Study of the thermal neutron activation of a gamma-ray detector for BNCT dose monitoring

We present a study on the neutron activation of a gamma-ray detector for a BNCT-SPECT dose imaging system. The detector is based on a LaBr3(Ce+Sr) scintillator crystal, coupled with a matrix of Silicon Photomultipliers (SiPMs), read by a dedicated electronics system. This detector has successfully demonstrated to be capable to identify the 10B compounds when irradiating borated vials with thermal neutrons. However, a background signal around 478 keV was detected, suggesting the activation of the detector itself. This study aims to determine the origin of this background signal by simulating the two main parts of the detector, which are the crystal and electronic boards, in order to assess their contribution to the background signal. The results of the FLUKA simulations show that the neutron capture reactions on both the crystal and electronic boards cause a relevant background nearby the BNCT signal, thereby limiting the detector's sensitivity. To address this issue, a customized cadmium shielding has been developed. This solution was tested at the TRIGA Mark II research nuclear reactor of Pavia University, where experimental measurements and corresponding FLUKA simulations proved its effectiveness.


Introduction
Boron Neutron Capture Therapy (BNCT) is a targeted radiotherapy technique in which patients are administered a 10 B-enriched compound that is preferentially taken up by tumor cells and are then irradiated with a thermal neutron beam.Energetic charged particles released by 10 B(n,) 7 Li capture reaction produce localized damage to the tumor cells, sparing healthy tissues [1].In recent years BNCT has re-gained more attention thanks to the development of accelerator-based BNCT facilities, already installed in Japan and under construction or planned also in Europe [2].This technique has the potential to be an effective cancer treatment solution, but it is not yet widely used due to several challenges, including the lack of a real-time dose monitoring method.Several groups are investigating the feasibility of using special SPECT systems to assess the dose delivered to the patient by detecting the prompt gamma rays at 478 keV emitted in 94% of boron-neutron capture reactions [3][4][5].We focused on the development of a scintillator-based detector for this application, which is read out by silicon photomultipliers (SiPMs) and custom-designed Application Specific Integrated Circuits (ASICs).The system has an excellent energy resolution (< 3% at 662 keV), in principle adequate to separate the 478 keV peak from other adjacent lines, mainly the 511 keV due to annihilation photons [6].A recent proof-of-principle prototype, based on a cylindrical LaBr 3 (Ce+Sr) scintillator crystal of 5 cm diameter and thickness, has successfully demonstrated its capability to identify the 10 B compounds when irradiating borated vials with thermal neutrons at the LENA (Laboratory of Applied Nuclear Energy) reactor in Pavia [7].However, a background signal at 478 keV was detected even with no boron inside the irradiated vial, indicating a possible neutron activation of the detector itself.Since this interference could limit the lowest detectable boron concentration, it is important to account for it in order to achieve the best sensitivity possible for the detector.Therefore, this study aims to investigate the origin of the observed background signal, combining both analytical investigation, simulations and -1 -experimentation, and to provide a feasible shielding solution for the detector.Further work will be devoted to testing directly the detection efficiency of this detector in a BNCT irradiation facility.
The outline of this paper is organized as follows.In section 2 the detector is described.Section 3 begins with the analysis and simulations of the electronic boards of the detector, followed by the scintillator crystal, to gain insight into their respective contributions to the background signal.Section 4 presents a possible shielding strategy, which is subsequently validated through the measurements presented in 5. Finally, section 6 provides the conclusions of the study.

