CsI(Tl) Detector Array

Full-spectrum PID

Machine learning enables automatic classification of α, p, d, t, and γ across the full energy range, overcoming traditional methods that only work well at high energies.

Xu, J. H. · Fang, Yongde · Wang, Sicheng · Ding, Bing · Ideguchi, Eiji · M. Kumar, Raju · Carpenter, Michael P. · Wang, Xuyang · Hong-Yi, Wu · Hua W. · Jianhong, Li · Jia Z.H. · Wen, Liang · Guo, Song · Li, G. · Qiang, Yunhua · Liu, Minliang · Zhang, Wenqiao · Jia, C. X. · Chen, Ruofu · Chen, Jie · Zheng, Yun · Li, Congbo · Wu, Xiaoguang · Min, Zheng · Zi-Hao, Zhao · Yun-Qiu, Li · Li, Jian · Rui, Hong · Zi-Yang, He · Li, Tianxiao · Petrache, Costel Marian · Zhou, Xiaohong · Gan, Zaiguo · Zhang Y.H.

Nuclear Science and Techniques/Hewuli 2026

Specifications

Particle-gated coincidence statistics increase

{'zh': '50', 'en': '50'} %

Advantages

Works across full spectrum

Fuzzy c-means and support vector machine do not rely on energy-dependent fitting, so they perform robustly even at low energies, overcoming the limitation of traditional methods that work only at high energies.

50% more statistics

Because identification works across all energies, more particle events are retained, leading to an approximate 50% increase in particle-gated coincidence statistics and improved data quality.

Handles weakly bound

For particles from weakly bound projectiles such as 7Li or 9Be, SVM shows enhanced separation capability compared to FCM, giving more accurate identification.

Applications

• Nuclear physics experiments：Boost statistics in fusion evaporation experiments by classifying particles over all energies.
• Radiation detection：Automatically identify particle types, reducing manual analysis and improving detection efficiency.
• Non-destructive testing：Leverage γ and charged-particle classification to assist in non-destructive material testing.