Statistical investigation of the large-area Si(Li) detectors mass-produced for the GAPS experiment

The lithium-drifted silicon (Si(Li)) detector developed for the General Antiparticle Spectrometer (GAPS) experiment features a thick (∼2.2 mm) sensitive layer, large (10 cm) diameter, and excellent energy resolution (∼4 keV for 20–100 keV X-rays) at a relatively high operating temperature (approximately −40 °C). Mass production of GAPS Si(Li) detectors has been performed to construct a large-volume silicon tracker for GAPS. We achieved the first success of the mass production of large-area Si(Li) detectors with a high (∼90%) yield rate. Valuable datasets related to detector fabrication, such as detector performance and manufacturing parameters, were recorded and collected during the mass production. This study analyzes the datasets using statistical methods with the aim of comprehensively examining the mass production and to gain valuable insight into the fabrication method. Sufficient uniformities of the performance parameters (leakage current and capacitance) between detectors and strips are found, demonstrating high-quality and stable mass production. We also search for correlations between detector performance and manufacturing parameters by using data-mining techniques. Conventional multivariate analysis (multiple regression analysis) and machine-learning techniques (regression tree analysis) are complementarily used, and it is found that the Li-drift process makes a significant contribution to the performance parameters of the finished detectors. Detailed investigation of the drift process is performed using environmental data, and physical interpretations are presented. Our results provide valuable insight into the fabrication methods for this kind of large-area Si(Li) detector, and encourages future projects that require large-volume silicon trackers.