Rapid and Accurate Estimation of Activation Free Energy in Hydrogen Atom Transfer-Based C–H Activation Reactions: From Empirical Model to Artificial Neural Networks

Abstract

A well-performing machine learning (ML) model is obtained by using proper descriptors and artificial neural network (ANN) algorithms, which can quickly and accurately predict activation free energy in hydrogen atom transfer (HAT)-based sp 3 C–H activation. Density functional theory calculations (UωB97X-D) are used to establish the reaction system data sets of methoxyl (CH 3 O·), trifluoroethoxyl (CF 3 CH 2 O·), tert -butoxyl (tBuO·), and cumyloxyl (CumO·) radicals. The simplified Roberts’ equation proposed in our recent study works here [ R 2 = 0.84, mean absolute error (MAE) = 0.85 kcal/mol]. Its performance is comparable with univariate Mulliken-type electronegativity (χ) with the ANN model. The ANN model with bond dissociation free energy, χ, α-unsaturation, and Nolan buried volume (% V buried ) successively improves R 2 and MAE to 0.93 and 0.54 kcal/mol, respectively. It reproduces the test sets of trichloroethoxyl (CCl 3 CH 2 O·) with R 2 = 0.87 and MAE = 0.89 kcal/mol and accurately predicts the relative experimental barrier of the HAT reactions with CumO· and the site selectivity of CH 3 O·.

Repurposing quantum chemical descriptor datasets for on-the-fly generation of informative reaction representations: application to hydrogen atom transfer reactions

“…“alkoxy dataset” entry in Table 2). 79 In previous work, an empirical model based on five computed parameters, respectively describing the electronegativities and delocalization of the reactants and products, as well as the thermodynamic driving force, was proposed. 59 Applying this empirical model to the full dataset, an MAE 0.85 kcal mol −1 ( R 2 = 0.84) is obtained on the training set, and 1.24 kcal mol −1 ( R 2 = 0.72) on the test set.…”

“…alkoxy" entry in Table 2). 79 In Fig. 7, the performances of the various model architectures are presented.…”

Substituting density functional theory in reaction barrier calculations for hydrogen atom transfer in proteins

“…12 Previous works oen reported respectable accuracies below 1 kcal mol −1 , but relied on inputs derived from DFT calculations. [13][14][15] While a low number of single point energy calculations are still less expensive than the otherwise required optimizations, ideally one would want to replace all DFT calculations at inference time, be it at the loss of some accuracy. And indeed, this has been attempted for, e.g., catalysis on metal surfaces and reactions between small molecules, reaching an MAE of 5 and 2.6 kcal mol −1 , respectively.…”

Polar Effects in Hydrogen Atom Transfer Reactions from a Proton-Coupled Electron Transfer (PCET) Perspective: Abstractions from Toluenes

“…17,25 Others have incorporated these effects into empirical models through statistical analysis and machine learning. 9,16,21,26,27 From one perspective, the proposal of asynchrony or imbalance in a PCET reaction harkens back to the classic physical-organic discussions of "asynchrony" that were often illustrated as off-diagonal paths in More O'Ferrall−Jencks diagrams such as in Figure 1. 28,29 In his Principle of Nonperfect Synchronization (PNS), Bernasconi related this imbalance in PT reactions to changes in their intrinsic barriers.…”

“…Analyses of HAT reactions have recently been expanded to include “imbalances” in the ET or PT components of a concerted e – /H + (H • ) transfer. − Computational and experimental studies alike have found that the relationship between Δ G ° PT and Δ G ° ET can affect barrier heights in ways not captured by Δ G ° HAT . These effects have sometimes been attributed to a so-called “asynchronous” or “imbalanced” transition state, wherein the PT and ET reaction coordinates have proceeded to different extents. − Srnec in particular has developed theoretical models for the effects of asynchronicity and discussed the possible connection with polar effects. , Others have incorporated these effects into empirical models through statistical analysis and machine learning. ,,,, …”