# Automated construction of molecular active spaces from atomic valence orbitals

## TL;DR

The paper presents AVAS, an automated method for constructing active orbital spaces in multireference electronic structure calculations, simplifying the process especially for transition metal chemistry and bond dissociation studies.

## Contribution

Introduces AVAS, a straightforward automated technique for defining active spaces based on atomic valence orbitals, improving ease and reproducibility of multireference calculations.

## Key findings

- Accurately calculates excitation energies for transition metal complexes
- Effectively follows bond breaking in a Fenton reaction
- Simplifies active space determination in MR calculations

## Abstract

We introduce the atomic valence active space (AVAS), a simple and well-defined automated technique for constructing active orbital spaces for use in multi-configuration and multireference (MR) electronic structure calculations. Concretely, the technique constructs active molecular orbitals capable of describing all relevant electronic configurations emerging from a targeted set of atomic valence orbitals (e.g., the metal d orbitals in a coordination complex). This is achieved via a linear transformation of the occupied and unoccupied orbital spaces from an easily obtainable single-reference wavefunction (such as from a Hartree-Fock or Kohn-Sham calculations) based on projectors to targeted atomic valence orbitals. We discuss the premises, theory, and implementation of the idea, and several of its variations are tested. To investigate the performance and accuracy, we calculate the excitation energies for various transition metal complexes in typical application scenarios. Additionally, we follow the homolytic bond breaking process of a Fenton reaction along its reaction coordinate. While the described AVAS technique is not an universal solution to the active space problem, its premises are fulfilled in many application scenarios of transition metal chemistry and bond dissociation processes. In these cases the technique makes MR calculations easier to execute, easier to reproduce by any user, and simplifies the determination of the appropriate size of the active space required for accurate results.