Stochastic Thermodynamics of Non-Linear Electronic Circuits: A Realistic Framework for Thermodynamics of Computation

The rigorous description of intrinsic thermal noise in complex, non-linear and out of equilibrium electronic circuits is a problem of fundamental as well as practical importance, of relevance for the design of new computing schemes that are energetically efficient. In this contribution we develop a formalism that allows us to construct thermodynamically consistent stochastic models of arbitrary circuits. It is thus able to accomodate large classes of technologically relevant circuits (for example, single electron and CMOS devices). The formalism is based on the theory of stochastic thermodynamics, which allows to provide a full thermodynamic characterization of the system and its entropy production, and to derive different fluctuation theorems. As a first application, we perform a full analysis of a CMOS inverter, or NOT gate. Based on this elementary example, we propose a design for a binary stochastic neuron, which employs intrinsic thermal noise as a resource. This can be considered as a generator of random bits with controllable rate and bias, and is the basis for the physical implementation of artificial neural networks or stochastic annealers.