Towards materials data science – where high-throughput computation will meet high-throughput experimentation

Constructing a materials discovery platform requires concerted efforts between multiple domains, including theory, experiments and the computational methods for synergetic exchanges in between [1]. Firstly, these exchanges require domain-informed, compact data representations and metrics to facilitate the high-throughput methodology and unite disparate fields associated with materials science. I discuss corresponding examples from electronic structure [2] and crystal structure [3] data. Secondly, existing materials characterization methods are not designed to scale up to macroscopic samples and batches, and the correspondence between multimodal measurements are often not exact [4], I discuss these existing limitations and propose solutions by co-designing experimental and computational workflows [5]. [1] M. Aykol et al. Matter 1, 1 (2019).[2] R. P. Xian, V. Stimper et al. arXiv:2005.10210.[3] C. J. Bartel et al. J. Am. Chem. Soc. 142, 5135 (2020).[4] T. L. Burnett and P. J. Withers, Nat. Mater. 18, 1041 (2019).[5] M. Du et al. under review.