Cluster percolation in the three-dimensional $\pm J$ random-bond Ising model

Basic Information

• Paper ID: 2511.05748
• Title: Cluster percolation in the three-dimensional $\pm J$ random-bond Ising model
• Authors: L. Münster, M. Weigel (Technische Universität Chemnitz, Emory University)
• Classification: cond-mat.dis-nn, cond-mat.stat-mech, physics.comp-ph
• Publication Date: November 11, 2025
• Paper Link: https://arxiv.org/abs/2511.05748v1

Abstract

This paper investigates the relationship between cluster percolation and equilibrium ordering phenomena in the three-dimensional $\pm J$ random-bond Ising model based on large-scale parallel tempering Monte Carlo simulations, systematically examining the effects of varying antiferromagnetic bond fractions. The study considers multiple cluster definitions, primarily constructed in the overlap space of two independent real replicas. For the pure ferromagnetic limit, the percolation point coincides with the thermodynamic ordering transition. In disordered ferromagnets and strong disorder spin glass phases with introduced antiferromagnetic bonds, this correspondence changes: the percolation transition occurs above the thermodynamic ordering point, accompanied by the appearance of two percolating clusters of equal density. Only at lower disorder ferromagnetic or spin glass transition points do these two clusters' densities begin to differentiate, thereby providing percolation characteristics of the thermodynamic transition.

Research Background and Motivation

Research Questions

This study aims to establish quantitative connections between geometric cluster percolation and thermodynamic phase transitions, particularly in disordered spin systems with frustration. The core question is: how do different types of cluster percolation in the three-dimensional $\pm J$ random-bond Ising model reflect the ferromagnetic and spin glass ordering transitions of the system?

Significance

1. Theoretical Importance: The cluster method provides an elegant geometric description of continuous thermodynamic phase transitions. In pure ferromagnets, the percolation transition of Fortuin-Kasteleyn-Coniglio-Klein (FKCK) clusters perfectly corresponds to the ferromagnetic transition (density equals magnetization).
2. Computational Value: Cluster-based Swendsen-Wang algorithms significantly reduce critical slowing down effects and represent powerful simulation tools.
3. Physical Challenge: For frustrated systems like spin glasses, the identity between spin-spin correlation functions and cluster connectivity probabilities ($|\langle s_x s_y \rangle| = \langle \gamma_{xy} \rangle$) weakens to an inequality ($|\langle s_x s_y \rangle| \leq \langle \gamma_{xy} \rangle$), causing FKCK clusters to percolate far above the ordering transition temperature, losing direct correspondence with the phase transition.

Limitations of Existing Approaches

1. Single-replica clusters (Ising, FKCK) percolate at temperatures far above phase transitions in frustrated systems
2. Two-replica clusters (Houdayer, CMRJ), while sensitive to overlap, lack systematic study of their precise relationship with thermodynamic transitions in three dimensions
3. Absence of comprehensive understanding of cluster behavior across different frustration strengths (antiferromagnetic bond fraction $\phi$)

Research Motivation

By systematically studying three representative cases—$\phi = 0$ (pure ferromagnet), $\phi = 0.125$ (disordered ferromagnet), $\phi = 0.5$ (spin glass)—the authors establish a complete picture of cluster percolation-thermodynamic phase transition correlations, providing foundations for understanding frustrated systems and developing efficient algorithms.

Core Contributions

1. Systematic Cluster Analysis: First comprehensive study of multiple cluster definitions (Ising, Houdayer, FKCK, CMRJ) in the three-dimensional $\pm J$ model across different frustration strengths
2. Key Findings:
   • Pure ferromagnet ($\phi=0$): CMRJ cluster percolation transition coincides with ferromagnetic transition, belonging to the Ising universality class
   • Disordered systems ($\phi>0$): CMRJ percolation transition occurs above thermodynamic transition, belonging to random percolation universality class, accompanied by two percolating clusters of equal density
   • Only at the thermodynamic transition point do the two clusters' densities begin to differentiate, providing percolation characteristics of the phase transition
3. Quantitative Results:
   • Determined positions of CMRJ percolation lines in the phase diagram (Figure 1)
   • Precisely extracted critical temperatures and critical exponents for various transitions (Tables I-III)
   • Verified qualitative correspondence between CMRJ cluster density difference and overlap
4. Theoretical Insights:
   • Proved equivalence of CMRJ cluster connectivity function to spin correlation function in pure ferromagnets (Appendix A)
   • Discovered "conserved-overlap transition" phenomenon (Appendix C)

