Orbital Optimization and Neural-Network-Assisted Configuration Interaction Calculations of Rydberg States

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Molecular Physics

Background:

• Rydberg excited states in molecules are difficult to model computationally.
• Standard atomic basis sets often fail to accurately represent the diffuse electron distribution of Rydberg states.
• This confinement leads to inaccuracies in electronic structure calculations.

Purpose of the Study:

• To develop a more accurate computational method for calculating molecular Rydberg excited states.
• To overcome the limitations of traditional basis sets in representing diffuse electron distributions.
• To improve the convergence and accuracy of many-body calculations for Rydberg states.

Main Methods:

• Utilized a plane wave basis set within a Hartree-Fock calculation for variational optimization of molecular orbitals for excited states.
• Employed configuration interaction (CI) calculations, including full CI and neural-network-based selective CI.
• Applied the method to H2, H2O, and NH3 molecules.

Main Results:

• Optimized orbitals significantly enhanced the convergence of many-body calculations.
• Calculations for H2, H2O, and NH3 yielded excitation energies in close agreement with experimental data.
• The new method demonstrated superior accuracy compared to calculations using standard basis sets lacking diffuse functions.

Conclusions:

• The presented approach of using excited-state optimized orbitals is effective for accurate Rydberg state calculations.
• This method overcomes the limitations of basis set confinement, providing reliable excitation energies.
• It offers a robust alternative for electronic structure calculations of challenging Rydberg states.