Machine Learning of Potential-Energy Surfaces Within a Bond-Order Sampling Scheme. In: Misra S, Gervasi O, Murgante B, Stankova E, Korkhov V, Torre C, Rocha AMA, Taniar D, Apduhan BO, Tarantino E, editors. Computational Science and Its Applications – ICCSA 2019. Cham: Springer International Publishing; 2019. pp. 388-400. [DOI: 10.1007/978-3-030-24311-1_28]

A density-matrix adaptation of the Hückel method to weak covalent networks

The coupled-monomers model is built as an adaptation of the Hückel MO theory based on a self-consistent density-matrix formalism. The distinguishing feature of the model is its reliance on variable bond and Coulomb integrals that depend on the elements of the density matrix: the bond orders and partial charges, respectively. Here the model is used to describe electron reactivity in weak covalent networks Xn±, where X is a closed-shell monomer. Viewing the electron as the simplest chemical reagent, the model provides insight into charge sharing and localisation in chains of such identical monomers. Data-driven modelling improves the results by training the model to experimental or ab initio data. Among key outcomes is the prediction that the charge in Xn± clusters tends to localise on a few (2-3) monomers. This is confirmed by the properties of several known cluster families, including Hen+, Arn+, (glyoxal)n-, and (biacetyl)n-. Since this prediction is obtained in a purely coherent covalent regime without any thermal excitation, it implies that charge localisation does not require non-covalent perturbations (such as solvation), decoherence, or free-energy effects. Instead, charge localisation is an intrinsic feature of weak covalent networks arising from their geometry relaxation and is ultimately attributed to the correlation between covalent bond orders and equilibrium bond integrals.

Chemical promenades: Exploring potential-energy surfaces with immersive virtual reality

The virtual-reality framework AVATAR (Advanced Virtual Approach to Topological Analysis of Reactivity) for the immersive exploration of potential-energy landscapes is presented. AVATAR is based on modern consumer-grade virtual-reality technology and builds on two key concepts: (a) the reduction of the dimensionality of the potential-energy surface to two process-tailored, physically meaningful generalized coordinates, and (b) the analogy between the evolution of a chemical process and a pathway through valleys (potential wells) and mountain passes (saddle points) of the associated potential energy landscape. Examples including the discovery of competitive reaction paths in simple A + BC collisional systems and the interconversion between conformers in ring-puckering motions of flexible rings highlight the innovation potential that augmented and virtual reality convey for teaching, training, and supporting research in chemistry.