Response of a Li-glass/multi-anode photomultiplier detector to collimated thermal-neutron beams

The response of a position-sensitive Li-glass scintillator detector being developed for thermal-neutron detection with 6 mm position resolution has been investigated using collimated beams of thermal neutrons. The detector was moved perpendicularly through the neutron beams in 0.5 to 1.0 mm horizontal and vertical steps. Scintillation was detected in an 8 X 8 pixel multi-anode photomultiplier tube on an event-by-event basis. In general, several pixels registered large signals at each neutron-beam location. The number of pixels registering signal above a set threshold was investigated, with the maximization of the single-hit efficiency over the largest possible area of the detector as the primary goal. At a threshold of ~50% of the mean of the full-deposition peak, ~80% of the events were registered in a single pixel, resulting in an effective position resolution of ~5 mm in X and Y. Lower thresholds generally resulted in events demonstrating higher pixel multiplicities, but these events could also be localized with ~5 mm position resolution.


Introduction
The scienti c program to be performed at the European Spallation Source (ESS) [1][2][3][4] requires position-sensitive, 3 He-free [5][6][7][8], thermal-neutron detectors with high countingrate capability. Small-angle neutron-scattering experiments requiring two-dimensional position sensitivity [9][10][11][12][13][14][15][16][17][18] will be performed with Solid-state Neutron Detectors (SoNDe, patent EP000003224652A1) [19][20][21]. The SoNDe concept employs an array of detector modules to instrument large areas with a reconstruction accuracy of ∼6 mm on the position of the detected neutron. A SoNDe "module" consists of a thin, thermal-neutron sensitive, Li-glass scintillator sheet (GS20) attached to a 64-pixel multi-anode photomultiplier tube (MAPMT). Signals resulting from the scintillation light are processed using custom electronics [21]. In the envisioned operation mode at ESS, known as "Time of ight" mode (TOF), these electronics timestamp all pixels having signals above threshold if any single pixel amplitude exceeds threshold. Events involving the ring of a single pixel (multiplicity M = 1 events) are thus straightforward to interpret. At the boundaries between pixels and in the corners, scintillation light su cient to trigger several pixels (multiplicity M > 1 events) is often registered. The behavior of clusters of bordering pixels in these regions is thus of interest. LEDs and laser light have been used extensively to study the detailed responses of several di erent MAPMTs [22][23][24][25][26][27][28][29][30][31]. Previously, scans of a ∼1 mm collimated beam of ∼4 MeV α-particles from an 241 Am source [32] and ∼100 µm diameter beams of 2.5 MeV protons and deuterons [33] have been used to study the position-dependent response of a SoNDe detector prototype in regions well-removed from the edges of the detector acceptance. Thermal neutrons have also been used to perform rst tests both on similar detectors [34] and on SoNDe modules [21]. The thermal-neutron interaction with the 6 Li of the Li-glass has a Q-value of 4.78 MeV and results in an α-particle (2.05 MeV) and a triton (2.73 MeV). In this work, a SoNDe module has been systematically scanned 2 through beams of thermal 2 neutrons. The goals were to: 1. complement the existing α-particle, proton, and deuteron studies of the positionsensitive behavior of the detector, for events triggering multiple pixels to establish the response at the pixel boundaries and the corners where four pixels meet 2. provide thermal-neutron data with ∼1 mm precision on the position sensitivity of the detector for events triggering only one pixel, as a single-pixel mode-of-operation is anticipated as the ESS default 3. map the response of the detector as a function of both threshold and beam position for events which only trigger one pixel 4. determine the detector threshold that maximizes the number of single-pixel events 5. study regions within the detector where the position-reconstruction accuracy for an event better than ∼6 mm may be obtained for M > 1 events 6. provide a thermal-neutron dataset at the edge of the detector, for adjacent pixels with the highest gain contrast, to aid our understanding of the SoNDe module at its periphery.

