Automated error control in divide-and-conquer self-consistent field calculations.

Area of Science:

• Computational Chemistry
• Materials Science
• Quantum Mechanics

Background:

• Linear-scaling methods, such as the divide-and-conquer (DC) approach, are crucial for large-scale electronic structure calculations.
• The DC approximation introduces errors controlled by a buffer region, requiring careful parameterization.
• Accurate error estimation is vital for reliable computational results in chemistry and materials science.

Purpose of the Study:

• To propose an energy-based error estimation scheme for the DC self-consistent field (SCF) method.
• To develop a procedure for automatically determining optimal buffer region sizes in DC calculations.
• To assess the performance of the proposed method across different molecular and material systems.

Main Methods:

• Implementation of a two-layer buffer region scheme within the DC-SCF method.
• Development of an energy-based criterion to estimate approximation errors.
• Automated determination of buffer region parameters based on the error estimation.

Main Results:

• The proposed energy-based error estimation and automatic buffer determination procedure demonstrated satisfactory performance for water clusters and proteins.
• The method's effectiveness was limited in calculations involving delocalized systems, such as graphene.
• The study highlights the importance of system-specific parameterization in DC electronic structure calculations.

Conclusions:

• The developed energy-based error estimation scheme provides a viable approach for optimizing buffer regions in DC-SCF calculations.
• The automatic procedure is effective for localized systems but requires further refinement for delocalized electronic structures.
• This work contributes to improving the efficiency and accuracy of large-scale electronic structure computations.