An efficient and near linear scaling pair natural orbital based local coupled cluster method

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Theoretical Chemistry

Background:

• The local pair natural orbital-coupled cluster single double (LPNO-CCSD) method provides efficient and reliable calculations for large molecules.
• The original LPNO-CCSD method faced computational expense for very large systems due to fifth-order scaling steps.
• Limitations included PNO expansion in canonical virtual orbitals and untruncated single excitations.

Purpose of the Study:

• To overcome the computational limitations of the original LPNO-CCSD method for large molecules.
• To develop a new method that maintains high accuracy while significantly improving computational efficiency.
• To enable reliable coupled cluster calculations on unprecedentedly large molecular systems.

Main Methods:

• A complete redesign of the LPNO-CCSD method combining pair natural orbitals (PNOs) and projected atomic orbitals (PAOs).
• PNOs are expanded in PAOs within electron pair-specific domains, exploiting locality.
• The new domain-based local pair natural orbital-coupled cluster single double (DLPNO-CCSD) method is nearly linear scaling.

Main Results:

• The DLPNO-CCSD method achieves computational savings exceeding one order of magnitude for larger systems compared to LPNO-CCSD.
• Correlation energy loss due to domain approximations is less than 0.05%.
• Calculations were performed on systems with over 450 atoms and 8800 basis functions, with DLPNO-CCSD steps often faster than Hartree-Fock.

Conclusions:

• DLPNO-CCSD is as accurate as LPNO-CCSD but significantly more computationally efficient for large molecules.
• The method is a black box, requiring no user adjustments, and is nearly linear scaling.
• This development enables reliable coupled cluster calculations on large molecules with unprecedented efficiency and accuracy.