The efficiency of a catalyst relies on the capability of promoting reactions directly on its surface. Thus, there have been many attempts to accurately determine the atomic structure at the surface when gasses are flown on the catalytic sample.
Obviously, in situ electron microscopy is the best candidate to provide that answer, due to the possibility of recording ultra high resolution information while flowing gasses on the sample at high temperature. However, due to the nature of the imaging process, only 2D projections can be captured, severely limiting our understanding of the catalytic process.
Researchers at EMAT, Antwerp, have combined the high stability of the Climate in situ gas&heating solution with their multi-year experience on developing algorithms to accurately retrieve 3D structures down to atomic resolution. Platinum nanoparticles were cyclically exposed to reducing and oxidizing gas mixtures to promote and study surface re-faceting; after each step, high resolution STEM images were acquired and they were analyzed by means of a novel methodology based on deep convolutional neural networks (CNN) and molecular dynamics simulations.
The results show that it is now possible to obtain very accurate 3D atomic models that enable researchers to “see and count” atoms which are sitting on the surface. These new findings will allow a much deeper characterization and understanding of the processes behind catalytic reactions.
Obviously, in situ electron microscopy is the best candidate to provide that answer, due to the possibility of recording ultra high resolution information while flowing gasses on the sample at high temperature. However, due to the nature of the imaging process, only 2D projections can be captured, severely limiting our understanding of the catalytic process.
Researchers at EMAT, Antwerp, have combined the high stability of the Climate in situ gas&heating solution with their multi-year experience on developing algorithms to accurately retrieve 3D structures down to atomic resolution. Platinum nanoparticles were cyclically exposed to reducing and oxidizing gas mixtures to promote and study surface re-faceting; after each step, high resolution STEM images were acquired and they were analyzed by means of a novel methodology based on deep convolutional neural networks (CNN) and molecular dynamics simulations.The results show that it is now possible to obtain very accurate 3D atomic models that enable researchers to “see and count” atoms which are sitting on the surface. These new findings will allow a much deeper characterization and understanding of the processes behind catalytic reactions.