REAlgo: Rapid and efficient algorithm for estimating MP2/CCSD energy gradients for large molecular clusters.

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Molecular Modeling

Background:

• Accurate calculation of energy gradients is crucial for molecular structure optimization.
• Correlated electronic structure methods (MP2, CCSD) provide high accuracy but are computationally expensive for large systems.

Purpose of the Study:

• To develop a rapid and efficient algorithm for evaluating energy gradients of large molecular clusters.
• To enable geometry optimization of large molecular systems using correlated methods.

Main Methods:

• Segregation of Hartree-Fock (HF) and correlation energy components for evaluation.
• Approximation of correlation energy using two-body interaction energies.
• Monomer-centric fragment approximation for correlation gradients.

Main Results:

• The algorithm was successfully implemented and tested with the BERNY optimizer in Gaussian.
• Demonstrated accuracy and efficiency for large molecular clusters up to 3000 basis functions, including water clusters.
• Achieved geometry optimization at the CCSD level for clusters with ~800 basis functions on modest hardware.

Conclusions:

• The developed algorithm offers a computationally feasible approach for accurate gradient calculations in large molecular systems.
• This method significantly reduces the computational cost associated with correlated methods for large clusters.
• Enables routine use of high-level correlated methods for geometry optimization of complex molecular assemblies.