Data sampling scheme for reproducing energies along reaction coordinates in high-dimensional neural network potentials

Reversible Restrain and Release of the Dynamic Valence Isomerization in a Shape-shifting Bullvalene by Complex Formation

In search for structural features that enable the control of the valence isomerization of the fluxional bullvalene, a bullvalene-bis(harmane) conjugate is identified that acts as chelating ligand in complexes with metal ions. Spectrometric titrations show that this ligand forms 1 : 1 complexes with Ag+, Cu+, Cu2+, and Zn2+. Most importantly, detailed NMR-spectroscopic analysis at different temperatures reveals that the complexation with Ag+ strongly affects the dynamic isomerization of the bullvalene unit of the ligand such that only one predominant valence isomer is formed, even at 5 °C. Detailed 1H-NMR-spectroscopic studies disclose an increased barrier (~11 kJ mol-1) of the Cope rearrangement. Furthermore, the addition of hexacyclene displaces the Ag+ from the complex, so that the valence isomerization is accelerated and an equilibrium with two predominant isomers is formed. In turn, repeated addition of Ag+ regains the complex with the restrained isomerization of the bullvalene unit. This method to control the valence isomerism by straightforward chemical stimuli may be used to simplify structural analysis at elevated temperatures, i. e. a feature not available so far with bullvalenes, and it may be employed as functional element in dynamic supramolecular assemblies.

Dataset's chemical diversity limits the generalizability of machine learning predictions

The QM9 dataset has become the golden standard for Machine Learning (ML) predictions of various chemical properties. QM9 is based on the GDB, which is a combinatorial exploration of the chemical space. ML molecular predictions have been recently published with an accuracy on par with Density Functional Theory calculations. Such ML models need to be tested and generalized on real data. PC9, a new QM9 equivalent dataset (only H, C, N, O and F and up to 9 “heavy” atoms) of the PubChemQC project is presented in this article. A statistical study of bonding distances and chemical functions shows that this new dataset encompasses more chemical diversity. Kernel Ridge Regression, Elastic Net and the Neural Network model provided by SchNet have been used on both datasets. The overall accuracy in energy prediction is higher for the QM9 subset. However, a model trained on PC9 shows a stronger ability to predict energies of the other dataset.