Automatic Generation of Accurate and Cost-Efficient Auxiliary Basis Sets

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Theoretical Chemistry

Background:

• Density fitting (DF) and resolution-of-the-identity (RI) methods significantly accelerate quantum chemical calculations.
• Auxiliary basis sets (ABSs) are crucial for the efficiency of DF/RI methods.
• Previous work established an algorithm for automatically generating standard-form ABSs.

Purpose of the Study:

• To investigate cost reduction strategies for automatically generated auxiliary basis sets (ABSs).
• To assess the accuracy of reduced ABSs for density fitting (DF) and resolution-of-the-identity (RI) calculations.
• To evaluate the impact of ABS reduction on Hartree-Fock (HF) and Møller-Plesset perturbation theory (MP2) energies.

Main Methods:

• Contracting ABSs using singular value decomposition (SVD) as proposed by Kállay.
• Truncating ABSs by removing high-angular momentum functions.
• Applying these reduction techniques to HF and MP2 calculations for various molecules.

Main Results:

• A combination of SVD contraction and high-angular momentum function removal significantly reduces ABS size.
• Reduced ABSs achieve accuracy comparable to original, unpruned ABSs (γ ≈ 5-6).
• Further reduction yields accurate results suitable for routine applications (γ ≈ 3-4).

Conclusions:

• The developed reduction schemes offer substantial computational savings for DF/RI calculations.
• The accuracy of reduced ABSs is maintained across HF and MP2 levels of theory.
• The generated ABSs demonstrate high transferability, applicable to complex property modeling with high-level methods.