Abstract
High-purity germanium (HPGe) detectors are widely used to measure gamma rays and identify radioactive materials, but accurate results depend on both reliable detector models and careful calibration. This work aims to help improve high precision HPGe spectrometry by verifying the newly implemented pulse height tally capability in OpenMC and analyzing how correlated uncertainties propagate to the best-fit efficiency curve during a calibration.
To verify OpenMC’s PHT capability, an MCNP6.3 (the industry standard) model of a real-world HPGe detector was developed and tested extensively to ensure agreement with experimental results. An equivalent OpenMC model was developed to exactly replicate the material and geometry specifications of the MCNP model. Using 5E10 photon histories, a set of calibration isotopes with realistic source geometries were run against these models to analyze qualitative and quantitative differences between the artificial spectra produced by both transport codes. The results demonstrate that the newly-implemented PHT capability of OpenMC very closely replicates the response function of MCNP. This supports the use of OpenMC as a viable open-source radiation transport tool for some applications in HPGe gamma spectrometry.
In the second half of this thesis, a detailed covariance analysis is applied to the task of uncertainty propagation of source activities to the best-fit coefficients of the efficiency curve. In a practical HPGe counting experiment, the efficiency curve of a detector must be calibrated against a set of gamma sources which have uncertain quantities of activity. Gamma lines emitted from the same isotope should be treated as fully statistically correlated, while separate sources are uncorrelated. The distinction between these two levels of correlation has a significant impact on the propagation of uncertainty to subsequent measurements made with the calibrated detector. While covariance analysis is not new to HPGe spectrometry, this topic is under-represented within scientific literature with many authors misreporting error or choosing to avoid reporting uncertainty all together. The mathematics and Python code developed in this section are verified against expected statistical results.