Nanoscale Thermal Neuromorphic Computing

Thermal computing is an emerging computational approach that involves using heat currents instead of electrical currents to perform logical operations. In this talk, I will describe theoretical advances toward the design of thermal neuromorphic computing devices that operate at the nanoscale. The field of neuromorphic computing has gained significant attention due to its potential applications in new computing architectures that increase computing efficiency by co-locating memory and processing units. Here, we apply a theoretical framework that employs stochastic energetics methods and a nonequilibrium Green's function approach to examine vibrational heat flow in a model nanoscale system. The model system consists of a lattice of interconnected particles bridging two heat baths with time-dependent oscillating temperatures. I will show that the oscillating thermal gradient generated across the lattice can induce phenomena such as heat current hysteresis and controllable energy storage which are absent in systems with static temperatures. These phenomena are then discussed in the context of designing thermal memristors and memcapacitors—important neuromorphic computing components—at the molecular level. The primary outcome of the work is the demonstration in a foundational model that nanoscale systems can exhibit the transport functionalities needed for thermal neuromorphic applications.