A Machine Learning study of the two-dimensional antiferromagnetic Q-state Potts model on the square lattice

arXiv:2512.06762 · hep-lat, cond-mat.dis-nn, cond-mat.stat-mech · Submitted 2025-12-07 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: High Energy Physics: Lattice. The day in lattice field theory: what the simulations put on the table.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "A Machine Learning study of the two-dimensional antiferromagnetic Q-state Potts model on the square lattice".

Mira: Here's the gist of the paper. The critical phenomena of two-dimensional (2D) antiferromagnetic q-state Potts model on the square lattice with q = 2, 3, 4, 5 and 6 are investigated using the technique of supervised neural network (NN).

Kai: Remarkably, despite the fact that the MLP is trained without any input from these considered models, it correctly identifies the critical temperatures of the studied physical systems.

Mira: Particularly, the MLP outcomes suggest convincingly that the q = 3 model is critical only at zero temperature and q = 4, 5, 6 models remain disordered at all temperatures.

Kai: Model Definition and Context The Hamiltonian H of the 2D antiferromagnetic q-state Potts model on the square lattice considered here can be expressed as.

Paper summary: Kai: We’ve been looking at how this machine learning model tackles these two-dimensional antiferromagnetic q state Potts models on a square lattice, and now we're coming to the conclusion of the paper "A Machine Learning study of the two-dimensional antiferromagnetic Q-state Potts model on the square lattice."

Mira: Yeah, it boils down to this neural network training without any actual experimental spin data, yet it correctly identifies critical temperatures for q values two through six.

Lev: That's a huge thing for us because we spend so much time building error correction codes that rely on understanding phase transitions in these types of systems.

Kai: Exactly. The paper suggests this simple machine learning approach can be quite effective at mapping out where the phase transitions are happening in these complex magnetic models.

Mira: It points toward a really interesting idea about universality, suggesting this kind of network architecture might work for many different physical models that differ a lot from each other.

Lev: But we have to remember the authors trained this thing on artificial configurations, so we gotta ask what the real world constraints are before we can trust these numbers.

Kai: So for now, it's a powerful tool for finding critical points in these specific q state Potts models, but it doesn't guarantee correctness outside of those exact parameters.

Mira: And the authors themselves flag that this model only works where the network can actually find the transitions correctly, which is a fair limitation to keep in mind.

Lev: It really shows a path forward for using machine learning to predict critical phenomena, but we still have to see if that method scales up to systems with less constrained physics.

Kai: That's what we need to figure out next, especially how they handle those situations where the training signal doesn't perfectly match the input from a real physical system.

Conclusion: Kai: So we've looked at how this machine learning model tackles these two-dimensional antiferromagnetic q state Potts models on a square lattice, and now we're coming to the conclusion of the paper "A Machine Learning study of the two-dimensional antiferromagnetic Q-state Potts model on the square lattice."

Mira: Yeah, it boils down to this neural network training without any actual experimental spin data, yet it correctly identifies critical temperatures for q values two through six.

Lev: That's a huge thing for us because we spend so much time building error correction codes that rely on understanding phase transitions in these types of systems.

Kai: Exactly. The paper suggests this simple machine learning approach can be quite effective at mapping out where the phase transitions are happening in these complex magnetic models.

Mira: It points toward a really interesting idea about universality, suggesting this kind of network architecture might work for many different physical models that differ a lot from each other.

Lev: But we have to remember the authors trained this thing on artificial configurations, so we gotta ask what the real world constraints are before we can trust these numbers.

Kai: So for now, it's a powerful tool for finding critical points in these specific q state Potts models, but it doesn't guarantee correctness outside of those exact parameters.

Mira: And the authors themselves flag that this model only works where the network can actually find the transitions correctly, which is a fair limitation to keep in mind.

Lev: It really shows a path forward for using machine learning to predict critical phenomena, but we still have to see if that method scales up to systems with less constrained physics.

Kai: That's what we need to figure out next, especially how they handle those situations where the training signal doesn't perfectly match the input from a real physical system.