System architecture
The gamma-ray detector we developed is called BeNEdiCTE (Boron NEutron CapTurE) module.It has already been used to perform BNCT measurements with the TRIGA Mark II research nuclear reactor of Pavia University, as described in [7].The first version was based on a cylindrical LaBr 3 (Ce+Sr) scintillator crystal of 5 cm diameter and thickness, read out by a matrix of 64 SiPMs and four custom ASICs.The scintillator crystal has a detection efficiency of 90% at 478 keV and the detector was able to identify a 10 B concentration of 62ṗpm in the target vial [6].The readout electronic is assembled on printed circuit boards (PCBs) positioned beneath the crystal.In particular, there is the motherboard that hosts the SiPMs, the ASICs and the ADCs for the digitalization of the ASICs'outputs; and then the powerboard that supplies the power to the entire system and hosts the FPGA, responsible for controlling data collection.The current version of the module is slightly different, since it now uses a different square LaBr 3 (Ce+Sr) crystal with dimensions 5 cm × 5 cm × 2 cm, to improve imaging capabilities.Since the scintillator material is the same, as well as the PCBs, it is possible to address the background contribution observed in the measurements by analyzing this new prototype.Overall, the detection module is well-suited for SPECT-BNCT measurements due to its excellent energy resolution (< 3% at 662 keV) and its compact size.The 2 cm thickness of the scintillator crystal allows the detector to achieve good detection efficiency, around 60% at 478 keV.Moreover, thanks also to the squared shape, it determines a more localized scintillation light distribution of incident gamma rays, improving the imaging capability.To obtain SPECT images, a lead channel-edge pinhole collimator has been manufactured [8].Its dimensions are 94 mm × 94 mm × 100 mm and the acceptance angle of 9.52 • provides imaging of a circle with diameter 5 cm by placing the detector and the object at 30 cm from its center.Through ANTS2 simulations [9], its geometrical efficiency was estimated to be 3.84 × 10 −6 and its spatial resolution 8 mm.The collimator has been introduced in the simulations of this work, since it is expected to be used in the final SPECT configuration.

Assessment of possible background contributions with Monte Carlo simulations
In order to evaluate the potential sources of background at the same energies of the BNCT signal, the detection module can be viewed as consisting of two main components: the electronic boards and the scintillator crystal.To determine their respective contributions, the signal generated by each component has been examined separately and then compared to the real signal.Monte Carlo simulations were performed with the FLUKA code version 4-2.2 through the graphical interface Flair [10][11][12][13][14].In this version of the software, the transport of low energy neutrons ( < 20 MeV) is carried out with a multigroup transport algorithm.However, the 10 B(n,) 7 Li reaction is simulated with a pointwise algorithm, thus allowing the correct generation of the 478 keV gamma rays [10].
-2 -Some approximations have been adopted to simplify the simulations of this part of the study.These include utilizing a monochromatic spectrum for the neutron source, employing vacuum surroundings instead of air, and excluding walls from the geometry setup Each simulation output value is accompanied by its relative error   =   / x, where   is the estimated standard deviation of the mean and x is estimation of the mean of such value.The simulation relative error  mentioned in this study corresponds to the average of each relative error   over the total number of values   .

PCB contribution
One of the possible sources of background gamma rays at 478 keV is the boron contained in the Flame Retardant 4 (FR4) substrate material of the PCBs.Indeed, FR4 is composed of glass fiber and epoxy resin in variable proportions.Boron is mainly present as boron trioxide B 2 O 3 in the glass fibre, with a weight percentage ranging from 5% to 10% [15].Other electronic components of the detector module, such as SIPMs and ASICs, contain 10 B in p-type semiconductor structures which use boron atoms as dopants.In the FR4 substrate the 10 B weight fraction is estimated to be between 0.2 and 0.3%, and assuming a FR4 density of around 1.85 g/cm 3 , it results in a concentration of at least 2.2 × 10 20 10 B atoms/cm 3 [16].The motherboard and the powerboard have a volume of 9.4 cm × 9.4 cm × 1.6 mm and 9.4 cm × 9.4 cm × 1.2 mm, respectively.The SIPMs cover an area equal to the crystal's surface (5 cm × 5 cm) and are fabricated with a p-over-n technology.The boron-doped area is present only in the implant defining the junction on the front surface of the sensors, which has a depth of a few μm and a concentration in the order of 10 20 atoms/cm 3 [17].The ASICs have a highly p-doped silicon substrate, with a concentration in the order of 10 21 atoms/cm 3 [18].The volume of each one of the four ASICs modules is ∼ 3 mm × 3 mm × 1 mm.
Since both the boron concentration and the volume of the electronic boards is higher than in the other electronic components, it is reasonable to assume that the FR4 material has the larger contribution of 10 B atoms.
A.1.Materials and method.The geometry of the simulations with PCBs and without PCBs are shown in figures 1(a) and 1(b), respectively.The 25 meV monoenergetic thermal neutron beam has a section of 0.25 cm 2 and is directed towards the target, which is a cylindrical vial of water containing 10 ppm of 10 B, which is the minimum concentration needed to perform BNCT-SPECT [19].The vial has a diameter d=1 cm and a height h=5 cm, thus it contains  10 = 2.34 × 10 18 10 B atoms.It is positioned horizontally, allowing the neutron beam to pass through its entire height, and thus enabling the assessment of boron self shielding effects inside the vial.The collimator and detector are positioned perpendicularly to the beam axis.The surrounding material is vacuum.
The collimator, shown in figure 2(a), was modeled according to the geometry described in [8].The geometry of the detector, including the PCBs, is shown in figure 2(b).The scintillator crystal was implemented as a 5 cm × 5 cm × 2 cm rectangular parallelepiped made of LaBr 3 .The simplification of not including the co-dopants is justified in sub-section 3.2.The two boards have an area of 9.4 cm × 9.4 cm and thicknesses of 1.6 mm and 1.2 mm, respectively.Given the complexity of closely reproducing the semiconductor structures present on such electronic boards, and in light of the considerations in section 3.1, we will approximate the PCBs as uniformly made of FR4 material.The FR4 material composition is 40% epoxy resin and 60% glass fiber by weight [16].The electronic components soldered on the boards were not implemented in the geometry since they are negligible in neutron reactions [16].-4 - The epoxy resin and glass fiber material composition implemented in the simulations are described in table 1 and table 2, respectively.The density of the FR4 was assumed to be 1.85 g/cm 3 .In order to simulate the output of the detector, a DETECT card was issued.In the context of FLUKA, the term "card" refers to a default text file designated for specific input commands.In particular, the DETECT card allows to score energy deposition events in a given region.The card was set to score the events in the scintillator with energies between 400 keV and 600 keV.

