Magnitude of pseudopotential localization errors in fixed node diffusion quantum Monte Carlo.

Area of Science:

• Computational Physics
• Quantum Chemistry
• Materials Science

Background:

• Real space diffusion quantum Monte Carlo (QMC) is increasingly applied to heavy elements due to growing computational power.
• Standard QMC techniques can lead to significant pseudopotential localization errors, especially for heavy elements like cerium (Ce).

Purpose of the Study:

• To formally investigate and quantify the pseudopotential localization error in QMC simulations of heavy elements.
• To explore methods for reducing localization error by improving the Jastrow factor and trial wavefunctions.
• To assess the performance of different QMC schemes (locality approximation and T-moves) and Jastrow factor types for cerium.

Main Methods:

• Application of real space diffusion quantum Monte Carlo.
• Development of a Jastrow sensitivity metric for pseudopotential design.
• Extrapolation schemes to determine bare fixed-node energies.
• Estimation of localization errors for locality approximation and T-moves schemes for Ce (3+ and 4+ charge states).

Main Results:

• Localization error can be substantial (around 1 eV for isolated Ce atom) with standard QMC techniques.
• Energy-minimized Jastrow factors with three-body terms most effectively reduce localization error for Ce.
• Locality approximation shows lower Jastrow sensitivity and generally smaller errors than T-moves.
• Further improvements in Jastrow factors and trial wavefunctions are needed for chemical accuracy with heavy elements.

Conclusions:

• Pseudopotential localization error is a critical issue for heavy element QMC calculations.
• Optimizing Jastrow factors is essential for mitigating these errors.
• Advanced Jastrow factors and trial wavefunctions are necessary to achieve chemical accuracy in heavy element simulations.