Neutron beams
The measurements were performed at the R2D2 beamline at the JEEP II reactor [35,36] at the Institute for Energy Technology (IFE) [37]. The setup is illustrated in Fig. 1.
The nominal central-beam ux was 10 5 /s/cm 2 with a ∼0.6 • divergence. Thermal-neutron beams (2.0 Å, ∼18 meV and 2.4 Å, ∼13 meV) were de ned using a composite Ge wafer monochromator [36]. The resulting thermal-neutron beams drifted ∼20 cm to the rst of a pair of JJ X-Ray IB-C80-AIR slits [38] which employed borated-aluminum blades to control the beam ux. The slit spacing was ∼100 cm, with the downstream slit located ∼20 cm upstream of the detector. A 5 mm thick HeBoSint mask [39] with pinholes was used to further collimate the beam to either ∼1 mm or ∼3 mm in diameter. A stack of three 2 mm thick Mirrobor sheets [40] with a 100 mm 2 square aperture acted as a nal barrier to 2 The 2.0 and 2.4 Å neutron beams employed in this work had energies lying at the upper end of the coldneutron energy window and are thus "not quite but nearly" thermal.
The 50 mm × 50 mm × 1 mm sheet with polished faces and rough-cut 1 mm edges was dry-tted to the MAPMT window and held in place with tape along the edges. The dry-t approach was chosen to avoid the degradation of any optical-coupling medium. A piece of standard white copy paper (136 g/cm 2 ) placed over the upstream face of the GS20 diffusely re ected scintillation light back towards the MAPMT, increasing the amount of scintillation light reaching the MAPMT by ∼40%. The (assumed uniform) density of 6 Li in GS20 is 1.58 × 10 22 atoms/cm 3 . The cross section for the n (25 meV) + 6 Li → 3 H (2.73 MeV) + α (2.05 MeV) capture reaction is 940 b, which yields a detection e ciency of ∼75% for the 1 mm sheet. The average ranges of the 3 H and α-particle in the GS20 are 34.7 µm and 5.3 µm, respectively [46]. The 4.78 MeV capture reaction results in a ∼6600 scintillation photon full-deposition peak [21] (roughly equivalent to 20-30% of anthracene) peaked at ∼390 nm [47]. Scintillation-light transport from the GS20 (refractive index 1.55 at 395 nm) across a ∼100 µm air gap (refractive index 1) due to the concavity of the MAPMT borosilicate-glass window and then into the MAPMT window (refractive index 1.53) is generally ine cient.

Multi-anode photomultiplier tube
An 8 × 8 pixel (∼6 mm × ∼6 mm per pixel) 10 stage Hamamatsu type H12700A MAPMT with a borosilicate glass window was employed. The sensitive bialkali photocathode area is 48.5 mm × 48.5 mm, while the outer dimensions are 52 mm × 52 mm, resulting in 87% of the surface being active. The peak quantum e ciency of the photocathode is ∼33% at ∼380 nm, nicely overlapping the ∼390 nm (peak) scintillation-light distribution produced by GS20. For the MAPMT used here, at a cathode-to-anode voltage of −1000 V, the Hamamatsu data sheet speci ed a gain of 2.09 × 10 6 , a dark current of 2.67 nA, a (worst-case) factor 1.7 gain di erence between pixels, and ∼2 % electronic crosstalk between pixels. Calvi et al. [29] report that electronic crosstalk is actually dependent upon both the pixel and the position within this pixel and that it uctuates di erently near horizontal and vertical edges (see below). The operating voltage was -900 V. Corrections for pixel-to-pixel gain variations were performed o ine using a gain map measured with the 3 mm, 2.4 Å neutron beam used to irradiate each pixel center consecutively, as described in [33].

Readout electronics
A compact readout module designed for the H12700A MAPMT by IDEAS [49] was employed for data acquisition. The 113 g module is 50 mm × 50 mm × 55 mm (deep).
It consists of two boards: front-end and controller [21]. The front-end board uses four 16-channel IDE3465 ASICs to digitize the MAPMT signals with a precision of 14 bits. The controller board accomodates an FPGA and a MiniIO port for ethernet communication.
The electronics can operate in two modes: the TOF mode previously discussed and "Allchannel Spectroscopy" (ACS) that was used here. In ACS mode, when any pixel-amplitude threshold was exceeded, the digitized signal amplitudes from all 64 pixels were read out.
ACS rate limitation is ∼10 kHz, which corresponds to ∼4 MHz/m 2 ). A hardware threshold of 500 ADC channels was employed, which corresponds to ∼5% of the mean channel of the 4.78 MeV full-energy deposition peak measured by a single pixel for irradiation at its center. Higher thresholds were applied o ine.

