[M-IS17] Interface- and nano-phenomena on crystal growth and dissolution

[MIS17-P06] Development of an early detection method of nucleation in in-situ liquid-cell TEM observation

*Yuki Kimura¹, Hiroyasu Katsuno¹, Tomoya Yamazaki¹, Ichigaku Takigawa^2,3 (1.Institute of Low Temperature Science, Hokkaido University, 2.AIP, RIKEN, 3.ICReDD, Hokkaido University)

Keywords:Nucleation, Machine learning, Transmission electron microscopy

One of our goal is direct observation at the moments of nucleation in atomic scale. Nucleation process seems easy: Growth units agglomerate and form a particle by overcoming a nucleation barrier, which depends on a difference of the chemical potentials of two phases. However, recent investigations using a liquid-cell transmission electron microscope (LC-TEM) have reported more complex processes in the actual case. For instance, oriented-attachment growth of nano-crystallites [1] and the two-step nucleation of calcium carbonate [2] are reported. We found that two types of amorphous particles with different roles and properties promote protein crystallization [3]. It is easy to observe a nucleation by our naked eye or under optical microscope because of lower magnifications, i.e., large solution volume. However, it is not easy to observe a nucleation process in atomic scale under TEM, because even if the nucleation rate is controlled to be very high, the solution volume in the field of view is much smaller and, therefore, expected number of nuclei becomes much lower. In addition, it is also difficult to increase nucleation rate dramatically after starting the observation. Recent achievements have been based on great efforts such as a bunch of experimental runs and long-term observations. Therefore, to observe whole a process of nucleation with higher magnification, we are trying to develop an early detection method of a nucleation event in a stochastic nature using a neural network model.


Acknowledgement
This work was supported by Grant-in-Aid for Scientific Research (S) of JSPS KAKENHI Grant Number 20H05657.

References
[1] D. Li et al., Science 336 1014 (2012).
[2] M. H. Nielsen et al., Science 345 1158 (2014).
[3] T. Yamazaki et al. PNAS 114 2154 (2017).