Frozen natural orbitals for projection-based embedding method and its application to quantum computation

Abstract

Projection-based embedding offers a robust framework for modeling chemical processes in large molecular systems. It partitions computations into low-level treatments for the environmental subsystems and high-level treatments for the active subsystems. To address the computational bottleneck of applying high-accuracy wavefunction theory to the active regions, we introduce frozen natural orbitals (FNOs) approach. This technique efficiently compresses the virtual orbital space to generate a compact set of natural orbitals, which significantly accelerates the convergence of correlation energy recovery. We evaluated the FNO-based embedding approach in comparison with two alternative virtual space truncation approaches for a diverse range of molecular systems. Our results demonstrate that the FNO-embedded method exhibits superior performance by delivering accurate correlation energies with a substantially fewer number of virtual orbitals. Its potential for quantum computing is also discussed, as this reduction in orbitals can be directly translated to lower qubit requirements and thereby facilitate the quantum simulations on near-term quantum hardware.