Design and Initial Tests of a Fast Neutron Radiography Detector Prototype with Silicon Photomultiplier Readouts

Abstract

Among non-destructive testing (NDT) techniques, fast neutron radiography with a higher penetration capability has achieved rapid advancements. However, the application of the radiography detector in many fast neutron imaging systems is limited by unfavorable detection efficiency and imaging spatial resolution. In this paper, a fast neutron radiography detector was designed, which was composed of a pixelated EJ200 scintillator array, a 16 × 16 silicon photomultiplier (SiPM) array, and capacitive multiplexing network readout electronics. The main parameters of the detector were optimized using Monte Carlo simulations. In addition, the prototype of the detector was fabricated and tested under a 14 MeV D-T neutron source. The preliminary test results demonstrated that the spatial resolution of the prototype reached 1.2 mm. Moreover, the conflict between spatial resolution and detection efficiency could be mitigated by using a pixelated scintillator structure. Overall, SiPMs enabled the extensive application of the imaging system because of their excellent photon detection performance, relatively low price, and joint possibility for large areas.

1. Introduction

Fast neutron imaging can effectively compensate for the shortcomings of X-ray and thermal neutron imaging; hence, it can be widely applied to the non-destructive testing (NDT) field. This technology has a higher mean free path compared with X-rays or gamma rays owing to the application of thick, dense, and high-Z materials [1]. The macroscopic interaction between fast neutrons and many nuclides is characterized by an extremely low cross-section, which endows these fast neutrons with strong penetration [2]. As a result, the imaging technique based on fast neutrons has emerged as a promising non-destructive probing tool for the inspection and imaging of various practical problems concerning large and thick samples composed of low-Z/high-Z materials or a mixture of both.

As a category of electrically neutral particles, neutrons are not blocked by the electric field outside nuclei. They penetrate through the electron layer and interact with the nuclei of the atoms of matter. However, high-energy neutrons have a small reaction cross-section for most materials. Neutrons can be detected using the (n, p) reaction, which has a large elastic scattering cross-section. The detection of fast neutrons is mainly achieved indirectly. More specifically, incident fast neutrons interact with hydrogen nuclei to produce recoil protons, and then the reaction caused by these protons is analyzed. Organic scintillators can be used for the detection of fast neutrons. They contain a high amount of hydrogen, and directly detectable scintillation photons are generated after the absorption of energy from recoil protons.

Conventional detector systems used in fast neutron imaging can be classified into the following categories: (1) photosensitive element charge-coupled device (CCD) cameras or complementary metal oxide semiconductor (COMS) cameras with neutron–photon converters, mainly hydrogen-rich plastic converters [3,4,5]; (2) microchannel plates (MCPs) with amorphous silicon arrays [6,7]; (3) position-sensitive gaseous electron multipliers (GEMs) with hydrogen-rich converters [8,9]; and (4) hydrogen-rich plastic converters with imaging plates [10,11]. These systems exhibit a contradiction between detection efficiency and spatial resolution. Ma et al. [12] used a 2 mm thick polyethylene plate coupled to an MCP to obtain images with a resolution of 500 μm under a 14 MeV neutron beam. However, the detection efficiency of this device was only 0.14%. To alleviate this contradiction, a pixelated structure scintillator array was employed in this study. It has been validated that the pixelated structure scintillator array is effective for obtaining improved spatial resolution in X-ray imaging and positron emission tomography (PET) systems [13,14]. It also works for the neutron detector system. Fast neutron radiography is composed of pixelated scintillators, optical guides, and photosensor arrays. The cut between each scintillator pixel is filled with reflective materials. The position of the incident neutron can be rebuilt by the signal distribution from the photosensor array. The probability of the interaction between fast neutrons and hydrogen nuclei increases with an increase in the thickness of the scintillator. Therefore, it can be speculated the fast neutron detection efficiency may be improved by increasing the thickness of the scintillator. However, if the scintillator has a monolithic structure, without the pixel cut filled with reflective materials, the luminescence will spread out to cause more signals from the photosensor array unit. This will significantly increase the difficulty in position reconstruction and reduce the position resolution of the detector. The pixel scintillator restricts the spread of the luminescence. The optical guide slightly diffuses the luminescence at the end of the pixel scintillator to three to five photosensor units. The position of a scintillation event can be estimated by Anger logic [15]. The pixelated structure can improve the detection efficiency under the premise of ensuring the position resolution.