Department of Physics, National Taiwan Normal University

hep-lat, cond-mat.dis-nn, cond-mat.stat-mech

Submitted: 2025-12-07

Updated: 2026-11-06

Comments: 11 pages, 10 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 69/100

The gist: The critical phenomena of two-dimensional (2D) antiferromagnetic q-state Potts model on the square lattice with q = 2, 3, 4…

Key concepts

Antiferromagnetic Potts Model
This is a physical system where spins on a square lattice interact in an antiferromagnetic manner. The spins can take on $q$ different discrete values. These models are complex because they have many low-energy ground states, making them challenging to study compared to simpler ferromagnetic versions.
Neural Network (MLP)
A simple multilayer perceptron was used as the machine learning tool. It was trained using artificially created 'staggered' configurations rather than real simulation data. The network learns to associate specific input patterns with outputs that correspond to critical behavior.
Pseudo-critical Temperature $T_c(L)$
This is a temperature derived from the MLP output magnitude ($R$). When the input configuration is similar to training data, $R o 1$. The pseudo-critical temperature is defined where this magnitude reaches a specific value ($1 + 1/ ext{ ext{sqrt}(2)} / 2$), helping to estimate the true critical temperatures.
Q-state Potts Model
This model describes spins on a lattice that can take one of $q$ discrete values. The specific value of $q$ (e.g., 3, 4, 5, or 6) determines the complexity and behavior of the system under different temperature conditions.

Terminology

Summary

The critical phenomena of two-dimensional (2D) antiferromagnetic q-state Potts model on the square lattice with q = 2, 3, 4, 5 and 6 are investigated using the technique of supervised neural network (NN). p. 0 Remarkably, despite the fact that the MLP is trained without any input from these considered models, it correctly identifies the critical temperatures of the studied physical systems. p. 0 Particularly, the MLP outcomes suggest convincingly that the q = 3 model is critical only at zero temperature and q = 4, 5, 6 models remain disordered at all temperatures. p. 0

Model Definition and Context

The Hamiltonian H of the 2D antiferromagnetic q-state Potts model on the square lattice considered here can be expressed as. The Potts variable σi at each site i takes an integer value from a set of discrete positive integers, namely, 1, 2, 3, q − 2, q − 1 and q. p. 1 Depending on the sign of the interactions between any pairs of nearest Potts spins, the q-state Potts model can either be a ferromagnetic or an antiferromagnetic system. p. 0 Antiferromagnetic Potts models are more challenge to study than their ferromagnetic counterparts due to the fact that the antiferromagnetic systems have extremely highly degenerated ground states. p. 0 On two-dimensional (2D) square lattice and q ≥ 2, while the critical temperatures Tc for q-state ferromagnetic Potts models exist and fulfill Tc > 0, only q = 2 (Tc = 1. p. 01346) and q = 3 (Tc = 0) antiferromagnetic Potts models can be ordered. p. 0

How the training signal is built

Instead of using real spin configurations obtained from the Monte Carlo simulations, two kinds of artificially made (stagger-types) configurations are considered as the training set. Specifically, the training set consists of N (N = 200 or 400 in this study) copies of both the left and the right panels of fig. p. 1 2. Here each building unit is made up of L1 by L1 sites and the value of L1 can be varied. p. 1 Moreover, the MLP output labels for the training configurations resulting from the left and the right panels of fig. p. 1 2 are (1, 0) and (0, 1), respectively. p. 1 The construction of testing set involves considering spins belonging to the first L1 by L1 squared sub-area for a given configuration on an L by L squared area. p. 1

NN Architecture and Training Procedure

The NN used in this investigation is a multilayer perceptron (MLP) which consists of one input layer, one hidden layer of two neurons or any other number of neurons, and one output layer. The activations used for the hidden layer and the output layer are ReLU and Softmax, respectively. p. 1 The procedure involves one-hot encoding between the input and hidden layers. p. 1 L2 regularization with the input parameter being 1 is considered to avoid overfitting. p. 1 The algorithm and optimizer employed are minibatch and adam, respectively. p. 1

