Predicting Electronic Screening for Koopmans Spectral Functional Calculations
Author Information
Author(s): Yannick Schubert, Sandra Luber, Nicola Marzari, Edward Linscott
Primary Institution: University of Zurich, École Polytechnique Fédérale de Lausanne, Paul Scherrer Institute
Hypothesis
Can a machine-learning model predict screening parameters for Koopmans spectral functionals to reduce computational costs?
Conclusion
The machine-learning model can predict screening parameters with minimal loss of accuracy, significantly speeding up calculations.
Supporting Evidence
- Using predicted screening parameters leads to orbital energies differing by less than 20 meV on average.
- The model requires minimal training data, making it efficient for practical applications.
- Koopmans spectral functionals can now be applied to problems that were previously too computationally expensive.
Takeaway
Scientists created a smart computer program that helps predict how materials behave, making it faster and easier to study them.
Methodology
A machine-learning model was developed to predict screening parameters from orbital densities calculated at the DFT level.
Limitations
The model's transferability is limited to similar systems and may not perform well across different chemical environments.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website