Mechanism for Forming B,C,N,O Rings from NH3BH3 and CO2 via Reaction Discovery Computations

What Does the Machine Learn? Knowledge Representations of Chemical Reactivity

In a departure from conventional chemical approaches, data-driven models of chemical reactions have recently been shown to be statistically successful using machine learning. These models, however, are largely black box in character and have not provided the kind of chemical insights that historically advanced the field of chemistry. To examine the knowledgebase of machine-learning models-what does the machine learn-this article deconstructs black-box machine-learning models of a diverse chemical reaction data set. Through experimentation with chemical representations and modeling techniques, the analysis provides insights into the nature of how statistical accuracy can arise, even when the model lacks informative physical principles. By peeling back the layers of these complicated models we arrive at a minimal, chemically intuitive model (and no machine learning involved). This model is based on systematic reaction-type classification and Evans-Polanyi relationships within reaction types which are easily visualized and interpreted. Through exploring this simple model, we gain deeper understanding of the data set and uncover a means for expert interactions to improve the model's reliability.

Predicting Hydride Donor Strength via Quantum Chemical Calculations of Hydride Transfer Activation Free Energy

We propose a method to approximate the kinetic properties of hydride donor species by relating the nucleophilicity (N) of a hydride to the activation free energy ΔG^⧧ of its corresponding hydride transfer reaction. N is a kinetic parameter related to the hydride transfer rate constant that quantifies a nucleophilic hydridic species' tendency to donate. Our method estimates N using quantum chemical calculations to compute ΔG^⧧ for hydride transfers from hydride donors to CO₂ in solution. A linear correlation for each class of hydrides is then established between experimentally determined N values and the computationally predicted ΔG^⧧; this relationship can then be used to predict nucleophilicity for different hydride donors within each class. This approach is employed to determine N for four different classes of hydride donors: two organic (carbon-based and benzimidazole-based) and two inorganic (boron and silicon) hydride classes. We argue that silicon and boron hydrides are driven by the formation of the more stable Si-O or B-O bond. In contrast, the carbon-based hydrides considered herein are driven by the stability acquired upon rearomatization, a feature making these species of particular interest, because they both exhibit catalytic behavior and can be recycled.

Thermodynamic and kinetic hydricities of metal-free hydrides

Thermodynamic and kinetic hydricities provide useful guidelines for the design of hydride donors with desirable properties for catalytic chemical reductions.

Uncovering reaction sequences on surfaces through graphical methods

ZStruct is a graph-based model that generates an ensemble of plausible reaction pathways starting from a given initial state, without requiring prior knowledge of reaction intermediates.