Predicting the synthesizability of hypothetical crystals is challenging due to the wide range of parameters governing materials synthesis. This paper presents a deep-learning model that uses three-dimensional pixel-wise images of crystal structures, color-coded by chemical attributes, to learn features of synthesizability. A convolutional encoder learns these features, enabling accurate classification of materials into synthesizable crystals versus crystal anomalies. The model's usefulness is demonstrated by predicting synthesizability for hypothetical crystals for battery electrode and thermoelectric applications.

Ali Davariashtiyani, Zahra Kadkhodaie, Sara Kadkhodaei