Department of Physics, Kyungpook National University
경북대학교 물리학과

KNU Physics Thesis Presentations

A Study of Silicon Photodetector at Cryogenic Temperatures and Deep Learning Data Analysis for PET Application

by Jin Jegal (Ph.D Candidate)

222호 (219동)




In this thesis, the characteristics of silicon photodetectors were studied and two applications using these silicon photodetectors were investigated. The temperature-dependent electrical characteristics of a silicon avalanche photodiode (APD) are studied at absolute temperatures from 26 K to 300 K in the cryostat. As the temperature decreases, not only the thermal noise of the semiconductor and the electron mobility decrease, but also the dark current noise is reduced dramatically. In addition, the voltage required to achieve the appropriate gain is lower. α, β, and γ spectra of 241Am, 207Bi, and 137Cs are obtained as a scintillation radiation detector by utilizing a methylammonium lead bromide (MAPbBr3) perovskite crystal coupled with the APD at cryogenic temperatures. The outstanding semiconducting and optical properties of MAPbBr3 perovskite crystal at cryogenic temperatures enable incident energy to be measured with high energy resolution and very low threshold energy through a combination with the low-temperature characteristics of the APD. With these advantages, the feasibility of utilizing the MAPbBr3 coupled with the APD for low-temperature experiments such as light-dark matter searches and space applications that require to achieve high quantum efficiency, picosecond timing resolution, and low-power consumption was investigated.

Furthermore, convolutional neural network (CNN)-based data reconstruction is performed by utilizing positron emission tomography (PET) application with silicon photomultipliers in the Kyungpook National University Advanced Positronium Annihilation Experiment (KAPAE) detector. This compact 4π detector can simultaneously detect γ-rays in all directions, enabling to study visible and invisible exotic decay processes and to locate a tumor position in PET applications for small animals. The usage of the CNN for the localization of positronium (Ps) annihilation synonymous with tumor localization is investigated. Two-γ decay systems of the Ps annihilation from 22Na and 18F radioactive sources are simulated using a GEANT4. The spatial error in the XY plane from the CNN is compared to those of the classical weighted k-means algorithm centroiding, Y-network, and ResNet. The proposed CNN-based data reconstruction will be helpful in studying Ps annihilation physics and effectively overcoming the spatial resolution limitations of PET in a nuclear medical imaging.



지도교수: 김홍주