Dissociative Electron Attachment Prediction of Halogenated Organic Molecules Using Machine Learning.

Area of Science:

• Physical Chemistry
• Computational Chemistry
• Materials Science

Background:

• Dissociative Electron Attachment (DEA) is crucial in environmental science, nanotechnology, biology, and astrochemistry.
• Experimental DEA studies are limited to gas-phase, high-vacuum conditions, restricting analysis of larger or less volatile molecules.

Purpose of the Study:

• Develop machine learning (ML) models to predict negative ion formation in halogenated organic molecules.
• Enable computational prediction of DEA processes for molecules difficult to study experimentally.

Main Methods:

• Created classification models for DEA resonance energy range and regression models for peak energy estimation.
• Utilized a relational database of experimental DEA data and molecular descriptors for 143 molecules.
• Compared various ML algorithms, including ensemble Voting Classifier and Random Forest, with 120 molecules for training and 23 for testing.

Main Results:

• The ensemble Voting Classifier achieved 94.9% accuracy in cross-validation.
• The Random Forest model demonstrated a mean absolute error of 0.301 eV (cross-validation) and 0.234 eV (test set) for regression.
• ML models accurately predicted key DEA parameters, comparable to experimental resolutions.

Conclusions:

• Machine learning offers a feasible computational approach for predicting DEA phenomena.
• This work lays the groundwork for expanding DEA research to a wider range of molecules.
• The developed models can guide future experimental investigations and theoretical studies.