A.2. Results.
The energy deposition spectra of the simulations are shown in figure 3. The broadening of the 478 keV signal is due to the Doppler effect, since the gamma rays are emitted in flight by the 7 Li after the boron neutron capture reaction [21].It is evident from the plot that the signal caused by the activation of the PCBs (red curve) significantly exceeds the useful signal, the one that originates from the boron present in the vial (blue curve).The number of counts under the two broadened peaks have been calculated, after linear background subtraction, and reported in table 3.By adding the PCBs in the setup, the number of events increases by a factor of 31.Moreover, the peak at 511 keV, mainly due to pair production of high energy gamma rays in the collimator, is visible in both spectra.

Scintillator activation
Another source of gamma rays at energies near 478 keV might be due to the neutron activation of the LaBr 3 scintillator crystal.Self-activity is also present due to the unstable nature of 138 La, but no particular gamma ray line falls inside the energy interval of interest [22].Among the most intense gamma rays generated by lanthanum and bromide neutron reactions, a 487 keV gamma emission can be highlighted.It is due to the decay of 140 La, which is produced via a neutron-capture reaction of the -5 -Figure 3. DETECT scoring in the simulation without PCBs (blue curve) and with PCBs (red curve).The spectrum ranges from 400 keV to 600 keV and the results are given in counts per primary particle (pri).It is possible to see how the broadened peak centered at 478 keV is higher in the simulation including the PCBs.Simulation relative error  < 6%.After the decay, the daughter nucleus 140 Ce is left in an excited state and decays to its ground state upon emission of gamma rays, among which there is the 487 keV gamma ray.The activity () of the 140 La isotopes during irradiation depends on time, and it can be written as in equation (3.2): where R is the reaction rate and  = ln 2  1/2 = 4.78 × 10 −6 s −1 is the decay constant of 140 La.It is possible to write the number of emitted gamma rays with energy   = 487 keV per second  478 keV during a neutron irradiation by multiplying the activity formula (equation (3.2)) by the absolute gamma intensity  487 keV : The absolute gamma intensity  487 keV = 0.455 represents the emission probability of a gamma ray at 487 keV following a 140 La decay [25].Cerium and stronzium co-dopants should be considered in the activation study.However, they are present in concentration in the order of few % mol in the crystal [26].Moreover, the only significant gamma rays emitted by either Ce and Sr isotopes in the energy range of interest are a 475 keV prompt gamma ray from 140 Ce and a 485 keV from 86 Sr, and their partial elemental capture cross section are 0.082 b and 0.0315 b respectively, much smaller than the 2.79 b of the 487 keV gamma ray from lanthanum [24].B.1.Materials and method.In the next set of simulations, both the contributions of the 478 keV prompt gamma rays from 10 B contained in the target vial and in the PCBs and the 487 keV delayed gamma rays from the decay of 140 La are assessed.A RADDECAY card, which commands the simulation of radioactive decays, was issued with option "Semi-Analogue", which treats each single radioactive nucleus in a Monte Carlo way [10].The setup is the same of figure 1(b) and the simulation involved two intermediate steps to account for the generation of gamma rays via neutron capture and their subsequent transportation and absorption in the detector.
In the first step, neutron capture events were simulated.A RESNUCLEi card, which scores the number of stopping nuclei that were created in the run on a region basis, was issued.The results are given in nuclei per primary particle.The number of residual nuclei indicates how many atoms are undergoing a capture process and so it gives information about how many gamma rays are emitted.In order to simulate more realistically the source distribution in the second step, the bodies were subdivided in arbitrary sections which are represented in figure 4. In particular, the vial was divided lengthwise in five 1 cm-thick sections, whereas the scintillator crystal was divided in two 1 cm-thick sections.Since in the vial and in the boards the reaction to be estimated is 10 B(n,) 7 Li, the residual nuclide of interest is 7 Li.Instead, in the scintillator the selected residual nuclide is 140 Ce.-7 -

