SiRO, a scintillator-based hodoscope for muography applications

Weighting over 206 times more than electrons and having relativistic travel speeds that increase their lifetime, cosmic muons, originating from cosmic rays induced air showers, are able to reach the ground and penetrate hundreds of meters of high-density materials before disintegrating. Moreover, due to their relatively weak interaction with matter—with mean energies of several GeVs as minimum ionizing particles—they exhibit a large mean free path and only suffer small angle deviations when interacting with matter. These characteristics make the trajectory of the muon to be mostly straight and allow us to backtrack its path along the line and scan the traversed material. By measuring the muons entering the detector and registering their specific directions, we can make assumptions on the structure of the scanned volume – fewer muons will come from directions where the mean density is larger and more muons from regions with lower density.

Muography^1,2 is an imaging technique conceptually similar to classical radiography, but using cosmic muons instead of x rays. Due to the high penetrability of muons, it is especially useful for the scanning of massive structures. There is a distinction between muon radiography (also referred to as simply muography or absorption/transmission muography) and muon tomography. Whereas muography is based on measurements of muon flux attenuation inside a high-density object, muon tomography relies on studying the muon deflection angle after interacting with the studied volume. In our research, we focus solely on absorption muography.

Muography may be used either as a complementary method along with other conventional techniques (gravimetry, seismic prospecting, etc.) or as a stand-alone method, in cases where conventional methods cannot be applied due to large distances between object and detector or risk of damaging the object while scanning. It has a wide area of applicability,³ being suitable in many fields, such as volcanology,^4–6 geology,^7,8 archaeology,⁹ mining,¹⁰ or civil constructions.¹¹ 

Muon detectors are in a continuous development process in order to achieve better accuracy and, considering the specific requirements of each application, to lower the cost, increase their maneuverability and sturdiness—to be easily transported in hard-to-reach places and endure rough environmental conditions.

In this regard, the scintillator-based muon detector SiRO (SiPM ReadOut) was built within the Astroparticle Physics Group Laboratory at IFIN-HH.

SiRO (SiPM ReadOut muon detector), as presented in Fig. 1, left, was designed to measure the directional muon flux. It consists of six detection layers, placed two by two in order to form three matrices of pixels. Each layer from a group is oriented perpendicular to its neighbor, to provide information for trajectory reconstruction. A rendered view of the detector can be seen in Fig. 1, right.

Each detection layer of $∼$1 m $2$ is composed of 24 independent channels, each consisting of a $100×4×1$ cm $3$ plastic scintillator bar (Polystyrol 80%, Methylmetacrylate 20%), with two 2 mm deep grooves along its length. These grooves house 1.5 mm optical fibers (BCF-91A from Saint Gobain) used to collect the scintillation light and guide it to the sensors.

The S10362-33-100C MPPCs from Hamamatsu¹² are the optical sensors of choice. They offer a series of advantages compared to classical photomultipliers, like small dimensions, low power consumption, low operating voltage (tens of V, unlike classical photomultipliers that require kV), better time resolution, robust, insensitive to magnetic fields, and sensitive to single photons. It was observed from previous tests¹³ that these sensors exhibit a strong dependence of the noise rate with temperature; therefore, all measurements need to be performed in a controlled environment.

Each detection layer is placed on an aluminum frame, for optical shielding.

Depending on the application, the matrices are placed at different distances from one another, thus providing a variable range in angular acceptance. An angular resolution of 1.15 $°$ (0.02 rad) can be achieved at a minimum solid angle of 0.66 sr, obtained for a maximum distance between the top and the bottom matrices of 2 m. For a maximum solid angle of 2.63 sr, obtained for the matrices placed next to each other, with the distance between the active layers given only by the thickness of the encasing boxes, the angular resolution becomes 3.27 $°$ (0.057 rad).

For each detection layer (L1, $…$⁠, L6), six identical PCBs (printed circuit boards) were designed to support 4 SiPMs each and the analog electronics for the amplification of the collected signals. The total voltage amplification of the front-end amplifier is composed of a multiplication of three stages: G = 11*11*2 = 242.

A 20-pin connector is used by each PCB to receive power from the NIM rack (⁠ $±$6 V), as well as the bias voltage for each SiPM, the output signals are carried out through the same connector.

