Introduction

Alpha-emitting radionuclides can cause severe damage to biological tissues when inhaled, swallowed, or when they enter the body through open wounds [1]. Owing to the very high concentration of energy deposited along their track, alpha particles have a high probability of inducing double-stranded deoxyribonucleic acid (DNA) breaks, almost completely irreparable damage, making them very cytotoxic [2]. Following a possible radiological emergency involving an accidental or deliberate dispersion of alpha-emitting radionuclides in the environment, tinny fallout particles could be easily deposited in common environmental surfaces (e.g., in soil, sand, vegetation, concrete, etc.) and ingested or inhaled by living beings. This scenario represents a very complex situation that can cause significant damage to humans and affect many aspects of peoples' lives, including their local economies. Alpha contamination could also be a serious concern during the decommissioning of nuclear fuel fabrication and reprocessing facilities.

The efficiency of responding to such radiological scenarios depends strongly on the available instrumentation for detecting alpha-emitting radionuclides. Due to the short range of alpha particles in air of only a few centimeters [3], this cannot be achieved by conventional detectors for alpha contamination such as detectors based on the scintillation technology (e.g. silver activated ZnS thin films [4]) and detectors based on the semiconductor devices (e.g. passivated implanted planar silicon and silicon gold surface-barrier detectors [5, 6]). These detectors must be positioned within a few centimeters of the source (typically, within 4 cm) in order for the alpha radiation to be detected and thus have major drawbacks [7] such as, for example, personnel will be exposed to other hazards and risks such as other types of radiation, fire, etc. and detectors may become contaminated if they touch the source.

The drawbacks of conventional detection techniques for alpha contamination can be overcome by using the alpha particle-induced ultraviolet (UV) luminescence of air, the so-called alpha radioluminescence [8, 9]. The ultraviolet signature of an alpha particle stopping in air was discovered in 1903 by Sir William and Lady Huggins when measuring the emission spectrum of a radium bromide source in air [10]. They noticed that spectral lines recorded in the ultraviolet (UV) region were in complete agreement with the band spectrum of molecular nitrogen excited by an electric discharge. The phenomenon of radioluminescence, characterized by the faint luminescence of molecular nitrogen in air caused by alpha particles, has attracted increasing attention both in academic research and radiation safety applications [7,8,9]. The main benefit of this technique arises from the inherent ability of UV light to propagate through the air over substantial distances, often spanning kilometers, which surpasses the limited range of alpha particles in air (a few centimeters) by several orders of magnitude. This enables the non-contact detection of contamination from a safe distance, thereby mitigating risks to personnel, reducing detection time, and minimizing costs. Radionuclides that emit both alpha and gamma radiation, such as 241Am, can also be detected with gamma spectrometry systems (for example, 241Am emits a strong gamma emission line at 59.5409 keV [11]). However, the advantage of the radioluminescence method is in full display for radionuclides that are almost exclusively alpha emitters, such as 210Po. The remote detection of this type of radionuclide presents a real challenge for gamma spectrometric techniques since 210Po has only two alpha transitions, the higher energy one at 5.407 MeV having a transition probability of 99.99876%, and a single gamma transition at 0.803052 MeV, which is practically negligible with a transition probability of only 0.00124% [11].

As illustrated in Fig. 1, the majority of air luminescence lines span the UV-A wavelength region between 315 and 400 nm. The high solar ultraviolet background in this wavelength region has been a limiting factor for the environmental application of this technique. At night, the UV-A background is also substantial due to city lights or other stray light sources. With advancements in solar-blind photomultiplier tube (PMT) technology based on cesium telluride photocathode, the UV-C region of the spectrum has become promising for outdoor measurements since very few solar UV-C photons reach the ground level, with virtually no photons below 280 nm [12].

