Efficient force calculations in strongly correlated materials within DFT+G using Portobello

Abstract

We derive a simplified formulation for computing the forces in density functional theory + Gutzwiller (DFT+G), implement them in the open-source quantum-embedding code Portobello, and validate by comparing numerical derivatives of the free energy against the analytical force derivatives in FeSe and NiO. This implementation is also used to predict the chalcogen height in FeSe and FeTe. In contrast with DFT in the local density approximation (LDA), which greatly overestimates chalcogen heights, DFT+G replicates the experimental chalcogen height using significantly fewer computational resources than DFT+DMFT requires.

New version program summary

Program Title: Portobello

CPC Library link to program files: https://doi.org/10.17632/5p4nggbktd.2

Developer's repository link: https://gitlab.com/ranadler/portobello-public

Licensing provisions: GPLv3

Programming language: Fortran, C++, Python

Journal reference of previous version: Computer Physics Communications, Volume 294, 2024, 108907

Does the new version supersede the previous version: Yes

Reasons for the new version: Implementation and release of DFT and DFT+G forces

Summary of revisions: Upgraded to a new version of the underlying DFT (and GW) code, FlapwMBPT, which includes DFT forces and an APW+lo+HDLO+HELO basis set. Implemented DFT+G forces and the extensions necessary to compute them within the Portobello framework. Implemented more sophisticated minimization techniques for the DFT problem.

Nature of problem: Strongly correlated material design requires a fast and flexible method for computing forces and relaxing structures. DFT+G captures strong correlations without excessive computational costs, but forces within this theory had not been previously evaluated.

Solution method: Reformulate the forces to efficiently use existing DFT force implementations in the evaluation of the DFT+G force terms.