Full Configuration Interaction Excited-State Energies in Large Active Spaces from Subspace Iteration with Repeated...

Area of Science:

• Quantum chemistry
• Computational physics
• Materials science

Background:

• Calculating excited-state energies is crucial for understanding molecular properties and reactions.
• Existing quantum Monte Carlo methods face challenges with stability and accuracy for large systems.

Purpose of the Study:

• To present a novel, stable, and systematically improvable quantum Monte Carlo (QMC) approach for calculating excited-state energies.
• To implement this method using a fast randomized iteration technique for the full configuration interaction (FCI) problem.

Main Methods:

• An asymmetric variant of subspace iteration is employed, avoiding dot products of random vectors.
• Trial vectors are used to maintain orthogonality and estimate eigenvalues.
• The method is applied to challenging molecular systems in large active spaces, including carbon dimer, oxo-Mn(salen), ozone, and butadiene.

Main Results:

• The approach successfully calculates ground- and excited-state energies for complex molecular systems.
• Results show agreement with state-of-the-art methods (heat-bath CI, DMRG, FCIQMC) to within sub-milliHartree accuracy for total energies.
• Estimated excitation energies agree to within approximately 0.1 eV across all test cases.

Conclusions:

• The presented QMC approach offers a stable and accurate method for computing excited-state energies.
• This method is systematically improvable and suitable for large, challenging molecular systems.
• The findings advance the capabilities of computational chemistry for electronic structure calculations.