Fig. 1
figure 1

Typical alpha radioluminescence spectrum measured in dry air at normal temperature and pressure (NTP) and N2 + NO gas atmosphere with 2 ppm NO. In air, more than 95% of the total intensity is contained in the UV-A and UV-B spectral region (within 310 nm to 400 nm), while the intensity of UV-C radioluminescence (below 280 nm) is very low (< 5%). The radioluminescence signal in the UV-C spectral region can be enhanced by more than three orders of magnitude by purging the space around the alpha source with N2 + NO gas mixture. The spectra are measured with a CAS140D spectroradiometer at the PTB Ion Acceleration Facility (PIAF) with a narrowly collimated beam of 5 MeV alpha particles. The intensities are normalized to the 337 nm spectral line of air radioluminescence

Full size image

The luminescence of air is largely attributed to the radiative relaxation of molecular nitrogen (N2) as displayed in Fig. 1. Even though nitrogen comprises 78% of air, oxygen and water vapor quench the luminescence, therefore the signal is enhanced in a pure nitrogen atmosphere, reaching 600% gain in UV-A region. To achieve even higher luminescence yields, trace amounts of nitric oxide—between 10 and 200 ppm—can be added to a nitrogen atmosphere leading to a drastic increase in the UV-C output, surpassing yields of other UV regions by a wide margin [13].

The EMPIR project 19ENV02 RemoteALPHA [14] has led to the development of novel instrumentation and methods for remote alpha source detection, including the metrological infrastructure to ensure the traceability of alpha activity measurements. The project has the objective of making the optical alpha sensing technique a useful contamination monitoring tool in nuclear safety activities and radiological emergencies which might involve the accidental or deliberate dispersion of radioactive sources in the environment.

Within the RemoteALPHA project, several radioluminescence detection systems aiming at different application scenarios have been developed [15]: heavy tabletop or tripod-based scanning systems for indoor and outdoor use, and lightweight systems optimized for drone-assisted environmental monitoring. Here, the scanning optical system with a 240 mm diameter UV fused silica (UVFS) lens and two-stage goniometer has been used with optomechanical components installed on a heavy-duty tripod. The present work shows an application of the scanning UVFS radioluminescence detector to quantify total surface activities of americium-spiked environmental samples (leaves, sand, and soil) and nuclear materials involving a standard 239Pu source, a set of depleted uranium sources with complex geometry, and a UO2 pellet. These samples were selected to test the feasibility of radioluminescence instrumentation and methods for scenarios encountered (a) when fallout containing alpha emitters (in the form of radioactive dust or washout during rain) is deposited in common environmental materials such as soil, sand, concrete, leaves, etc., and (b) during decommissioning of nuclear facilities where alpha emitters are produced, handled, used, and stored.

Experimental

The UVFS lens system (Fig. 2a) was positioned at a reference distance of 2 m from the alpha particle sources to scan the scene at pitch and yaw angles (see Fig. 2b). The goniometric scanner allows high-fidelity imaging with guaranteed repeatability down to 0.025° resolution, which translates to a pixel size of less than 1 mm at a distance of 2 m. The optical part is encompassed in an aluminum frame and includes a UVFS lens, collimator, filter assembly, and PMT according to a selected wavelength range:

  • in UV-A, Hamamatsu H10682-210 photon counting head is used with a 2-filter assembly consisting of Edmund Optics #65-189 and Semrock FF01-340/12-25 centered at 337 and 340 nm, respectively;

  • in UV-C, Hamamatsu H11870-09 photon counting head is used with 2 Semrock FF01-260/16-25 filters centered at 260 nm.

Fig. 2
figure 2

a Radioluminescence imaging system built around a 240 mm diameter lens telescope. The system is mounted on a two-stage goniometer (Newport MBGM160PE and RVS80CC) and has an Intel RealSense D435 depth camera that captures an RGB color image of the scene. b Schematic diagram of radioluminescence mapping. The samples placed in the airtight enclosure are scanned by the optical system located 2 m away from the sample. The 2D image of the scene is obtained by scanning the optical system with the two-stage goniometer in pitch and yaw directions

Full size image

The scanning system also features an Intel RealSense D435 depth camera which takes an RGB color image and appends it with the depthFootnote 1 information allowing superimposing the ultraviolet scan onto the RGBD image given the known offset of optical elements. The camera also facilitates selecting the area of interest which determines the scan extent in pitch (Y-coordinate) and yaw (X-coordinate). As illustrated in Fig. 4, the scanning is performed in a snake pattern from the bottom left corner with the full yaw range (fast rotational stage) traveling every pitch step (slow cradle). The scan speed is limited at 10 deg s−1 and 0.5 deg s−1 for the yaw and pitch stages, respectively.

