Thermodynamically Informed Priors for Uncertainty Propagation in First-Principles Statistical Mechanics

TL;DR

This paper introduces a Bayesian approach incorporating zero Kelvin ground state knowledge to quantify and propagate uncertainties in thermodynamic calculations, demonstrated through a phase diagram of FCC Zirconium Nitride.

Contribution

It presents a novel thermodynamically informed prior within a Bayesian framework for uncertainty quantification in first-principles statistical mechanics.

Findings

Created a phase diagram with confidence intervals for Zirconium Nitride.

Effectively propagated uncertainties in thermodynamic calculations.

Incorporated zero Kelvin ground state knowledge into prior selection.

Abstract

This work demonstrates how first-principles thermodynamic research within a Bayesian framework can quantify and propagate uncertainties to downstream thermodynamic calculations. To address the issue of Bayesian prior selection, knowledge of zero Kelvin ground states in the material system of interest is incorporated into the prior. The effectiveness of this framework is shown by creating a phase diagram for the FCC Zirconium Nitride system, including confidence intervals on phase boundary regions of interest.