Machine-learning prediction of infrared spectra of interstellar polycyclic aromatic hydrocarbons

Abstract

We design and train a neural network (NN) model to efficiently predict the infrared spectra of interstellar polycyclic aromatic hydrocarbons (PAHs) with a computational cost many orders of magnitude lower than what a first-principles calculation would demand. The input to the NN is based on the Morgan fingerprints extracted from the skeletal formulas of the molecules and does not require precise geometrical information such as interatomic distances. The model shows excellent predictive skill for out-of-sample inputs, making it suitable for improving the mixture models currently used for understanding the chemical composition and evolution of the interstellar medium. We also identify the constraints to its applicability caused by the limited diversity of the training data and estimate the prediction errors using a ensemble of NNs trained on subsets of the data. With help from other machine-learning methods like random forests, we dissect the role of different chemical features in this prediction. The power of these topological descriptors is demonstrated by the limited effect of including detailed geometrical information in the form of Coulomb matrix eigenvalues.

arXiv comment8 figures

References in corpus (17)

• Scikit-learn: Machine Learning in Python
• Fast and Accurate Modeling of Molecular Atomization Energies with Machine Learning
• Infrared Emission from Interstellar Dust. IV. The Silicate-Graphite-PAH Model in the Post-Spitzer Era
• Detection of the Aromatic Molecule Benzonitrile ($c$-C${}_6$H${}_5$CN) in the Interstellar Medium
• Machine Learning Molecular Dynamics for the Simulation of Infrared Spectra
• How to represent crystal structures for machine learning: towards fast prediction of electronic properties
• Early Science with SOFIA, the Stratospheric Observatory for Infrared Astronomy
• Finding unprecedentedly low-thermal-conductivity half-Heusler semiconductors via high-throughput materials modeling
• Unidentified Infrared Emission bands: PAHs or MAONs?
• Mapping PAH sizes in NGC 7023 with SOFIA
• The Carriers of the Interstellar Unidentified Infrared Emission Features: Aromatic or Aliphatic?
• High-resolution IR absorption spectroscopy of polycyclic aromatic hydrocarbons: the realm of anharmonicity
• The PAH hypothesis after 25 years
• Laboratory Determination of the Infrared Band Strengths of Pyrene Frozen in Water Ice: Implications for the Composition of Interstellar Ices
• Adsorption of Organic Molecules on Onion-Like-Carbons: Insights on the Formation of Interstellar Hydrocarbons
• Formation of interstellar complex polycyclic aromatic hydrocarbons: Insights from molecular dynamics simulations of dehydrogenated benzene
• Polycyclic aromatic hydrocarbon emission toward the Galactic bulge

Cited by in corpus (8)

• Computational Infrared Spectroscopy of 958 Phosphorus-bearing Molecules
• Infrared spectra of neutral polycyclic aromatic hydrocarbons by machine learning
• Machine-learning identified molecular fragments responsible for infrared emission features of polycyclic aromatic hydrocarbons
• Machine Learning interpretation of the correlation between infrared emission features of interstellar polycyclic aromatic hydrocarbons
• Evolution of Fullerenes in Circumstellar Envelopes by Carbon Condensation: Insights from Reactive Molecular Dynamics Simulations
• Neural Network Approach for Predicting Infrared Spectra from 3D Molecular Structure
• Density Functional Theory Calculations on the Interstellar Formation of Biomolecules
• Catalytic role of HI in the interstellar synthesis of complex organic molecule