### Current browse context:

cond-mat.stat-mech

### Change to browse by:

### References & Citations

# Condensed Matter > Statistical Mechanics

# Title: Inclusive Thermodynamics of Computational Machines

(Submitted on 2 Jun 2022)

Abstract: We introduce a framework to analyze the thermodynamics of a physical system that first implements an abstractly defined computational machine and then is reinitialized, completing a full thermodynamic cycle. Our starting point is to derive a lower bound on the dissipated work that occurs in a complete cycle of any classical or quantum physical process that implements a given computational machine. It turns out that this lower bound on dissipated work exactly equals the expected entropy production arising in the Hamiltonian formulation of stochastic thermodynamics. This leads us to derive an integral fluctuation theorem for the dissipated work of any system modeled using the Hamiltonian formulation, along with a mismatch cost theorem. Explicitly connecting our framework with computer science, we analyze how the lower bound on dissipated work depends on the size complexity of a deterministic finite automaton. We prove that given a set of equivalent automata, the one with minimal size complexity is the one with minimal dissipation. This result is valid for all deterministic finite automata and at all iterations. We also provide a modified version of the exchange fluctuation theorem for computational machines. In addition to considering a deterministic finite automaton, we present results concerning the thermodynamics of a Markov information source (i.e., a hidden Markov model), and those concerning the possible use of rate distortion theory in Shannon communication channels.

Link back to: arXiv, form interface, contact.