Optimal control of dissipation in stochastic logic circuits

Landauer's principle and developments from stochastic thermodynamics provide insight into the energetic costs of logical operations. As transistors on chips become smaller, one must begin to consider the effects of thermal fluctuations that approach the magnitude of the correspondingly smaller operating voltages. We use a thermodynamically consistent model of complementary metal-oxide-semiconductor (CMOS) logic circuits to explore energetic costs. Leveraging a numerically robust framework built on mean field dynamics and automatic differentiation, we optimize time-dependent control protocols that reveal physical design principles for device operation. Using single bit erasure as a building block, we consider how to generalize design principles for more complex noisy, nanoscale circuits.