Response of a Li-glass/multi-anode photomultiplier detector to focused proton and deuteron beams

The response of a position-sensitive Li-glass based scintillation detector to focused beams of 2.5 MeV protons and deuterons has been investigated. The beams were scanned across the detector in 0.5 mm horizontal and vertical steps perpendicular to the beams. Scintillation light was registered using an 8 by 8 pixel multi-anode photomultiplier tube. The signal amplitudes were recorded for each pixel on an event-by-event basis. Several pixels generally registered considerable signals at each beam location. The number of pixels above set thresholds were investigated, with the optimization of the single-hit efficiency over the largest possible area as the goal. For both beams, 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.

e modular SoNDe concept will facilitate the instrumentation of large areas with a position reconstruction accuracy of ∼6 mm for the detected neutron. A SoNDe "module" consists of a thin Li-glass scintillator sheet (GS20) that is sensitive to thermal neutrons coupled to a 64-pixel multi-anode photomultiplier tube (MAPMT) used to collect the scintillation light. Signals are read out using custom electronics.
Laser light has previously been employed to study the responses of several different MAPMTs in detail [20][21][22][23][24][25][26][27][28]. ermal neutrons have been used to perform rst tests both on similar detectors [29] and on SoNDe modules [9,30]. A thermal-neutron interaction with the 6 Li of the Li-glass results in an α-particle (2.05 MeV) and a triton 2.73 MeV). Scans of a collimated beam of ∼4 MeV α-particles from a 241 Am source have been used to study the position-dependent response of a SoNDe detector prototype [31]. Here, in the absence of tritium beams, beams of 2.5 MeV protons and deuterons have been scanned across the face of a SoNDe module. e goals were to: 1. complement the existing α-particle studies on the position-sensitive response of the detector for events triggering up to four pixels 2. provide data with precision of be er than 1 mm on the position sensitivity of the detector, particularly for events triggering only one pixel (Sec. 2.2.3 and

Sec. 4)
3. investigate the areal response of the detector to events which trigger only one pixel 4. determine the optimal threshold for such events 2. Apparatus

Proton and deuteron beams
e Lund Ion Beam Analysis Facility [32] of the Division of Nuclear Physics at Lund University employs a single-ended 3 MV (max) Pelletron electrostatic accelerator supplied by the National Electrostatics Corporation (NEC) [33]. is machine was used to deliver continuous beams of protons and deuterons with energies of 2.5 MeV to the module under investigation.  ing, object and aperture slits for adjusting the beam size and intensity, quadrupole magnets for focusing, and electrostatic steerers for ne tuning of the beam position [34,35]. A ∼200 nm thick Si 3 N 4 vacuum window [36] separated the highvacuum beamline from the detector chamber operated at room pressure and temperature. e detector chamber contained a motorized XYZ translator on which a SoNDe module ( Fig. 3(b)) was mounted. e beam spots at the location of the SoNDe module were estimated to be ∼100 µm in diameter using a uorescent glass plate. e sizes of the beamspots were due largely to multiple sca ering in the vacuum window. Beam intensity was adjusted using the aperture slits so that the average counting rate on the SoNDe module was 5 kHz.
e amplitude of proton and deuteron signals was measured as the thickness of the traversed air gap between the vacuum window and the GS20 was increased in 1 mm steps up to 6 mm ( Fig. 2). Figure 2(a) shows SRIM [37,38] calculations of the proton and deuteron energy loss in the vacuum window and air. ese predict that a 2.5 MeV proton loses ∼6 keV in the window and 14 keV/mm in air, while for deuterons the equivalent numbers are ∼10 keV and 23 keV/mm. Figure 2(b) shows the measured scintillation-light yield from the GS20 as the air gap is varied, along with a GEANT4 simulation [41]. e simulation, in addition to energy deposition, models scintillation emission and transport. It ts the data best when the Birks constant [42,43] for GS20 is set to 0.021 mm/MeV. e light yield predicted by the simulation was normalized to the measured data so the deviation between measurements and simulations was minimized (at most 5%). e deviation could stem from a combination of uncertainty in the measured air gap (±0.2 mm) and e ects not yet covered in the simulation. e correlation between the data and simulations con rms that protons of all energies produce more scintillation light than deuterons of the same energy. is is because for a given energy, the speci c ionization density of deuterons is higher than that of protons, resulting in a higher level of saturation of the local scintillation-production mechanisms.