Since measurements were performed in the laboratory environment which allowed the dark operation, most scans used the UV-A detector configuration in air. This setup has been already tested measuring uranium mineral samples (pitchblende) reaching the detection limit below 1 kBq at a high integration time per pixel [16]. Whenever the activity of the source was too low to be detected in air, the chamber was purged with N2 + NO gas with 200 ppm nitric monoxide, and the UV-C detector configuration was applied. The NO concentration was chosen to be above saturation (10–50 ppm [13, 15]) due to the direct displacement of air (purging) in the chamber instead of flashing the pre-vacuumed chamber.

The radioluminescence mapping capability of the UVFS scanner was tested with two groups of low-activity alpha particle-emitting samples: (a) 241Am-spiked environmental samples (leaves, sand, and soil) and (b) nuclear materials involving a standard 239Pu source, a set of depleted uranium sources with complex geometry and a UO2 pellet.

Environmental samples were prepared by distributing a gravimetrically determined amount of liquid 241Am standard solution over the material surface, with the deposited activity between 0.6 and 12 kBq (see Table 1). The composition of these samples was chosen to represent a continental European environment with leaves of the plan tree, loam soil for soil samples, and quartz sand for sand samples. Sample containers are made of plexiglass and designed as equilateral prisms with 45 mm side length and 12 mm height required to fit inside an optical chamber of IFIN-HH Liquid Scintillation Counter (LSC) for sample evaluation using the triple-to-double coincidence ratio (TDCR) technique, see Fig. 3. In the TDCR technique, radioluminescence from the low-activity samples is measured in a triangular UV tight box (i.e., no external UV light can penetrate the box) containing three PMT arranged in each box face (120° to each other), placed very near to the sample (ca. 10 cm). The activity of the sample is then determined from the double and triple coincidence counting rates [17].

Table 1 The deposited activity and the mass of the environmental samples
Full size table
Fig. 3
figure 3

Airtight plexiglass case with a window made of quartz glass and sample containers inside

Full size image

The depleted uranium samples and the UO2 pellet were selected to test the developed instrumentation and procedures for scenarios involving the decommissioning of nuclear facilities (e.g., facilities where alpha-emitting material is manufactured, handled, used, and/or stored). These objects are storage wastes stockpiled in the IFIN-HH nuclear repository.

The environmental samples were placed in an airtight enclosure to enable optical measurements of environmental samples while conforming to the radiation protection rules concerning the handling of open sources and chemical safety due to NO [18]. The box has dimensionsFootnote 2 of 21 cm × 27 cm × 25 cm and features a 20 cm × 20 cm fused silica window at an angle of 45° for the optimal transmission in UV-A and UV-C, with other sides made of Plexiglas (see Fig. 3). The gas line connectors are installed to allow purging the volume with N2 + NO gas mixture to maximize the detection sensitivity of the optical setup. The fused silica window has very high transmission in the UV spectral range of (91 ± 2)% and thus affects very little the radioluminescence intensity. Without the chamber, radioluminescence mapping would be about 9% more intense. Although some measurements lasted many hours, the likelihood of radon accumulation is very low (if not impossible) because during the measurement in air both the inlet and outlet ports of the enclosure (chamber) were open and the entire chamber was housed in a ventilation box, and during the measurement with gases the chamber was constantly flushed.

The measuring distance of 2 m is a reference distance. The results measured at this reference distance (Iref, xref = 2 m) can then be used to deduce expected counts at some other distance x (in meters) by using the inverse square root law [3]: \({I}_{x} = {I}_{\mathrm{ref}}{\left({x}_{\mathrm{ref}} / x\right)}^{2}\).

