Phase diagrams of polymer-containing liquid mixtures with a theory-embedded neural network

TL;DR

This paper introduces a theory-embedded deep neural network that accurately predicts phase diagrams of polymer-containing liquid mixtures, leveraging physical insights to improve efficiency and predictive power.

Contribution

The study presents a novel DNN architecture with an embedded theoretical layer that enhances accuracy and reduces model size for phase diagram predictions.

Findings

Accurately predicts phase diagrams of polymer solutions.

Demonstrates effectiveness for salt-doped diblock copolymer melts.

Reduces neural network complexity while maintaining performance.

Abstract

We develop a deep neural network (DNN) that accounts for the phase behaviors of polymer-containing liquid mixtures. The key component in the DNN consists of a theory-embedded layer that captures the characteristic features of the phase behavior via coarse-grained mean-field theory and scaling laws and substantially enhances the accuracy of the DNN. Moreover, this layer enables us to reduce the size of the DNN for the phase diagrams of the mixtures. This study also presents the predictive power of the DNN for the phase behaviors of polymer solutions and salt-free and salt-doped diblock copolymer melts.