Novel directional detector for radioactive material search and monitoring

Abstract

A detector which in addition to automatic radiation alarming and gamma-ray spectroscopy enables directional sensing for fixed, relocatable, and vehicle-mounted detection systems, is demonstrated. A robust and reliable directional sensing relies on inter-detector shadowing effect of attached asymmetric NaI(Tl) and CsI(Na) scintillators. The scintillator package is connected to a single photomultiplier tube. Demonstrated directional sensing is based on background-subtracted NaI(Tl) and CsI(Na) total counts. When a Cs-137 source is positioned in front of the NaI(Tl) detector at 0^o, the NaI(Tl)/CsI(Na) total count ratio is 1.238(9). At 90^o, 180^o, and 270^o, the ratios are 1.000(7), 0.623(6), and 0.743(6), respectively. Scintillator crystal asymmetry causes the observed difference in ratios between 90^o and 270^o. The differences can be further increased by enlarging the sizes and/or modifying the shapes of the asymmetric phoswich crystals. This ratio difference, combined with the time behavior of the total counts, forms the basis of directional sensing. Compact single-channel digital nuclear electronics is used for the instrument’s signal processing. Demonstrated technology enables the development of user-friendly and highly integrated detection systems with extended performance.

Introduction

Detection of ionizing radiation is widely employed, for example, in safety, security, safeguards, and non-proliferation applications^1–5. Despite sophisticated instruments, in-field radiation detection can sometimes be challenging. Some radionuclides, such as fissile U-235 emit only low-energy gamma rays and can be easily shielded⁶. Especially in mobile search measurements, even the shortest source-to-detector distance can be long, complicating the detection. The relative velocity between the source and the detector also affects detection likelihood, as do the detector size, shape, and orientation. When an alarm is triggered in an unknown situation, it is often crucial to quickly determine the direction of the excess radiation. Rapid alarm assessment is greatly facilitated if the alarming detector is a spectrometer that enables immediate nuclide identification.

In our recent publication⁷, we introduced a handheld spectroscopic multipurpose detector that has a directional gamma-ray sensing capability. It employes a symmetric phoswich detector configuration. When an excess gamma radiation is being detected, the instrument can only uniquely determine whether the radiation arrives from the front or back of the detector. If the excess radiation comes from the side, the instrument provides three-dimensional cone, and in two dimensions, two possible directions on a detection plane. The actual direction can then be determined by pointing the detector toward one of the proposed directions. A two-dimensional detection plane refers to a flat surface where both the detector and the source are located. This mode of operation is suitable for handheld instruments that are easy to rotate. However, the approach is less effective for fixed, relocatable, or vehicle-mounted detection systems. Currently, directional sensing for these systems is often achieved using multiple detector modules, sometimes with additional collimators. Directional sensing with a two-detector array was recently studied in⁸. In^9, two collimated thallium-doped sodium iodide NaI(Tl) detectors were mounted inside a van to monitor the sides. When the vehicle passes an unshielded source, for example, on the left side, the left detector registers more counts above background than the right detector, indicating the direction. When the count rate reaches its maximum, the source is at its shortest distance from the detector, perpendicular to the vehicle’s movement. Further information and examples of mobile detection systems can be found in^10,11. Radiation monitoring of pedestrians, vehicles, and/or cargo at security checkpoints associated with major public events or official border crossing points is typically conducted lane by lane using relocatable or fixed detectors^12,13. Monitoring detectors may be placed on one or both sides of the lane, and collimators are sometimes employed to enhance directionality.

In this publication, we introduce and demonstrate the performance of the asymmetric phoswich detector. We aim to show that it enables a robust and highly integrated directional sensing capability for fixed, relocatable, and vehicle-mounted detection systems. A single detector module housing two asymmetric phoswich crystals attached to each other is sufficient for alarming, directional sensing, and gamma-ray spectroscopy. Using only one physical detector module with a single data output channel significantly saves space and simplifies the operation. This approach also improves cost-effectiveness.

