Transporting Baguettes With Minimal Action: The Geometry of Optimal Nonequilibrium Processes in Stochastic Thermodynamics

Abstract: What are the fundamental limitations placed by the laws of thermodynamics on the energy expenditure needed to carry out a given task in a nonequilibrium environment in finite time? In this thesis, we investigate "optimal nonequilibrium processes": how nonequilibrium state changes in a thermodynamic system may be performed most efficiently, in the sense of requiring the least amount of thermodynamic work. Surprisingly, there is a hidden, fundamental geometric structure in this optimization problem that is related to the mathematics of optimal transport theory: how to optimally send, e.g., baguettes from bakeries to caf\'es, given supply and demand constraints, that requires the least amount of total distance traveled by the baguettes. After giving a brief overview on the mathematical framework of stochastic thermodynamics for the overdamped Langevin equation, we present a trio of works: (1) applying optimal control theory to the Fokker-Planck equation to calculate exact optimal protocols for low-dimensional systems, which reproduces the intriguing previously-discovered discontinuities in globally optimal protocols and reveals new non-monotonic optimal protocols for a certain system; (2) exploiting the importance sampling of Langevin trajectories under different protocols, to adaptively optimize protocols by controlling the thermodynamic state trajectory, which is useful for efficiently calculating free energy differences between different Hamiltonians; and finally, (3) deriving an exact geometric description of optimal nonequilibrium processes and a geodesic-counterdiabatic decomposition for the optimal protocols that enact them, which satisfyingly explains the highly non-intuitive properties of discontinuities and possible non-monotonicity in globally optimal protocols.