Geometry of the energy landscape of the self-gravitating ring

TL;DR

This paper investigates the geometric structure of the energy landscape in a self-gravitating ring model, revealing how curvature properties characterize phase transitions and estimating chaos through geometric methods.

Contribution

It introduces a geometric approach to analyze the energy landscape of the self-gravitating ring, linking curvature to phase transitions and chaos.

Findings

Curvature indicates a phase transition from flat to curved landscape.

Curvature fluctuations peak at a secondary transition energy.

Geometric estimates of Lyapunov exponents match numerical results.

Abstract

We study the global geometry of the energy landscape of a simple model of a self-gravitating system, the self-gravitating ring (SGR). This is done by endowing the configuration space with a metric such that the dynamical trajectories are identified with geodesics. The average curvature and curvature fluctuations of the energy landscape are computed by means of Monte Carlo simulations and, when possible, of a mean-field method, showing that these global geometric quantities provide a clear geometric characterization of the collapse phase transition occurring in the SGR as the transition from a flat landscape at high energies to a landscape with mainly positive but fluctuating curvature in the collapsed phase. Moreover, curvature fluctuations show a maximum in correspondence with the energy of a possible further transition, occurring at lower energies than the collapse one, whose existence had been previously conjectured on the basis of a local analysis of the energy landscape and whose effect on the usual thermodynamic quantities, if any, is extremely weak. We also estimate the largest Lyapunov exponent $\lambda$ of the SGR using the geometric observables. The geometric estimate always gives the correct order of magnitude of $\lambda$ and is also quantitatively correct at small energy densities and, in the limit $N\to\infty$, in the whole homogeneous phase.