Methods

Detector system

The components of the asymmetric phoswich detector used in this work are shown in Fig.1a. A NaI(Tl) and sodium-doped cesium iodide CsI(Na) scintillation crystals with signal decay times of 250 ns¹⁴ and 630 ns¹⁵ are used in the asymmetric phoswich detector module. The combined cylindrical NaI(Tl) and CsI(Na) scintillation crystal package measures 38 mm in diameter and 37.5 mm in thickness. Dimensions of the individual asymmetric scintillation crystals are provided in Fig.1b. The scintillation crystals are optically coupled to a Hamamatsu R6231 photomultiplier tube (PMT). The PMT is enclosed in a mu-metal shielding to protect it from static and/or low-frequency magnetic fields, while the phoswich scintillator package is housed in a 0.5 mm thick aluminum casing. The protective aluminum layer is thin enough to avoid significant attenuation of gamma-rays above 50 keV in energy¹⁶. The detector was custom-made by Scionix Holland B.V.

Fig. 1.

a) Exploded view of the asymmetric phoswich detector module. The main parts are: (1) usbBase, (2) 14-pin socket, (3) Mu-metal shield, (4) Photocathode of the PMT, (5) Diffuse reflector, (6) CsI(Na) crystal, (7) NaI(Tl) crystal, and (8) Aluminum housing. b) Detailed dimensions of the asymmetric (1) NaI(Tl) and (2) CsI(Na) crystals.

As in the previous study with the symmetric phoswich detector⁷, a USB-powered and controlled Bridgeport usbBase multichannel analyzer (MCA) was employed. It was directly attached to the 14-pin socket of the PMT (see Fig. 1 a). The usbBase provides a high-voltage supply and includes the voltage divider for the connected PMT. The high voltage used in the experiments with the asymmetric phoswich detector was 860 V. The Bridgeport usbBase was connected to a standard laptop computer, which is also used for list-mode data storage and analysis.

The same custom firmware as in⁷ was used to produce the required list-mode data. Each recorded event includes a timestamp and values from pulse integration over a long (4 µs) and a short time (0.4 µs) window. Integration over the long window corresponds to the event energy. Signals from the NaI(Tl) and CsI(Na) crystals can be separated based on their different signal decay times using two-dimensional pulse shape discrimination (PSD) analysis. Basically, the ratio between the short- and long-time integrations provides the Y-axis value, while integration over the long-time window provides the corresponding X-axis value (see Fig. 2). A modified charge-comparison method is applied for PSD. In the standard charge-comparison method, the tail of the pulse (Q_tail) is integrated in the short window, and the total pulse (Q_total) is integrated in the long window¹⁷. The PSD parameter is then calculated as Q_tail/Q_total. In our modified approach, the short window is integrated from the start of the pulse, and Q_long is the total pulse. Additionally, a constant A was incorporated into the calculation of the PSD parameter, as shown in the Eq. (1) below:

1

.

Fig. 2.

A two-dimensional PSD histogram associated to Cs-137 source.

The constant A, set to 13 for the presented detector, modifies the PSD parameter and facilitates the discrimination of signal pulses. Two-dimensional gates, similar to those used in⁷ for NaI(Tl) and CsI(Na) crystals, were applied in this study. For the Y-axis gates (shown as blue and orange lines in Fig. 2), the following Eq. (2) was used to compensate for the shape of the PSD spectrum:

2

In Eq. (2), C represents the constant value of the Y-gate, k is a positive constant, and x denotes the pulse integration in the longer time frame.

Events within the corresponding gates were projected onto the X-axis to produce conventional NaI(Tl) and CsI(Na) gamma-ray spectra for further processing. The list-mode data format enables the generation of optimized gamma-ray spectra. Both PSD and list-mode data acquisition are fully implemented using a field-programmable gate array (FPGA) integrated in the Bridgeport usbBase and custom Python code.

