Quantitative Evaluation of Supported Catalysts Key Properties from Electron Tomography Studies: Assessing Accuracy Using Material-Realistic 3D-Models

Abstract

Electron Tomography (ET) reconstructions can be analysed, via segmentation techniques, to obtain quantitative, 3D-information about individual nanoparticles in supported catalysts. This includes values of parameters out of reach for any other technique, like their volume and surface, which are required to determine the dispersion of the supported particle system or the specific surface area of the support; two figures that play a major role in the performance of this type of catalysts. However, both the experimental conditions during the acquisition of the tilt series and the limited fidelity of the reconstruction and segmentation algorithms, restrict the quality of the ET results and introduce an undefined amount of error both in the qualitative features of the reconstructions and in all the quantitative parameters measured from them. Here, a method based on the use of well-defined 3D geometrical models (phantoms), with morphological features closely resembling those observed in experimental images of an Au/CeO2 catalyst, has been devised to provide a precise estimation of the accuracy of the reconstructions. Using this approach, the influence of noise and the number of projections on the errors of reconstructions obtained using a Total Variation Minimization in 3D (TVM-3D) algorithm have been determined. Likewise, the benefits of using smart denoising techniques based on Undecimated Wavelet Transforms (UWT) have been also evaluated. The results clearly reveal a large impact of usual noise levels on both the quality of the reconstructions and nanometrological measurement errors. Quantitative clues about the key role of UWT to largely compensate them are also provided.