Memory-induced excitability in optical cavities

Neurons and other excitable systems can release energy suddenly given a small stimulus. Excitability has recently drawn increasing interest in optics, as it is key to realize all-optical artificial neurons enabling speed-of-light information processing. However, the realization of all-optical excitable units and networks remains challenging. Here we demonstrate how laser-driven optical cavities with memory in their nonlinear response can sustain excitability beyond the constraints of memoryless systems. First we demonstrate different classes of excitability and spiking, and their control in a single cavity with memory. This single-cavity excitability is limited to a narrow range of memory times commensurate with the linear dissipation time. To overcome this limitation, we explore coupled cavities with memory. We demonstrate that this system can exhibit excitability for arbitrarily long memory times, even when the intercavity coupling rate is smaller than the dissipation rate. Our coupled-cavity system also sustains spike trains – a hallmark of neurons – that spontaneously break mirror symmetry. Our predictions can be readily tested in thermo-optical cavities, where thermal dynamics effectively give memory to the nonlinear optical response. The huge separation between thermal and optical timescales in such cavities is promising for the realization of artificial neurons that can self-organize to the edge of a phase transition, like many biological systems do.