Methodology Details

Task Definition

Input: $\pm J$ random-bond Ising model on three-dimensional simple cubic lattice $\hat{H}_J(S) = -\sum_{\langle x,y \rangle} J_{xy} s_x s_y$ where spins $s_x \in \{\pm 1\}$ and bond distribution is: $P_J(J_{xy}) = \phi \delta(J_{xy} + 1) + (1-\phi) \delta(J_{xy} - 1)$

Tasks:

1. Determine percolation transition temperatures $T_c$ for different cluster definitions
2. Extract critical exponents $\nu, \beta/\nu, \gamma/\nu$
3. Establish relationships between cluster densities and thermodynamic order parameters (magnetization $m$, overlap $q$)

Cluster Definitions

1. Ising Cluster

Bond occupation probability: $p_{xy}^{(\text{Ising})} = \begin{cases} 1 & \text{if } s_x = s_y \\ 0 & \text{else} \end{cases}$

2. Houdayer Cluster (Two-replica)

Based on overlap $q_x = s_x^{(1)} s_x^{(2)}$: $p_{xy}^{(H)} = \begin{cases} 1 & \text{if } q_x = q_y \\ 0 & \text{else} \end{cases}$

3. FKCK Cluster (Single-replica)

$p_{xy}^{(\text{FKCK})} = \begin{cases} 1 - \exp(-2J_{xy}s_x s_y/T) & \text{if } J_{xy}s_x s_y > 0 \\ 0 & \text{else} \end{cases}$

4. CMRJ Cluster (Two-replica core)

Based on vector spin $\tilde{s}_x = (s_x^{(1)}, s_x^{(2)})$: $p_{xy}^{(\text{CMRJ})} = \begin{cases} 1 - \exp(-2J_{xy}\tilde{s}_x \cdot \tilde{s}_y/T) & \text{if } J_{xy}\tilde{s}_x \cdot \tilde{s}_y > 0 \\ 0 & \text{else} \end{cases}$ where $\tilde{s}_x \cdot \tilde{s}_y = s_x^{(1)}s_y^{(1)} + s_x^{(2)}s_y^{(2)}$. Key property: Bonds are occupied only when simultaneously satisfied in both replicas.

Simulation Methods

Monte Carlo Updates

Combination of three update types:

1. Single-spin flip: Ensures ergodicity
2. Swendsen-Wang cluster update: Fast equilibration at high temperatures
3. CMRJ cluster update: Enhanced sampling at low temperatures
4. Parallel tempering: 20-40 temperature points with exchange acceptance rates >10%

Finite-Size Scaling Analysis

Percolation order parameter (largest cluster density): $L^{\beta/\nu}\rho_1 = \Psi_\rho(tL^{1/\nu})$

Wrapping probability: $R(t,L) = \Psi_R(tL^{1/\nu})$

Average cluster size: $\chi_\rho(T_c) \sim L^{\gamma/\nu}$

Thermodynamic order parameters (magnetization/overlap): $L^{\beta/\nu}m/q = \Psi_{m/q}(tL^{1/\nu})$

Correlation length (second-moment estimate): $\xi_{m/q}/L = \Psi_\xi(tL^{1/\nu})$

Parameter Extraction Strategy

1. Determine $T_c$ and $\nu$ through data collapse of wrapping probability/wrapping cluster number
2. Fit power law $\chi_\rho \sim L^{\gamma/\nu}$ at $T_c$
3. Fix $1/\nu$ and extract $\beta/\nu$ through order parameter collapse
4. Calculate statistical errors using bootstrapping
5. Estimate systematic errors from standard scaling corrections by varying minimum system size

