Dynamics of catalyst nanoparticles quantified from in situ TEM video

Quantification of the restructuring and migrating behaviors of working nanocatalysts at high spatiotemporal resolution is of a rigorous challenge but of vital significance to provide insights into the microstructural intrinsic of catalytic stability under the stimuli of the reaction environment. In this work, a deep learning-driven in situ TEM video quantification has been developed, capable of identifying and tracking every nanoparticle within the multi-particles video recorded during catalytic reaction. Through this methodology, evolutionary tracks of NiAu particles during catalyzing CO2 hydrogenation and CuPd particles in a redox environment have been resolved. These quantitative behaviors of reconstruction and migration derived from in situ TEM data, for the first time, unravel the surface-anisotropic catalytic reaction over individual particle, which is consistently measured as multiple changing descriptors including particle diameter/area, circularity, and migration velocity. Such reaction and microstructure inhomogeneity deconstructed from working nanocatalyst offers convincing elucidation about the micro-dynamic mechanism of catalyst coalescence and migration. This paper highlights the merits of interdisciplinary study rooting in artificial intelligence-driven in situ TEM analysis.