A quadratically convergent VBSCF method.

Area of Science:

• Quantum Chemistry
• Computational Chemistry
• Theoretical Chemistry

Background:

• Valence Bond Self-Consistent Field (VB-SCF) methods are crucial for describing electron correlation.
• Efficient optimization of orbitals and configuration coefficients is essential for accurate VB-SCF calculations.

Purpose of the Study:

• To introduce a quadratically convergent Newton-Raphson scheme for VB-SCF.
• To compare its efficiency and convergence with the Super-Configuration Interaction (Super-CI) method.
• To develop a hybrid approach for improved computational economy and convergence.

Main Methods:

• Simultaneous optimization of orbitals and configuration coefficients using a Newton-Raphson scheme.
• Implementation and testing of the Newton-Raphson VB-SCF method in actual calculations.
• Development and application of a combined Newton-Raphson and Super-CI approach.

Main Results:

• The Newton-Raphson method demonstrates applicability and quadratic convergence.
• The Hessian's positive definiteness is a necessary condition for Newton-Raphson convergence.
• The combined Super-CI/Newton-Raphson method offers computational advantages over individual methods when the Hessian is not positive definite.

Conclusions:

• The quadratically convergent Newton-Raphson scheme provides an efficient route for VB-SCF optimization.
• A hybrid Super-CI/Newton-Raphson strategy is computationally more economical for optimizing nonorthogonal orbitals.
• This combined approach enhances convergence and efficiency in quantum chemical calculations.