Technical Innovations

1. Dual-cluster Phenomenon Identification: First systematic report of CMRJ clusters exhibiting two percolating giant clusters of equal density at the percolation transition for $\phi>0$, a unique geometric feature of frustrated systems
2. Power-Law Extrapolation Method (Appendix B): For cases with significant scaling corrections, use power-law extrapolation of effective exponents: $\theta_{\text{eff}}(L_{\min}) = \theta_0 + \frac{c}{L_{\min}^{-\tilde{\theta}_1} + 1}$ improving precision of critical exponent extraction
3. Conserved-Overlap Dynamics (Appendix C): Define auxiliary Hamiltonian with frozen local overlap $q_x = \text{const}$, discovering new phase transition above spin glass transition temperature $T_{\text{fr}} = 2.045(23)$
4. Multi-replica FKCK Extension (Appendix D): Explore FKCK clusters with $K>2$ replicas, providing direction for future search of cluster definitions exactly corresponding to thermodynamic transitions

Experimental Setup

System Parameters

• Lattice: Three-dimensional simple cubic with periodic boundary conditions
• Size Range: $L = 8$ to $L = 256$ (logarithmic scaling $L_{k+1} = 2^{1/z}L_k$, $z=4$)
• Antiferromagnetic Bond Fractions: $\phi \in \{0, 0.125, 0.5\}$

Disorder Samples

• $\phi=0$: 500,000 samples at $L=16$, 26,000 samples at $L=128$
• $\phi>0$: At least 1,000 samples for all sizes, maximum 20,597 samples ($L=32$)

Equilibration Criteria

• Bidirectional convergence from high-temperature and ground-state initial conditions
• Chain overlap reaches steady state ($\phi>0$)
• Measurement intervals $\geq 2\tau_{\text{int}}$ (integrated autocorrelation time)

Evaluation Metrics

1. Percolation quantities:
   • Largest/second-largest cluster densities $\rho_1, \rho_2$
   • Number of wrapping clusters $w_R$
   • Average cluster size $\chi_\rho$
   • Correlation length $\xi_\rho$
2. Thermodynamic quantities:
   • Magnetization $m = [\langle |\hat{m}| \rangle_S]_J$
   • Overlap $q = [\langle |\hat{q}| \rangle_{S^2}]_J$
   • Magnetic susceptibilities $\chi_m, \chi_q$
   • Correlation lengths $\xi_m, \xi_q$

Comparison Methods

• Ising cluster: Simple same-spin connectivity
• Houdayer cluster: Same-overlap domain connectivity
• FKCK cluster: Fortuin-Kasteleyn weights
• CMRJ cluster (core research object): Two-replica FKCK generalization

Experimental Results

Main Results

1. Pure Ferromagnet ($\phi=0$)

CMRJ Cluster Coincides with Ferromagnetic Transition (Figures 3-5, Table I):

• Critical temperature: $T_{\text{CMRJ}} = 4.511527(16)$ vs. $T_I = 4.511523(11)$ (literature value)
• Critical exponents:
  • $\nu = 0.6300(17)$ vs. 0.629912(86) (Ising universality class)
  • $\gamma/\nu = 1.9638(18)$ vs. 1.96390(45)
  • $\beta/\nu = 0.516(4)$ vs. 0.518 (theory)

Key Verification (Figure 4):

• Correlation length $\xi_\rho$ and correlation length $\xi_m$ show perfect collapse under identical parameters
• Proves $\langle s_x s_y \rangle_S = \text{Prob}(x,y \text{ connected})$ (theoretical derivation in Appendix A)

Density Relationship (Figure 5): $\rho_1 = m = \sqrt{q} \quad (L \to \infty)$ Second-largest cluster density $\rho_2 \to 0$, with only one macroscopic cluster.

2. Disordered Ferromagnet ($\phi=0.125$)

Dual Percolation Transitions (Figures 6-8, Table II):

Geometric Features (Figures 7-8):

• $T > T_{\text{CMRJ}}$: No percolation
• $T_f < T < T_{\text{CMRJ}}$: Two percolating clusters of equal density ($\rho_1 \approx \rho_2 \approx 0.35$)
• $T < T_f$: Density differentiation ($\rho_1 - \rho_2 \sim \sqrt{q}$)

Peak Behavior (Figure 9):

