Numerically Discovered Inherent States are Always Protocol Dependent in Jammed Packings

TL;DR

This study reveals that the inherent states in jammed packings are highly protocol-dependent, with the ability to find these states diminishing as system size increases, due to the energy landscape's saddle points.

Contribution

It demonstrates that the process of numerically discovering inherent states is inherently protocol-dependent and becomes unreliable for larger systems, highlighting limitations in current computational methods.

Findings

The optimal time step for finding inherent states scales as N^{-3}.

Reliability of finding inherent states decreases with system size, especially beyond 64 particles.

The energy landscape's saddle points influence the difficulty of locating inherent states.

Abstract

The energy landscape for soft sphere packings exists in a high-dimensional space and plays host to an astronomical number of local minima in a hierarchical and ultrametric arrangement. Each point in the landscape is a configuration that can be unambiguously mapped to its inherent state, defined as the local minimum that the configuration will flow to under perfectly overdamped continuous dynamics. Typically, discrete in time dynamics are used to computationally find local minima, but it is not known whether these algorithms are capable of reliably finding inherent states. Here, we use steepest descent dynamics to find the distribution of the largest time step, $\delta_\textrm{best}$, which finds the inherent state. We find that for systems of $N$ particles, $\delta_\textrm{best}$ is approximately proportional to $N^{-3}$, and weakly dependent on d and $\varphi$. We argue that the proportionality is due to saddle points in the energy landscape. Our results suggest that it is impossible, in practice, to reliably find inherent states for systems of about 64 particles or more.