Divide-and-Conquer Hartree-Fock Calculations on Proteins

Area of Science:

• Theoretical Chemistry
• Computational Chemistry
• Quantum Chemistry

Background:

• Achieving linear scaling in ab initio electronic structure calculations is a major goal in theoretical chemistry.
• Standard Hartree-Fock (HF) calculations are computationally expensive due to their cubic scaling with system size.
• The divide-and-conquer (DC) approach offers a potential solution for linear-scaling HF calculations.

Purpose of the Study:

• To implement and validate the divide-and-conquer (DC) algorithm for linear-scaling Hartree-Fock (HF) calculations.
• To assess the efficiency of the DC-HF approach on various molecular systems, including proteins.
• To investigate the impact of fragment-based initial guesses on DC-HF convergence.

Main Methods:

• Implementation of the divide-and-conquer (DC) algorithm within the Hartree-Fock (HF) framework.
• Validation of the DC-HF method on polyglycines, polyalanines, and eleven real protein structures (up to 608 atoms).
• Comparison of a fragment-based initial guess (molecular fractionation with conjugated caps - MFCC) against the superposition of atomic densities (SAD) initial guess.

Main Results:

• The DC-HF algorithm was successfully validated on systems up to 608 atoms.
• The fragment-based MFCC initial guess significantly reduced the number of self-consistent field (SCF) cycles required for convergence.
• MFCC enabled convergence for globular proteins where the SAD initial guess failed.

Conclusions:

• The divide-and-conquer (DC) algorithm is a viable method for achieving linear-scaling Hartree-Fock (HF) calculations.
• The MFCC initial guess is a powerful tool for accelerating convergence in DC-HF, particularly for complex protein systems.
• This approach enhances the feasibility of large-scale electronic structure calculations in computational chemistry.