• Third-largest cluster peak density: $\rho_{\max}^{(3)} \sim L^{-0.426} \to 0$, peak position $\to T_{\text{CMRJ}}$
• Second-largest cluster peak density: $\rho_{\max}^{(2)} \to 0.35271(23)$, peak position $\to T_f = 3.235(4)$

3. Spin Glass ($\phi=0.5$)

Three-Level Transitions (Figures 11-12, Table III):

Density Evolution (Figure 14):

• High temperature: $\rho_1 = \rho_2 \approx 0.5$ (two equal-density Houdayer sub-regions)
• $T_{\text{sg}} < T < T_{\text{CMRJ}}$: $\rho_1 = \rho_2 \approx 0.48$ (two equal-density percolating clusters)
• $T < T_{\text{sg}}$: $\rho_1 - \rho_2 = q$ (density difference directly corresponds to overlap)

Key Differences:

• At spin glass transition, Houdayer cluster surface shows no critical behavior (chain overlap non-singular)
• Contrasts with ferromagnetic case (surface criticality)

Ablation Studies

Cluster Type Comparison (Figure 2, Figure 7)

• Ising cluster: Density differentiation temperature $T_{\text{cl}}^I = 4.3027(3) < T_I$, does not reflect transition
• Houdayer cluster: Density differentiation accurately corresponds to transition, but always percolates at high temperature
• FKCK cluster: For $\phi>0$, percolation temperature far above transition, loses correspondence
• CMRJ cluster: Density differentiation accurately corresponds to transition, percolation transition provides geometric precursor

Replica Number Dependence (Appendix D, Table IV)

Multi-replica FKCK clusters ($K$ replicas):

• $\phi=0, K=2$: $T_c = 2.80981(4)$ (random percolation)
• $\phi=0.125, K=2$: $T_c = 2.39619(18) < T_f$ (below ferromagnetic transition)
• $\phi=0.5, K=2$: $T_c = 1.7315(7) > T_{\text{sg}}$ (above spin glass transition)
• Conclusion: No case found with $K>1$ exactly corresponding to phase transition

Case Studies

Conserved-Overlap Transition (Appendix C, Figures 16-17)

Auxiliary dynamics with frozen local overlap $q_x(t_{\text{fr}}) = \text{const}$:

• Order parameter: $q_{\text{fr}} = [\langle |\hat{q}_{\text{fr}}| \rangle_V]_J$
• Transition temperature: $T_{\text{fr}} = 2.045(23)$ ($\phi=0.5$)
• Critical exponents: $\nu = 1.00(8)$, $\beta/\nu = 0.28(6)$, $\gamma/\nu = 2.33(10)$
• Physical Significance: Vector spin $\tilde{s}_x$ exhibits "rigidity" before entering spin glass phase

Nishimori Line Characteristics (Figure 10 inset)

On Nishimori line $T_N(\phi) = 2/\ln[(1-\phi)/\phi]$: $m = q$ Verifies theoretical prediction of correlation identities.

Experimental Findings

1. Universality Classification:
   • $\phi=0$: CMRJ percolation = Ising universality class
   • $\phi>0$: CMRJ percolation = Random percolation universality class ($\nu \approx 0.876$)
2. Dual-Cluster Mechanism: CMRJ clusters in frustrated systems are geometric sub-regions of Houdayer clusters; percolation transition corresponds to sub-region connectivity transition
3. Density-Order Parameter Correspondence:
   • Ferromagnetic: $\rho_1 - \rho_2 \sim \sqrt{q}$
   • Spin glass: $\rho_1 - \rho_2 = q$
4. Algorithm Implications: CMRJ clusters in three-dimensional spin glass already percolate near $T_{\text{sg}}$ (rigidity), limiting cluster update efficiency

Related Work

Cluster Method Foundations

1. FKCK Clusters (Fortuin & Kasteleyn 1972, Coniglio & Klein 1980):
   • Foundational work on percolation-phase transition mapping in pure ferromagnets
   • Theoretical basis of Swendsen-Wang algorithm (1987)
2. Percolation in Frustrated Systems:
   • De Arcangelis et al. (1991): First discovery of high-temperature FKCK cluster percolation in spin glasses
   • Cataudella et al. (1994): Proof that frustration breaks connectivity-correlation identities

