Speaker:  Jean-Charles Delvenne (University of Louvain)

Date: Oct. 6th (Mon.) 15:30- 

Place: 413 Building No.5

Title: Moments of entropy production

Abstract: 

Is there a real-valued random variable whose first, second and third moments are 1, 3 and 2 respectively? What about a positive random variable? This is an example of the 'moment problem' in probability theory, which was essentially solved by Stieltjes, yielding a complete list of inequalities satisfied by all moments of a (real or positive) random variable.

In this talk we will solve the following variant: What are all inequalities satisfied by all moments of entropy production of a system verifying local detailed condition?

We will then focus on the case of white noise. We show that the third moment is an upper bound on the violation of fluctuation-dissipation in purely dissipative devices. We exemplify our bounds on real-life devices such as transistors and diodes.

The case of cross-moments, such as higher-order auto-correlations for overdamped Markov dynamics, is wide open and highly relevant for  real-life nonlinear devices.

Reference: Delvenne, J. C., & Van Brandt, L. (2025). Moments of Entropy Production in Dissipative Devices. Physical Review Letters, 134(23), 237101.

c