A Machine-Learning-Driven Dataset of 140,000 PAH Infrared Spectra

This ZIP archive contains the infrared (IR) spectral dataset of polycyclic aromatic hydrocarbons (PAHs) presented in the companion paper. The dataset comprises 144,111 IR spectra across 48,037 closed-shell, even-carbon benzenoid PAH structures in neutral, cationic, and anionic charge states, accompanied by per-molecule uncertainty estimates. The spectra were computed using a machine-learning-based double-harmonic (ML-DH) model, as described in the paper below. ---------------------------Data Contents---------------------------This archive contains the ML-DH predicted spectra for the COMPAS PAH dataset, together with embedded molecular metadata and JSD-based confidence indicators. All spectra are provided in IR_COMPAS.zip. 1. File FormatThe spectra are stored in extended .xyz format. Each file contains the optimized molecular geometry and the corresponding ML-DH IR spectral data, and is named after the original molecule in the COMPAS-1, COMPAS-2, or COMPAS-3 dataset. 2. File StructureEach .xyz file follows the structure below: Line 1: Total number of atoms. Line 2: Comment line containing molecular metadata, e.g.: Source=COMPAS2 Formula=C18H12 Charge=0 JSD=0.02728where:Source: original COMPAS subset;Formula: molecular formula;Charge: charge state (0, +1, or -1);JSD: Jensen-Shannon divergence-based confidence indicator. For molecules without a direct validation spectrum, the JSD value is inherited from the most structurally similar validation molecule. Lower JSD values indicate higher expected reliability of the ML-DH predicted spectrum. Lines 3 to before spectral header: Optimized equilibrium geometry, with atomic symbols and Cartesian coordinates. Spectral header: scaled_Freq(cm^-1) Inten(KM/mol) scale_factor Following lines: IR spectral data in three columns:1. Scaled harmonic vibrational frequency (cm^-1);2. IR intensity (KM/mol);3. Frequency scaling factor. ---------------------------Citation--------------------------- If you use this dataset or the associated code in your research, please cite the following article: A Machine-Learning-Driven Dataset of 140,000 PAH Infrared Spectra for Interpreting Aromatic Infrared BandsAuthors: Xinghong Mai and Zhao Wang ---------------------------License---------------------------This dataset is distributed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License (CC BY-NC-SA 4.0). You are free to share and adapt the material for non-commercial purposes, provided that appropriate credit is given and that any derivative works are distributed under the same license.