Multi-Replica Clusters

1. Houdayer Cluster (2001):
   • Efficient algorithm for two-dimensional spin glasses
   • This paper proves its high-temperature total percolation in three dimensions
2. CMRJ Cluster:
   • Chayes, Machta & Redner (1998, 2008): Graph representation theory
   • Jörg (2005): Cluster algorithm implementation
   • Machta et al. (2008): $\rho_1 - \rho_2 = q$ in SK model
   • This paper's contribution: First systematic study of three-dimensional short-range models
3. Newman-Stein Multi-Replica FKCK (2007):
   • Theoretical tool proving spin glass symmetry breaking
   • This paper's Appendix D: No $K$ value found exactly corresponding to phase transition

Three-Dimensional ±J Model Phase Diagram

• Hasenbusch et al. (2007-2008): Disordered Ising, SG universality class confirmation
• Baity-Jesi et al. (2013): Precise SG critical exponents
• This paper supplements CMRJ percolation line (Figure 1 green crosses)

Conclusions and Discussion

Main Conclusions

1. Pure Ferromagnet ($\phi=0$):
   • CMRJ cluster percolation transition completely equivalent to ferromagnetic transition
   • Connectivity function = spin correlation function, belongs to Ising universality class
   • Single macroscopic cluster only, $\rho_1 = m$
2. Frustrated Systems ($\phi>0$):
   • CMRJ percolation transition above thermodynamic transition, belongs to random percolation universality class
   • Percolation transition produces two giant clusters of equal density (unique geometric precursor)
   • Thermodynamic transition corresponds to cluster density differentiation:
     • Ferromagnetic: $\rho_1 - \rho_2 \sim \sqrt{q}$
     • Spin glass: $\rho_1 - \rho_2 = q$
3. Universality:
   • Disordered ferromagnet ($\phi=0.125$): $T_{\text{CMRJ}} = 3.715$, $T_f = 3.241$
   • Spin glass ($\phi=0.5$): $T_{\text{CMRJ}} = 3.510$, $T_{\text{sg}} = 1.102$
   • All CMRJ percolation for $\phi>0$: $\nu \approx 0.876$, $\gamma/\nu \approx 2.045$

Limitations

1. Cluster-Correlation Correspondence:
   • CMRJ cluster connectivity function $\neq$ overlap correlation function in frustrated systems
   • No cluster definition found maintaining the identity
2. Algorithm Efficiency:
   • CMRJ clusters in three-dimensional spin glass already percolate near $T_{\text{sg}}$
   • Cluster rigidity limits acceleration effects (less pronounced than in two dimensions)
3. System Sizes:
   • Maximum $L=256$ ($\phi=0$), $L=128$ ($\phi>0$)
   • Scaling corrections require power-law extrapolation treatment
4. Phase Diagram Coverage:
   • Only three representative $\phi$ values studied
   • Behavior near multicritical point ($\phi^* = 0.23180$) not explored in detail

Future Directions

1. Improved Cluster Definitions:
   • Explore $K>2$ replicas or machine learning-constructed clusters
   • Find approaches recovering connectivity-correlation identities in frustrated systems
2. Algorithm Improvements:
   • Combine non-local structures (loops, strings)
   • Develop specialized updates for percolating clusters
3. Physical Mechanisms:
   • Deeper understanding of "conserved-overlap transition" universality
   • Study microscopic origins of dual-cluster appearance
4. Generalization Applications:
   • Extend to other frustrated models (random-field Ising, Potts model)
   • Behavior in four and higher dimensions

In-Depth Evaluation

Strengths

1. Systematicity and Completeness (★★★★★):
   • Covers complete spectrum from no frustration to strong frustration
   • Cross-comparison of multiple cluster definitions
   • Dual representation of thermodynamic and geometric quantities
2. Methodological Rigor (★★★★★):
   • Large-scale parallel tempering simulations (>20,000 samples)
   • Complete finite-size scaling analysis (8 sizes)
   • Power-law extrapolation treatment of scaling corrections (Appendix B)
   • Bootstrap error estimation
3. Physical Insights (★★★★★):
   • Dual-cluster phenomenon first systematic report (core finding)
   • Establishment of cluster density differentiation as geometric phase transition characteristic
   • Unexpected discovery of conserved-overlap transition
4. Theoretical Contributions (★★★★☆):
   • Rigorous proof of CMRJ-FKCK equivalence in pure ferromagnets (Appendix A)
   • Numerical verification of connectivity-correlation inequality in frustrated systems
5. Writing Clarity (★★★★★):
   • Clear logical structure (organized by increasing $\phi$)
   • Information-rich figures (14 main figures + 4 appendix figures)
   • Comprehensive appendices (theoretical derivations, technical details, extended results)

