Accurate and efficient linear scaling DFT calculations with universal applicability.

Area of Science:

• Computational Chemistry
• Materials Science
• Quantum Mechanics

Background:

• Density functional theory (DFT) calculations face significant computational expense due to cubic scaling with system size.
• Linear scaling algorithms have been developed to enable ab initio calculations for larger systems, but often compromise accuracy or require extensive parameter tuning.
• Existing linear scaling methods may introduce small errors or demand complex fine-tuning, limiting their universal applicability.

Purpose of the Study:

• To develop a highly accurate and broadly applicable linear scaling approach for ab initio calculations.
• To overcome the accuracy limitations and parameter tuning issues associated with traditional linear scaling algorithms.
• To enable efficient simulation of large systems using DFT within reasonable computational resources.

Main Methods:

• Implementation of a novel linear scaling approach within the BigDFT package.
• Utilizing an ansatz based on localized support functions.
• Employing an underlying Daubechies wavelet basis for accurate linear scaling calculations.

Main Results:

• Achieved high accuracy in linear scaling calculations, overcoming previous limitations.
• Demonstrated universal applicability across diverse systems and boundary conditions.
• Enabled efficient simulations of large systems with linear scaling computational time and moderate resource demands.

Conclusions:

• The developed linear scaling method offers a significant advancement for large-scale ab initio simulations.
• The approach provides a balance between computational efficiency, accuracy, and ease of use.
• Validated effectiveness for single-point calculations, geometry optimizations, and molecular dynamics.