Various environmental factors, including temperature variations, magnetic fields, and variations in background radiation, may affect the reliability of the directional sensing. PMTs are susceptible to external magnetic fields, which can negatively impact their gain and other characteristics. Even minor changes in magnetic flux density (on the order of a fraction of mT) can significantly affect PMT performance if magnetic shielding is not employed¹⁸. Although complete shielding is not generally feasible, a common approach is to use a cylindrical mu-metal shield around the PMT, as implemented in the detector used in this study. A single-layer solid mu-metal shield with a thickness of 0.64 mm is used to protect the PMT inside the detector. Despite magnetic shielding, caution is necessary when operating the detector near strong magnets or magnetic fields generated by transformers or power lines. Complete elimination of magnetic field effects would require replacing the PMT with a silicon photomultiplier (SiPM), which is beyond the scope of this study.

Temperature variations can affect the detector in several ways. Both the scintillators and the PMT are sensitive to temperature changes, which influence their performance. For instance, the sensitivity and dark current of PMTs depend on ambient temperature¹⁸, as does the light output of the scintillators^14,15, all of which impact the stabilization of energy calibration. Additionally, the decay time of the scintillator is temperature-dependent, thereby affecting the PSD of the detector. The directional sensitivity, which relies on the NaI(Tl)/CsI(Na) total count ratio, is influenced by changes in decay time and light output. Although decay time changes are small (for NaI(Tl), the decay changes by less than approximately 3 ns per Celsius degree¹⁴⁾, compensation is necessary. Temperature stabilization can be achieved by installing a temperature sensor inside the detector enclosure. Using a look-up table of relative light output values per temperature for NaI(Tl) and CsI(Na), the gain can be adjusted accordingly. To compensate for the decay time, the positions of the Y-axis gates can be adjusted based on the measured temperature. Temperature reading can also be used to adjust the PMT sensitivity. Also important is to insulate the detector to help prevent rapid temperature changes and ensure thermal equilibrium between the PMT and the scintillators.

Background radiation variations can also affect the reliability of directional sensing, particularly in mobile measurement scenarios. To take these variations into account, the radiation background can be determined using the moving average of the series of measurements conducted employing the first-in, first-out (FIFO) principle. Background averaging should automatically be disabled if a radiation alarm is generated to prevent unnecessary elevation of the background average. During the alarm, the previously collected background should be used. Once the alarm has cleared, background averaging should resume after a brief waiting period.

Results

Static measurements

All experiments were conducted in the irradiation hall of the Finnish Radiation and Nuclear Safety Authority. Static measurements were performed by placing the asymmetric phoswich detector module on a rotating table. A Cs-137 source with an activity of 164.1 MBq was placed on a tripod 5 m from the detector. Figure 3a and b show the recorded gamma-ray spectra of NaI(Tl) and CsI(Na) detectors when the Cs-137 source was in front of the NaI(Tl) detector at a 0^o angle. In all static measurements with the Cs-137 source the acquisition time was 120 s, and the corresponding background measurement lasted 300 s. The full width at half maximum values for Cs-137 662 keV peak were 6.1% for NaI(Tl) and 9.0% for CsI(Na), respectively. The spectrometric nuclide identification capability is maintained across all angles, with NaI(Tl) serving as the primary detector due to its superior energy resolution. Additional 60 s long measurements were performed using 10.4 GBq Am-241 and 77 MBq Co-60 sources to confirm detector performance over a broad energy range from 60 to 1332 keV. The source-to-detector distance was 5 m for Am-241 source and 0.5 m for Co-60 source.

Fig. 3.

a) A Cs-137 spectrum measured with a NaI(Tl) detector when the source is in front of NaI(Tl) detector at 5 m distance. b) A CsI(Na) spectrum measured in the same conditions.

