Neural network variational Monte Carlo for positronic chemistry

Area of Science:

• Quantum Chemistry
• Computational Physics
• Materials Science

Background:

• Calculating properties of positron-molecule complexes is difficult due to basis set limitations.
• Positron interactions with molecules are crucial for understanding annihilation processes and material properties.

Purpose of the Study:

• To develop and validate a novel method for accurate quantum chemical calculations of positron-molecule complexes.
• To assess the performance of the Fermionic Neural Network (FermiNet) wavefunction for these challenging systems.

Main Methods:

• Utilized the Fermionic Neural Network (FermiNet) wavefunction, a basis-set-independent approach.
• Performed quantum chemical calculations for ground-state properties of various atoms and small molecules with positrons.
• Calculated binding energies and annihilation rates for positron-molecule systems.

Main Results:

• FermiNet achieved highly accurate, state-of-the-art ground-state energies for diverse positron-molecule systems.
• Accurate binding energy for positron-benzene complex calculated, showing good agreement with experimental data.
• Annihilation rates computed by FermiNet favorably compared with results from explicitly correlated Gaussian wavefunctions.

Conclusions:

• The FermiNet wavefunction offers a significant advancement for quantum chemical calculations of positron-molecule interactions.
• Basis-set-independent neural network methods provide a versatile and accurate approach for complex quantum systems.
• This work expands the applicability of neural network wavefunctions beyond standard molecular Hamiltonians.