3D high resolution nanoparticles imaging technology development through AI

(From left side) Professor Ye Jong-cheol in Bio and brain engineering Doctorate Han Yo-seob Doctors degree Cha En-joo masters degree Jeong Hyeong-jin [Source = KAIST]

Faculty research team of Ye Jong-cheol in Bio and Brain Engineering in KAIST said that they developed AI technology that enhances 3D figure and restoration performance of composition distribution through collaboration with Samsung Advanced Institute of Technology. Collaboration team utilizes system that combines Energy Dispersive X-ray Spectroscopy (EDX) with Scanning Transmission Electron Microscope (STEM).

By restructuring figure of substance and composition distribution that form nanoparticles through the research it can gives a help to concrete analysis of semiconductor particle likes quantum dot that comprises of common used display.

The study analysis result that is participated with doctorate Han Yo-seob in faculty professor team doctorate Cha Eun-joo masters degree Jeong Hyeong-jin professional research team Jang Jae-deok professional researcher Lee Joon-ho as co-author at this time is posted in online Nature Machine Intelligence in 8th February. (Name of thesis : Deep learning STEM-EDX tomography of nanocrystals)

Energy Dispersive X-ray Spectroscopy (EDX) is used in analysis of substance for nanoparticles and is possible for chemical analysis in that it seems unique release spectrum following substance of product that is responded by X-ray. In order to find out between the fault and degradation mechanism of various nano material like quantum dot and battery etc importance and necessity of spectroscopy that can be analyzed with figure and composition distribution analysis has been surged.

However to enhance resolution of EDX measurement signal when nano material is exposed to electronic beam for a long time permanent damage of material would be caused. In this reason data acquisition time of projection for 3D imaging of nano particle is restrained and the way to reduce angle which measures the scan time for one angle or decreases angle that measures is used. When the 3D imaging is restored through projection data that is acquired by existing way measurement of atom signal that is existed of small amount could not be done or resolution and restoration imagings degree of accuracy will be low.

However kernel regression of AI base that is development by co-research team by themselves and projection enhancement is developed of accuracy and resolution dramatically. Research team developed network that provides data which enhances signal-to-noise ratio (SNR) to signal of projection data lessen the scan time through kernel regression of AI that can learn by itself and then based on the improved high resolution EDX projection data it is succeed to restore accurate 3D restoration imaging through small amount of projection data that can not be done by existing way.

Algorithm that is developed by research team can distinguish the figure of atom that forms nanoparticles compared with existing EDX 3D reconfiguration technique based on measurement signal from boundary line and co-relation with optical characteristic of sample is verified in 3D imaging quantum dot of restored various core-shell structure.

Picture 1. 3D restoration result that is restored with developed algorithm and projection data of quantum dot [Source=KAIST]

Picture 2. When synthesize the quantum dot two kind of synthesis restoration comparison result is made of difference to shell coding processing [Source=KAIST]

Website: : http://www.wip-news.com/news/articleView.html?idxno=5244