The 360^o angular response of the detector was determined by rotating the detector in 15^o increments. Figure 4 shows the angular behavior of NaI(Tl)/CsI(Na) count-rate ratio. Background-subtracted total counts were used in the analysis. Asymmetry is clearly evident, for example, the NaI(Tl)/CsI(Na) ratio is 1.000(7) at 90^o and 0.743(6) at 270^o. This graph also includes the count rate ratios for Co-60. Energy dependence is clearly apparent in the ratios, although the overall trends remain similar. The front and side angular responses were also studied using the Am-241 source. Because Am-241 emits primarily low-energy gamma-rays, the corresponding NaI(Tl)/CsI(Na) ratios vary more significantly. The ratios for the NaI(Tl)/CsI(Na) are 11.9(2) at 0^o, 2.01(2) at 90^o, and 0.562(5) at 270^o. Notably, the measured side ratios correspond closely to the NaI(Tl)/CsI(Na) surface area ratios, which are 2 at 90^o and 0.5 at 270^o. At lower gamma-ray energies, the NaI(Tl)/CsI(Na) ratios become more sensitive to the imperfections in source-to-detector alignment. Figure 4 also presents the statistical 1σ uncertainty boundaries of NaI(Tl)/CsI(Na) ratios. As the statistical uncertainties are small, the range of angles that may produce a certain NaI(Tl)/CsI(Na) ratio is also narrow. This is especially true at the detector sides (90^o and 270^o) where the NaI(Tl)/CsI(Na) ratio changes rapidly as a function of angle. Therefore, directional uncertainty is often dominated by measurement’s statistical uncertainty instead of uncertainty introduced by the reference angular calibration (Fig. 4).

Fig. 4.

Angular response behavior of the asymmetric detector using Cs-137 and Co-60 sources. Associated statistical 1σ uncertainties are also presented.

Detector employed in this work was designed to be evenly efficient to all directions in the detection plane, see Fig. 5. For this reason, the diameter of the detector is approximately the same as its thickness, see Fig. 1. The detector’s ability to detect, especially from the sides (90^o and 270^o), can be enhanced by increasing the detector’s thickness more than its diameter. Also note that the shadowing effect between NaI(Tl) and CsI(Na) detectors intensifies as the detector sizes increase. This furthermore enhances the directional performance of the detector. Figure 5 clearly shows the electronics shielding effect at around 180^o.

Fig. 5.

A 360^o polar plot representing NaI(Tl), CsI(Na), and NaI(Tl) + CsI(Na) total efficiencies for Cs-137 at 5 m source to detector distance. 1-sigma uncertainties are included, but they are smaller than the efficiency markers.

Dynamic measurements

In the dynamic tests, the same Cs-137 source was placed on a remotely controlled movable table, while the asymmetric detector remained on a separate rotating table. One-second spectra were generated and used in the primary data analysis. The corresponding background measurement lasted 300 s. The measurements were performed in 45^o steps, starting from 0^o and completing a full circle. The Cs-137 source moved at a speed of 10 cm/s. In each experiment, the minimum source-to-detector distance was 2.5 m. Figure 6 shows two cases where the Cs-137 source approached from the direction of the NaI(Tl) crystal and passed the detector on the left and right sides. Vertical black lines in the figures indicate the times when the source-to-detector distance reached its minimum. At these points, the combined NaI(Tl) and CsI(Na) background-subtracted total counts also reached their maximum. The corresponding NaI(Tl)/CsI(Na) ratios were 0.97(4) and 0.77(3), which agree with the static measurements shown in Fig. 4. The slope of the NaI(Tl)/CsI(Na) ratio indicates the direction of movement. Figure 7 presents the NaI(Tl)/CsI(Na) ratios associated with minimum source-to-detector distances in other measurement orientations. These ratios also correspond quite closely to the results from the static measurements in Fig. 4. In all measurements except at 180^o, the count rate peaked when the source-to-detector distance was minimized. At 180^o, this was not the case due to the radiation shielding effect of the electronics.

Fig. 6.

A Cs-137 source arrives from the direction of NaI(Tl) scintillator and passes the detector from the a) left and b) right side. Vertical black lines correspond to times when the source to detector distance has its minimum. Savitzky-Golay filter was used to smooth the data for visualization purposes only.