This study was conducted to develop a large-area scintillator detection system with higher detection efficiency and favorable spatial resolution. In this study, the pixelated structure was composed of pixelated scintillator arrays and pixelated photosensitive elements—SiPM arrays. More specifically, the pixelated scintillator arrays comprised hydrogen-rich plastic scintillators, which were designed and optimized by Monte Carlo simulation. They were finally cut, polished, and optically isolated to form arrays with a certain size and shape. A SiPM array with a pixel size of 3.14 mm × 3.14 mm was designed to replace conventional photomultiplier tubes (PMTs) to record the output scintillation photons from the scintillator. A total of 256 SiPM units were encapsulated into a 5 cm × 5 cm array. Self-designed multiplexing readout electronics and a Data AQuisition (DAQ) system were developed.

In this study, Monte Carlo simulation results were obtained. The size of the scintillators was optimized to achieve the optimal balance between detection efficiency, spatial resolution, and processing difficulties. Moreover, a detector prototype was fabricated and tested under a 14 MeV fast neutron source.

2. Detector Working Principle

There are two kinds of interactions between neutrons and nuclei, namely, scattering and absorption. In the scattering interaction, a neutron collides with a nucleus and transfers some portion or all its energy to the recoiling nucleus. In the absorption interaction, a nucleus absorbs neutrons and enters an excited state. Because of the high hydrogen content and large cross-section of (n, p) elastic scattering of the organic scintillator, the EJ200 plastic scintillator developed by Eljen Tech [16] was selected in this study. It had a scintillation yield of 10,000 photons per 1 MeV of deposited electron energy, a light attenuation length of 380 cm, rise and decay times of 0.9 and 2.1 ns, and a maximum emission wavelength of 425 nm. These characteristics made it suitable for the fast neutron detection.

The SiPM is a solid-state photon-sensitive device based on the reverse-biased P-N diode. It features high quantum efficiency, low operating voltage, immunity to magnetic field interference, low cost, and easy integration into arrays. It is suitable for all applications where high measurement precision and quantification of low light/radiation levels are required. Owing to their better single-photon resolution capability, SiPMs have widely substituted for PMTs in scintillation photon readouts [17,18].

In this study, the fast neutron radiography detector consisted of an EJ200 array and a SiPM array. The schematic view of the detector is shown in Figure 1. Incident fast neutrons collided with hydrogen nuclei in the scintillators to produce recoil protons. Then, these protons deposited energy along their trajectory with the production of isotropic scintillation photons. Most of these photons were limited in the cuboid scintillator and transferred toward the SiPM array through the optical guide. Subsequently, these photons were distributed by the optical guide to SiPMs. Next, they were converted to electrical signals and recorded by the electronics.

The EJ200 organic scintillator (Eljen Technology, Sweetwater, TX, USA) had a refractive index of 1.58. The K9 optical glass with a refractive index of 1.5 was selected as the optical guide placed between the EJ200 array and the SiPM array [19]. It had a low scattering cross-section for neutrons. The optical guide could reduce the total internal reflection of scintillating photons at the end face of EJ200. Meanwhile, it could enhance the transverse diffusion of outgoing scintillating photons to more SiPM pixels, which would ultimately improve the detection efficiency and position resolution of the detector for fast neutrons.

3. Detector Simulation and Optimization

Theoretically, the spatial resolution of the pixel-readout-type detector increases with a decrease in the pixel size of the scintillator and the SiPM array [20]. To determine the influence of the main parameters on the imaging quality and to obtain better spatial resolution, the structure and specific parameters of this detector were optimized by Monte Carlo simulation using Geant4 10.7 software [21]. The simulation results were regarded as a guideline for fabricating the detector prototype.