Methods

The radioluminescence scanning system used in this study can be considered a single-pixel detector. The radioluminescence image is obtained by mechanically aligning the lens system using two goniometric stages to scan the area around the source (or expected contamination), resulting in a radioluminescence image in terms of pitch and yaw angles. Each pixel in the radioluminescence image represents the UV intensity in the detector's field of view (FOV) at a specific pitch and yaw angle. The FOV (i.e., the angular extent of the object seen by the detector) depends on the active area of the PMT and influences both the number of photons detected and the filtering efficiency of the radioluminescent light [15]. Since FOV is approximately 1 degree, while the resolution in the pitch and yaw directions is between 0.1° and 0.5°, there is an overlap in the FOVs of neighboring scan points. This will cause some radioluminescence image blur. The blurring is mathematically characterized by the so-called point spread function (PSF) or impulse response, which represents the system's response to a point source [19]. Finally, the radioluminescence image is juxtaposed with the traditional image to outline the shape and location of the source. The measured image consists of a radioluminescence signal and a background signal, which in turn consists of the PMT dark rate and the ambient UV component that reaches the PMT. Source signal discrimination is therefore necessary for proper quantification of alpha source activity.

The final image of the alpha source is a convolution of the true source image (its physical shape) with the radioluminescence kernel and detector’s PSF. Since radioluminescence kernels are not error-free,Footnote 3 each deconvolution step will introduce inaccuracies in the deconvoluted (true) image which will propagate to the source activity quantification; therefore, a simpler method has been formulated to determine the source activity from the radioluminescence image.

The method works with the raw data and uses the full extent of the source radioluminescence profile. For quantitative analysis, the initial step involves identifying the source region. When dealing with highly active samples, such as those with activity on the order of a few hundred kilobecquerels or higher, the identification of the source region is straightforward. Conversely, for low-activity samples, typically on the order of a few kilobecquerels, smoothing, and other image processing methods must be used to highlight the source effect.

Figure 4 demonstrates how simple image processing facilitates discrimination of low-activity sources when the source signal is grossly obscured by noise. After the source has been outlined, the corresponding background area is defined to estimate the noise in the source region. Given the scan pattern displayed in Fig. 4, the best image parts for estimating the background lay to the side of the signal since these data points are the closest to the source in both time and space. Among other factors, this background selection accounts for the relaxation of PMT after exposure to intense light.Footnote 4

Fig. 4
figure 4

The radioluminescence mapping procedure of the 4.9 kBq 239Pu source measured in air from a distance of 2 m. The depth camera RGBD image a is used to select the scanning extent in pitch and yaw axes. The raw UV image c is smoothed using an averaging filter with a kernel size of 11 pixels. The smooth image d is then used to establish the signal and background domains: the source is outlined in red, and the background area (white) is selected according to the scan pattern b of the detector. The smooth data is further denoised through thresholding determined by the background signal, and the processed image e is superimposed onto the color image a to produce the final radioluminescence image f

Full size image

After the source and the background domains are defined, the computations are performed on raw data, with the total surface activity \({A}_{s}\) calculated as

$$A_{s} = k\frac{{{\Sigma }_{{{\text{net}}}} }}{s}$$
(1)

where \({\Sigma }_{\mathrm{net}}\) is the sum of net count rates in the source domain, \(s\) is the scanner sensitivity (in units of s−1 MBq−1) at the same measurement conditions determined using the reference source, and \(k\) is the dimensionless factor indicating the scan resolution and transmission corrections (see Supplementary Information).

Results and discussion

The measurement campaign was carried out as a joint effort between IFIN-HH, PTB, and MATE. All samples were provided by IFIN-HH, with measurements split into 3 phases: (1) the reference 239Pu source which is the standard source for calibrating contaminometers; (2) environmental samples of sand, soil, and leaves spiked with the standard 241Am solution and (3) a set of nuclear materials featuring a depleted uranium bullet, two pieces of depleted uranium shielding used to collimate industrial radiography sources, and a non-irradiated (fresh) UO2 pellet.

All measurements were performed at a distance of 2 m between the lens and the source, with the scan resolution set at 0.1°. The pixel time varied between 0.4 and 40 s depending on the source activity and the time allotted for the measurement. The open (granular) environmental sources were always measured in an airtight chamber (see Fig. 3) with a quartz window transmission coefficient of (91 ± 2)% in UV-A and UV-C wavelength regions. The measurements with the N2 + NO gas mixture were done at a constant pressure of (45 ± 5) × 103 hPa at the inlet, with the outlet connected to a bubbler before evacuating the gas into the lab ventilation system. All uncertainties are reported with k = 1.

