Algorithms and Electronics for Processing Data from Pixelated Semiconductor Gamma-Ray Detectors

Abstract

Pixelated semiconductor gamma-ray detectors are a versatile tool for the characterization of radiation in a wide range of environments. The latest low noise, waveform sampling electronics coupled with innovative reconstruction algorithms allow for year-on-year improvements in energy and position resolution. Meanwhile, CdZnTe’s unparalleled combination of high energy resolution and 3D position sensitivity enables many advanced algorithms for imaging and other applications. The development of radiation detectors based on new materials further ensures a bright horizon for this technology. This dissertation presents new work covering all three aspects of pixelated semiconductor radiation detection in the Orion Group of the University of Michigan. New electronics development is presented which aims to condense all the front end control logic and some of the digital processing currently done by field programmable gate arrays and computers down onto a single silicon chip. This can enable significant further reductions in radiation detector size, weight and power while maintaining the high performance and imaging capabilities characteristic of CdZnTe. Subpixel position reconstruction accuracy and precision was improved by a series of new algorithms. Machine learning models trained to predict subpixel positions from signal waveforms demonstrated improvement over the resolution that could be achieved with classical techniques. The radial position was predicted to less than 200 µm precision using only the collecting pixel waveform. Additionally, an improvement of 50 µm position resolution is demonstrated for the XY subpixel coordinates. Meanwhile, a range of algorithms are considered for correcting position distortions due to space charge, nonideal weighting potentials and other effects. An iterative technique is presented that successfully corrects some of the distortions in the cathode/anode ratio depth reconstruction. A technique is presented to image pair production events in CdZnTe. This is typically accomplished with electron tracking. Here, a method to reconstruct images without electron tracking is presented, which exploits asymmetries in pair production and charge induction physics. The imaging capability is demonstrated both in simulation and in experiment with the 4.44 MeV photons from a PuBe source. Extensive work is done to characterize and improve the performance of perovskitebased radiation detectors. Thanks to waveform digitization and principal component analysis techniques, world-record performance of less than 1% full width at half maximum for the Cs-137 662 keV line is achieved with pixelated CsPbBr3-based detectors. Hybrid organic-inorganic solution grown perovskites are also demonstrated to work, with a resolution of less than 3% achieved for the same Cs-137 line.