3.1. Spatial Resolution

Spatial resolution is an important factor in the evaluation of an imaging system. In this study, the best spatial resolution of the detector system was simulated. The incident neutrons with 4 MeV energy in the simulation were perpendicular incident to the SiPM array plane strictly. In addition, the neutron position reconstruction was realized by the center of gravity method [22]. For the pixeled detector, the center of gravity method obtained the best position resolution. The edge method [23] was used to evaluate the position resolution of this detector. In the implementation of this method, an object with a certain blocking efficiency for fast neutrons, such as a boron-containing polyethylene plate, was placed in front of the detector. A neutron source with a uniform incidence to the detector was placed in front of the detector. Then, the detector obtained a neutron image with a sharp edge, as shown in Figure 2a. The edge-spread function (ESF) was obtained by scanning the edge image, and it was represented as the projection of neutron count distribution in the X direction, as shown in Figure 2b. The corresponding line distribution function (LSF) was obtained by the derivation of the obtained ESF, as shown in Figure 2c. A position resolution of much less than 1 mm was obtained.

For an imaging system with a pixel structure, the basic pixel arrangement and the size would influence the quality of the final image. The detector structure is illustrated in Figure 1. The pixel size of the SiPM array was 3.14 mm × 3.14 mm, which is a common size for the SiPM commercially available in the imaging field. The thickness of the optical guide was set to 3 mm. The pixel size of the cuboid scintillator array was changed in the simulation. Based on that, the relationship between the spatial resolution and the pixel size of the scintillator array was studied. The simulation results are presented in Figure 3.

The results revealed that the spatial resolution decreased with an increase in the pixel size of the scintillator array. For the cuboid scintillator array with a pixel size of 0.5 × 0.5 × 25 mm³, the best spatial resolution was 0.26 mm. For that with a pixel size of 1.5 × 1.5 × 25 mm³, the best spatial resolution was 0.68 mm. Of note, the pixel size of 1 mm corresponded to a limiting spatial resolution of 0.46 mm. This was the minimum size of the process that could be achieved by the current research team.

3.2. Neutron Conversion Efficiency

The EJ200 organic scintillator is widely used in fast neutron detection. The technical specification of the EJ200 is listed in Table. The fast-timing properties make it suitable for large-area and fast neutron imaging systems. In this study, the conversion efficiency ratio was defined as the ratio of the number of neutrons that collided with a hydrogen nucleus to produce a recoil proton to the total number of incident neutrons. The fast neutron conversion efficiency within the scintillator was proportional to the thickness of the scintillator. The EJ200 with a cross-section area of 1 mm × 1 mm was irradiated by a 14 MeV neutron source in the simulation. The variation in the neutron detection efficiency with the thickness of the scintillator was recorded.

The energy of recoil protons produced by the interaction between neutrons and hydrogen nuclei was deposited and caused scintillating photons. When the thickness of the scintillator increased, the amount of recoil protons increased. This was the main reason for an increase in the detection efficiency. However, the increase in the detection efficiency may be limited because the number of scintillating photons would decrease when the thickness of the EJ200 exceeded a certain value. This can be attributed to the light loss of the thick scintillator. Figure 4 shows the relationship between neutron detection efficiency and the thickness of the EJ200. Detection is defined as the fraction of the incident neutrons producing more than 100 photons on the SiPM. When the thickness of the EJ200 exceeded 25 mm, the fast neutron detection efficiency started to drop. This may be because the scintillating photons at the front of the EJ200 were attenuated or self-absorbed during transmission. The signals of the SiPM could not trigger the readout electronics effectively.

The scintillator attenuation length given in the manual (

Table 1. Technical specification of the EJ200 [24].

EJ200 Physical and Scintillation Constants
Light output in % relative to anthracene	64
Density, g/cm3	1.023
Polymer base	Polyvinyl toluene
Wavelength of max. emission, nm	425
Scintillation efficiency, photons per 1 MeV e−	10,000
Rise time, ns	0.9
Decay time, ns	2.1
Pulse width, FWHM, ns	2.5
Organic flours, %	3
Number of H atoms per cm3	5.17 × 1022
Number of C atoms per cm3	4.69 × 1022
Number of electrons per cm3	3.33 × 1023

) was the intrinsic light transmission attenuation length. It was determined by only the self-absorption characteristics of the scintillator. The geometric factors of the scintillator were excluded. For a scintillator constrained to a finite size, because of the effects of its geometry, such as reflection, refraction, and total reflection, as well as with the action of external conditions, the attenuation length for light transmission became shorter. The light attenuation length of the EJ200 with a section of 1 mm × 1 mm was also simulated, and it was related to its geometry and light shielding craft. Figure 5 shows the change in the photon number from one side of the cuboid scintillator to the other side with an increase in the scintillator thickness. This result indicated that the number of photons decreased when the thickness of the EJ200 exceeded 20 mm. This was consistent with the simulation result of the detection efficiency.

