# Analyzing Spectral Similarities for Structural Identification Using a New Benchmark Database

## TL;DR

This paper introduces a new database and methods to improve the accuracy of identifying molecular structures using vibrational spectra.

## Contribution

A new benchmark database and mode-dependent scaling factors are introduced to enhance spectral analysis accuracy.

## Key findings

- Mode-dependent scaling factors provide the highest accuracy in spectral analysis.
- Euclidean and Manhattan distances effectively reveal structural variations in spectra.
- The database and methods can serve as benchmarks for future predictive models.

## Abstract

Vibrational spectra, characterized by structurally sensitive
features,
offer critical insights into molecular structures, bonding, and dynamics.
Yet, interpreting measured spectra and identifying corresponding structures
require theoretical equivalents and quantitative analysis. Here, we
introduce a new experimental database that includes broad-range ionization-detected
stimulated Raman scattering signatures besides harmonic Raman frequencies
calculated with widely used density functional methods/basis sets.
By comparing experimental fundamental bands and computed data, we
derive single global and multiple range- and mode-dependent scaling
factors and analyze the resulting error distributions, showing that
mode-dependent scaling provides the greatest accuracy. Additionally,
we explore various methods for evaluating similarities between measured
fundamental spectra of different compounds and calculated data sets
of conformers. Our findings indicate that Euclidean and Manhattan
distance metrics for frequencies and intensities uncover subtle structural
variations, yielding spectral similarity rankings that facilitate
structural identifications. This new database and methodology address
key challenges in spectral assignment, and we anticipate that they
will serve as benchmarks for future predictive models and foster the
development of advanced strategies.