Thermochemistry of Tungsten-3p Elements for Density Functional Theory, Caveat Lector!

Abstract

There are two primary foci in this research on WE (E = Si, P, and S) bonds: prediction of their bond dissociation enthalpies (BDEs), including σ- and π-bond energy components, and assessing the uncertainty of these BDE predictions for levels of theory commonly used in the literature. The internal standards for computational accuracy include metal-element bond lengths (mean absolute error = 1.8 ± 1.2%), main group homolog BDEs versus higher levels of ab initio theory (W1U and G4 BDEs, R2 = 0.98), and DLPNO-CCSD(T)/def2-QZVPP calculations for metal-ligand BDEs (R2 = 0.88). The W═Si first π-bond is underreported for density functional theory (DFT)/MP2 methods versus DLPNO-CCSD(T), while the latter shows negligible strength for the W;Si second π-bond, consistent with the literature. This research highlights clear issues with the underlying assumptions required for the use of perturbation theory methods for the fragments derived from W-P homolysis. The difficulties associated with modeling the metal thermochemistry with DFT (and MP2) levels of theory are manifest in the broad standard deviations observed. However, the average BDEs found using 48 popular DFT and MP2 levels of theory are reliable, 10.8 ± 6.8% mean absolute error (with W-P removed) versus DLPNO-CCSD(T), with the caveat that the individual basis set/pseudopotential/valence basis set combination can vary wildly. Analysis of the absolute error percentages with respect to the level of theory indicates little benefit to going higher on Jacob's Ladder, as simpler methods have lower error versus high-level ab initio techniques such as G4 and DLPNO-CCSD(T).