Understanding the unique structural motifs in nanosized materials and their transformations remains a challenge, because the choice of experimental techniques that provide atomic-level information and are also suitable for in-situ and in-operando studies is still limited. Among those few, an invaluable tool is X-ray absorption spectroscopy (XAS), which provides rich information about the atomic arrangements around the absorbing metal atoms and their electronic state. The accuracy of conventional approaches for XAS data analysis is, however, limited, when they are applied to such intrinsically heterogeneous, disordered materials as nanoparticles (NPs). Information about the strong asymmetry of bond-length distributions, distant coordination shells and many-atom distribution functions can contain the key answers regarding the NPs 3D structure but cannot be accounted for adequately in conventional fitting of extended X-ray absorption fine structure (EXAFS). Even less developed is the quantitative analysis of another part of XAS signal, X-ray absorption near edge structure (XANES), which contains complimentary information, but has better signal-to-noise ratio, which means that the data can be collected with better time-resolution, for more diluted samples, on strongly attenuating support materials and/or in complex experimental setups. Recently we have demonstrated that machine learning methods, coupled with ab-initio XAS calculations can be successfully used to correlate XANES1 and EXAFS2 features with the descriptors of 3D local structure of metallic materials. Here we apply this method to studies of NPs size and shape effect in XANES and EXAFS spectra in mono- and bimetallic NPs.
1 J. Timoshenko, D. Lu, Y. Lin, A. I. Frenkel, J. Phys. Chem. Lett., 8, 5091-5098 (2017)
2 J. Timoshenko, A. Anspoks, A. Cintins, A. Kuzmin, J. Purans, A. I. Frenkel, Phys. Rev. Lett., 120, 225502 (2018).
 will give a talk entitled “Learning nanomaterials structure using neural networks”.){kind=link}