An efficient hybrid orbital representation for quantum Monte Carlo calculations

Area of Science:

• Computational physics
• Quantum chemistry

Background:

• Real-space quantum Monte Carlo (QMC) methods are powerful for studying quantum systems.
• The efficiency of QMC is limited by the memory requirements of trial wavefunctions, particularly when using B-splines.
• Existing methods face trade-offs between computational speed, memory usage, and accuracy.

Purpose of the Study:

• To develop a novel hybrid orbital representation for trial wavefunctions in QMC.
• To reduce the memory footprint of QMC calculations without sacrificing accuracy or speed.
• To expand the applicability of QMC to larger and more complex quantum systems.

Main Methods:

• Introduced a hybrid orbital representation combining localized atomic basis sets and B-splines.
• Implemented this hybrid representation for single-particle orbitals in QMC.
• Performed benchmark calculations on Nickel Oxide (NiO).

Main Results:

• The hybrid representation significantly reduces memory requirements (by 8x for NiO) compared to conventional B-splines.
• Achieved superior accuracy in benchmark calculations.
• Maintained or improved the high speed of evaluation characteristic of B-splines.

Conclusions:

• The hybrid orbital representation offers a practical solution to memory limitations in QMC.
• This method enhances the accuracy and efficiency of QMC simulations.
• It broadens the scope of quantum systems amenable to study using QMC.