JINST 19 P05047
The results of the first simulation are summarized in table 4. The number of residual nuclei in each region is reported, along with the relative error .The vial slicing highlighted that 80% of the residual 7 Li atoms scored in the vial are in the first section vial1, showing that the majority of the reactions happen in the first section because of the boron self-shielding.Therefore, in the next simulation the source will be set coincident with the vial1 region.Instead, in the two scintillator regions scintillator1 and scintillator2 the residual 140 Ce nuclides are distributed with a ∼ 60-40 percentage.Hence, in the second step, the simplifying assumption of making coincide the volumetric source with the whole scintillator will be applied.The total number of residual 7 Li nuclides, summed over the five vial sections is 2.972 × 10 −3 pri −1 , whereas for the scintillator the total number of residual 140 Ce nuclides is 4.80 × 10 −6 pri −1 .
Table 4. Scoring of the RESNUCLEi card in the first simulation.For each region section are indicated the residual nuclei of interest, the scoring of the RESNUCLEi card N RES in atoms per unit primary, the percentage of the scoring with respect to the total of the region and the Monte Carlo relative error .

Region
Nuclei N RES [pri In the second step, the transport of the gamma rays from the point of their creation to the detector was simulated in order to model also the geometric efficiency and the detection efficiency of the detector.In order to simulate the gamma-ray sources, the assumption of uniformly distributed isotropic emission was set.The energy of the gamma-ray source was monoenergetic in the case of the 487 keV source in the scintillator.Instead, the sources in the vial, motherboard and powerboard were broadened uniformily between 470.5 keV and 485.5 keV in order to simulate the ∼ 15 keV Doppler broadening [21].
The scorings are performed with a DETECT card in the scinitillator region.The total number of events under the 478 keV Doppler-broadened peak and under the 487 keV monoenergetic peak were counted and the results are shown in table 5.

B.2. Results.
In order to compare the contributions of the prompt gamma rays and the delayed gamma rays, the time evolution needs to be considered.For the prompts, the production rate is constant over the measurement time   , whereas for the delayed it follows the equation (3.2).Therefore, in order to find the total number of gamma rays emitted during the measurement time   , the activity should be integrated over time, as in equation (3.4).
The final formulas that should be used to compare the contributions are given by equation (3.5) and (3.6).
where  RES and  DET are the results given by the RESNUCLEi and DETECT card respectively, and  10 = 94% is the branching ratio of the 10 B(n,) 7 Li reaction with the emission of the 478 keV gamma ray [27].
The unit of the results is pri −1 s, which corresponds to the number of counts registered by the module normalized by the neutrons per second of the beam.The results are summarized in table 6, assuming an irradiation time   =30 min.The motherboard and the powerboard are considered together and their contributions are summed.The contribution of the 487 keV gamma rays from the scintillator is 6 times smaller than the one of the 478 keV gamma rays from the vial.Furthermore, the contribution of the 478 keV gamma rays from the PCBs is 31.56times higher than the vial one, which is consistent with the result found in the previous set of simulations.Table 6.Total number of gamma rays   from each region combining the two sets of simulations, according to equations.(3.5) and (3.6) and assuming a measurement time   = 30 min.In the third column the ratio with respect to the number of gamma rays in the vial is calculated.

