講演者： 大賀成朗（東大物理学科）

日時：6月20日（火）16:00-

形式：対面、5号館401

タイトル：Universal thermodynamic bounds on two-time correlations

アブストラクト：

Two-time correlations (auto- and cross-correlations) are widely used to study system dynamics in many areas of physics. They are measurable in most real-world systems, and at the same time, they are informative and quantitatively related to a variety of physical functions, such as mixing, oscillation, response, velocity, and information transfer. Therefore, finding universal relations between thermodynamic costs and correlation functions is crucial in two ways: it reveals a fundamental

thermodynamic cost of these physical functions, and it provides a tool to infer thermodynamic quantities from measurable correlations. In this context, several thermodynamic bounds have been discovered for various aspects of the shape of correlation functions, including auto-/cross-correlations, short-/finite-/long-time regime, and spectral density.

In this seminar, after a general introduction to two-time correlations, I discuss two lines of research. The first is the recent

series of thermodynamic bounds on the asymmetry of cross-correlations [1–3]. The asymmetry is a fundamental and measurable signature of nonequilibrium, as it quantifies the violation of microscopic reversibility, and it captures physical processes such as directed information flow, nonequilibrium circulation, non-reciprocal motion, and anomalous response. I discuss the results, with particular emphasis on the relation to previous thermodynamic bounds. The second is a set of thermodynamic bounds on the eigenvalues of the generator, which (in part) characterizes the long-time behavior of all two-time

correlations. Following the recently conjectured bounds [4–6], we recently proved one of them [1,4]. We also newly found and proved a thermodynamic bound on the spectral perturbations, i.e., the difference of eigenvalues between a non-detailed-balanced system and its detailed-balanced counterpart [7]. I discuss these results with the mathematical background of the eigenvalues of stochastic matrices.

[1] N. Ohga, S. Ito, A. Kolchinsky, PRL in press (arXiv:2303.13116) (2023).

[2] N. Shiraishi, arXiv:2304.12775 (2023).

[3] T. Van Vu, V. T. Vo, K. Saito, arXiv:2305.18000 (2023).

[4] A. C. Barato, U. Seifert, PRE 95, 062409 (2017).

[5] M. Uhl, U. Seifert, J. Phys. A 52, 405002 (2019).

[6] L. Oberreiter, U. Seifert, A. C. Barato, PRE 106, 014106 (2022).

[7] A. Kolchinsky, N. Ohga, S. Ito, arXiv:2304.01714 (2023).