Machine learning materials physics: Integrable deep neural networks enable scale bridging by learning free energy functions

The free energy of a system is central to many material models. Although free energy data is not generally found directly, its derivatives can be observed or calculated. In this work, we present an Integrable Deep Neural Network (IDNN) that can be trained to derivative data obtained from atomic scale models and statistical mechanics, then analytically integrated to recover an accurate representation of the free energy. The IDNN is demonstrated by training to the chemical potential data of a binary alloy with B2 ordering. The resulting DNN representation of the free energy is used in a mesoscopic, phase field simulation and found to predict the appropriate formation of antiphase boundaries in the material. In contrast, a B-spline representation of the same data failed to resolve the physics of the system with sufficient fidelity to resolve the antiphase boundaries. Since the fine scale physics harbors complexity that emerges through the free energy in coarser-grained descriptions, the IDNN represents a framework for scale bridging in materials systems. •Introduces machine learning as an approach to scale bridging in materials physics.•Advances machine learning for physics via new Integrable Deep Neural Networks: IDNNs.•The IDNNs enable scale-bridging and the correct physics in computations.