Automated Active Space Selection with Dipole Moments.

Area of Science:

• Quantum chemistry
• Computational materials science

Background:

• Multireference calculations are crucial for strongly correlated systems in molecular and material development.
• Selecting appropriate active spaces for these calculations is challenging and often requires extensive human expertise, sometimes yielding non-physical results.

Purpose of the Study:

• To develop and evaluate automated protocols for active space selection in multireference calculations.
• To utilize dipole moments (ground and excited state) as physical observables for guiding active space selection.

Main Methods:

• Developed two protocols for automated active space selection based on ground-state and excited-state dipole moments.
• Constructed a dataset of 1275 active spaces from 25 molecules, considering 51 active space sizes per molecule.
• Analyzed the relationship between active space, dipole moments, and vertical excitation energies.

Main Results:

• Protocols successfully identify suitable active spaces for reasonable vertical excitation energies, particularly for the first three excitations.
• Accuracy is maintained while reducing computation time by over tenfold by excluding large active spaces.
• Demonstrated applicability to potential energy surface scans and spin state determination in transition metal oxides.

Conclusions:

• Automated active space selection using dipole moments offers a reliable and efficient alternative to manual methods.
• These protocols reduce computational cost and human input, facilitating the study of complex chemical systems.
• The developed methods enhance the predictive power of multireference calculations for molecular and material design.