# Evaluation of Density Functionals for Si–O–C–H Molecule Thermochemistry

## TL;DR

This paper evaluates how well different density functionals predict the thermochemistry of Si–O–C–H molecules compared to high-level quantum calculations and experiments.

## Contribution

The study provides new benchmark data for silicon-containing molecules and identifies the best-performing DFT functionals for various thermochemical properties.

## Key findings

- M06-2X functional has the lowest error for enthalpy of formation.
- SCAN functional performs best for vibrational frequencies and zero-point energies.
- B2GP-PLYP functional shows smallest errors in reaction stability evaluations.

## Abstract

The energies and vibrational frequencies of molecular
species with
Si–O–C–H compositions have been calculated at
the CCSD­(T) level (coupled cluster with single and double excitations
and a perturbative treatment of triple excitations). The CCSD­(T) results
compare well with experimental data where the difference in enthalpy
of formation is typically only about 1–2 kJ/mol in most cases.
In addition, the same molecules have been calculated with density
functional theory (DFT) calculations using nine commonly used density
functionals and two different basis sets. The performance of the DFT
calculations is compared with the CCSD­(T) benchmark values in terms
of enthalpy of formation, reaction energy, vibrational frequencies,
and zero-point energies. The results show that the M06-2X functional
provides the lowest mean absolute error (MAE) in terms of the enthalpy
of formation, whereas, for the vibrational frequencies and zero-point
energies, the SCAN functional gives the lowest MAE values. The results
were also grouped according to the types of bonds that are present
in the molecules. Moreover, an elaborate set of possible reactions
within the molecular species in the Si–O–C–H
system is included to evaluate the performance of the different DFT
functionals with respect to the relative stability of species within
the same reaction system. In this case, the B2GP-PLYP functional shows
the smallest errors. PW6B95 is the functional that most consistently
performs well for the studied properties of the included molecules.
The coupled cluster data sets provide new benchmark data, several
of which are not previously available for silicon chemistry.