Quantitative estimation of localization errors of 3d transition metal pseudopotentials in diffusion Monte Carlo

Area of Science:

• Computational Chemistry
• Quantum Mechanics
• Materials Science

Background:

• Non-local pseudopotentials are essential for quantum mechanical simulations but introduce localization errors.
• Diffusion Monte Carlo (DMC) is a powerful method for electronic structure calculations, susceptible to these pseudopotential errors.

Purpose of the Study:

• To estimate and compare localization errors for two families of non-local pseudopotentials for first-row transition metals (Sc-Zn).
• To assess the impact of Jastrow factors and different approximation schemes (locality approximation, T-moves) on energy accuracy.
• To determine the dominant error source (locality vs. fixed-node) in DMC calculations for these atoms.

Main Methods:

• Employed an extrapolation scheme and multideterminant wavefunctions to estimate pseudopotential errors.
• Analyzed the sensitivity of DMC energies to Jastrow factors.
• Compared the locality approximation and T-moves scheme for total energy calculations.

Main Results:

• Newer pseudopotentials (Krogel et al.) reduce locality errors by ~40% compared to older ones (Burkatzki et al.) using the locality approximation.
• Locality errors are similar for both pseudopotential sets when using the T-moves scheme.
• Jastrow factor sensitivity decreases across the Sc-Zn series.
• Fixed-node error is more significant than locality error for single-determinant calculations.

Conclusions:

• Pseudopotential development has successfully reduced localization errors in DMC.
• The choice of pseudopotential and approximation scheme impacts energy accuracy.
• For Sc-Zn atoms, fixed-node errors are a primary concern in single-determinant DMC calculations.