Predicting the FCI energy of large systems to chemical accuracy from restricted active space density matrix renormalization group calculations

Abstract

We present a theoretical analysis and a new theory-based extrapolation method for the recently introduced restricted active space density matrix renormalization group (DMRG-RAS) method [arXiv:2111.06665] in electronic structure calculations. Large-scale numerical simulations show that our approach, DMRG-RAS-X, reaches chemical accuracy for challenging strongly correlated systems such as the Chromium dimer or dicarbon up to a large cc-pVQZ basis with moderate computational demands. The method is free of empirical parameters, performed robustly and reliably in all examples we tested, and has the potential to become a vital alternative method for electronic structure calculations in quantum chemistry, and more generally for the computation of strong correlations in nuclear and condensed matter physics.

Toward Large-Scale Restricted Active Space Calculations Inspired by the Schmidt Decomposition

“…Note that the RAS scheme can also be combined either with orbital optimization in a self-consistent manner or with perturbation theory to further improve the accuracy of the model. , In our proof of concept implementation of the DMRG-RAS scheme we demonstrate in section that high-quality solutions can be obtained even without such corrections just by treating large enough completely correlated active spaces. In addition, in a recent work some of us have proved and demonstrated by large-scale DMRG calculations that the DMRG-RAS energies obtained for increasing A spaces can be extrapolated to the FCI solution within chemical accuracy.…”