Revealing the Spectrum of Unknown Layered Materials with Super-Human Predictive Abilities

We use semi-supervised learning to discover over 1000 new two-dimensional layered materials that have yet to be discovered or synthesized. We accomplish this by combining physics with machine learning on experimentally obtained data and verify a subset of candidates using density functional theory. Our model accelerates the discovery of layered materials by 13 times compared to random trial-and-error approaches. Even compared to expert scientists working in the field of two-dimensional materials, it is five times better than practitioners in the field at identifying layered materials and is comparable or better than professional solid-state chemists. We also find that our model is orders of magnitude faster than any human.

To achieve super-human performance, we employ semi-supervised learning techniques for the first time in materials discovery. Semi-supervised learning utilizes unlabeled data in addition to labeled data, which is powerful in cases where labels are expensive to obtain or are noisy. We find that semi-supervised learning provides benefits over supervised learning in identifying layered materials, and it may be applicable to a wide range of problems in materials science.