Uncertainty and Exploration of Deep Learning-based Atomistic Models for Screening Molten Salt Properties and Compositions

TL;DR

This paper demonstrates how deep learning models can predict molten salt properties with quantified uncertainty, enabling more reliable screening of compositions and conditions relevant to nuclear reactors.

Contribution

It introduces ensemble learning methods to quantify neural network uncertainty, improving confidence in DL-based property predictions for complex molten salts.

Findings

Neural network uncertainty can be effectively quantified using ensemble methods.

DL models can extrapolate to new compositions and conditions with confidence.

Uncertainty predictions highlight limitations in density changes, guiding model deployment.

Abstract

Due to extreme chemical, thermal, and radiation environments, existing molten salt property databases lack the necessary experimental thermal properties of reactor-relevant salt compositions. Meanwhile, simulating these properties directly is typically either computationally expensive or inaccurate. In recent years, deep learning (DL)-based atomistic simulations have emerged as a method for achieving both efficiency and accuracy. However, there remain significant challenges in assessing model reliability in DL models when simulating properties and screening new systems. In this work, structurally complex LiF-NaF-ZrF$_4$ salt is studied. We show that neural network (NN) uncertainty can be quantified using ensemble learning to provide a 95% confidence interval (CI) for NN-based predictions. We show that DL models can successfully extrapolate to new compositions, temperatures, and timescales, but fail for significant changes in density, which is captured by ensemble-based uncertainty predictions. This enables improved confidence in utilizing simulated data for realistic reactor conditions, and guidelines for training deployable DL models.