2.1 Crystal growth

The CLYC crystal was grown using the vertical Bridgman method at BGRI. In this method, stoichiometric amounts of CsCl, LiCl, YCl₃, and CeCl₃ (with a purity of 99.99%) were mixed in a quartz crucible. Natural LiCl with a ⁶Li isotope content of approximately 7% was used. The dopant (Ce³⁺) concentration was 0.5 mol% as a standard doping level. After the powders were mixed in the quartz crucible, the pressure was pumped to a vacuum of 10⁻⁵ Pa to avoid oxidization. The quartz crucible was sealed, and the mixed powders were fused in a stove at a temperature above 640 °C. As the quartz crucible was steadily lowered at a velocity of 2 mm/h, a CLYC crystal ingot was grown.

After cutting and polishing, the CLYC crystal was packed in an aluminum casing with a reflective inner coating and sealed with a quartz window at one end. The packed crystal has a diameter of 25.4 mm and a thickness of 15 mm, as shown in Fig.1. The density of the CLYC is 3.31 g/cm³. The radioluminescence spectrum was measured by a Zolix Omni-λ300 monochromator under Cs gamma-ray irradiation, as shown in Fig.2. The radioluminescence has two main components, one in the 290 nm range due to CVL, and one in the 360–440 nm range due to Ce³⁺ ions, in good agreement with the spectra of many commercially available bialkali photomultiplier tubes (PMTs). In the range of 360–440 nm, two peaks at 374 and 400 nm were found, which are due to re-emission after self-absorption.

Fig.1Full-screen View

Photograph of CLYC crystal 25.4 mm in diameter and 15 mm in thickness grown at BGRI. The crystal is packaged in an aluminum enclosure.

Fig.2Full-screen View

Radioluminescence spectrum of CLYC crystal. The two main emission components are due to CVL and Ce³⁺ions, respectively.

2.2 Data acquisition

The CLYC crystal was coupled with a 2-in.-diameter 9815B blue-green-sensitive bialkali PMT (Electron Tubes Inc.) with optical grease. The 9815B PMT’s quantum efficiency is ~30% at the peak emission wavelength of CLYC. To estimate the gamma-ray energy resolution of the CLYC crystal, a ¹³⁷Cs source was used. The signals from PMT anodes with different heights were analyzed by a multichannel analyzer. A shaping time of 4 µs was used on the spectroscopy amplifier.

An Am–Be source that emits approximately 1 × 10⁴ neutrons per second was used as a neutron and gamma-ray source to estimate the PSD capability of CLYC. A schematic diagram of the experimental setup is shown in Fig.3. A Tektronix DPO7104 oscilloscope was set to FastFrame mode (1 G samples per second and 2-μs recording length) to record the PMT anode signals from both neutrons and gamma rays with a 50-Ω terminator. The CLYC crystal coupled with the 9815B PMT was placed in a polyethylene tube. A polyethylene block 2.5 cm in thickness and a lead block 2.5 cm in thickness were placed between the CLYC crystal and the Am–Be source. Approximately 5000 signals were recorded and then analyzed off-line as described in Sect. 3.2.

Fig.3Full-screen View

(Color online) Experimental setup with Am–Be source. The CLYC crystal was coupled with a 9815B PMT, which was connected directly to a digital oscilloscope (Tektronix DPO7104).

3.1 Gamma-ray energy resolution

Figure 4 shows the energy spectrum measured with the CLYC crystal under irradiation from the ¹³⁷Cs gamma-ray (662 keV) source. The position of the 662 keV energy peak was used to scale the horizontal axis in energy units (keV). The CLYC energy resolution at the 662 keV full energy peak is ~4.5% (FWHM), which is better than those of scintillators commonly used for gamma detection, such as NaI(Tl) and CsI(Tl), which have energy resolutions of 6.5% and 6% at 662 keV, respectively.

Fig.4Full-screen View

¹³⁷Cs energy spectrum measured with CLYC crystal. An energy resolution of ~4.5% (FWHM) at 662 keV was obtained.

3.2 Neutron and gamma-ray discrimination

PSD was applied to separate gamma-ray waveforms from neutron waveforms using the traditional charge integration method. In this method, two integration windows of the signal pulses, the prompt integration window and delayed integration window, are selected and integrated. The prompt window includes the sharp rise and peak portion, and the delayed window includes the decaying portion of the signals. There are generally infinite combinations of possible integration window widths. It is essential to achieve the best possible PSD by optimizing the window widths. An automated waveform analysis algorithm can optimize the window width through the following procedure:

(1) Align the waveforms to a common zero. The point where the amplitude was 20% of the peak value was chosen as the common zero.

(2) Perform charge integration according to the two integration windows. The width of the prompt integration window ranged from 5 to 300 ns in 5 ns steps. The delayed integration window started immediately after the end point of the prompt window, and its width ranged from 10 to 1300 ns in 10 ns steps.

(3) Calculate the PSD ratio. For each unique combination of the prompt and delayed integration windows, the PSD ratio was calculated for each waveform using the charge integrals Q_prompt and Q_delayed. The PSD ratio was defined as Q_delayed/Q_prompt.

(4) Calculate the figure of merit (FoM). The FoM was utilized to evaluate the neutron and gamma-ray discrimination performance. It was defined as

where D is the distance between the centroids of the neutron and gamma peaks in the PSD ratio histogram, and n_FWHM and γ_FWHM are the corresponding FWHMs. To obtain the centroid positions and FWHMs, Gaussian fits were applied to the neutron and gamma-ray peaks in the histogram.

