Variational Density Fitting with a Krylov Subspace Method.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Numerical Analysis

Background:

• Variational density fitting methods often encounter ill-conditioned and indefinite systems of linear equations.
• These challenges are exacerbated by finite precision arithmetic and large system dimensions.
• Existing methods, like truncated eigenvalue decomposition, may struggle with robustness and scalability.

Purpose of the Study:

• To implement and evaluate a novel variational density fitting methodology using iterative linear algebra.
• To address the challenges posed by ill-conditioned, indefinite, and large-scale equation systems.
• To enhance the computational efficiency and scalability of density fitting calculations.

Main Methods:

• Implementation of a variational density fitting approach utilizing iterative linear algebra.
• Application of a preconditioned Krylov subspace method designed for indefinite and ill-conditioned systems.
• Integration with double asymptotic electron repulsion integral expansions within the deMon2k package.

Main Results:

• The proposed method effectively solves systems with up to 130,000 auxiliary functions.
• Demonstrated effectiveness in alleviating issues related to ill-conditioning and matrix definiteness.
• Exhibited excellent robustness and scalability, particularly in a parallel computing environment.

Conclusions:

• The developed iterative linear algebra methodology offers a robust and scalable solution for variational density fitting.
• This approach overcomes limitations of traditional methods, enabling larger and more complex electronic structure calculations.
• The findings suggest significant improvements for computational chemistry software like deMon2k.