Automated Workflow for Accurate High-Throughput GW Calculations

The GW approximation represents the state-of-the-art ab-initio method for computing excited-state properties. Its execution requires control over a larger number of (often interdependent) parameters, and therefore its application in high-throughput studies is hindered by the intricate and time-consu...

Full description

Saved in:
Bibliographic Details
Published inarXiv.org
Main Authors Varrassi, Lorenzo, Ellinger, Florian, Flage-Larsen, Espen, Wolloch, Michael, Kresse, Georg, Marzari, Nicola, Franchini, Cesare
Format Paper
LanguageEnglish
Published Ithaca Cornell University Library, arXiv.org 21.10.2024
Subjects
Automation
Convergence
Datasets
Elementary excitations
Mathematical analysis
Parameters
Workflow
Online AccessGet full text

Cover

More Information
Summary:The GW approximation represents the state-of-the-art ab-initio method for computing excited-state properties. Its execution requires control over a larger number of (often interdependent) parameters, and therefore its application in high-throughput studies is hindered by the intricate and time-consuming convergence process across a multi-dimensional parameter space. To address these challenges, here we develop a fully-automated open-source workflow for G\(_0\)W\(_0\) calculations within the AiiDA-VASP plugin architecture. The workflow is based on an efficient estimation of the errors on the quasi-particle (QP) energies due to basis-set truncation and the pseudo-potential norm violation, which allows a reduction of the dimensionality of the parameter space and avoids the need for multi-dimensional convergence searches. Protocol validation is conducted through a systematic comparison against established experimental and state-of-the-art GW data. To demonstrate the effectiveness of the approach, we construct a database of QP energies for a diverse dataset of over 320 bulk structures. The openly accessible workflow and resulting dataset can serve as a valuable resource and reference for conducting accurate data-driven research.
ISSN:2331-8422