Quantitative analysis of non-equilibrium systems from short experiments

Non equilibrium systems abound in nature on all scales. However, general principles that guide their working are not known. A common feature of ALL non-equilibrium systems is however their  constant production of entropy.

This entropy production rate (EPR) is a fundamental quantitative measure of the irreversibility of the dynamics andin addition carries information about efficiencies of processes, heat generated by computations, hidden degrees of freedom, rates of energy transfer and optimal operating conditions.

Reliably measuring the EPR for a system using currently available methods however, requires an enormous amount of data and hence, despite rapid advancements in the field, there has not yet been any breakthroughs in the science of nanoscale artificial or living matter directly linked to entropy production.

In this project, we address exactly this issue by proposing a method recently found by us, that requires only very short time series data  to measure entropy production.When used in conjunction with a machine learning algorithm that we have developed, no details need be known about the system of interest.

Hence quantitative information can be obtained about a non-equilibrium system without the need for simplifying models or assumptions.

We propose to include experimental and theory collaborations to test this scheme, develop the theory of the functioning of microscopic machines at short-time scales as well as create repositories of machine-learning algorithms.