Complete Realization of Energy Landscape and Non-equilibrium Trapping Dynamics in Spin Glass and Optimization Problem

TL;DR

This paper introduces a novel method to fully analyze the energy landscapes and non-equilibrium dynamics of small spin glasses and satisfiability problems, revealing unexpected behaviors like decreased ground state identification at lower temperatures.

Contribution

It presents a new approach to unveil complete energy landscapes and long-time dynamics of glassy systems, including a variant for larger systems, advancing understanding of their complex behaviors.

Findings

Decreased likelihood of finding ground states at lower temperatures due to local minima trapping.

Multiple abrupt jumps in ground-state probability over time.

Similar phenomena observed in large systems with a partial landscape extraction approach.

Abstract

Energy landscapes are high-dimensional surfaces representing the dependence of system energy on variable configurations, which determine crucially the system's emergent behavior but are difficult to be analyzed due to their high-dimensional nature. In this article, we introduce an approach to reveal the complete energy landscapes of small spin glasses and Boolean satisfiability problems, which also unravels their non-equilibrium dynamics at an arbitrary temperature for an arbitrarily long time. In contrary to our common belief, our results show that it can be less likely to identify the ground states when temperature decreases, due to trapping in individual local minima, which ceases at different time, leading to multiple abrupt jumps with time in the ground-state probability. Simulations agree well with theoretical predictions on these remarkable phenomena. Finally, for large systems, we introduce a variant approach to extract partially the energy landscapes and observe both analytically and in simulations similar phenomena. This work introduces new methodology to unravel the non-equilibrium dynamics of glassy systems, and provides us with a clear, complete and new physical picture on their long-time behaviors inaccessible by modern numerics.