Figure 2 shows the electronic data acquisition system. A custom NIM module for independently adjusting the Bias Voltage for each of the 144 SiPM devices was developed. The 6634B 100 Watt System PowerSupply, from Agilent/HP was used as the main high voltage power supply (65–80 V).

Six custom NIM modules compare the 24 signals corresponding to each detection layer with a preset voltage threshold $V T H$ of 50 mV and transform them to a TTL logic, with a fixed length of 100 ns, suited for the FPGA (field-programmable gate array) input.

All 144 signals from the 6 discriminator modules are routed to an FPGA NIM module. The module is composed of a commercial Darnaw1 unit with a Spartan-3E FPGA from Xilinx,^14,15 connected to a custom PCB motherboard. A commercial MINI-32 USB small development board containing a PIC32MX534F064H microcontroller from Microchip has the role of interfacing the FPGA to a computer via the USB bus and also to transfer the acquired data to the PC.

The data acquisition is controlled via LabVIEW interface. Two modes of operation were programmed for the FPGA unit. One mode, called “Alignment,” was developed for calibration purposes, leveling out the channels’ response by counting all their triggered events in a time gate of 10 s and aligning their triggering rate through the adjustment of the bias voltage.

The other mode, called “Measurement” mode, was designed for data acquisition. For this purpose, the 144 channels of the detector are grouped into 6 groups of 24 channels, each corresponding to one layer. Every group of 24 channels is passed through an OR gate and the outputs, corresponding to each layer, are sent to an AND gate with 6 inputs and a time frame of 100 ns.

In order to set the coincidence trigger between the six layers in any desired combination, six ON/OFF switches are hardwired to the FPGA to program the AND gate. All boolean operations are coded into the FPGA as can be seen in block diagram from Fig. 2. The resulting signal from the AND gate is the trigger signal for the acquisition software to write data on storage (PC). The recorded information contains the channel number of the detected signal and a time stamp with millisecond precision. Trajectory reconstruction and event validation are performed afterward using the ROOT toolkit¹⁶ before any other analysis. Other future reconstruction procedures will be applied on the validated data and are dependent on the type of application required.

One important tool utilized on a large scale by muography applications is represented by Monte Carlo routines used to simulate the interaction of muons with materials of interest, such as the composing materials of the objects to be scanned, or the sensitive volumes of the detectors involved in the measurement process. The Monte Carlo method was used to simulate the interaction of muons with the detector.

An accurate description of the directional muon flux at sea level is of the utmost importance for muography applications.

The first necessary step in any simulation is emulating the incoming particles, in our case, the directional muon flux. We have used the CORSIKA toolkit (COsmic Ray SImulation for KAscade), a Monte-Carlo software developed to simulate the generation and propagation of secondary particle cascades in the atmosphere,¹⁷ to calculate the angular and energetic distributions of secondary muons at the ground level. The input for CORSIKA simulations was the flux of primary protons and He nuclei measured by the AMS experiment.^18,19 This method was previously used by Wentz et al.²⁰ 

The resulting muon flux was divided into eight energetic intervals, from 0.2 to 658 GeV, and seven angular intervals, between 0 $°$ and 70 $°$⁠, for the zenith angle. The angular and energetic distribution of muons for energies below 60 GeV is depicted in Fig. 3.

Subsequently, the interaction of the muons obtained from CORSIKA with the active volume of the detector was simulated using the Geant4 package.²¹ The energy and zenith angles of the incident muons were distributed according to Fig. 3, while the azimuth angle was random. The detector geometry used in the GEANT4 simulations was the same as the one described in Sec. II including the metal boxes.

Two sets of simulations were performed. The first one aimed to emulate the detector response when exposed to “open-sky” muons, i.e., atmospheric muons that reach the detector without passing through any object. For the second simulation, a $20×20×10$ cm $3$ lead object was placed on top of the detector, centered at 48 cm (in between channels 12 and 13) on the x axis, and at 52 cm (in between channels 13 and 14) on the y axis, with its axes aligned with the detector. The distance between the top and bottom layers was 150 cm.

For each simulated set, 20 $× 10 6$ muons were randomly generated from a 16 m $2$ surface with energies and directions distributed according to the parameterization in Fig. 3. The emission source was centered with the detector and placed 1 m above it to ensure that the generated muons covered the entire solid angle of the detector.