Fig. 7.

Two-sided arrows present the direction of source movement in different experiments. Presented NaI(Tl)/CsI(Na) ratios are related to the instances when the source to detector distance minimizes. Associated statistical 1σ uncertainties are also presented.

As with the symmetric phoswich detector, the direction can be uniquely determined using NaI(Tl)/CsI(Na) ratio only when the source is either in front (maximum ratio) or behind (minimum ratio) the detector. If the source is located on the side, the detector always proposes two angles, one of which is correct. When using an asymmetric phoswich detector, these angles are not symmetric with respect to the detector’s center axis.

Discussion

The angular asymmetry shown in Figs. 4, 5, 6 and 7 is highly useful in practical applications. In many detection operations, the direction of movement is controlled and often appears one-dimensional. For example, consider radiation monitoring at security checkpoints where pedestrians, vehicles and/or cargo pass the detectors along straight lanes or corridors, or mobile searches where the detector is aligned with the direction of vehicle movement (see Fig. 8). It is recommended to mount the asymmetric phoswich detector so that the NaI(Tl) crystal faces the primary direction of source arrival or detector movement. That is partly because the detector electronics, which will scatter and/or absorb some incoming gamma-rays, is located on the CsI(Na) side, reducing detection efficiency in that direction. To ensure reliable source localization with the introduced detector, it is important to align the movement and the detector orientation as well as possible. Based on Fig. 4, an alignment accuracy of plus-minus 15 degrees is sufficient to obtain reliable directional results. In the applications described above, the source-to-detector distance reaches a minimum and a gamma-ray flux [1/(s cm²)] reaches a maximum when the source is at either side of the detector on the thick orange line shown in Fig. 8. Detection likelihood at this optimal position can be further improved by increasing the side detection efficiency of the detector, for example, by enlarging its cross-sectional area toward the sides. The NaI(Tl)/CsI(Na) count rate ratio indicates the side (approximately 0.97 or 0.77), and the slope confirms the relative direction of source movement if necessary.

Fig. 8.

Schematic illustration of the use of asymmetric phoswich detector for checkpoint monitoring or mobile search. The rest coordinate system is attached to the detector. Orange arrows indicate the possible directions of source movement. The NaI(Tl) [CsI(Na)] scintillator is shown in black [red] color. When the source is not moving on the detector’s centerline (dashed lines), the source-to-detector distance always reaches a minimum and count rate a maximum when the source is at either side of the detector on the thick orange line. The corresponding NaI(Tl)/CsI(Na) ratio explicitly identifies the side. A NaI(Tl)/CsI(Na) ratios from the dynamic measurements are presented in the figure.

Successful directional sensing requires that the acquisition time sufficiently matches with the relative velocity between the source and the detector. A sensitivity analysis can be performed using the data presented in Fig. 6. The ANSI N42.35 standard specifies a reference source velocity of 2.2 m/s for passing through a radiation portal monitor¹⁹. This means that during a 1 s acquisition time, the source moves 2.2 m. We can simulate this situation by creating a time series in which 22 consecutive original 1 s background-subtracted measurements are summed and normalized to 1 s acquisition time. In the original measurements, the source moved at a velocity of 0.1 m/s. The outcome is shown in Fig. 9a and b. Similar total counts and NaI(Tl)/CsI(Na) ratio trends as in Fig. 6 remain visible, but shortening the acquisition time from 1 s could already be considered under these conditions to maintain directional accuracy. However, shortening the acquisition time will also reduce the counting statistics of individual measurements. This can be partially compensated by increasing the detector’s size and efficiency. A similar analysis can be performed using a standard walking speed of 1.2 m/s, as defined in ANSI N42.35. These data are shown in Fig. 9c and d. Under these conditions, a 1 s acquisition time should easily be sufficient for reliable conclusions.

Fig. 9.

