Abstract
We investigate the Lévy glass, a mean-field spin-glass model with power-law distributed couplings characterized by a divergent second moment. By combining extensively many small couplings with a sparse random backbone of strong bonds, the model is intermediate between the Sherrington-Kirkpatrick and the Viana-Bray models. A truncated version where couplings smaller than some threshold $\epsilon$ are neglected can be studied within the cavity method developed for spin glasses on locally treelike random graphs. By performing the limit $\epsilon \to 0$ in a well-defined way, we calculate the thermodynamic functions within replica symmetry and determine the de Almeida-Thouless line in the presence of an external magnetic field. Contrary to previous findings, we show that there is no replica-symmetric spin-glass phase. Moreover, we determine the leading corrections to the ground-state energy within one-step replica symmetry breaking. The effects due to the breaking of replica symmetry appear to be small, in accordance with the intuitive picture that a few strong bonds per spin reduce the degree of frustration in the system.
Introduction: Bridging the Gap in Spin Glass Theory
Spin glasses are complex magnetic systems where random, competing interactions lead to "frustration"—a state where no single spin configuration can satisfy all local constraints. Traditional models like the Sherrington-Kirkpatrick (SK) model assume weak, Gaussian-distributed interactions between all pairs. However, many physical systems, such as magnetic impurities in metals interacting via the RKKY interaction, are better described by Lévy distributions. These "heavy-tailed" distributions feature a few exceptionally strong interactions amidst a sea of weak ones.
This research re-examines the Lévy spin glass, a model that sits between the fully connected SK model and the sparse Viana-Bray model. By applying the cavity method, the authors challenge previous assumptions about the stability of the replica-symmetric (RS) phase and provide a rigorous thermodynamic description of these systems.
Methodology: The Truncated Cavity Approach
To address the mathematical challenges posed by the divergent second moment of Lévy distributions, the authors utilize a truncated model. By introducing a threshold $\epsilon$, interactions weaker than this value are ignored, effectively creating a sparse graph that can be analyzed using the cavity method. This method focuses on the distribution of local magnetic fields rather than the global partition function.
The full Lévy glass is recovered by taking the limit $\epsilon \to 0$. To solve the resulting self-consistent equations for local field distributions, the authors employ population dynamics, a numerical algorithm that evolves a set of fields toward a steady state. To account for the complexities of the low-temperature phase, they also calculate corrections using one-step replica symmetry breaking (1-RSB).
Key Results and Insights
The Phase Diagram and Critical Temperature
A central finding is the mapping of the transition temperature $T_c$ against the Lévy exponent $\alpha$. As $\alpha$ increases toward 2 (the Gaussian limit), the critical temperature rises sharply. The model demonstrates that the "backbone" of strong bonds provides a structural framework, while the background of weak bonds creates a mean-field environment.
Local Field Distributions and the Instability of Symmetry
Contrary to earlier theories, this study proves that there is no stable replica-symmetric phase at low temperatures. The de Almeida-Thouless (AT) line, which marks the boundary of RS stability, coincides with the freezing temperature in zero field. This implies that the system immediately enters a complex phase characterized by replica symmetry breaking upon cooling.
Furthermore, the local magnetic fields in the Lévy glass are distinctly non-Gaussian, a departure from the SK model. This non-Gaussianity is crucial for capturing the correct physics of the heavy-tailed interactions, as shown in the distribution of local fields $P(h)$.
Frustration and Ground State Energy
Interestingly, the study finds that smaller values of $\alpha$ (heavier tails) actually reduce the overall frustration in the system. The few dominant strong bonds effectively "lock" the spins into specific orientations, making the conflicting weak bonds less influential. While the RS solution is technically unstable, 1-RSB calculations show that the energy corrections are small (approximately 1-2%), suggesting that the symmetric solution remains a reasonable first-order approximation for the ground-state energy.
Conclusion and Future Directions
This work provides a definitive thermodynamic treatment of the Lévy spin glass, correcting long-standing misconceptions about its phase stability. While the 1-RSB corrections are small, future research could explore the full RSB hierarchy to achieve an exact solution at absolute zero. Additionally, extending this mean-field analysis to finite-dimensional systems remains a significant challenge for understanding real-world magnetic materials.
Summary updated