Grassmann Extrapolation for Accelerating Geometry Optimization.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Molecular Modeling

Background:

• The self-consistent field (SCF) procedure is crucial for electronic structure calculations.
• Geometry optimization is essential for determining molecular structures and properties.
• Existing methods for SCF convergence can be slow, particularly for large systems.

Purpose of the Study:

• To adapt and apply the Grassmann extrapolation (G-Ext) method for accelerating geometry optimization.
• To investigate the effectiveness of G-Ext in improving SCF convergence speed.
• To identify optimal parameters and computational strategies for G-Ext in molecular geometry optimization.

Main Methods:

• Extension of the Grassmann extrapolation (G-Ext) method from Born-Oppenheimer molecular dynamics to geometry optimization.
• Utilizing density matrices from preceding optimization steps.
• Application of a nonlinear, structure-preserving mapping onto the Grassmann manifold for initial guess generation.

Main Results:

• Demonstrated significant acceleration of the self-consistent field (SCF) convergence.
• G-Ext method shows excellent performance improvements, particularly for large molecular systems.
• Identification of optimal parameters and computational strategies through diverse molecular testing.

Conclusions:

• The adapted G-Ext method is effective in accelerating geometry optimization.
• G-Ext provides a robust approach for improving SCF convergence in computational chemistry.
• This method offers a valuable tool for efficient molecular modeling, especially for complex systems.