Diffusive memristors in the edge of chaos

Biological nervous systems are theorized to operate in the edge of chaos (EOC), a stable and locally active regime enabling energy-efficient information processing. Electronic implementations of such behavior have mainly relied on electrothermal phase transitions, which are power-intensive and unlike biological dynamics. Diffusive memristors, by contrast, harness ionic drift and diffusion resembling biological dynamics, yet their potential for EOC operation has remained unexplored. Here, we experimentally demonstrate ionically moderated EOC behavior in diffusive memristors. At the hundred-nanoampere level, the device exhibits sharp negative differential resistance (NDR) enhanced by a Ridley entropy production minimization transition, noise amplification and spontaneous oscillatory dynamics consistent with operation near the EOC. These emergent behaviors arise from coupled ionic transport, filament evolution and circuit dynamics, as confirmed by compact modeling. This system achieves power efficiency over three orders of magnitude greater than electrothermal EOC devices, introducing low-power neuromorphic elements that emulate brain-like dynamics through nanoionics. Biological nervous systems operate at the edge of chaos for efficient information processing. Kim et al. realize such biological dynamics in diffusive memristors, achieving three orders of magnitude higher power efficiency than other devices with comparable functionalities.