A free energy landscape analysis of resistance fluctuations in a memristive device.

Resistance noise in memristive devices is often attributed to simple thermally activated processes, such as fluctuations across single energy barriers. However, this picture may underestimate the complexity of the underlying atomic dynamics, which can be described as transitions between many local minima in a high-dimensional free energy landscape shaped by energetic and entropic contributions, yet such landscapes are difficult to access experimentally. Using a hidden Markov model, we analyse resistance fluctuations in a nanoscopic volume of the phase-change material germanium telluride. We quantify the transition rates between discrete resistance states over a wide temperature range. The rates follow an Arrhenius-like behaviour, but the extracted attempt frequencies span several orders of magnitude and include values far below typical phonon frequencies. This spread reflects substantial entropic contributions to the free energy barriers, which we quantify by tracking individual transitions across temperatures. This approach should be broadly applicable to memristive materials, where significant resistance changes are linked to atomic-scale transitions.