The result of the PSD optimization is shown in Fig.5. This figure shows that the FoM is more sensitive to the prompt window width than to the delayed window width. With increasing prompt window width, the FoM shows a short, rapid increase, followed by a long, slow decrease. Changing the delayed window width has a minimal effect on the FoM. For the CLYC crystal in this study, the best FoM we found was 2.6, at a prompt window width of 20 ns and delayed window width of 900 ns.

Fig.5Full-screen View

(Color online) FoM under different prompt and delayed integration window combinations. The prompt window width ranges from 5 to 300 ns in 5 ns steps. The delayed window width ranges from 10 to 1300 ns in 10 ns steps.

Using the optimal prompt and delayed integration windows, a two-dimensional PSD scatter plot was generated for the neutron and gamma-ray events from an Am–Be source (Fig.6). In Fig.6, gamma-ray events and neutron events exhibit two distinct distribution patterns. Gamma-ray events have a lower PSD ratio across a wide energy range, whereas neutron events are isolated and have a higher PSD ratio. The neutron events can be divided into two groups representing different reactions^[14]: fast neutrons and thermal neutrons, as indicated in Fig.6. The thermal neutrons are produced by the ⁶Li(n,α)t reaction, and the fast neutrons are from either the ³⁵Cl(n,p)³⁵S reaction or the ³⁵Cl(n,α)³²P reaction. Figure 7 (dashed line) shows the one-dimensional projection of the neutron and gamma-ray events to the Y axis in Fig.6. A Gaussian function was fitted to the neutron and gamma-ray peaks in Fig.7(solid line). The centroids and FWHMs obtained from the Gaussian fits were used to calculate the FoM.

Fig.6Full-screen View

Two-dimensional PSD scatter plot based on the data collected under irradiation from an Am–Be source. The X axis is the PSD ratio (Q_delayed/Q_prompt), and the Y axis is the total integral Q_total corresponding to the energy of the events. Gamma-ray events (bottom) and neutron events (top) are clearly separated. Fast neutrons (³⁵Cl) and thermal neutrons (⁶Li) are indicated.

Fig.7Full-screen View

One-dimensional projection of neutron and gamma-ray events to the Y axis in Fig.6. A Gaussian function was fitted to the neutron and gamma-ray peaks (solid line).

3.3 Pulse shape analyses

According to the optimal PSD results, neutron signals and gamma-ray signals were averaged separately to produce the standard pulses. The averaging process reduces the noise observed in the individual signals, making it easier to analyze them. The standard pulses, normalized by the maximum, are shown in Fig.8. The characteristics of the neutron pulse and gamma-ray pulse exhibit significant differences owing to their different scintillation mechanisms.

Fig.8Full-screen View

(Color online) Overlay of neutron and gamma-ray standard pulses. The red and green lines are the gamma-ray and neutron standard pulse exponential fitting curves, respectively. Inset shows magnified view of the pulses at 90–270 ns.

For a Ce³⁺-doped CLYC crystal, up to four mechanisms generally contribute to the scintillation: Ce³⁺ (direct electron–hole capture by Ce³⁺), V_k (binary V_k–electron diffusion), self-trapped exciton (STE) emission, and CVL. Detailed descriptions of those mechanisms can be found in Refs.[3], [10], and [16].

To verify the scintillation mechanisms under neutron and gamma-ray excitation described in the literature, we used the exponential decay curves to fit the standard pulses ( Fig.8). Equation (2) was used to fit the gamma-ray standard pulse. There are four exponential decay components and thus four decay time constants. The decay time constant of each exponential component represents the decay time of the corresponding scintillation mechanism.

where I_gamma is the amplitude of the gamma-ray standard pulse from the maximum point to the end; τ_g1, τ_g2, τ_g3, and τ_g4 are the scintillation decay times for the CVL, Ce³⁺, V_k, and STE mechanisms, respectively, and A_g1, A_g2, A_g3, and A_g4 are the amplitudes of each component.

For the neutron standard pulse, an exponential decay function with two exponential decay components was used, as shown in Eq. (3).

where I_neutron is the amplitude of the neutron standard pulse from the maximum point to the end; τ_n3 and τ_n4 are the scintillation decay times for the V_k and STE mechanisms, respectively, and A_n3 andA_n4 are the amplitudes of each component.

Table 1 lists the fitting results for gamma- and neutron-induced CLYC emission. For the gamma waveform, the decay times are 2, 50, 420, and 3400 ns for the CVL, Ce³⁺, V_k, and STE mechanisms, respectively. For the neutron waveforms, only V_k and STE are observed, with decays times of 390 and 1500 ns, respectively.

The decay times measured in our work, Ref.[10], and Ref.[16] are listed in Table 2. The decay times in our work are essentially in agreement with the results from Ref.[10], except for the STE/neutron result, whereas the results from Ref.[16] were quite different from ours. Because the Ce³⁺ doping concentration was the same (0.5 mol%), these differences may be caused by different signal sample rates and signal recording lengths. Ref.[16] applied a sample rate of 250 Ms/s, which might be inadequate for obtaining accurate decay times. Thus, Ref.[16] showed longer decay times than our work (sample rate: 1 Gs/s) and Ref.[10] (sample rate: 2 Gs/s). The signal length recorded by the oscilloscope is 2 μs in this work, and in Ref.[10] it is 10 μs. The signal length of 2 μs is suitable for the gamma pulse, because its amplitude dropped almost to the baseline in 2 μs (Fig.8, solid line). However, the neutron pulse, which has a longer decay time, requires a longer recording length than the gamma pulse to obtain the neutron decay times, especially for the STE mechanism.