Orbital Optimization of Large Active Spaces via AI-Accelerators.

Area of Science:

• Quantum Chemistry
• Computational Chemistry
• Materials Science

Background:

• Accurate electronic structure calculations are crucial for understanding molecular behavior.
• Large active space calculations are computationally demanding, limiting their application.
• Efficient methods are needed to tackle strongly correlated systems.

Purpose of the Study:

• To present an efficient orbital optimization procedure for quantum chemistry.
• To combine GPU-accelerated DMRG with CAS-SCF for large active spaces.
• To enable accurate calculations for unprecedented system sizes.

Main Methods:

• Utilized a highly GPU accelerated, spin-adapted Density Matrix Renormalization Group (DMRG) method.
• Integrated DMRG with the Complete Active Space Self-Consistent Field (CAS-SCF) approach.
• Performed CAS-SCF based orbital optimizations on NVIDIA DGX-A100 and DGX-H100 hardware.

Main Results:

• Achieved unprecedented CAS sizes of up to CAS(82,82) and beyond.
• Demonstrated the criticality of accurate DMRG calculations for converged CAS-SCF energies.
• Observed sensitive dependence of optimized orbitals on DMRG parameters for iron-sulfur complexes.

Conclusions:

• The developed DMRG-SCF approach enables accurate calculations for large active spaces.
• This method significantly reduces the time required for complex molecular system studies.
• Opens new avenues for tackling strongly correlated molecular systems.