Spiking oscillators empower heat dissipation to compute

Heat dissipation is a universal behavior commonly existing in our world, particularly within semiconductor chips. However, heat dissipation imposes significant limitations on the downscaling size of the semiconductor chips. In our research, we leverage the concept of heat dissipation between spiking oscillators for computational purposes. Instead of using electricity, information is transmitted through the medium of heat. Our investigation delves into the evolving synchronization patterns of spiking nano-oscillators arising from thermal interactions. Additionally, we showcase a wide range of reconfigurable electrical dynamics that mimic those of biological neurons, all achieved through heat dissipation. Notably, we achieve inhibitory functionality using a single oxide device, and the flow of information is exclusively driven by thermal interactions in a cascading manner. This study lays the groundwork for the development of scalable and energy-efficient thermal neural networks, thereby advancing the field of brain-inspired computing.