Different machine learning (ML) models are proposed in the present work to predict density functional theory-quality barrier heights (BHs) from semiempirical quantum mechanical (SQM) calculations. The ML models include a multitask deep neural network, gradient-boosted trees by means of the XGBoost interface, and Gaussian process regression. The obtained mean absolute errors are similar to those of previous models considering the same number of data points. The ML corrections proposed in this paper could be useful for rapid screening of the large reaction networks that appear in combustion chemistry or in astrochemistry. Finally, our results show that 70% of the features with the highest impact on model output are bespoke predictors. This custom-made set of predictors could be employed by future Δ-ML models to improve the quantitative prediction of other reaction properties.

In this work we perform molecular dynamics simulations of mixtures of a prototypical protic ionic liquid, 1-butyl-3-methylimidazolium tetrafluoroborate ([BMIM][BF 4 ]), with lithium tetrafluoroborate (LiBF 4), confined between two borophene walls of three different surface charges,-1, 0 and +1 e/nm 2 , where e is the elementary charge. The properties of the system are analyzed by means of ionic density profiles, angular orientations of [BMIM] + cations close to the wall and vibrational densities of states for the salt cations close to the walls. Lateral structure of the first layer close to the surface is also studied on one hand, calculating Minkowski parameters and the Shannon entropy of the patterns of the 2D density maps of the anions placed there and, on the other hand, computing the 2D-Fourier transform of the positions of these anions. Our results are compared with those obtained previously for the same mixtures confined between two graphene walls. Although similarities exist between both cases, interesting differences are observed in the lateral structure that the ionic liquid

No abstract