STM Journals Peer-reviewed research from Consortium e-Learning Network

International Journal of Photochemistry and Photochemical Research Review Article

Using Machine Learning to Guess Photochemical Reaction Pathways

  1. V. Basil Hans Department of management and science, Srinivas University, Mangalore
  2. Aaron Kunal Tatpati Department of management and science, Srinivas University, Mangalore

Abstract

Photochemical reactions are crucial to many activities in the fields of energy conversion, environmental cleanup, and synthetic chemistry. However, predicting their causes and results effectively is still very hard since they entail excited electronic states, nonadiabatic transitions, and complicated potential energy surfaces. Machine learning (ML) has been a powerful technique to go along with classic quantum chemistry methods in the last few years. It offers better prediction capability and lower processing costs. This study investigates the amalgamation of machine learning algorithms with photochemical data to simulate and forecast reaction pathways under light irradiation. We use a carefully chosen set of known photochemical reactions, which includes transition states, absorption spectra, and reaction outcomes. Random forests, support vector machines, and graph neural networks are examples of supervised learning models that are trained on quantum-chemically derived descriptors. These descriptors include HOMO-LUMO gaps, oscillator strengths, and excited-state energies. Unupervised clustering methods are also used to sort response types and find new mechanistic patterns. Our results show that ML models can accurately forecast if a reaction will work and what intermediates are expected to generate, which is far better than baseline statistical methods. Also, interpretability methods show important electrical properties that affect photochemical behaviour. This work shows how useful data-driven methods can be in photochemistry and paves the way for faster discovery and smarter design of light-driven processes in materials science and synthetic chemistry.

Keywords

References (17)

  1. Meuwly M. Transformative Uses of Machine Learning in Chemical Reactions. 2021. doi:10.48550/arXiv.2101.03530
  2. Verma S, Rivera M, Scanlon DO, Walsh A. Machine-learned calibrations for high-throughput molecular excited state calculations. 2022. J. Chem. Phys.156, 134116 (2022)
  3. Gyo Jung S, Jung G, M. Cole J. Automatic Prediction of Peak Optical Absorption Wavelengths in Molecules Utilising Convolutional Neural Networks. 2024. ncbi.nlm.nih.gov
  4. Wigh D, Arrowsmith J, Pomberger A, C. Felton K et al. ORDerly: Data Sets and Benchmarks for Chemical Reaction Data. Twenty-four. ncbi.nlm.nih.gov
  5. Guan Y, W. Coley C, Wu H, Ranasinghe D et al. Predicting regio-selectivity using a machine-learned reaction model and quantum mechanical descriptors that are created on the fly. 2020. ncbi.nlm.nih.gov
  6. Friederich P, dos Passos Gomes G, De Bin R, Aspuru-Guzik A et al. Machine learning dihydrogen activation in the chemical environment of Vaska's complex. 2020. ncbi.nlm.nih.gov
  7. Wei J, Duvenaud D, Aspuru-Guzik A. Neural Networks for Predicting Organic Chemistry Reactions. 2016. ncbi.nlm.nih.gov
  8. Jorner K, Brinck T, Norrby PO, Buttar D. Machine learning and mechanistic modelling come together to make accurate predictions about activation energies in experiments. 2020. ncbi.nlm.nih.gov
  9. Schreiner M, Bhowmik A, Vegge T, Busk J, Winther O. Transition1x - a dataset for building generalizable reactive machine learning potentials. Scientific Data. 2022;9(1). doi:10.1038/s41597-022-01870-w
  10. Choi S. Using machine learning to guess the architectures of transition states in gas-phase chemical processes. Twenty twenty-three. ncbi.nlm.nih.gov
  11. Zhao Q, M. Anstine D, Isayev O, M. Savoie B. Δ(2) machine learning for predicting reaction properties. 2023. ncbi.nlm.nih.gov
  12. Baiardi A, A. Grimmel S, Steiner M, L. Türtscher P et al. Extensive Quantum Mechanical Investigation of Chemical Reaction Pathways. 2021. doi:10.48550/arXiv.2112.04944
  13. Schütt KT, Gastegger M, Tkatchenko A, Müller KR, Maurer RJ. Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions. Nature Communications. 2019;10(1). doi:10.1038/s41467-019-12875-2
  14. Schütt KT, Arbabzadah F, Chmiela S, Müller KR, Tkatchenko A. Quantum-chemical insights from deep tensor neural networks. Nature Communications. 2017;8(1). doi:10.1038/ncomms13890
  15. Beker W, Roszak R, Wołos A, H. Angello N et al. Machine Learning Might Merely Reflect Trends in Literary Popularity: A Case Study of Heterocyclic Suzuki–Miyaura Coupling. 2022. ncbi.nlm.nih.gov
  16. Mercado R, M. Kearnes S, W. Coley C. Data Sharing in Chemistry: Lessons Learnt and a Case for Mandating Structured Reaction Data. 2023. ncbi.nlm.nih.gov
  17. Zahrt A, Mo Y, Y. Nandiwale K, Shprints R et al. Machine Learning Helps Find Electrochemical Reactions. 2022. ncbi.nlm.nih.gov