_{BKT} = 0.8928(2)$ (from $\alpha=\pi$) and $0.8927(5)$ (from $\alpha=\pi/6$).
  • These values are consistent with previous literature and Monte Carlo simulations.
• Insight: The method successfully captures the scale-invariant behavior of the low-temperature conformal phase and the gapped high-temperature phase.

5. Significance and Future Outlook

• Competitiveness with Monte Carlo: The paper demonstrates that TRG, when combined with symmetry-twisted partition functions, is a powerful alternative to Monte Carlo for studying critical phenomena, particularly for systems where local order parameters are insufficient or where overlap problems plague MC.
• Topological Phases: The authors suggest that this approach extends to detecting topological order and symmetry fractionalization. In topological phases, twisted partition functions exhibit distinct behaviors compared to SSB phases, potentially serving as order parameters for symmetry-protected topological (SPT) states and topologically ordered phases.
• Future Directions:
  • Application to generalized $O(2)$ models with competing interactions.
  • Extension to lattice gauge theories to study confinement-deconfinement transitions and topological phases.
  • Improvement of TRG algorithms (e.g., GPU acceleration, improved coarse-graining) to reduce errors in critical exponent calculations.

In summary, this work establishes symmetry-twisted partition functions as a robust, universal order parameter for detecting phase transitions and provides a computationally efficient TRG framework to calculate them, successfully reproducing known critical data for Ising, $O(2)$, and BKT systems.