Efficient and stochastic multireference perturbation theory for large active spaces within a full configuration...

Area of Science:

• Quantum Chemistry
• Computational Chemistry
• Stochastic Methods

Background:

• Full Configuration Interaction Quantum Monte Carlo (FCIQMC) is a powerful method for large configuration interaction (CI) problems.
• FCIQMC has been adapted for active space calculations and orbital optimization.
• Stochastic methods offer potential for tackling computationally intensive quantum chemistry problems.

Purpose of the Study:

• To develop an efficient stochastic approach for sampling multireference perturbation theories within the FCIQMC framework.
• To investigate the feasibility of applying FCIQMC to strongly contracted N-Electron Valence second-order Perturbation Theory (sc-NEVPT2).
• To enable accurate calculations for large active spaces in complex chemical systems.

Main Methods:

• Adaptation of FCIQMC for efficient sampling of multireference perturbation theories.
• Development of schemes for controlled stochastic sampling of higher-body intermediates.
• Application of the stochastic FCIQMC approach to compute sc-NEVPT2 energies.

Main Results:

• Strongly contracted N-Electron Valence second-order Perturbation Theory (sc-NEVPT2) is found to be stable within the stochastic FCIQMC framework.
• The stochastic approach requires a similar number of walkers to converge sc-NEVPT2 expectation values as the underlying CI problem.
• Accurate sc-NEVPT2 energies were obtained for a biologically relevant system with a large (24, 24) active space using a small number of walkers.

Conclusions:

• The developed stochastic FCIQMC approach provides an efficient and accurate method for computing sc-NEVPT2 energies.
• This method is suitable for large active spaces that are intractable for traditional computational approaches.
• The work opens avenues for efficient stochastic solvers for multireference problems in quantum chemistry.