So, the thickness of the EJ200 should be large enough to maintain high detection efficiency and small enough to maintain the scintillating photon number to the SiPM readout. Eventually, a thickness of 25 mm was chosen for the EJ200.

4. Detector Prototype and Readout Electronics

4.1. SiPM Array

In this study, SiPMs of the J30035 series from SensL (Cork, Ireland) were selected as the photon sensors [25]. The SiPM is a pixel-type photon detector with many advantages, such as compact dimensions, easy integration into arrays, even gain response, and low cost. To achieve a large area of imaging targets, four arrays with a size of 8 × 8 were arranged to form a complete array with an effective area of about 50 mm × 50 mm, as shown in Figure 6. The size of a single SiPM unit was 3.14 mm × 3.14 mm, and the gap between adjacent SiPMs was 0.36 mm.

Figure 6a shows a complete array containing 16 × 16 arrays. Figure 6b presents a microcell forming a basic SiPM unit. This microcell consisted of a photodiode connected in series to a burst resistor operating in the avalanche mode.

4.2. Electronics

The 8 × 8 SiPM array described above provided 64 fast output signals. After the encapsulation of the complete array, there were a total of 256 SiPM output signals. A large number of high-speed electronic channels were needed for the individual readout of SiPM signals.

For the readout electronics matched with the SiPM array, a capacitive multiplexing network was adopted to reduce the readout channel number from 256 (individual readout scheme) to 4 [26,27].

The final output four-channel signals were amplified, shaped, and transmitted to the data acquisition system. Figure 7 shows a schematic diagram of the self-design capacitive multiplexing network. In this network, each anode of the SiPMs was connected to one or more weighted capacitors depending on the position. The anode signal from the SiPMs was split into one or more signals depending on the number of weighted capacitors that were connected to each anode.

The capacitive multiplexed networks encoded the 256 SiPM output signals into four position signals (P, Q, R, and S). The energy E of the incident fast neutron and its two-dimensional position (X, Y) was determined by the following equations

$$\begin{gathered} \mathrm{E}=\mathrm{P}+\mathrm{Q}+\mathrm{R}+\mathrm{S} \\ \mathrm{X}=\frac{\left(\mathrm{R}+\mathrm{S}\right)-\left(\mathrm{P}+\mathrm{Q}\right)}{\mathrm{P}+\mathrm{Q}+\mathrm{R}+\mathrm{S}} \\ \mathrm{Y}=\frac{\left(\mathrm{P}+\mathrm{S}\right)-(\mathrm{Q}+\mathrm{R})}{\mathrm{P}+\mathrm{Q}+\mathrm{R}+\mathrm{S}} \end{gathered}$$

(1)

4.3. Scintillator Array and Detector Prototype

As discussed above, an EJ200 scintillator array with a pixel size of 1 mm × 1 mm × 20 mm was cut and optically isolated to form the fast neutron-sensitive scintillator. The scintillator array is shown in Figure 8. The effective area of the EJ200 array was about 5 × 5 cm². A mixture of titanium oxide and glue with a thickness of less than 0.2 mm was applied between adjacent pixels as photo isolation. The 16 × 16 SiPM array and the capacitive multiplexed networks were packaged on the same PCB board, as shown in Figure 9a part 1. Figure 9a part 2 shows the preamplifier of the four position signals (P, Q, R, and S). Figure 9a part 3 shows the data acquisition module. A gigabit network port was used in the data output module to connect the detector to a personal computer. The acquired data were accessed online via a computer or transferred to a personal computer for offline processing. Figure 9b shows the SiPM array after the assembly of all components.

A K9 glass with a thickness of 3 mm as the optical guide was placed between the scintillator array and the SiPM array to enhance the diffusion of scintillation light and transmit it to the SiPM array. Silicon oil was used as the coupling medium between the EJ200 array, the optical guide, and the SiPM array.

The K9 glass exhibited high transparency and high radiation resistance. The 3 mm thickness ensured a suitable balance while spreading scintillating light from EJ200 over more SiPM pixels and not excessively decreasing the light intensity at each signaled SiPM pixel. Moreover, K9 was a favorable material for the fabrication of optical guides owing to its low cross-section of neutron scattering.