GEANT4 simulation
A detailed computer model of a SoNDe module is nearing completion [50]. This C++ model employs the GEANT4 Monte Carlo toolkit [51] version 4.10.6 [52]. It includes the GS20 sheet together with the glass window and photocathode of the MAPMT. Opticalcoupling media may be placed between the GS20 and the MAPMT window. The model simulates the interactions of ionizing radiation in the GS20 to the level of the emission of scintillation light and includes the transport of the scintillation photons to the MAPMT cathode. It also includes a model for electronic crosstalk. Electronic crosstalk results from voltage divider biasing, stray capacitances leading to AC coupling between pixel anodes, and charge sharing across neighboring dynode chains, all known to a ect the performance of the H12700 MAPMT. It results in signal from the illuminated pixel leaking into neighboring pixels. De ned for each neighboring pixel as the ratio of the induced signal to the signal registered in the illuminated pixel, it has been reported to be up to ∼3% in vertically adjacent pixels and up to ∼7% in horizontally adjacent pixels [29]. The probability of signal leakage has been shown to be lowest at a pixel center and highest at pixel edges. Based 7 on these measurements, electronic crosstalk was modelled on an individual scintillationphoton basis with the crosstalk probability increasing linearly as the pixel edge was approached. For uniform pixel illumination, the crosstalk model was con gured so that once all scintillation photons were detected, adjacent pixels each registered 5% of the signal detected in the illuminated pixel. Figure 3 presents some results from the GEANT4 simulation of the scintillation light. In Fig. 3(a) In the rst 0.1 mm of the GS20 sheet, ∼18% of the incoming neutrons are absorbed by the 6 Li in the scintillator. This process continues exponentially so that ∼5% of the incident neutrons are absorbed in the last 0.1 mm of the 1 mm thick GS20. There is a 22% chance of a neutron passing through the GS20 without interacting. In Fig. 3

Measurement
Collimated thermal-neutron beams were used to irradiate the SoNDe module at wellde ned positions. After passing through the hole in the Mirrobor sheet, neutrons entered the black box which was positioned on an XY coordinate scanner instrumented with two translation stages (M-IMS600 and M-IMS300) and a motor controller (ESP301), all from MKS Newport Corporation [53]. The SoNDe module was located inside the black box and positioned so that its face was parallel to the upstream side of the black box, and both were perpendicular to the neutron beam. The beam struck the upstream face of the GS20 sheet after passing through a thin layer of tape and and white paper. The SoNDe module was stepped through the neutron beams with a stepsize of 0.5-1 mm in the X and Y directions. The anode signals from each of the MAPMT pixels were processed using the dedicated SoNDe electronics. Negative polarity analog pulse heights for each event with at least one pixel producing a signal above the 500 ADC channel threshold were recorded.

Control, visualization, and data logging were provided by the ESS Event Formation Unit
(EFU) [54][55][56]    As previous work [22-27, 30, 32, 33] has clearly demonstrated that MAPMT pixelgain maps depend strongly upon the method of illumination, all of the results presented below have been pedestal subtracted and gain corrected with pixel-gain maps produced from 3 mm, 2.4 Å neutron-beam irradiations of the pixel centers.  The scan shows that signal leakage to adjacent pixels is ∼7-12% when the neutron beam strikes the center of either pixel. This represents a larger spread of scintillation signal into the adjacent pixel than was the case for previous investigations of relatively central pixels with charged-particle beams [32,33] and may be related to the di usely re ecting white paper placed at the front face of the GS20 sheet. α-particle scan results for (non-edge) P36, P37, P44, and P45 [32] demonstrated summed gain-corrected charge distributions that were at across the pixels and boundary regions. Proton-and deuteron-scan results for (non-edge) P37 and P38 [33] demonstrated summed gain-corrected charge distributions that were slightly convex and centered at the pixel edge. This was because the pixels together collected slightly more of the scintillation light produced from an event at the boundary between them than they collected from an event at the center of either pixel, with the missing light collected by the surrounding pixels. Here, the measured distributions may indicate a light-collection enhancement when P34 is irradiated. Simulations including the nominal 5% level of electronic crosstalk underestimate the amount of signal leaking into the adjacent pixel. When the level of electronic crosstalk is increased to 7%, the agreement between the simulation and the data is best (see below). Figure 6(b) shows the light-sharing ratio between P33 and P34 de ned as (P33−P34)/(P33+P34). For the nominal 5% level of electronic crosstalk, the simulation results in too much signal in the irradiated pixel relative to the adjacent pixel. Agreement at the border between pixels is very good. Overall agreement is again best when the level of electronic crosstalk in the simulation is increased to 7%. This level of crosstalk is larger than that measured by [29].

