Density functionals for high-throughput materials discovery*

While the design of materials from first principles has long seemed unattainable, simultaneous advancements in practical DFT and high-performance computing have brought this goal into view. Recent meta-generalized gradient approximations (meta-GGAs) like r²SCAN [1] have shown wide-ranging accuracy across solid-state compositional spaces [2], and sufficient numerical stability to permit hight-througput sweeps [3]. To this end, I'll discuss how the Materials Project has adopted r²SCAN as a primary engine for materials discovery, and how predicted thermodymic properties are changed by ascending the density functional ladder from the GGA to meta-GGA level. I'll pay particular attention to under-explored chemical spaces as we update, refine, and grow the Materials Project database using r²SCAN. Last, I will discuss the development of meta-GGAs that synthesize first principles and semi-emprical approaches.

[1] J.W. Furness, A.D. Kaplan, J. Ning, J.P. Perdew, and J. Sun, J. Phys. Chem. Lett. 11, 8208 (2020).

[2] M. Kothakonda, A.D. Kaplan et al., ACS Mater. Au 3, 102 (2022).

[3] R. Kingsbury et al., Phys. Rev. Mater. 6. 013801 (2022).