Direct and universal bounded entropy evaluation in complex simulations

Complex physical simulations are ubiquitously employed to characterize thermodynamics in diverse systems; free-energy, or related quantities, are often their goal. While enthalpy is calculable using the apriori choice of interactions (i.e., force-field, coupling parameters), entropy remains a challenge to quantify directly. Typically, methods for free-energy estimation are based on thermodynamic relations (e.g., Jarzynski’s equality) rather than the independent statistics of the target ensembles.

We have previously demonstrated a universal entropy calculation scheme using loss-less compression algorithms which delivers a tight upper-bound estimate [1]. In this talk, I will describe another novel information-based approach for estimation of entropy. This new approach establishes tight lower and upper entropy bounds for complex simulations. Our novel method, based on counting statistics, is inherently universal and provides a direct estimate for entropy along with margins of error. Given this method’s universality and statistical guarantees, we expect it to be highly useful for both simulated and experimentally recorded systems.

[1] R. Avinery, M. Kornreich, R. Beck (2017) arXiv:1709.10164