Compared to a cylindrical detector, the square shape of our matrices induces a non-uniformity in the azimuthal distribution of the validated events. To correct this effect, a truncation of the sensitive volume of the detector was performed by removing from the analysis the events that crossed each matrix of pixels outside of a circle with a diameter of 1 m, equal to one side of the active volume. In this way, the square detection area is reduced to a disk. This truncation procedure was applied to both simulations and measurements presented throughout the text.

Figure 4, left, displays the simulated detector response to “open-sky” muons. The right side of the figure shows that the shape of a $20×20×10$ cm $3$ lead object, placed on the detector, can be reconstructed if sufficient events are accumulated. Based on the number of simulated muons and the muon flux measured at the ground level, of $∼$120 muons $$ m $− 2 s − 1$⁠, an exposure time of 2.89 h was estimated to be necessary for the real detector to collect sufficient statistics so that the reconstructed image is similar to the simulated one, as seen in Fig. 4.

To calculate the efficiency of the detector, 20 $× 10 6$ muons were generated from 1 m $2$ surface placed at 1 mm above the first layer. The efficiency was calculated as the ratio between the muons that interacted with all layers of the detector (considered valid event) and the number crossing the first detection layer. A 15.5 $%$ detection efficiency was obtained for a detector configuration with 150 cm between the top and bottom active layers, increasing as the distance between the active layers decreases. The maximum efficiency, 47.1 $%$⁠, was obtained for the layers placed on top of each other. It must be noted that simulations provide only the influence of the geometry of the detector and the energy deposit of particles on the efficiency. Light-collection and other detector-specific effects (e.g., dead time) are not included. They further degrade detection efficiency.

Test campaigns were undertaken in order to evaluate the performance of the SiRO detection system.

We started with “open-sky” measurements at the ground level in the Astroparticle Physics Laboratory from IFIN-HH. We continued in the same location with the identification of the shape of a $20×20×10$ cm $3$ lead object placed on top of the detector.

Changing the location, we mapped the directional muon flux in the IFIN-HH Underground Laboratory from Unirea Mine, Slănic Prahova.

The calibration was obtained for the 144 independent channels by aligning their counting rates. To eliminate the dependence of the MPPC 10362-33-100C sensors on temperature, the measurements were carried out in rooms with a controlled temperature of 25  $°$C, in both above and underground laboratories.

Measurements were performed in the Astroparticle Physics Laboratory (92  m a.s.l., 44.3511 $°$ N, 26.0450 $°$ E). Once the detector was calibrated, the data acquisition software was set to “Measurement” mode, continuously recording the signals from the detection channels and saving the data into a text file.

After the selection of the valid events, i.e., events that gave a single signal in each of the six active layers), the next step was the truncation of data, to eliminate the anisotropy given by the corners of the matrices of pixels in a similar way as performed for the simulations.

In the left plot of Fig. 5, the directional flux of cosmic muons is observed, as measured with the SiRO detector, with detection planes placed at 70 cm one from one another, with no objects placed above the detector. This type of measurement, without the presence of the object that needs to be scanned and performed in the same location and conditions, is referred to as “open-sky” measurement. The data acquisition time for this measurement was 21.5 h. After adding a $20×20×10$ cm $3$ lead object on top of the detector, centered on the 11th pixel (corresponding to 42 cm) on the x axis and on the 11th pixel (corresponding to 42 cm) on the y axis, with its axes aligned with the detector, another 21.7 h set of measurements was performed.

The data obtained in both campaigns were normalized to their own total number of recorded tracks, and the contribution of the “open-sky” data was subtracted to eliminate the effects of the environment and detector, thus highlighting only the contribution given by the object, as seen in Fig. 5, right plot. A homogeneous distribution can be observed, except for the central region where the lead object was placed, with dimensions similar to the object (⁠ $5×5$ pixels, equivalent to $20×20$ cm $2$⁠). To better quantify the observed object, complex reconstruction methods and longer exposure times are required. They will be the subject of another study.

Directional muon flux measurements were performed in the IFIN-HH Underground Laboratory in the Unirea Mine,²² Slănic Prahova (428 m a.s.l., 45.2360 $°$ N, 25.9409 $°$ E), situated at 208 m (⁠ $∼$610 m.w.e.—meters-water-equivalent) below ground.^23–25 The temperature in the mine is approximately constant at 13  $°$C.

