Applying Machine Learning to Vibrational Spectroscopy

Abstract

The low-energy region of the potential energy surface (PES) of the protonated phenylalanine/serine dimer is mapped using the basin-hoping search algorithm, and 37 isomers are identified within 180 kJ·mol of the global-minimum structure. Cluster structures are grouped using hierarchical clustering to partition the PES in terms of nuclear configuration. Calculated IR spectra for the various isomers are then compared with the isomer-specific IR spectra by means of the cosine distance metric to facilitate spectral assignment and identify which regions of the PES are populated in the electrospray ionization process.

Microhydration of protonated 5-hydroxyindole revealed by infrared spectroscopy

“…For larger hydrates (and more flexible organic chromophores), it may be rather difficult to find geometries by hand and, as a consequence, systematic computational sampling techniques such as basin-hopping or molecular dynamics should be employed. [28][29][30] Geometries, energies, and harmonic IR spectra of stable 5HIH + -W n (n = 1-3) and 5HIH + -W-Ar/N 2 (ESI †) structures are calculated at the B3LYP-D3/aug-cc-pVTZ level using GAUSSIAN09. 27,[31][32][33][34] This hybrid density functional with additive dispersion correction has proven to yield reliable results for related aromatic clusters.…”

Augmenting Basin-Hopping With Techniques From Unsupervised Machine Learning: Applications in Spectroscopy and Ion Mobility

“…To determine which (if any) of the computed [Ser 2 + H] + isomers are observed experimentally, calculated harmonic vibrational spectra were compared against the experimental IRMPD spectrum using the methodology outlined by Fu and Hopkins (2018) The experimental spectrum employed was a concatenation of the spectra recorded by Seo et al in the 1,000–1,900 cm −1 region and by Sunahori et al in the 3,200–3,800 cm −1 region (Sunahori et al, 2013; Seo et al, 2018). These spectra were digitized using a custom-written python script from figures in their respective publications, interpolated in 2 cm −1 intervals, then normalized such that the maximum intensity in each region was set to 1.…”

“…A second option is to represent the molecular nuclear configuration as a vector, R ⇀ (Fu and Hopkins, 2018), containing the mass-weighted distance between each atom and the molecular center-of-mass:…”

Sodium cationization can disrupt the intramolecular hydrogen bond that mediates the sunscreen activity of oxybenzone

“…The fit between the experimental IRMPD spectra and spectra generated by DFT calculations was assessed by comparing the cosine similarity of the lowest-energy structures. This similarity measure has been shown to be a successful tool for evaluating molecular-based systematic relationships, [39][40][41] including IRMPD spectral fits. 42 Fit similarity is assessed as the cosine of the angle y between two spectral vectors A and B, which is calculated as the dot product of these vectors divided by the product of their Euclidian norms, according to:…”