Identifying multi-stage nanocrystal growth using in situ TEM video data

The in situ transmission electron microscopy technique is receiving considerable attention in material science research, as its in situ nature makes possible discoveries that ex situ instruments are unable to make and provides the capability of directly observing nanocrystal growth processes. As incresing amounts of dynamic transmission electron microscopy (TEM) video data become available, one of the bottlenecks appears to be the lack of automated, quantitative, and dynamic analytic tools that can process the video data efficiently. The current processing is largely manual in nature and thus laborious, with existing tools focusing primarily on static TEM images. The absence of automated processing of TEM videos does not come as a surprise, as the growth of nanocrystals is highly stochastic and goes through multiple stages. We introduce a method in this article that is suitable for analyzing the in situ TEM videos in an automated and effective way. The method learns and tracks the normalized particle size distribution and identifies the phase-change points delineating the stages in nanocrystal growth. Using the outcome of the change-point detection process, we propose a hybrid multi-stage growth model and test it on an in situ TEM video, made available in 2009 by Science.