Title: Exact, complete expressions for the thermodynamic costs of circuits

Abstract: The generalized Landauer's bound gives the minimal amount of heat generated by any physical process that implements a given computation, when there are no constraints on the process. However, common engineered systems implement computations using circuits, which introduces constraints on the physical process. So the minimal amount of heat that a circuit generates is greater than the minimal amount of heat required to implement its computation.
To analyze this increase in the minimal heat, we first decompose the heat generated by any physical process that implements a given dynamics, into a sum of three terms. The first term, the drop of entropy of the state distribution, is independent of the details of the physical process, beyond the fact that it implements the desired dynamics. The second and third terms instead reflect the "nitty gritty details" of the process. The second term has a purely information-theoretic character, while the third term is a linear function of the initial distribution of states. (As an important example, we show that all of the heat generated by wires in circuits arises in this third term.) Next, we introduce a special kind of process, "subsystem processes", which formalize what it means to run a particular gate in a circuit.
We then use these results to derive an exact equation for the amount of heat generated by any physical process that implements a specified circuit, i.e., that runs all the gates in the circuit, in succession. This equation provides a new optimization problem for circuit designers: how to design a circuit that both implements a desired computation and generates minimal heat. Interestingly, the objective function of this optimization problem involves functions that are information-theoretic in character but are apparently new to the information theory literature.