這本次研究中藉由半導體技術(Technology Computer Aided Design, TCAD)模擬來研究並實際生產低增益雪崩放大二極體(Low Gain Avalanche Detector, LGAD)粒子感測器，這是目前許多高能對撞機預計用在更新軌跡偵測器的高時間解析度元件，我們試著用模擬來建立整個生產的過程與不同的光罩設計，以了解各製程步驟及結構對LGAD的特性的影響，從結構的模擬結果中發現在光罩設計中將核心的增益區與接面邊界延伸(Junction Termination Extension, JTE)分開10到15微米的間距可以減輕增益電場在邊緣因JTE而產生的尖端電場將崩潰電壓從260伏特提升至280伏特，同時不會影響有效區域的範圍及邊界，因此可在相同的占比內獲得更高的崩潰電壓。也藉由增益值模擬為製程參數訂出LGAD的合適參數範圍，並根據後來製程中的回饋加大測試幅度，最終將硼的劑量訂在每平方公分5×1013 ~ 2×1014並經過至少400分鐘的長時間擴散，搭配擴散200分鐘或30分鐘的磷來形成不同深淺的增益層。然而縱使已加重硼的劑量，最終並沒有形成預想中的增益層。根據二次離子質譜儀分析(Secondary Ion Mass Spectrometer, SIMS)的結果比對，推測在製程上有許多因素使量測劑量、深度等無法與製程參數匹配，導致不適合從此次結果建立模擬到實際製程的連結。從此次結果比對中顯示若下次遇到類似情況，從改變溫度著手或許更能穩定的改動濃度曲線來形成增益層。;In this study, technology computer aided design (TCAD) simulations are used to study and build a process flow for producing low gain avalanche diode (LGAD) particle sensors. The LGAD sensor is expected to be used in many high-energy colliders to upgrade the trajectory detectors and improve time resolution. The entire manufacturing process and various mask designs are simulated to understand how each step in the process step and the structure of the LGAD affect its properties. The structural simulation shows that leaving the 10 ~ 15 μm distance between the gain region and the junction termination extension (JTE) can reduce the intensity of the electric field protruding from the edge, which is formed by the gain electric field limited by the JTE. This reduction in intensity can increase the breakdown voltage from 260V to 280V while maintaining the same space utilization, active area, and boundary. Gain simulation provides the definitive standard for determining the range of process parameters. 5×1013 ~ 2×1014 cm-2 boron dose is diffused at least 400 minutes to co-locate with phosphorus diffusion for either 200 or 30 minutes, resulting in the formation of layers with different depths of gain layers. However, even when the boron dose is very high, the gain layer still does not form. Based on the results of the secondary ion mass spectrometer (SIMS), which indicate a discrepancy between the doses and concentration depths in the process flow, we suspect that these mismatched relationships are caused by other factors during the manufacturing process. Therefore, it is inappropriate to establish a relationship between simulation and manufacturing based on this result. The failed modification suggests that modifying the diffusion temperature could be a more dependable approach to resolving similar issues in the future.