SoNDe module
As described in the following sections, the core components of a SoNDe module ( Fig. 3)  sheet was held in place on the MAPMT window using tape along the thin edges.
Consistent with the planned con guration at ESS, no optical coupling medium was employed between the GS20 and MAPMT and no optical re ector was placed over the front face of the GS20. e density of 6 Li in GS20 (assumed to be uniform) is 1.58 × 10 22 atoms/cm 3 . At thermal energies (25 meV), the n + 6 Li → 3 H + α capture cross section is 940 b, resulting in a detection e ciency of ∼75% for a 1 mm sheet.
For 2.5 MeV protons and deuterons, the GEANT4 simulation predicts ∼125 and ∼100 scintillation photons reaching the photocathode, respectively. Light transport from the GS20 (refractive index 1.55 at 395 nm) through a thin air gap (refractive index 1) into the MAPMT borosilicate glass window (refractive index 1.53) is rather ine cient. Each of the 64 pixels has an area of ∼6 mm × ∼6 mm. e peak quantum e ciency of the bialkali photocathode, ∼33% at ∼380 nm, is well matched to the scintillation emission spectrum from GS20, which peaks at ∼390 nm. e Hamamatsu data sheet for the H12700A MAPMT used in this study gives a gain of 2.09 × 10 6 and a dark current of 2.67 nA at a cathode-to-anode voltage of −1000 V, and a factor 1.7 (worst-case) pixel-to-pixel gain di erence.

Readout electronics
Produced by IDEAS [52], the readout electronics for the SoNDe module [9] consist of a front-end board and a controller board. for production running at ESS at average rates of 20 MHz/m 2 and "All-channel Spectroscopy" mode (ACS), used in this work, with a rate limitation of ∼10 kHz for one SoNDe module, equivalent to ∼4 MHz/m 2 . In TOF mode, when any pixel-amplitude threshold is exceeded, the controller board is signaled to identify the trigger pixel (the pixel with the largest signal), perform the time stamping, and then pass the resulting data to the ethernet interface. In ACS mode, when any pixel-amplitude threshold is exceeded, the digitized signals from all 64 pixels are read out. In the ACS-mode investigations, a low hardware threshold of 750 ADC channels was employed, which corresponds to 12.5% of the mean channel of the distribution of the full energy deposition of 2.48 MeV protons (Fig. 4). Higher thresholds were applied o ine, as were corrections for di ering pixel gains.

Measurement
Proton and deuteron beams were used to systematically irradiate the SoNDe module at well-de ned positions. A er leaving the vacuum window, the beams passed through ∼1.0 mm of air before striking the upstream face of the GS20 sheet at normal incidence. e SoNDe module was translated with its face perpendicular to the direction of the beams using an XYZ-coordinate scanner instrumented with Physik Instrumente M-111 micro translation stages and C-862 motor controllers [53]. e scanning assembly was con gured to allow for regular scans in two dimensions with a stepsize of 0.5 mm in both the X and Y directions. e assembly could also move in the Z direction away from the vacuum window. e temperature (∼25°C), pressure (∼101.3 kPa), and humidity (∼30%) within the detector chamber that housed the scanning assembly were logged at the beginning and end of each scan. e anode signals from each of the pixels in the MAPMT were processed using the purpose-built SoNDe electronics. e negative polarity analog pulses for each event with at least one pixel showing a signal above the threshold were measured. e threshold se ing corresponded to an ADC value of about 750. e data were recorded on disk using an ESS Event Formation Unit (EFU) [54][55][56] running on a Centos 7 PC connected through the MiniIO port to Ethernet using the UDP protocol [57]. e EFU data-acquisition system is designed for use by ESS instruments and the acquisition closely resembles the mode of operation anticipated at ESS. Data were recorded for ∼2 s (10000 events) at each point on a scan, followed by a motor translation, so that a complete scan of 2 × 2 pixels with 0.5 mm spacing took several hours. e data were subsequently analyzed using the Python-based [58] pandas [59] analysis tools.