A Cs-137 source approaches from the direction of NaI(Tl) scintillator and passes the detector from the left in a) and c) and from the right in b) and d). Summed acquisition time per data point in a) and b) is 22 s, while in c) and d) the summed acquisition time per data point is 12 s. Vertical solid black lines indicate the times when the source-to-detector distance reaches its minimum and dashed lines mark the instances of the maximum total count rates for the given time resolutions. Each data point is located to the end of the corresponding integration window. The figure also shows the NaI(Tl)/CsI(Na) ratios with 1 σ statistical uncertainties associated with the maximum total counts.

To ensure reliable source localization, it is important to choose the acquisition time so that even in the case of the fastest and closest possible pass of the source, either from the left (90^o) or the right (270^o), the detector makes several measurements while the source is in the vicinity of the detector (near count-rate maximum), see Fig. 9. Also note that theoretically (inverse square law) the moved distance required to reduce the gamma-ray flux from the maximum value to half is equal to the minimum distance between the source and the detector.

The presented directional sensing method is robust because the NaI(Tl)/CsI(Na) ratio strongly depends on the radiation arrival angle (see Figs. 4, 5, 6 and 7). The inter-detector shadowing effect responsible for this behavior becomes even more pronounced when the scintillator crystal sizes are increased and/or the shapes of the asymmetric phoswich crystals are modified. The same ratio-enhancing effect occurs when gamma-ray energy decreases from 1332 keV (Co-60) and/or 662 keV (Cs-137) to 60 keV (Am-241).

When using asymmetric phoswich technology, the same detector unit that alarms and provides gamma-ray spectra for nuclide identification also delivers directional information required for source localization. This significantly reduces the system’s mass, size, and complexity. Directional sensing becomes relevant only after the instrument alarms. Note that in case of an unshielded high-activity source, the alarm may be triggered well before the source is on the detector’s side in the optimal position. Conversely, if the alarm is triggered only when the source is in the optimal position with minimal statistics, directional sensing may be inconclusive due to statistical uncertainties. This limitation can be mitigated by, for example, increasing the size of the detector, slowing down the relative velocity between the source and the detector, extending the acquisition times, minimizing the source-to-detector distance, and using efficient background subtraction algorithms.

Conclusion

A single asymmetric phoswich detector enables one-sided monitoring of two adjacent pedestrian, vehicle, or cargo lanes simultaneously, or monitoring both sides of the road during mobile operations. If only one pedestrian lane needs to be monitored, the compact size and inherent directionality of the instrument make it practical to place the asymmetric phoswich detector in the center of the lane to minimize the maximum source-to-detector distance. The directional sensing capability facilitates alarm response and enhances immediate situational awareness and confidence. In some applications, such as mobile use or unattended monitoring, data fusion with the closed-circuit television (CCTV) systems can be considered.

Because the asymmetric phoswich detector functions like a standard spectrometric radiation detection instrument, the online data it generates can be used for automatic alarming and nuclide identification. Its gamma-ray energy range is comparable to that of conventional spectrometric detection instruments. Finally, the asymmetric phoswich detector’s size, shape, orientation, and measurement timing can be optimized for the intended application. The detector electronics should always be positioned to minimize interference with the detection operation.

Author contributions

K.P. made the conceptual design related to the detector use and mostly wrote the manuscript. V.B. did the experimental part of the work, processed the collected data, and made the figures. Both authors contributed to finalizing the manuscript.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Data availability

The datasets used and/or analyzed during this study are available under a collaboration agreement with the Radiation and Nuclear Safety Authority of Finland. Please contact the corresponding author for more information.

Declarations

Competing interests

University of Jyväskylä has applied patent WO2024213819A1 Application number: PCT/FI2023/050204 Status: Pending Inventors: Ville P. Bogdanoff, Kari A. Peräjärvi The aspect covered with the patent is the asymmetric phoswich package configuration. The aspect covered with the patent is the asymmetric phoswich package configuration.