Predicting binding energies of astrochemically relevant molecules via machine learning

Abstract

Context. The behaviour of molecules in space is to a large extent governed by where they freeze out or sublimate. The molecular binding energy is therefore an important parameter for many astrochemical studies. This parameter is usually determined with time-consuming experiments, computationally expensive quantum chemical calculations, or the inexpensive yet relatively inaccurate linear addition method. Aims. In this work, we propose a new method for predicting binding energies (BEs) based on machine learning that is accurate, yet computationally inexpensive. Methods. We created a machine-learning (ML) model based on Gaussian process regression (GPR) and trained it on a database of BEs of molecules collected from laboratory experiments presented in the literature. The molecules in the database are categorised by their features, such as mono- or multilayer coverage, binding surface, functional groups, valence electrons, and H-bond acceptors and donors. Results. We assessed the performance of the model with five-fold and leave-one-molecule-out cross validation. Predictions are generally accurate, with differences between predicted binding energies and values from the literature of less than ±20%. We used the validated model to predict the binding energies of 21 molecules that were recently detected in the interstellar medium, but for which binding energy values are unknown. We used a simplified model to visualise where the snow lines of these molecules would be located in a protoplanetary disk. Conclusions. This work demonstrates that ML can be employed to accurately and rapidly predict BEs of molecules. Machine learning complements current laboratory experiments and quantum chemical computational studies. The predicted BEs will find use in the modelling of astrochemical and planet-forming environments.

Simulation of CH$_3$OH ice UV photolysis under laboratory conditions

“…Appendix C contain the list of newly added surface photolysis and Appendix D shows the bimolecular surface reactions and binding energies used in this paper cross-checked against Wakelam et al (2017). We highlight that we assume the binding energy as a single value, and not as a distribution as discussed by Shimonishi et al (2018), Grassi et al (2020), Bovolenta et al (2022), Villadsen et al (2022), andMinissale et al (2022). In addition to the methanol formation and destruction pathways included in these lists, we have added the methanol ice photodissociation routes (see Table 1).…”

Graphs and Kernelized Learning Applied to Interactions of Hydrogen with Doped Gold Nanoparticle Electrocatalysts

“…Moreover, a similar size of the training data as here, only thousands of observations, has been used with multilabel classification problems. − Both shallow and deep neural networks (NNs) are popular tools in catalysis. ,− However, NNs are known to be “data hungry”, often requiring at least a few tens of thousands of training data points to reach their optimal performance, thus further justifying the method choice. Furthermore, several non-NN-based methods have also shown to be able to predict binding behavior accurately for molecules and hydrogen on various surfaces and doped graphene. ,,− …”

A statistical and machine learning approach to the study of astrochemistry

“…Across the literature, there is oen signicant disagreement when it comes to the values of binding energies. [24][25][26] While there exist many different methods of estimating these values, [27][28][29] we utilise a Bayesian inference approach.…”