Region
[pri The conclusions drawn are based on a measurement time of 30 minutes, but since the 487 keV gamma rays emission from 140 La heavily depends on the duration of the measurement, it's worth considering what would happen in the case of longer measurement times   .Having the 140 La a half life  1/2 = 1.6757 d, thus much longer than the usual   , the integral in equation (3.4) can be approximated as  ≃ • 2  2 .Figure 5 shows the 478 keV contribution from the vial and the 487 keV contribution from the scintillator as a function of irradiation time, in hours.After roughly 3 h of irradiation, the contribution due to the scintillator activation starts to dominate over the signal from the borated vial.Moreover, if multiple successive irradiations are performed, without sufficient cooling time, the activity of 140 La would not decay to zero before the start of the next measurement, further increasing its contribution to the signal.

Shielding material for the detector
The simulations revealed that there are background signals present around the BNCT signal of interest due to both the neutron activation of the PCBs, because of their boron content, and the neutron activation of the scintillator crystal, because of lanthanum activation.Therefore, in order to reduce this background contribution and improve the accuracy and reliability of our detection system, it is necessary to shield both the PCBs and the scintillator from thermal neutrons.
The necessary features which make a material suitable for thermal neutron shielding in our application are: • High thermal neutron cross section in order to reduce the neutrons interacting on the PCBs and on the crystal; • No intense prompt or delayed gamma rays following neutron capture nearby 478 keV; • Low gamma absorption at 478 keV in order to collect the largest signal possible coming from the irradiated tumour.
Other desirable features are compactness, ease of handling and manufacturing, availability and a relatively low cost.
Commonly used neutron shielding materials exploit the naturally high capture cross sections of elements such as cadmium, gadolinium, boron and lithium.To initially assess and verify our simulations we have utilized cadmium as our primary shielding material.This is due to its high thermal neutron cross section (19852 b for 113 Cd [27]), relative affordability, and the fact that it can be manufactured in easily-handled foils that can be cut to size as needed.However, one drawback of cadmium is its strong gamma emission up to a few MeV mainly due to the neutron radiative capture -10 -reaction 113 Cd(n,) 114 Cd, with the two strongest lines at 558 keV and 651 keV [24].Fortunately, this does not significantly impact the useful signal, thanks to the high energy resolution of the detector, but it is still an issue in terms of counting rate to the detector and also in terms of clinical application and biological shielding [28].We plan to further explore the use of 6 Li enriched material as potential alternative for shielding purposes.Although lithium has a high cross section (940 b [27]) and does not produce high energy gamma radiation upon neutron absorption, it is also considerably more expensive than cadmium and not easily available on the market in suitable formats.
The chosen cadmium format were sheets with a thickness of 0.5 mm.This thickness allows to absorb the 99.721% of the 25 meV neutrons [29], thus effectively shielding the detector, and it only blocks a small fraction (less than 4% [30]) of gamma rays at 478 keV, thus preserving the BNCT signal detection.

Experimental results
In this final section, some experimental measurements aimed at assessing the impact of a cadmium neutron shielding are first analyzed (sub-section 5.1).Then, in sub-section 5.2 a further set of FLUKA simulations is presented, with the aim of confirming the above mentioned experimental results.

