Multireference configuration interaction and perturbation theory without reduced density matrices

Area of Science:

• Quantum Chemistry
• Computational Physics
• Theoretical Chemistry

Background:

• Calculating dynamic correlation for multireference wave functions in large active spaces is computationally intensive.
• High-rank reduced density matrices (RDMs) pose a significant bottleneck due to their evaluation and storage demands.

Purpose of the Study:

• To present a stochastic formulation of multireference configuration interaction and perturbation theory.
• To overcome the limitations imposed by expensive RDMs in computational chemistry.

Main Methods:

• Developed a stochastic algorithm that bypasses the need for high-rank RDMs.
• The method is flexible, accommodating various active space reference wave functions like selected configuration interaction and matrix product states.
• Leveraged Monte Carlo methods for parallelizability and reduced memory footprint.

Main Results:

• The stochastic algorithm demonstrates competitiveness with deterministic methods even for small active spaces (14 orbitals).
• The formulation successfully avoids the computationally expensive evaluation and storage of RDMs.
• Benchmark applications showcase the practical utility of the stochastic approach.

Conclusions:

• The proposed stochastic method offers a computationally efficient alternative for dynamic correlation calculations.
• This approach significantly reduces the bottleneck associated with RDMs in multireference quantum chemistry.
• The flexibility and efficiency make it suitable for large active space problems.