Approaching Chemical Accuracy for Neutral and Ionic Water Clusters with the bn-r2SCAN Meta-GGA
Abstract
Gas-phase neutral and ion–water clusters are fundamental models for hydrogen bonding, many-body interactions, proton transfer, and ionic hydration in liquid water and aqueous environments, and they are stringent tests for density functional approximations: semilocal functionals without exact exchange or an explicit self-interaction correction have typically described their interaction and binding energies poorly. The non-self-consistent SCAN@HF approach was recently shown to describe water clusters and liquid water with high accuracy, but subsequent analysis demonstrated that this performance arises largely from cancellation between a substantial functional-driven error and an opposing error introduced by the overlocalized Hartree–Fock density. Here we assess a new self-consistent meta-generalized gradient approximation, bonded-norm r2SCAN (bn-r2SCAN), for neutral, protonated, and deprotonated water clusters and for alkali-and halide–water complexes. bn-r2SCAN approaches chemical accuracy across these diverse systems, with overall mean unsigned binding-energy errors of 0.33− 1.04 kcal/mol for the benchmark sets considered, and substantially improves upon r2SCAN. Error decomposition shows that this accuracy does not rest on the near-total cancellation of large opposing errors underlying SCAN@HF: for the BEGDB dataset, the mean functional-and density-driven errors of bn-r2SCAN are 0.34 and −0.89 kcal/mol, and even for the largest clusters both remain a small fraction of the total binding energy, whereas for SCAN@HF the functional-driven error of SCAN and the nonvariational density over-localization introduced by the Hartree–Fock density each exceed 5 kcal/mol and cancel to within a few tenths. Notably, bn-r2SCAN was parametrized using band gaps and lattice constants (not weak bonds) of simple solids, making its performance for aqueous clusters a demanding test of transferability and identifying it as a self-consistent and computationally efficient route to modeling water and aqueous systems.
Related
- Classical quantum chemical solution of the FeMo-cofactor model to chemical accuracy and its implications
- Chemical potentials from structure factors: I. Neutral multi-component mixtures
- A New Family of Nonempirical Range-Separated Functionals: RSX‑r2SCAN, RSX‑r2SCAN0, and RSX‑r2SCAN-QIDH
- Chemistry-resolved disagreement between Materials Project-compatible PBE(+U) and r2SCAN stability verdicts in the GNoME materials database
- Approaching the limit of quantum clock precision
BibTeX
@misc{shahi2026approaching,
title = {{Approaching Chemical Accuracy for Neutral and Ionic Water Clusters with the bn-r2SCAN Meta-GGA}},
author = {Chandra Shahi and Adrienn Ruzsinszky and John P. Perdew},
howpublished = {ChemRxiv},
year = {2026},
month = sep,
doi = {10.26434/chemrxiv.15009326/v1},
url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009326/v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Approaching Chemical Accuracy for Neutral and Ionic Water Clusters with the bn-r2SCAN Meta-GGA
AU - Chandra Shahi
AU - Adrienn Ruzsinszky
AU - John P. Perdew
PY - 2026
DA - 2026/09/22
PB - ChemRxiv
DO - 10.26434/chemrxiv.15009326/v1
UR - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009326/v1
AB - Gas-phase neutral and ion–water clusters are fundamental models for hydrogen bonding, many-body interactions, proton transfer, and ionic hydration in liquid water and aqueous environments, and they are stringent tests for density functional approximations: semilocal functionals without exact exchange or an explicit self-interaction correction have typically described their interaction and binding energies poorly. The non-self-consistent SCAN@HF approach was recently shown to describe water clusters and liquid water with high accuracy, but subsequent analysis demonstrated that this performance arises largely from cancellation between a substantial functional-driven error and an opposing error introduced by the overlocalized Hartree–Fock density. Here we assess a new self-consistent meta-generalized gradient approximation, bonded-norm r2SCAN (bn-r2SCAN), for neutral, protonated, and deprotonated water clusters and for alkali-and halide–water complexes. bn-r2SCAN approaches chemical accuracy across these diverse systems, with overall mean unsigned binding-energy errors of 0.33− 1.04 kcal/mol for the benchmark sets considered, and substantially improves upon r2SCAN. Error decomposition shows that this accuracy does not rest on the near-total cancellation of large opposing errors underlying SCAN@HF: for the BEGDB dataset, the mean functional-and density-driven errors of bn-r2SCAN are 0.34 and −0.89 kcal/mol, and even for the largest clusters both remain a small fraction of the total binding energy, whereas for SCAN@HF the functional-driven error of SCAN and the nonvariational density over-localization introduced by the Hartree–Fock density each exceed 5 kcal/mol and cancel to within a few tenths. Notably, bn-r2SCAN was parametrized using band gaps and lattice constants (not weak bonds) of simple solids, making its performance for aqueous clusters a demanding test of transferability and identifying it as a self-consistent and computationally efficient route to modeling water and aqueous systems.
ER -