How Does HF-DFT Achieve Chemical Accuracy for Water Clusters?

Area of Science:

• Computational Chemistry
• Quantum Chemistry
• Materials Science

Background:

• Recent calculations identified functional-driven errors (FEs) and density-driven errors (DEs) in semilocal density functionals for water dimer binding energy.
• Understanding these errors is crucial for accurately predicting properties of water clusters and condensed phases.

Purpose of the Study:

• To investigate approximate FEs and DEs in various water clusters (neutral, charged, alkali-, halide-bound) up to 20 monomers.
• To evaluate the accuracy of density functional approximations, particularly SCAN and its hybrid variants, for water cluster binding energies.
• To explore the phenomenon of error cancellation in composite methods.

Main Methods:

• Utilized r²SCAN 50, a 50% global hybrid of exact exchange with r²SCAN, as a proxy for the exact density.
• Analyzed FEs and DEs for SCAN and r²SCAN functionals in different water cluster systems.
• Investigated composite methods like SCAN@HF (SCAN on Hartree-Fock density) and HF-r²SCAN-DC4 (r²SCAN with dispersion correction on HF density).

Main Results:

• SCAN exhibits larger FEs than r²SCAN for neutral water clusters, while DEs are similar.
• SCAN@HF demonstrates chemical accuracy for absolute binding energies of water clusters due to error cancellation.
• HF-r²SCAN-DC4 achieves near-perfect cancellation of FE and DE, improving upon r²SCAN@HF.
• r²SCAN is highly accurate for water hexamers and likely for liquid water near the boiling point.

Conclusions:

• The accuracy of composite methods like SCAN@HF and HF-r²SCAN-DC4 stems from error compensation, not necessarily a more accurate reference density.
• Unconventional error cancellation, driven by density localization, appears reliable for diverse molecular properties.
• The findings provide insights into improving density functional approximations for condensed-phase water systems.