Influence of temperature and bias voltage on the performance of a high resolution PET detector built with position sensitive avalanche photodiodes

We evaluate the performance of an 8 × 8 array of 0.9 × 0.9 × 1 mm 3 cerium doped lutetium oxyothosilicate (LSO) crystals coupled to a position sensitive avalanche photodiode (PS-APD) as a function of bias voltage and temperature. We use this detector to develop a general methodology to optimize bias voltage, temperature, and gain for PET detectors using semiconductor photodetectors. This detector module will be used in a novel high resolution positron emission tomography (PET) camera dedicated to breast imaging under construction in our lab. Due to the tight packing of many PSAPDs in the system a thermal gradient is expected across the imaging heads. Data were collected for 11 PSAPD temperatures between 5 • C and 40 • C using a thermo-electric (Peltier) device. At each temperature the bias voltage was varied in steps of 5 V over a 50 V range. We present three methods to predict the optimal bias voltage at every temperature: one based on optimizing the coincidence time resolution, the others based on the relative change in PSAPD gain and leakage current due to the onset of hole multiplication. Optimal gain could also be predicted based on the quality of the flood histogram. At optimal bias voltage, the energy resolution degrades as (10.5 ± 0.1) + (0.038 ± 0.006) / • C · T %. Time resolution stays constant at 2.37 ± 0.02 ns below 15 • C. Above this temperature, time resolution deteriorates as (1.67 ± 0.06) + ((0.042 ± 0.002) / • C · T) ns. Even at high temperatures , all 64 crystal position peaks in the flood histogram are still clearly visible. The width of the peaks in the flood histogram show a quadratic degradation with temperature: (2.6 ± 0.1) · 10 −2 + (1.6 ± 0.2) · 10 −5 /(• C) 2 · T 2. We conclude that both the quality of the flood histogram as well as the coincidence time resolution are better parameters to estimate the optimal bias voltage, than energy resolution. Optimal bias voltage is found to be dependent on the value of k, the ratio between hole and electron multiplication. We achieve optimal bias at a similar gain at all temperatures. The optimal bias voltage changes linearly across the observed range.