Weaknesses

1. Theoretical Explanation Depth (★★★☆☆):
   • Microscopic mechanism of dual-cluster appearance insufficiently clarified
   • Lack of theoretical justification for why CMRJ percolation belongs to random percolation universality class
   • Physical origins of density-order parameter correspondence ($\sim\sqrt{q}$ vs. $=q$) not deeply discussed
2. Limited Practical Value (★★★☆☆):
   • No new efficient algorithms proposed
   • CMRJ cluster algorithmic advantages in three-dimensional spin glass not apparent
   • Phase transition location determination still relies on traditional thermodynamic quantities
3. Insufficient Generalization Exploration (★★★☆☆):
   • Only three $\phi$ values; phase diagram interpolation behavior unknown
   • Other disorder distributions (e.g., Gaussian) not considered
   • Dimensional dependence only briefly mentioned in introduction
4. Technical Details:
   • Conserved-overlap transition sample size relatively small (maximum 30,875)
   • Multi-replica FKCK analysis (Appendix D) somewhat rough

Impact Assessment

Short-term Impact (1-2 years):

• Provides percolation benchmark data for three-dimensional spin glass simulations
• Guides research seeking improved cluster definitions
• Inspires further exploration of conserved dynamics phase transitions

Medium-term Impact (3-5 years):

• Likely becomes standard reference for geometric description of frustrated systems
• Dual-cluster concept possibly extends to other disordered systems
• Promotes machine learning methods for cluster construction

Long-term Impact (>5 years):

• If cluster recovering connectivity-correlation identities found, represents major breakthrough
• Provides foundation for understanding geometric nature of spin glass transitions

Citation Potential: Estimated medium-high (50-150 citations/5 years)

• Essential reference for condensed matter theory (spin glasses, percolation)
• Selective citation in computational physics (Monte Carlo methods)
• Teaching reference for statistical mechanics (critical phenomena)

Applicable Scenarios

Direct Applications:

1. Numerical studies of three-dimensional ±J Ising model (phase transition location, critical exponent extraction)
2. Benchmark testing for cluster algorithm development
3. Methodological reference for finite-size scaling analysis

Potential Extensions:

1. Other short-range spin glass models (Edwards-Anderson, random-field)
2. Geometric phase transitions in optimization problems (MAX-SAT)
3. Disorder phase transition analogies in neural networks

Inapplicable Scenarios:

1. Long-range interaction systems (require rederivation)
2. Quantum spin glasses (cluster definitions need modification)
3. Dynamical phase transitions (paper focuses on equilibrium)

Selected References

Methodological Foundations:

• 6,7 Fortuin & Kasteleyn (1972), Coniglio & Klein (1980): FKCK clusters
• 26,27 Chayes et al. (1998), Jörg (2005): CMRJ clusters
• 31 Houdayer (2001): Two-dimensional spin glass cluster algorithm

Phase Diagram Data:

• 41-45 Hasenbusch et al. (2007-2008): Three-dimensional ±J model phase diagram
• 49 Baity-Jesi et al. (2013): Spin glass critical exponents

Theoretical Background:

• 12 Binder & Young (1986): Spin glass review
• 52,53 Nishimori (1980-1981): Nishimori line theory

Overall Assessment: This is a high-quality computational physics research paper making substantial contributions to geometric description of frustrated spin systems. The discovery and systematic characterization of the dual-cluster phenomenon demonstrates originality, with solid and reliable numerical work. Main limitations lie in theoretical explanation depth and practical algorithm development. Suitable for publication in Physical Review E or similar computational condensed matter physics journals, with significant reference value for spin glass and percolation theory communities.