Measurement at LENA reactor
The experimental measurements were performed at the Prompt Gamma Neutron Activation Analysis (PGNAA) facility of the TRIGA Mark II research nuclear reactor of Pavia University.
The goal was to confirm the background contribution near the BNCT signal due to the detector activation, especially because of the PCBs, and also to prove the effectiveness of the cadmium shielding.Two different measurements were conducted, to compare the performance of the unshielded and shielded detector.A vial with a volume of 5 mL and filled with distilled water (i.e., no boron) was irradiated with a thermal neutron flux of approximately 2 × 10 5 n/cm 2 /s.The detector was placed perpendicular to the neutron flux, and at a distance of 16.9 cm from the vial.For the first measurement it was left unshielded (6(a)), while for the second one it was wrapped in cadmium foils (6(b)).Due to the limited availability of cadmium foils at the time, they were used to cover the frontal surface, the lateral surfaces and the upper surface of the detector, as well as most of the back surface, leaving the down-facing surface unshielded.
The two measurements lasted 10 minutes each and the acquired spectra are visible in figure 7. From the plot emerges that when the detector was left unshielded, a clear peak at 478 keV was observed, which was consistent with the simulations and attribuited to the activation of the PCBs.On the other hand, when the detector was shielded, the peak almost disappeared, and the cadmium neutron capture gamma ray lines at 558 keV and 651 keV became visible, leading to a higher overall background.Moreover, the peak at 511 keV increased with the addition of the shielding.This might be due to pair production events caused by high-energy prompt gamma rays (E>1.022MeV) from cadmium neutron capture, which are listed in [24,31].Moreover, from the number of cadmium neutron capture prompt gamma rays, a gamma line at 522 keV can be identified [31], which is close enough to 511 keV to be superimposed under the peak.The total number of counts per second under the peak at 478 keV has been calculated after Matlab fitting of the photopeak (inset of figure 7), which was obtained fitting the spectrum with a double Gaussian function superimposed to a linear background.The value has -11 -  been found to be 77.42 ± 0.52 cps for the unshielded setup, and it decreased to 0.75 ± 0.31 cps with the cadmium foils.This residual background can be due to the incomplete shielding of the detector due to the limited availability of cadmium foils.However, these results confirmed the findings of the simulations and the effectiveness of the cadmium shielding.The value of 0.75 ± 0.31 cps can be taken as a reference value, and subtracted to the counts obtained when irradiating borated vials (e.g., vials filled with 500 ppm of 10 B), which will be of the order of tens cps.Without shielding (blue curve), the 478 keV peak is clearly visible, whereas in the shielded measurement (green curve) it almost disappears.On the other hand, the cadmium neutron capture peaks at 558 keV and 651 keV appear.The fitting of the peaks performed with Matlab is visible in the insets.

Simulation results
In this section, two further FLUKA simulations will be described and their results analyzed.The aim of the simulations is to approximately replicate the experimental setup of the measurements performed at the PGNAA facility of the LENA research reactor, described in section 5.1.
C.1.Materials and method.The following simulations were performed with the FLUKA code version 4.3-1.This version allows a fully pointwise treatment of low energy neutron interactions through the JEFF3.3database [32], thus better reproducing the experimental results [10].In both simulations, the irradiation room is approximated with a parallelepiped of internal dimensions 80 cm × 200 cm × 220 cm and wall thickness of 50 cm.The material is barite concrete with density 3.3 g/cm 3 , as reported in [33].The composition of the barite concrete was taken from [34] and it is reported in table 7. The irradiation room is filled with air.
The detector module geometry is the same as the one described in section 3.1.The water vial has radius 0.75 cm and height 3.5 cm.It is oriented vertically and positioned 20 cm away from the wall and 16.9 cm from the scintillator surface.The neutron beam has an annular shape with radius 3 cm, and it is directed towards the vial.The neutron energy spectrum is approximated using a Maxwell-Boltzmann distribution with  = 25 meV in order to better represent a thermal spectrum.The first simulation was performed with no shielding around the module, with the geometry shown in figure 8(a).In the second simulation, a cadmium shielding with a thickness of 0.5 mm completely surrounds the module, as shown in figure 8(b).The output of the detector is simulated by using a DETECT card, with an energy range between 300 keV and 1 MeV.

C.2. Results.
In order to better represent the experimental results, a convolution of the detector spectrum obtained from the simulations with a gaussian filter having a 20 keV FWHM was performed, in order to approximately reproduce the finite energy resolution of the system.A 20 keV FWHM corresponds to an energy resolution  = 4% at 500 keV.It should be stressed that in reality the energy resolution is not constant with the energy, but instead it approximately follows a ∼ 1 √  trend [35].Moreover, the values given as output from the DETECT estimator are in counts per primary particle (counts/pri).In order to compare them to the experimental results, they should be multiplied by the neutron flux , in neutrons per second (n/s or pri/s).The flux is given by  =  • , where  is the -13 -  The unshielded spectrum (blue line) resembles the corresponding experimental measurement in figure 7. The counts per second under the 478 keV peak were calculated after Matlab fitting (inset of figure 9) They result to be 117.60 ± 10.10 cps, compared to 77.42 ± 0.52 cps from the measurement.The spectrum simulated in the shielded setup (green line in figure 9) presents photopeaks at the energies corresponding to the strongest cadmium neutron capture gamma ray lines.The peak at 558 keV reaches a peak value of around 26 cps, almost 3 times higher than the peak in the experimental measure.Another difference is that in the simulation a peak at 725 keV emerges more than in the experimental measurement.The differences might arise due to the approximations made in the irradiation room geometry, in which all the other materials present in the experimental setup were neglected.Finally one can see that, even if the background is higher, there is no visible peak at 478 keV in the shielded spectrum.This indeed might suggest that the small peak present in the experimental measurement was due to the shielding partial coverage of the detector module.
To conclude, the simulations indeed confirm that the peak at 478 keV is due to the FR4 material in the PCBs and that a cadmium shielding effectively eliminates it, with the disadvantage of increasing the background.This would prove useful in the application of the module during a real BNCT treatment, since the offset caused by the boards would not probably be constant and controllable, thus not easily subtracted as a simple offset from the dose acquisitions.
-14 -Figure 9. DETECT scoring of the simulation without shielding (blue curve) and with 0.5 mm Cd shielding (green curve), superimposed to the measurement results without (blue dashed curve) and with shielding (green dashed curve).In the blue spectrum, it is possible to see a peak at 478 keV which is due to the emission from the PCBs.In the green curve, the peak at 478 keV is not present but the cadmium neutron capture prompt gamma rays appear.Moreover, the number of 511 keV gamma rays increases, possibly due to the pair production events from high-energy cadmium capture gamma ray.Simulation relative error  < 8%.The fitting of 478 keV peak performed with Matlab is visible in the inset.

