Search for: All records

A synergistic user-AI study for designing molecules with high CO₂affinity. We demonstrate how LLMs can increase chemists' creativity and highlight the vital role of domain knowledge. 

An active learning workflow that combines electronic structure and molecular topology was developed for the in silico design of organic ligands for actinium capture in solution. 

Abstract Carbon capture, sequestration and utilization offers a viable solution for reducing the total amount of atmospheric CO₂concentrations. On an industrial scale, amine‐based solvents are extensively employed for CO₂capture through chemisorption. Nevertheless, this method is marked by the high cost associated with solvent regeneration, high vapor pressure, and the corrosive and toxic attributes of by‐products, such as nitrosamines. An alternative approach is the biomimicry of sustainable materials that have strong affinity and selectivity for CO₂. Bioinspired approaches, such as those based on naturally occurring amino acids, have been proposed for direct air capture methodologies. In this study, we present a database consisting of 960 dipeptide molecular structures, composed of the 20 naturally occurring amino acids. Those structures were analyzed with a novel computational workflow presented in this work that considers certain interaction sites that determine CO₂affinity. Density functional theory (DFT) and symmetry‐adapted perturbation theory (SAPT) computations were performed for the calculation of CO₂interaction energies, which allowed to limit our search space to 400 unique dipeptide structures. Using this computational workflow, we provide statistical insights into dipeptides and their affinity for CO₂binding, as well as design principles that can further enhance CO₂capture through cooperative binding. 

A comprehensive computational study on the underlying reactivity of iron tetra-NHC complexes for C₂+ N₁aziridination catalysis is presented.