Accelerating SCF Orbital Optimization with S-GEK/RVO: Efficient Subspace Compression and Robust Convergence

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Self-consistent field (SCF) calculations are fundamental in quantum chemistry.
• Orbital optimization within SCF methods can be computationally intensive and prone to convergence issues.
• Existing methods like r-GDIIS have limitations in efficiency and robustness.

Purpose of the Study:

• To enhance the S-GEK/RVO method for more efficient and robust SCF orbital optimization.
• To introduce specific modifications addressing computational bottlenecks and convergence failures.
• To provide a competitive alternative to existing SCF optimization techniques.

Main Methods:

• Implemented subspace expansion using r-GDIIS or BFGS predictions.
• Introduced a strategy for mitigating undershooting in flat energy regions.
• Applied rigorous coordinate and gradient transformations for orbital rotation parametrization.

Main Results:

• The enhanced S-GEK/RVO variants demonstrated superior performance compared to the default r-GDIIS method.
• Improvements were observed in iteration count, convergence reliability, and wall time across diverse molecular systems.
• The method showed consistent outperformance on organic molecules, radicals, and transition-metal complexes.

Conclusions:

• The modified S-GEK/RVO method offers significant improvements in computational efficiency and robustness for SCF optimization.
• This approach presents a competitive alternative for electronic structure calculations.
• The enhancements suggest potential for broader application in orbital optimization and localization problems.