Over the past decades nanostructured materials have changed the way we design, develop and deploy new technologies in optics and electronics, bio-physics, medicine and catalysis; either because they exhibit superior performance, or because they have increased the number of degrees of technological freedom to make decisions. In particular metallic nanoparticles are ideal for numerous catalytic reactions, such as hydrogenation and dehydrogenation, and CO oxidation in catalytic convertors, and so extensive studies have focused on developing more robust process/structure/property relationships. Typically the catalytic performance is linked to more than one structural feature, which are controlled through multiple processing steps. This makes optimizing performance a challenge, and robust and informative process/structure/property predictions are in high demand. Combining computational data generated using molecular dynamics (MD), domain knowledge and machine learning (ML) can provide unique and cost effective insights into complex multi-variable correlations that are obscured when using conventional methods. In this research we implement feature analytics with hierarchical clustering, principal components analysis, and tree based classification and regression models with <i>k</i>-fold cross-validation, to correlate the synthesis conditions, structural disorder and the catalytic performance of metallic nanoparticles. As we will show, the analysis can be equally applied to the output of any computational or experimental characterisation method, and all data sets that can be adequately described by a consistent set of meaningful descriptors that can inform further research and development.

From process to properties: Correlating synthesis conditions, structural disorder and the catalytic performance of metallic nanoparticles

255th National Meeting (2018)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.