The speciation of phosphates adsorbed on γ-alumina revealed by 31P NMR, AIMD and machine learning

Abstract

The chemical nature of adsorbed inorganic additives such as phophates used in the preparation of heterogeneous catalysts is suspected to impact their resulting activity. Predominant phosphate species located on the surfaces of the γ-alumina catalytic support are identified by using one-dimensional ³¹P NMR spectra as the only experimental input. The detailed insight is made possible by combining machine learning (ML) ³¹P chemical shift prediction and ab initio molecular dynamics (AIMD) to sample conformers of 10 representative possible structures and generate theoretical spectra, which were then used to decompose mathematically the broad experimental peak. At low P concentration, several types of monomeric species are found to co-exist on the γ-alumina (110) facets. Increasing the P concentration yields a marked increase in one monomeric species and one dimeric species both located on the (110) facets, whereas phosphates are mainly absent from the (100) facet. The NMR spectra broadening is interpreted by two levels of structural disorders: the various types of P species and the conformational distribution of each species. We finally propose some implications for the catalytic properties.