Adaptively compressed exchange in LAPW
Davis Zavickis, Kristians Kacars, J\=anis C\=imurs, Andris Gulans

TL;DR
This paper introduces an implementation of the adaptively compressed exchange (ACE) operator within the LAPW method, enabling precise all-electron total-energy calculations with hybrid functionals and efficient gap predictions in solids.
Contribution
The paper presents a novel implementation of ACE in LAPW that maintains high precision in all-electron calculations with hybrid functionals, demonstrated on atoms, molecules, and solids.
Findings
ACE achieves high-precision total energies comparable to multi-resolution analysis.
High-energy local orbitals are crucial for accuracy in Fock exchange calculations.
The method accurately predicts PBE0 gaps in solids, aligning with other all-electron results.
Abstract
We present an implementation of the adaptively compressed exchange (ACE) operator in the LAPW formalism. ACE is a low-rank representation of the Fock exchange that avoids any loss of precision for the total energy. Our study shows that this property remains in the all-electron case, as we apply this method in non-relativistic total-energy calculations with a hybrid exchange-correlation functional PBE0. The obtained data for light atoms and molecules are within a few Ha off the precise multi-resolution-analysis calculations. Aside from ACE, another key ingredient to achieve such a high precision with Fock exchange was the use of high-energy local orbitals. Finally, we use this implementation to calculate PBE0 gaps in solids and compare the results to other all-electron results.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAtomic and Subatomic Physics Research · Quantum, superfluid, helium dynamics · Advanced NMR Techniques and Applications
