Quantum Monte Carlo for noncovalent interactions: an efficient protocol attaining benchmark accuracy

Area of Science:

• Quantum Chemistry
• Computational Chemistry
• Materials Science

Background:

• Accurate prediction of noncovalent interaction energies is crucial for drug design and hydrogen storage.
• Current methods like coupled-cluster CCSD(T)/CBS are computationally expensive for large systems.
• Fixed-node diffusion Monte Carlo (FN-DMC) offers a promising alternative due to its accuracy and favorable scaling.

Purpose of the Study:

• To analyze protocols and trade-offs for FN-DMC estimations of noncovalent interaction energies.
• To propose an efficient and accurate computational protocol for FN-DMC.
• To enable practical interaction energy calculations for large noncovalent complexes.

Main Methods:

• Development and analysis of FN-DMC protocols.
• Incorporation of simplified explicit correlation terms.
• Evaluation of computational scaling with O(N(3)) complexity.
• Benchmarking against CCSD(T)/CBS data.

Main Results:

• An efficient FN-DMC protocol with O(N(3)) scaling was developed.
• The protocol achieved excellent agreement with CCSD(T)/CBS data (mean unsigned error ~0.2 kcal mol(-1)).
• Accurate results were obtained for various complexes including benzene/hydrogen, benzene dimer, and adenine-thymine.

Conclusions:

• The proposed FN-DMC protocol provides a highly accurate and computationally efficient method for noncovalent interaction energy calculations.
• This approach is suitable for large molecular systems where CCSD(T)/CBS is computationally prohibitive.
• The method advances the theoretical prediction of intermolecular forces in chemistry and materials science.