Integration of high-throughput, provenance and dissemination: the case of novel 2D materials

Computational discovery of novel materials often requires high-throughput searches relying on complex workflows. However, these can be difficult to detail and reproduce. Frameworks like AiiDA [1] allow automated calculations and storage preserving full provenance, with no additional effort required. I will argue why this is essential using as case study the search for novel 2D materials [2] where, starting from ~110,000 unique experimentally-known 3D compounds we identify with high-throughput van-der-Waals DFT calculations 1825 potentially-exfoliable compounds. This portfolio can be further refined and used to compute vibrational, electronic, magnetic, and topological properties. For instance, we are able to identify 56 promising ferro- and antiferromagnetic systems. By uploading all data (automatically stored by AiiDA) to our platform materialscloud.org, results can be disseminated seamlessly, with DOIs assigned to the datasets and interactive browsing of the provenance to understand, reproduce and reuse the results.

[1] G. Pizzi et al., Comp. Mat. Sci. 111, 218 (2016)
[2] N. Mounet et al., arXiv:1611.05234