Thermodynamics of information processing in small systems (Q2895884)

In ``Thermodynamics of Information Processing in Small Systems'', based on the authors PhD-thesis, T. Sagawa reviews latest developments in this active research field.NEWLINENEWLINE The paper starts with a short historical sketch about thermodynamics and then relates to modern developments such as the fluctuation theorem and Jarzynski equality. Readers will find many valuable references here. In the second chapter, Maxwell's demon is reviewed, and the importance of information is explained with the example of a Szilard engine. Afterwards, the arguments by Brillouin, Landauer, and Bennett aiming to resolve the paradox of Maxwell's demon are discussed. One main result of the paper is that neither only the measurement (Brillouin) nor only the erasure (Landauer, Bennett) work are sufficient to resolve the paradox of apparent entropy reduction via feedback control. To support this result, in chapters 3 and 4 mathematical concepts of classical and quantum dynamics, measurement and information are reviewed, respectively. This includes Shannon entropy, Kullback-Leibler divergence, mutual information and some examples for erroneous measurements on the classical side and their quantum generalizations. In the 5th chapter, the second law is proven for a unitarily evolving system that starts from an initial canonical distribution. As the most important contribution, chapter 6 discusses how the second law is modified by the presence of feedback control. For this purpose, the evolution is partitioned into an unconditioned unitary propagation, a measurement, a following unitary propagation that is conditioned on the measurement, and a final unitary propagation which is again unconditioned. For a single sequence of these operations it turns out that the mutual information defined in the previous chapters now enters the second law. Finally, chapter 7 discusses the thermodynamics of memories, and it is argued that the sum of work required for information erasure and measurement has to exceed the excess work related to the mutual information.NEWLINENEWLINE The paper is well-written with a pedagogical style and self-contained. Though it might have profited a bit from a more detailed discussion of physically realistic models, it is well suitable for readers aiming to understand the thermodynamics of feedback control with little background in thermodynamics and quantum theory.