Supplementary code to "Leveraging interactions for energy-efficient swarm-based Brownian computing"

Supplementary code for Leveraging interactions for energy-efficient swarm-based Brownian computing by A. Pignedoli, A. Majumdar, and K. Everschor-Sitte, Physical Review Research (2026), doi:10.1103/k8wj-xsl7 (preprint: arXiv:2601.22874).

This archive contains the complete simulation and analysis pipeline used to produce the results and figures of the manuscript. A C++ implementation of the Gillespie algorithm (kinetic Monte Carlo) generates stochastic trajectories of N interacting Brownian quasiparticles on a two-dimensional 20×20 lattice with nearest-neighbor interactions, subject to a spatially varying temperature field.

We simulate two scenarios: a static mode with a fixed temperature profile, measuring the success ratio with which the swarm locates the global temperature minimum, and a dynamic mode in which the profile switches at a prescribed time, measuring adaptation accuracy and timescale. Accompanying Python code processes the raw trajectories into the manuscript results.

The repository includes a Docker environment that reproduces the configuration used to obtain the published results, and a reduced test configuration that runs the same pipeline in minutes. 

Released under the MIT License.