Results
Previous work [20-25, 27, 31] clearly indicates that MAPMT pixel-gain maps are highly dependent upon the method of photon production. us, all of the results presented below have been pedestal and gain corrected with pixel-gain maps generated from the average of the proton-and deuteron-beam irradiations of the pixel centers. and smallest when produced at the corner. Light produced at the corner of four pixels is shared equally by all four pixels. As before, the simulations underestimate the amount of scintillation light spreading to the pixels in close vicinity. Figure 6 shows how the scintillation light was shared by adjacent pixels P37 and P38 as the SoNDe module was scanned horizontally across the proton (Fig. 6(a)) and deuteron beams (Fig. 6(b)). more scintillation light than the deuteron beam. e scan from P37 to P38 shows that light leakage to neighbouring pixels is relatively low close to pixel centers. Moving the particle beam from the center of P37 towards P38, ∼4% of the total light yield is lost to P38 in the rst mm. Across the boundary between the pixels, the lightloss gradient increases to 35%/mm. Based upon the α-particle scan results [31], it was anticipated that the sums of the gain-corrected charge distributions would be at across the pixels and the boundary regions. Instead, the results have a slightly convex distribution centered at the pixel edge. is is because P37 and P38 together collect slightly more of the scintillation light produced from an event at the boundary between them than they collect from an event at the center of either pixel and the missing scintillation light is collected by the surrounding pixels (Fig. 7). Figure 6(c) shows the light-sharing ratio between P37 and P38 (de ned as the di erence between the means of the signal distributions in the pixels divided by the sum) for both protons and deuterons. e overlap between the proton and deuteron data indicates that the light spreading mechanism is very similar for both particles. e absolute di erence between the data and the simulation is up to 20%, greatest in the region between the center of a pixel and the edge. is di erence could be due to scintillation-light spreading mechanisms which are not yet addressed in the simulation or even electronic crosstalk.
e standard mode of operation of SoNDe at ESS will be TOF mode in which every pixel exceeding its individual threshold will be time stamped, resulting in a data set of time-stamped pixel IDs without the underlying ADC information. us, knowledge of the behavior of adjacent pixels when the scintillation is registered in two or more of them is important. Figure 7 shows the division of the signal in the SoNDe module as the proton beam was stepped across the boundary between adjacent pixels. e ∼100 µm diameter proton beam was simulated [41] using GEANT4to produce a distribution of scintillation light incident on the photocathode with a FWHM of ∼2 mm.    is ADC channel 2300 corresponding to 47% of the mean of the pixel-centered fulldeposition deuteron peak. Note that a threshold of at least 2500 ADC channels is necessary to completely discrimate against ∼1 MeV γ-rays from a 60 Co source, which is indicative of possible background contributions.

Summary and Discussion
e position-dependent response of a SoNDe module, which consists of a 1 mm thick sheet of GS20 scintillating glass coupled to a 64 pixel H12700A MAPMT has been measured using highly focused beams of protons and deuterons. e signal amplitudes from individual pixels were investigated as a function of beam position by stepping the module through the beams using a precision XY coordinate translator.
e ∼100 µm diameter beams facilitated highly localized response mapping with a step size of 0.5 mm. A detailed GEANT4 model of the SoNDe module greatly aided in the interpretation of these data and facilitated the calibration of the scintillation-light yield in GS20 as a function of beam energy for both beams (Fig. 2).
Spectra were gain corrected on a pixel-by-pixel basis using the data obtained when the beams were positioned at the center of each pixel. e signal amplitudes were highly dependent on the beam position (Figs. 4 and 5). e single-pixel signal was strongest when the beam was located at the pixel center. Moving the beam by ∼1 mm from the center of a pixel towards a neighbouring pixel resulted in a ∼4% leakage of the scintillation light to that pixel. As the pixel boundary was approached, the leakage gradient increased to ∼25%/mm (Fig. 6). While the simulations underestimated the total amount of scintillation light shared across a pixel boundary, the overall agreement between the GEANT4 model and the data is very good. e amount of scintillation light produced in the GS20 sheet by 2.48 MeV protons was a factor of ∼1.25 greater than that produced by 2.47 MeV deuterons. e spreading of light from protons and deuterons was indistinguishable. e GEANT4 simulation produced a visualization of the scintillation-photon distributions as a function of beam position (Fig. 7). e proton beam directed towards the central pixel region resulted in li le signal in an adjacent pixel. However, within ∼1 mm of the boundary, at least 40% of the scintillation light was registered in the adjacent pixel. When in use at ESS, SoNDe will record only the time-stamped pixel IDs for every pixel exceeding its individual threshold. us, in the pixel-boundary region, double counting can occur. e e ect of the pixel threshold on double and even higher-order counting (the hit multiplicity) was studied as a function of beam position and threshold (Fig. 8). At a threshold of ∼50% of the mean of the proton full-deposition peak, ∼80% of the beam protons were registered in a single pixel. Of the remaining protons, ∼10% were double counted and ∼10% were not detected. e double-counted protons were con ned to regions within ∼0.5 mm of pixel edges, while undetected protons were con ned to regions within ∼1 mm of pixel corners. us, operated in this mode, the active area of the SoNDe module (87% of the MAPMT surface) provides a position resolution of ∼5 mm in X and Y and a detection e ciency of ∼80% for 2.49 MeV protons. Increasing the threshold further simply resulted in a further reduction in the single-pixel event-detection e ciency and sensitive area.