Bayesian methods to extract cold curves from shockless compression experiments on the Z machine

Statistical methods are becoming ubiquitous in the data sciences and many of the advanced techniques are now being leveraged to better understand the results from dynamic material experiments. Here, we examine how velocimetry measurements in multi-megabar shockless compression experiments can be coupled to hydrocode simulations to constrain the reference isotherm in equation of state development. Bayesian inference is used to incorporate all known sources of error across multiple data sets for rigorous uncertainty quantification, while Monte Carlo methods are used to understand relative sensitivities. Unexpected problems with high-pressure sensitivity emerge when using conventional parametric model forms for the cold curve; a novel non-parametric model form is suggested as a better approach. Examples from materials with and without phase transformations including platinum compressed to 600 GPa and tin loaded to 120 GPa will be presented and compared to traditional analysis methods.