Efficient calculations of equation-of-state data in the warm-dense matter regime
ORAL
Abstract
Equation-of-state (EoS) data — relating the pressure and internal energy to material density and temperature — is a key quantity in the warm dense matter regime, for example as input to hydrodynamics codes used to guide inertial confinement fusion experiments. The first-principles methods, density-functional theory and path-integral Monte–Carlo, are considered state-of-the-art approaches to calculate EoS data. However, both methods are computationally expensive, which motivates the development of low-cost approaches such as average-atom models. In the first part of this talk, we benchmark EoS results from an average-atom model against the extensive first-principles dataset from Militzer et al. (Phys. Rev. E 103, 013203). In the second part, we develop a neural-network surrogate model as a numerically feasible alternative to calculating EoS data. We train two neural networks to interpolate this dataset, with one being trained using average-atom outputs and the other without. We also compare the accuracy of the machine-learned and average-atom models using out-of-distribution data from other sources.
*We acknowledge funding by the Center for Advanced Systems Understanding (CASUS) which is financed by the German Federal Ministry of Education and Research (BMBF) and by the Saxon Ministry for Science, Culture, and Tourism (SMWK) with tax funds on the basis of the budget approved by the Saxon State Parliament.
–
Publication: Planned paper: "Efficient calculations of equation-of-state data using average-atom and neural-network surrogate models" [T. J. Callow, E. Kraisler and A. Cangi]
Presenters
-
Timothy J Callow
- Center for Advanced Systems Understanding (CASUS)