Speaker
Description
Motivation: We present a gamma detector concept for high-energy prompt-gammas based exclusively on Cherenkov light for use in proton range verification (PRV) in proton and heavy ion therapy. The radiation backgrounds in these environments are very harsh, with a high abundance of 511 keV photons due to positron activation, fast neutrons, and pulsed sequences that lead to high emission rates of the prompt-gammas themselves. We propose to use exclusively the Cherenkov light emitted in a pure Cherenkov emitter, PbF$_2$ in this case. The motivation is three-fold: First, the detected Cherenkov intensity by 511 keV gammas in PbF$_2$ is less than 5 photons, therefore the detector is practically insensitive to this source of background by just placing the threshold above this level. Second, Cherenkov emission happens within picoseconds, therefore the main contribution to dead time is on the photodetector and electronics, thus allowing for much higher count rates than any scintillation material. And third, the production costs of PbF$_2$ are low compared to other gamma-detector materials, thus allowing to produce a medium-size imager cost-effectively.
Methods: The detector consisted of a monolithic 25$\times$25$\times$10~mm$^3$ PbF$_2$ crystal coupled to a 8x8 SiPM array readout in a column/row fashion. We evaluated its performance in the laboratory with a $^{228}$Th gamma source, which has an energy line at 2.6 MeV, and a slit of 1 mm width. The detector was mounted on a 2D translation stage, and we acquired 35 positions in 1 mm increments along one dimension. The range was intentionally set well outside the size of the detector (25 mm). Figure 1 shows the acquisition setup. Each event consisted of 16 waveforms: 8 rows and 8 columns recorded with a record length of 1 $\mu s$ and a sampling rate of 1 Gs/s. The signals from the 8 columns were multiplexed, combined into a single one and used to trigger the events. Two algorithms were used to determine the impact point for each event: the center of gravity (CoG) and rise to the power (RTP). CoG is a linear combination among of the rows (or channels) with a weighted probability that is linearly proportional to the fraction of signal detected in that element. The RTP follows a similar principle but the probability for each element is obtained by elevating the fraction of the signal detected to the k exponent. Two k values were evaluated: k=2 and k=3.
Results: A histogram was obtained for each position and algorithm for one of the detector axes. Figure 2 shows three representative distributions for 8 mm, 18 mm and 28 mm using the CoG algorithm. For each of such histograms, a gaussian git was applied and the center of the gaussian and the full width at half maximum (FWHM) were recorded as figures of merit. Figure 3 shows the position scan for the CoG, RTP(k=2) and RTP (k=3). The active area of the detector was identified between positions 5 and 29, which matches the width of the crystal (25 mm). A linear fit was applied to this range and the FWHM/slope of the fit was taken as the figure of merit for the resolution. The average of the resolutions across this range was 7.4 mm, 5.6 mm, and 4.8 mm for the CoG, RTP(k=2) and RTP (k=3) algorithms, respectively.
Discussion and Future Work: In the first part of the evaluation of this detector concept, we conclude the Cherenkov light generated by prompt-gammas of up to 2.6 MeV can be used to extract position information of the impact point of the prompt-gamma within the crystal. Specifically, we tested the spatial resolution along one dimension. We found significant sources of background, such as the plateaus outside the gaussian distributions in Figure 2. We are currently studying their nature and strategies to remove them.
The resolution values obtained are 2 to 3 times greater than one would desire for the PRV application. Dealing with such background might help reduce that value. Very preliminary data acquired at the proton cyclotron at Davis hinted resolutions below 3 mm. An improvement of resolution in the beam line compared to the benchtop setup could be explained by the greater energy of prompt gammas (up to 6.1 MeV vs up to 2.6 MeV), which leads to a greater Cherenkov light generation and thus significantly higher signal-to-noise ratio when detecting such prompt-gammas.
We continue to study the behavior of this detector concept using the same $^{228}$Th gamma source under different setup configurations (SiPM bias, trigger strategy, detector segmentation) and event reconstruction strategies. Moreover, by the time of the meeting, we expect to present data equivalent to Figures 2 and 3 acquired at the cyclotron at Davis, with a proton anergy of 67.5 MeV and beam currents of several nA.