Where the gains come from

The magnitude R of the MLP output vectors is defined such that if the input configuration is highly similar to the training set, then the output vector is close to (1, 0) or (0, 1) which would result in R ∼ 1. If the input configuration is in strong contrast to the training set, then the output vector is close to (0. p. 25, 0.5) which would lead to R ∼ 1/√2. p. 2 Inspired by this, for raw spin configurations from MC simulations with a given L, the temperature where the associated R take the value of 1 + 1/√2 / 2 is defined to be the pseudo-critical temperature Tc(L) corresponding to that L. p. 2 This approach in conjunction with certain semi-experimental finite-size formulas lead to accurate determination of the critical temperatures Tc for many different physical models. p. 2

Model Behavior and Conclusions

The MLP outcomes clearly demonstrate that a certain order is established for q = 3 model close to zero temperature. Moreover, evidence showing that q = 4, 5 and 6 models are always disordered at any temperatures including the zero temperature is revealed from the MLP results. p. 0 For q = 3 model, the rising of R from 1/√2 to 0. p. 59 when one approaches the very low-T region from the high-T region indicates that certain kind of order similar to the training set, namely the staggered order, establishes at very low-T region. p. 5 This is consistent with the theoretical prediction that q = 3 model has Tc = 0. p. 5 The behaviors of R as functions of T shown in the figure suggest strongly that for q = 4 model, no order is established at any finite T. p. 5 Finally, the outcomes in the figure clearly imply strongly that only q = 3 model is ordered at very low temperature, and q = 4, 5 and 6 models are disordered at any finite T. p. 5

Pruning redundant reasoning tokens

The MLP with such a simple architecture and is trained without any real physical configurations can be applied to compute the critical temperature of the 2D antiferromagnetic 2-state Potts models on the square lattice. It also uncover the critical behaviors correctly for the 2D antiferromagnetic 3-, 4-, 5-, and 6-state models on the square lattice. p. 6 This MLP is universal to a very large extent. p. 7

The MLP with such a simple architecture and is trained without any real physical configurations can be applied to compute the critical temperature of the 2D antiferromagnetic 3-, 4-, 5-, and 6-state models on the square lattice. It also uncover the critical behaviors correctly for the 2D antiferromagnetic q-state ferromagnetic Potts models on the square lattice. p. 6

The MLP with such a simple architecture and is trained without any real physical configurations can be applied to compute the critical temperature of the 2D antiferromagnetic q-state ferromagnetic Potts models on the square lattice.

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Enhanced capability to predict critical temperatures for antiferromagnetic q-state Potts models using a simple multilayer perceptron trained only on artificially made stagger-like configurations. This allows for quick identification of critical behaviors without needing real physical spin configurations or extensive Monte Carlo simulations for training.

  2. Ability to correctly identify the critical temperatures of 2D antiferromagnetic q=3 models as being critical only at zero temperature, and q=4, 5, and 6 models as remaining disordered at all temperatures. This provides a powerful tool for filtering out physically impossible or non-critical regimes quickly.

  3. Potential to determine phase transition orders for the studied models by analyzing the magnitude R of MLP output vectors in relation to temperature and system size L. For example, the rising of R from 1/sqrt(2) to 0.9 when one approaches the very low temperature region indicates that certain kind of order similar to the training set, namely the staggered order, establishes at very low-T region for q=3.

  4. Application to determine critical points and even critical exponents of many models differing significantly among themselves by training a similar MLP with ferromagnet-like configurations. This suggests a universal method applicable beyond the specific antiferromagnetic systems studied in this paper.

  5. Use of standard deviations of R to identify critical temperatures for ferromagnetic models, as STD is a quantity with small magnitude compared to that of R, the computed Tc is less accurate and can be influence by several NN-related parameters. This suggests a more robust metric for temperature estimation in ferromagnetic contexts.

Sources

Related papers