Collective dynamics and memory-induced long-range order in spiking oscillator arrays*

Recent work has experimentally demonstrated a new class of spiking oscillators known as "thermal neuristors," [1] which operate through thermal interactions between adjacent vanadium dioxide resistive memories. Here, we show large-scale simulations of the dynamics of both 1D and 2D arrays of thermally coupled neuristors. These simulations reveal a rich phase structure, tunable through memory strength and input voltage. Importantly, we observe a robust phase of memory-induced long-range order, which stands in stark contrast to the fragile long-range order typically observed at phase transition points. As an application, we demonstrate that a 2D array of thermal neuristors can function as a reservoir for reservoir computing, opening up new possibilities for tasks like image recognition.

 

[1] Qiu, E., Zhang, Y. H., Di Ventra, M., & Schuller, I. K. (2023). Reconfigurable cascaded thermal neuristors for neuromorphic computing. Advanced Materials, 2306818.