Conclusions
In this paper we have analyzed and simulated the possible sources of background near the BNCT signal caused by neutron capture in the gamma-ray detector planned to be used for dose monitoring purposes.Our findings indicate that both the scintillator crystal and the PCBs contribute to this background.In particular, the PCBs were found to be the highest contribution, giving 31 times the gamma rays than a 10 ppm 10 B vial.Thus, it is necessary to shield the entire detector from thermal neutrons to enhance the detection of the real signal coming from the boron in the patient (or in the vial).We have also demonstrated the effectiveness of our shielding with experimental measurements at the TRIGA Mark II research nuclear reactor of Pavia University.By using the cadmium shielding we were able to almost eliminate the background peak.Finally, two further FLUKA simulation reproducing the experimental setup confirmed that the background peak at 478 keV originates from the PCBs and corroborate the effectiveness of the cadmium shielding.
As the next step in our research, we have developed a new prototype of the detector based on smaller PCBs, with dimensions of 6.2 cm × 6.2 cm, which helps to reduce the amount of boron and minimize activation.Additionally, the compact size of the PCBs is beneficial for the future design of the final BNCT-SPECT system with multiple modules.Along with the new module, we are also implementing a smaller shielding case, with a labyrinth-like structure that will enable airflow, to keep the electronics cool, while still providing total shielding.

Figure 1 .
Figure 1.Schematic representation of the two simulated geometries.The neutron beam hits the borated vial, which is 30 cm away from the center of the lead collimator.In turn, the face of the scintillation crystal is at 30 cm from the collimator center.

Figure 2 .
Figure 2. Simulated geometry of the channel-edge pinhole collimator (a) and of the BeNEdiCTE detector, including the electronic boards (b).The motherboard and the powerboard have a thickness of 1.6 mm and 1.2 mm, respectively.

Figure 5 .
Figure 5. Cumulative number of gamma rays from the vial and from the scintillator as a function of time.

Figure 6 .
Figure 6.Shielded experimental setups at the TRIGA Mark II reactor of Pavia University.The vial contains distilled water and it is positioned at 16.9 cm from the detector's face.The thermal neutron beam exits from the back wall.

Figure 7 .
Figure 7. Spectra at 0 ppm acquired with the BeNEdiCTE module at TRIGA Mark II reactor of Pavia University.Without shielding (blue curve), the 478 keV peak is clearly visible, whereas in the shielded measurement (green curve) it almost disappears.On the other hand, the cadmium neutron capture peaks at 558 keV and 651 keV appear.The fitting of the peaks performed with Matlab is visible in the insets.

Figure 8 .
Figure 8. Schematic representation of the simulation geometries which reproduce the unshielded (a) and shielded (b) setups at the TRIGA Mark II reactor of Pavia University.The vial is positioned at 16.9 cm from the scintillator's face.The thermal neutron beam has a radius of 3 cm.The cadmium shielding has a thickness of 0.5 mm.

Table 3 .
Results of the simulations assessing the PCBs contribution. 478,tot is the number of counts from the DETECT scoring under the broadened 478 keV peak.The ratio between the counts of the two simulations is reported in the last row.

Table 5 .
Total counts under the peaks of the DETECT card  DET in the four simulations.The relative error  in the bins corresponding to the peaks is also reported.