[Paper Review] On the TAP approach to the spherical p-spin SG model

A. Crisanti, H.-J. Sommers|arXiv (Cornell University)|Jun 13, 1994

Quantum Chromodynamics and Particle Interactions1 references3 citations

TL;DR

This paper applies the TAP (Thouless-Anderson-Palmer) approach to the spherical p-spin spin glass model, deriving the TAP free energy via infinite-N diagrammatic expansion. It reveals a first-order, 'geometrical' phase transition at a critical temperature Tc higher than the replica symmetric solution's T_RSB, but consistent with dynamical predictions, driven by the dominance of high-free-energy states with maximal phase space volume.

ABSTRACT

In this letter we analyze the TAP approach to the spherical $p$-spin spin glass model in zero external field. The TAP free energy is derived by summing up all the relevant diagrams for $N o\infty$ of a diagrammatic expansion of the free energy. We find that if the multiplicity of the TAP solutions is taken into account, then there is a first order transition in the order parameter at the critical temperature $T_{ m c}$ higher than that predicted by the replica solution $T_{ m RSB}$, but in agreement with the results of dynamics. The transition is of ``geometrical'' nature since the new state has larger free energy but occupies the largest volume in phase space. The transition predicted by the replica calculation is also of ``geometrical'' nature since it corresponds to the states with smallest free energy with positive complexity.

Motivation & Objective

• To investigate the thermodynamic behavior of the spherical p-spin spin glass model using the TAP approach.
• To derive the TAP free energy through a systematic diagrammatic expansion in the N → ∞ limit.
• To resolve discrepancies between replica symmetric solutions and dynamical predictions regarding the phase transition temperature.
• To understand the nature of the phase transition in terms of free energy and phase space volume.
• To clarify the role of solution multiplicity in determining the true thermodynamic state.

Proposed method

• The TAP free energy is derived by summing all relevant diagrams in a diagrammatic expansion of the free energy for the spherical p-spin model.
• The analysis is performed in the thermodynamic limit (N → ∞), ensuring the validity of the mean-field approximation.
• The method accounts for the multiplicity of TAP solutions, which is crucial for determining the dominant phase.
• The free energy landscape is evaluated by considering both the energy and the entropy (complexity) of TAP solutions.
• The transition is analyzed by comparing the free energy and phase space volume of different solution branches.
• The approach distinguishes between thermodynamically stable states and those selected by volume dominance, even if higher in free energy.

Experimental results

Research questions

• RQ1 What is the correct critical temperature for the phase transition in the spherical p-spin spin glass model according to the TAP approach?
• RQ2 How does the multiplicity of TAP solutions influence the selection of the thermodynamically dominant state?
• RQ3 Does the TAP approach predict a first-order transition, and if so, what is its nature?
• RQ4 How does the TAP prediction for Tc compare with the replica symmetric solution and dynamical results?
• RQ5 Why does the new phase have higher free energy yet dominate due to geometrical factors?

Key findings

• The TAP approach predicts a first-order phase transition at a critical temperature Tc that is higher than the replica symmetric solution's T_RSB.
• The transition is of 'geometrical' nature, as the new phase has higher free energy but occupies the largest volume in phase space.
• The TAP solution correctly reproduces the dynamical transition temperature, resolving a discrepancy with the replica method.
• The replica symmetric solution also corresponds to a geometrical transition, arising from states with positive complexity and minimal free energy.
• The dominance of high-free-energy states is due to their exponentially larger phase space volume, not thermodynamic stability.
• The analysis confirms that solution multiplicity is essential for identifying the true thermodynamic state in spin glass systems.