[Paper Review] Is valence CCSD(T) enough for the binding of water clusters? The isomers of (H$_2$O)$_6$ and (H$_2$O)$_{20}$ as a case study

Golokesh Santra, Margarita Shepelenko+2 more|arXiv (Cornell University)|Aug 11, 2023

Molecular spectroscopy and chiralityChemistry3 citations

TL;DR

This study evaluates the adequacy of valence CCSD(T) for water cluster binding by examining post-CCSD(T) correlation effects in (H₂O)₆ and (H₂O)₂₀ isomers. It finds that core-valence correlation is the most significant correction beyond valence CCSD(T), while higher-order excitations favor more compact isomers, and relativistic and diagonal Born-Oppenheimer corrections largely cancel out.

ABSTRACT

Benchmark calculations on noncovalent interactions typically exclude correlation effects beyond valence CCSD(T) owing to their steep computational cost scaling. In this work, we consider their importance for water clusters, specifically, eight isomers of (H$_2$O)$_6$ and four Wales-Hodges isomers of (H$_2$O)$_{20}$. Higher order connected triples, $T_3$--(T), reduce dissociation energies of the latter by about 0.4 kcal/mol, but this is more than compensated by an increase of up to 0.85 kcal/mol due to connected quadruple excitations. In general, higher-order correlation effects favor more compact isomers over more `spread-out' ones. We also consider additional small effects for balance: scalar relativistics reduce binding in (H$_2$O)$_{20}$ by ca. --0.4 kcal/mol, which fortuitously is compensated by the ca. 0.55 kcal/mol diagonal Born-Oppenheimer correction. Core-valence correlation has the greatest impact, at ca. 1.3 kcal/mol for the icosamer.

Motivation & Objective

• To assess whether valence CCSD(T) is sufficient for accurate binding energy predictions in water clusters.
• To quantify the impact of higher-order electron correlation effects—specifically connected triples (T₃–(T)) and quadruples—on isomer stabilities in (H₂O)₆ and (H₂O)₂₀.
• To evaluate the role of core-valence correlation, scalar relativistic effects, and diagonal Born-Oppenheimer corrections in noncovalent binding energies.
• To determine whether two-body approximations of post-CCSD(T) effects are sufficient for accurate binding energy predictions in larger clusters.

Proposed method

• Benchmark CCSD(T) calculations were performed using MOLPRO 2022.3, with complete basis set extrapolation using cc-pVnZ and aug-cc-pVnZ basis sets.
• Post-CCSD(T) contributions were evaluated using CCSDT, CCSDT(Q), and CCSDTQ(5) methods via CFOUR and MRCC, with NCC module for CCSDT(Q) calculations.
• Core-valence correlation was assessed using aug-cc-pwCVnZ basis sets, and scalar relativistic corrections were computed using the 2nd-order Douglas-Kroll-Hess approach.
• Diagonal Born-Oppenheimer corrections (DBOC) were computed using CFOUR 2.1, and many-body expansion (MBE) techniques were used to analyze contributions.
• Structures for (H₂O)₆ and (H₂O)₂₀ were taken from the BEGDB and WATER27 datasets, respectively, with energy differences analyzed across isomers.
• Basis set convergence and extrapolation parameters were derived from established literature, including Karton (2022) and Ref. 43.

Experimental results

Research questions

• RQ1 How do higher-order electron correlation effects—specifically connected triples and quadruples—affect the relative stability of water cluster isomers?
• RQ2 To what extent does core-valence correlation influence binding energies in (H₂O)₆ and (H₂O)₂₀ beyond valence CCSD(T)?
• RQ3 Do scalar relativistic and diagonal Born-Oppenheimer corrections significantly alter binding energy differences between isomers, and do they compensate each other?
• RQ4 Can two-body approximations of post-CCSD(T) effects capture the majority of beyond-CCSD(T) contributions in larger water clusters?

Key findings

• Higher-order connected triples (T₃–(T)) reduce dissociation energies of (H₂O)₂₀ by approximately 0.4 kcal/mol, while connected quadruple excitations increase them by up to 0.85 kcal/mol, favoring more compact isomers.
• The net effect of post-CCSD(T) valence correlation is to enhance binding energy, with quadruple excitations favoring compact isomers such as the prism structure over more spread-out ring forms.
• Core-valence correlation contributes 1.25–1.36 kcal/mol to the binding energy of (H₂O)₂₀, making it the largest correction beyond valence CCSD(T), and is significantly larger than in previous systems like the benzene dimer.
• Scalar relativistic corrections reduce binding energies by ~0.4 kcal/mol for (H₂O)₂₀, which is nearly fully compensated by diagonal Born-Oppenheimer corrections increasing binding by ~0.56 kcal/mol.
• The two-body approximation captures over 90% of all post-CCSD(T) contributions, indicating that higher-order terms can be realistically modeled in larger clusters using 2-body expansions.
• Three-body contributions to core-valence correlation are negligible (0.009–0.020 kcal/mol), while three- and four-body terms for scalar relativistic and DBOC corrections are insignificant.