Noncovalent Interactions in Density Functional Theory: All the Charge Density We Do Not See

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Accurate electronic densities are crucial for understanding molecular interactions via the Hellmann-Feynman theorem.
• Current density-functional approximations (DFAs) often yield unreliable electron densities, limiting predictive accuracy.
• Long-range van der Waals (vdW) dispersion forces are known to influence molecular systems but their impact on electron density is not fully characterized.

Purpose of the Study:

• To investigate and quantify the polarization of electron density induced by long-range vdW dispersion interactions.
• To evaluate the impact of dispersion-driven density shifts on electrostatic potentials and noncovalent interactions (NCIs).
• To bridge the gap between energy-based dispersion models and density-functional theory for improved computational methods.

Main Methods:

• Development and application of a newly developed fully coupled and optimally tuned variant of the many-body dispersion model (MBD@FCO).
• Benchmarking of vdW-induced density shifts against highly accurate coupled-cluster reference densities.
• Application of the MBD@FCO model to supramolecular datasets (S12L, L7) and a protein model (Fip35-WW).

Main Results:

• Dispersion interactions induce significant polarization in the electron density, with effect magnitude increasing with system size.
• Dispersion-driven polarization alters long-range electrostatic potentials by up to 4 kcal/mol.
• Reshaping of noncovalent interaction (NCI) isosurfaces leads to smoother, chemically interpretable interaction regions.

Conclusions:

• Dispersion interactions leave a measurable imprint on electron density, affecting electrostatics and NCIs.
• Findings have implications for biomolecular modeling and density-based chemical analysis.
• Results pave the way for dispersion-consistent DFAs and improved machine-learned models based on electron densities.