Comprehensive Assessment of GFN Tight-Binding and Composite Density Functional Theory Methods for Calculating...

Area of Science:

• Computational Chemistry
• Spectroscopy
• Quantum Mechanics

Background:

• Vibrational spectroscopy is crucial for chemical substance characterization.
• Accurate calculation of infrared (IR) spectra is essential for experimental analysis.
• Evaluating computational methods is key to improving spectral prediction.

Purpose of the Study:

• To assess the performance of semiempirical quantum mechanical (GFN tight-binding) and force-field methods for gas-phase IR spectra.
• To compare these methods against experimental data and low-cost density functional theory (DFT).
• To investigate the utility of atomic mass scaling as an alternative to frequency scaling.

Main Methods:

• Utilized a dataset of 7247 experimental references for automatic spectra comparison.
• Employed quantitative spectral similarity measures to compare theoretical and experimental spectra.
• Evaluated GFN1- and GFN2-xTB tight-binding methods, PMx competitors, and B3LYP-3c DFT.
• Investigated the impact of conformational changes on simulated spectra.

Main Results:

• Atomic mass scaling demonstrated an accurate and simple alternative to global frequency scaling in DFT and semiempirical calculations.
• The B3LYP-3c DFT composite method proved highly suitable for general IR spectra calculations.
• GFN1- and GFN2-xTB tight-binding methods outperformed PMx methods.
• Conformational changes had a moderate influence on simulated spectra, suggesting sampling may be simplified.

Conclusions:

• Semiempirical GFN-xTB methods and B3LYP-3c DFT offer efficient and accurate approaches for IR spectra prediction.
• Atomic mass scaling is a viable strategy for improving spectral accuracy in computational chemistry.
• Automated compound identification workflows may benefit from reduced conformational sampling.