5. Prototype Performance Test

The fast neutron beam test was carried out under a 14 MeV D-T neutron generator of the China Academy of Engineering Physics. It was a moveable imaging facility with a compact accelerator D-T neutron source. It consisted of a compact accelerator, two different collimators (one for thermal neutrons and the other for fast neutrons), a sample-bearing platform, and a detection system, which was mounted on the platform for mobility. Figure 10 shows the sketch map of the facility [29]. The maximum neutron yield was 1.7 × 10¹¹ neutrons/s under a deuterium beam current of 1.5 mA.

In the test, a stainless-steel sample was selected as the imaging object, as shown in Figure 11. The sample consisted of a substrate with a thickness of 1 cm, which incorporated four stepped pieces with a thickness of 3, 5, 5, and 15 mm. The distance between the outer four-step piece and the edge of the substrate was 4 mm. These four-step pieces formed three slits with a width of 1, 3, and 10 mm. The detector system was shielded by two shield materials, namely, the lead shield with a thickness of about 5 mm and boron with a thickness of 3 mm containing polyethylene in the outer layer. This structure was designed to shield gamma and stray neutron backgrounds. The detector prototype system was powered by a 5 V/2 A power supply. The SiPM array operated under 29.5 V. The acquired data were transferred to a personal computer via the RJ45 interface for data processing.

The detector system was placed on a platform with the stainless-steel sample placed in front of the detector. After the irradiation for 5 min, a two-dimensional image was obtained, as shown in Figure 12. Figure 12a shows a raw 2-D image of stainless-steel samples with the sample shape overlay on it (open beam). The four gaps in the sample were basically visible. It can be observed that there were four peaks in Figure 12b related to different slit spacings of the stainless-steel sample. The second, third, and fourth peaks were related to the slits in the sample with a gap of 1, 3, and 10 mm, respectively. In addition, the three-peak valleys were related to three pieces of stainless-steel samples with a thickness of 3, 5, and 5 mm, respectively. The edge of the final pieces of stainless-steel samples with a thickness of 15 mm was not observed because the position of the sample was placed too far to the right. The neutron intensity contrast in the picture corresponding to different thicknesses of four steps of the steel sample cannot be observed as expected. This may be caused by the non-collimation of the neutron source, the inconsistency in the gain of each SiPM pixel, and the inhomogeneity intensity distribution of the neutron source. The edges of the 2-D sample image also showed some imaging compression. These image distortions should be eliminated through position calibration and correction. The sample image quality could also be improved by the normalization based on the no-sample neutron exposure data. However, the no-sample data cannot be used because of the neutron count saturation of the detector system.

The spatial resolution of the detector was evaluated using the edge diffusion function of the sample. The minimum spatial resolution was 1.2 mm.

6. Summary and Discussion

In this study, an array-type scintillator detector was designed for fast neutron imaging. Through Monte Carlo simulations, the main parameters of the detector were optimized. In addition, the prototype of this detector was fabricated and tested, which was composed of scintillator array basic units with a dimension of 1 mm × 1 mm × 25 mm, 16 × 16 SiPM array, and self-design capacitive multiplexing network electronics. Using the D-T fast neutron source, stainless-steel sample imaging was obtained with a minimum spatial resolution of 1.2 mm.

The radiation damage of these SiPMs was also considered in the test [30]. After the radiation of the 14 MeV D-T neutron beam for 24 h, the gain in these SiPMs was measured and compared with that without radiation. The results demonstrated that the gain decreased and the dark current increased apparently. However, it could still work normally in the avalanche mode, which indicated that it had certain radiation resistance. Nevertheless, the initial research data were not comprehensive enough. Subsequent separate tests will be conducted, and corresponding articles will be published to discuss and explain relevant irradiation damage problems.

To avoid the long-time irradiation damage of these SiPMs, a new detector with a long bending scintillating fiber array can be designed. In addition, the SiPM array can be placed in the position without direct exposure to the neutron.

SiPMs have an increasingly wide range of applications owing to their high photon gain, low operating voltage, and relatively low cost. SiPMs can be assembled into photon detection modules with arbitrary sizes. After the spatial resolution is improved to a sufficient level, these modules can be used for non-destructive detection of larger volume objects.