Nanoporous metal-organic framework (MOF) materials are strong candidates for energy efficient carbon capture and storage (CCS) technologies. A total of ∼20,000 hypothetical MOFs were ab initio screened for CO2 adsorption using grand canonical Monte-Carlo (GCMC) simulations. Novel radial distribution function (RDF) scores were modified for periodic systems to predict the CO2 adsorption of MOFs using chemoinformatic models. The test set predictions yielded accuracies of 0.76 and 0.85 at 0.1 bar and 1 bar, respectively. The models were used to screen a large database for high performing MOFs and the top 100 structures were successfully validated by GCMC simulations. The chemoinformatic predictors of the CO2 adsorption of MOFs are available online at http://titan.chem.uottawa.ca/woolab/MOFIA/#carbondioxide.

There is a long history of using neural networks for function approximation in computational physics and chemistry. Despite their conceptual simplicity, the practitioner may face difficulties when it comes to putting them to work. This small guide intends to pinpoint some neural networks pitfalls, along with corresponding solutions to successfully realize function approximation tasks in physics, chemistry or other fields.

Metal organic frameworks (MOFs) are porous solids that are potential high performance carbon capture materials. We have mined a hypothetical MOF database for structures that have exceptional low-pressure CO2 adsorption properties. We have applied the REPEAT method to generate accurate atomic charges that regenerate the ab initio electrostatic potential. We show that large scale screening at high accuracy is feasible for thousands of structures. We identify promising synthesis targets, like a simple combination of chrysene linker and vanadium inorganic unit, and examine in detail structural features that make better performing MOFs from those that would not be synthesisable. We find that, although screening large numbers of hypothetical structures is necessary to provide experimental targets, there are limitations to the suggestion of using the this database directly for synthesis targets and propose improvements and constraints that should be incorporated into the design of further generations of such a building-block algorithm to reach the accuracy required for high-quality CO2 adsorption simulation.

We have studied a procedure to determine Tight-Binding (TB) parameters automatically, by which the band structure of the crystalline solid can be reproduced so as to be good agreement with that of first-principles molecular dynamics calculation. According to this procedure, we determine TB parameter sets for silicon and diamond accurately, and a fairly good set for their compound SiC.

Computational screening is becoming increasingly useful in the search for new materials. We are interested in the design of new semiconductors to be used for light harvesting in a photoelectrochemical cell. In the present paper, we study the double perovskite structures obtained by combining 46 stable cubic perovskites which was found to have a finite bandgap in a previous screening-study.1 The four-metal double perovskite space is too large to be investigated completely. For this reason we propose a method for combining different metals to obtain a desired bandgap. We derive some bandgap design rules on how to combine two cubic perovskites to generate a new combination with a larger or smaller bandgap compared with the constituent structures. Those rules are based on the type of orbitals involved in the conduction bands and on the size of the two cubic bandgaps. We also see that a change in the volume has an effect on the size of the bandgap. In addition, we suggest some new candidate materials that can be used as photocatalysts in one- and two-photon water splitting devices.