Spin localization in intermolecular complexes: A challenge for semi-local approximants for the embedding potential.

Area of Science:

• Quantum Chemistry
• Computational Chemistry
• Electronic Structure Theory

Background:

• Frozen-density embedding theory (FDET) methods approximate the non-additive kinetic potential bi-functional (vtnad).
• Accurate treatment of open shell systems, where spin is localized, is challenging due to potential charge redistribution errors.
• Existing semi-local approximations for vtnad can lead to qualitatively incorrect spin densities.

Purpose of the Study:

• To systematically analyze spin densities from various semi-local approximations of vtnad.
• To determine the applicability domains of these approximations for open shell systems.
• To introduce a new non-decomposable approximant for improved spin density calculations.

Main Methods:

• Evaluation of spin densities using multiple semi-local approximations to vtnad.
• Analysis of the impact of gradient-dependent corrections on spin densities.
• Development and testing of a novel non-decomposable vtnad approximant (vtnad(NDCS)).

Main Results:

• Semi-local vtnad approximations yield either qualitatively incorrect or reasonably accurate spin distributions.
• Gradient-dependent corrections do not resolve deficiencies in spin density predictions.
• A simple orbital energy-based criterion can predict when semi-local approximants are likely to fail.
• The new non-decomposable vtnad(NDCS) extends FDET applicability to embedded radicals and improves spin densities.

Conclusions:

• The choice of vtnad approximation significantly impacts spin density accuracy in FDET.
• A predictive criterion for approximant failure is proposed.
• The non-decomposable vtnad(NDCS) offers a more robust approach for calculating spin densities in challenging open shell systems.