(676b) Improvement of Energy Resolution Via Correction On Non-Uniform Light Collection in Large Scintillation Detectors

Improvement of energy resolution via correction on non-uniform light collection in large scintillation detectors

Abstract

Because of their low cost, high efficiency, moderate energy resolution and robustness, large scintillation detectors are quite popular in field applications. In addition to the relatively large band gap between the valence band and the conduction band (thus large amount of energy required to produce each signal carrier), non-uniform light collection efficiency throughout the crystal is also a reason of energy degradation in large scintillation detectors. At Canberra Industries, knowing the manufacturing details, we can accurately model the construction and materials of our scintillation detectors (the MD series stabilized NaI detectors). This has been done in the ISOCS (In Situ Object Counting System) characterization process for each detector type. In this work, these models were input into DETECT2000 for calculation of light collection efficiency distribution throughout the whole crystal.  Monte Carlo simulation was then performed for a specific counting geometry, at interested energies. Energy deposition together with interaction positions was recorded in list mode. Combining the knowledge of light collection efficiency at these interaction positions, we can form response functions of the scintillation detector at interested energies, for a specific counting geometry. The response function can then be used to deconvolve the measured spectrum to improve energy resolution.

Summary

        I.            Models for Scintillation Detectors

Being the manufacturer, we have access to all the detailed drawings of our scintillation detectors. During the ISOCS characterization, accurate Monte Carlo models are created for each detector type. The models are further verified by comparison between predicted and measured results. These models reflect the best knowledge of dimensions and construction materials of our scintillation detectors. An example of our newly developed stabilized NaI probe is shown in Figure 1.

      II.            Calculation of Distribution Map of Light Collection Efficiency

DETECT2000 is a Monte Carlo model of the behavior of optical systems with a special emphasis on scintillation detectors. Individual scintillation photons are generated in user specified regions inside the scintillator. The code then follows each photon in its passage through various components of the model and various interactions with surfaces. Absorption and re-emission by a wave-shifting component are allowed in the simulation. The fate of the photon is recoded. In this work, we imported our Monte Carlo model into DETECT2000 and calculated a distribution map of light collection efficiency throughout the scintillation crystal. It is assumed that emission of scintillation photons at a specific position is isotropic. Such a map is shown in Figure 2.

   III.            Construction of Detector Response Function

The detector peak response to a mono - energetic incident photon can be closely represented using a Gaussian shape peak. Detector energy resolution is described by the Full Width at Half Maximum (FWHM) parameter of the Gaussian function. In addition to other factors (statistics fluctuation in creation of signal carriers, electronics noise and so on), non-uniform distribution of light collection efficiency makes contribution to the degradation of energy resolution as well. Monte Carlo simulation was setup for a specific counting geometry. Incident photons with interested energies were traced through its lifetime inside the scintillation crystal. Energy deposited at each interaction position was recorded in list mode. Knowing the average energy required to produce a signal carrier (electron-hole pair in valence band and conduction band) inside the scintillator, we can calculate the amount of scintillation photons generated at each interaction position.  Based on the distribution map of light collection efficiency generated in Section II, the ?collected energy' can be calculated for each incident photon. Histogram of this ?collected energy' for a mono-energetic photon source can be used as the detector response function, which includes the contribution of non-uniform distribution of light collection. An example of this response function is shown in Figure 3.

    IV.            Deconvolution of Measured Spectra

Maximum Likelihood ? Expectation Maximization Method was used to deconvolve the measured spectrum. The formula used is shown below.

f_j is the incident energy spectrum that we wanted to deconvolve from the measured spectrum g_i . t_ij is the probability that an incident photon with energy j being recorded with energy i in the measured spectrum. This probability is directly related to the response function calculated in Section III. Starting from a uniform distribution f⁰, the result after 10 iterations is shown in Figure 4 together with the originally measured spectrum. Improvement in energy resolution is obvious. The reason of the slightly shift in the peak position is that the measured spectrum can't be perfectly modeled by a Gaussian function.

      V.            Conclusion

Accurate Monte Carlo models of Canberra scintillation detectors were created. Light collection efficiency distribution was calculated using DETECT2000 based on these models. Detector response function was calculated for a specific counting geometry at interested energies. More accurate representation of the incident gamma energy spectrum was deconvolved out of the measured spectrum using ML-EM method.

    VI.            Appendix

Figure 1 Monte Carlo model for Canberra IPROS-3 Stabilized NaI probe

Figure 2 DETECT2000 Calculated Light Collection Efficiency Distribution inside a 3? by 3? NaI Crystal

Figure 3 Simulated Detector Peak Response Function of IPROS-3 to 662 keV Incident Photns and its fitting function

Figure 4 Comparison between Measured Spectrum and Peak After Deconvolution