Results
Competing mechanisms for the spreading of scintillation light currently not addressed in the simulation may instead be responsible.

13
The pixel-hit multiplicity (M = 1, M = 2, etc.) for events as a function of the beamspot position has previously been studied with scans of ∼1 mm FWHM beams of αparticles [32] and ∼100 µm diameter beams of protons and deuterons [33]. A hit occurred if a pixel amplitude exceeded a threshold which was varied o ine. Here, the 3 mm FWHM 2.4 Å neutron beam was employed in a complementary study. Neutron-beam irradiations with a stepsize of 1 mm in X and Y were performed resulting in a 13 × 13 matrix of data. for software thresholds of 1360 ( Fig. 7(a)) and 4545 ( Fig. 7(b)) ADC channels, which correspond to ∼15% and ∼50% of the mean of the P37 pixel-centered full-deposition neutron

Summary and Discussion
Collimated beams of 13 meV and 18 meV neutrons from the IFE reactor ( Fig. 1) have been used to investigate the position-dependent response of a pixelated neutron detector known as a SoNDe module (Fig. 2). A SoNDe module consists of a 1 mm thick sheet of GS20 scintillating glass coupled to a 64 pixel H12700A MAPMT with dedicated readout electronics. The amplitudes of the pixel signals were investigated for di erent irradiation positions by scanning the module through the beam in steps of 0.5-1 mm using a motordriven XY table. A GEANT4 model of the SoNDe module greatly aided in the interpretation of the data (Fig. 3). The amount of scintillation light detected by the MAPMT was increased by ∼40% by placing a sheet of di usely re ecting white paper at the front face of the GS20. γ-rays and neutrons could generally be discriminated with a simple threshold cut (Fig. 4). The amplitudes of the gain-corrected signals were highly dependent on where the neutron beam struck the detector (Fig. 5). When directed towards a centralpixel region, ∼5% of the signal was detected in an adjacent pixel. However, within ∼1 mm of the boundary, ∼30% of the signal was registered in the adjacent pixel. At the boundary, the signal was evenly split between pixels. Overall agreement between the data and GEANT4 simulations was good when a 5% level of interpixel electronic crosstalk was considered, and excellent when a 7% level was assumed. The signal in a pixel adjacent to an edge pixel when the edge pixel was irradiated was underestimated (Fig. 6). For di erent beam positions, the e ect of raising the pixel threshold on the hit multiplicity was studied ( Fig. 7). When the threshold was set at ∼50% of the mean of the neutron full-deposition peak, ∼78% of the data had M = 1 and were localized to within 5 mm, ∼4% were M = 2, and ∼18% were undetected. The M > 1 data were con ned to within ±5 mm of the pixel edges and corners, and were thus within the 6 mm position resolution required for SoNDe operation at ESS. Increasing the threshold to higher values resulted in M = 1 event loss and a reduction of the sensitive area of the detector. Decreasing the threshold to ∼15% of the mean of the neutron full-deposition peak resulted in ∼2% event loss, ∼22% M = 1 data localized to within ±1 mm, and ∼66% M > 1 data. The M = 2 and 3 data could all be localized to within 5 mm, again within the 6 mm position resolution required for the operation of SoNDe at ESS.