An analysis of interest in the underground is the distribution of arrival directions of incident muons with respect to azimuth and zenith angles. Anisotropies in the cosmic muon flux may suggest the existence of regions with different densities compared to the rest of the scanned volume.

After performing “open-sky” measurements of the directional muon flux at the ground level at the location, the detector was placed underground, intentionally close to a mine wall (⁠ $∼$50 cm), oriented as shown in Fig. 6, center. The solid angle viewed by the detector is emphasized in Fig. 6, right. The duration of the open-sky measurements was 20 h while the exposure time for underground measurements was 46.3 h.

It can be easily observed in Fig. 7 how the presence of the wall decreases the number of incident muons, while the muons coming from the open side bring a consistent contribution to the flux.

The same anisotropy is also observed if we represent the angular distribution of muons. In Fig. 8, the zenith intervals were chosen so that each bin views equal solid angles. By eliminating the “open-sky” contribution (Fig. 8, left), we highlight the differences in density of the scanned volume, in our case, the mine wall and its cavity (Fig. 8, right). The anisotropy generated by the presence of the wall near the detector and the cavity on the opposite side is visible, as more muons (higher count rate) are recorded from the direction of the cavity (13 $°$–38 $°$ zenith angle and 50 $°$–145 $°$ azimuth), while less muons arrive from the direction of the wall (0 $°$–25 $°$ zenith and 220 $°$–320 $°$ azimuth).

Using plastic scintillator bars and SiPM devices, a detection system with the purpose of measuring trajectories of the incident cosmic muons has been developed.

Both the angular resolution of the detection system and the solid angle covered by the detector can be modified according to the requirement of each experiment by adjusting the distance between the three matrices of pixels. A maximum solid angle of 2.63 sr can be covered, when the matrices are placed one on top of each other, without any space in between, corresponding to an angular resolution of 3.27 $°$ (0.057 rad). When the distance between the top and bottom matrices of pixels is maximum (2 m), a resolution of 1.15 $°$ (0.02 rad) is obtained. In this configuration, the viewing solid angle is minimal (0.66 sr).

Because the counting rates of SiPM devices are correlated with the ambient temperature of the medium, the tests reported in this paper were performed at a fixed temperature of 25  $°$C in a controlled environment.

The response of the detector to the interaction with atmospheric muons was simulated using the GEANT4 toolkit. A procedure was developed to reproduce the natural directional muon flux using CORSIKA package. A 2D reconstruction of the “open-sky” muon flux was obtained. Also, a lead object was included in the simulations to analyze the detector response in its presence. Similar studies were performed in the laboratory, at the ground level, with the real detector. Results are in agreement with simulations, the reconstructed dimensions of the object are similar to the real object.

When mapping the arrival directions distribution of the atmospheric muons in the Underground Laboratory of IFIN-HH from Unirea Mine, Slănic Prahova, a strong anisotropy was observed, due to the presence of the mine wall close to the detector and the cavity on the opposite side.

Hereby, the good functioning of the detector and its ability to be successfully used in muography applications was demonstrated. Based on its performance, the SiRO detector qualifies for volcano scanning, geological evaluation, exploration of mineral deposits and underground cavities, safety assessment of civil infrastructure, or investigation of the internal structure of any large volume where other methods cannot be applied.

This research was funded by the Romanian Ministry of Research, Innovation and Digitization, grant PN 23 21 01 02 project within the National Nucleu Program, and the TE127/2020, TE128/2020, and PED289/2020 projects within PNCDI III.

The authors have no conflicts to disclose.

M. Niculescu-Oglinzanu: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Supervision (equal); Validation (equal); Writing – review & editing (equal). D. Stanca: Conceptualization (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Software (equal); Supervision (equal); Validation (equal); Writing – original draft (equal); Writing – review & editing (equal). A. Bălăceanu: Investigation (equal). M. Dobre: Formal analysis (equal); Software (equal). A. Gherghel-Lascu: Formal analysis (equal); Software (equal). A. Saftoiu: Software (equal); Writing – review & editing (supporting). R. Smău: Formal analysis (equal); Software (equal). C. Vancea: Formal analysis (equal); Investigation (equal); Validation (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.