In the first phase, the 4.9 kBq 239Pu source was used to evaluate the sensitivity coefficients of the scanning system for air and N2 + NO gas atmospheres corresponding to UV-A and UV-C detector configurations. It is always beneficial to assess sensitivities for local measurement conditions to account for the scintillation medium differences, whether it is pressure, humidity, gas impurities, or other factors. Measurements with N2 + NO gas are of particular emphasis since NO emissions are efficiently quenched by minute amounts of oxygen or water vapor in the chamber.

Table 2 compiles sensitivity coefficients for used measurement configurations derived with IFIN-HH reference 239Pu source as well as PTB activity standard of 210Po. The reported values include the systematic uncertainty of 10% to account for the unknown energy spectrum and angular dependence of evaluated sources. While air sensitivity values agree well for two reference sources, 239Pu-derived (local) N2 + NO sensitivity is almost 2 times lower due to the gas administration procedure: while 210Po measurements were done in a vacuum chamber with air evacuated before flashing it with N2 + NO, the IFIN-HH chamber was designed for the direct displacement of air by purging the volume with gas.

Table 2 Sensitivity coefficients for the used radioluminescence imaging system configurations derived with reference sources of 210Po and 239Pu
Full size table

In the second phase, measurements were performed on environmental samples of sand, soil, and leaves. Each material featured 3 units: high activity (code 1), low activity (code 2), and a neutral reference (see Fig. 5).

Fig. 5
figure 5

Processed radioluminescence images of environmental samples of sand, soil, and leaves measured in an airtight chamber. Raw UV images are smoothed and denoised through thresholding before overlapping with the corresponding color images. Image (a4) shows a separate scan of the lower activity leaf sample (0.67 kBq)

Full size image

Measurements of environmental samples were done in sets of 3 as displayed in Fig. 5. While all samples were tried in air, not all measurements proved successful as the scanning system was operated near its detection limit. At 40-s pixel time, the scan of the leaf set took 64 h, while soil and sand sets took 30 h each. The dark UV-A measurement of sand samples was contaminated with light entering the lab and was not repeated. The air scans of leaf and soil sets were able to resolve high-activity samples with total surface activities quantified at 1.4 and 0.3 kBq, respectively. The pixel time (i.e., PMT integration time) is usually chosen so that the measured signal at some optimal orientation of the scanner is at least 50% larger than the background (UV leakage and shot noise). The total measurement time depends on the strength of the radioluminescence signal and the resolution of the image. For example, an image with 2000 pixels takes a factor of 10 longer than one with only 200 pixels. The resolution of the image depends on the radioluminescence intensity and for samples with low activity (in the order of a few kBq) should not be less than 100 pixels.

Further scans of environmental samples, especially those below the detection limit in air, were done in N2 + NO using the UV-C configuration of the detector. With the pixel time ranging from 0.4 to 2.5 s, these measurements were significantly faster with 20 min to 2 h scan time. Also, the UV-C detector configuration allowed for conducting measurements with ambient light in the lab. The N2 + NO purging allowed quick detection of samples in each set; the low-activity leaf sample was also quantified with a separate measurement with the calculated total surface activity of just 51 Bq. The discrepancies with corresponding air measurements can be attributed to the gas flow instability.

Table 3 lists the total surface activity measured with the scanning method and compares it with the values obtained using the TDCR technique. All samples show good agreement between measurement techniques indicating the fitness of the remote imaging approach for the metrological evaluation of alpha sources. Low-activity soil and sand samples with TDCR-measured surface activities of approximately 8 Bq were below the detection limit of the used UV-C configuration. The uncertainties associated with the radioluminescence technique depend strongly on the uncertainty of the net count rate and were estimated by propagating it, along with the uncertainties of the distance to the source and properties of a scintillating gas propagating with the scanner sensitivity \(s\), and transmission uncertainties encoded in the factor \(k\), in the activity function given by Eq. (1).

Table 3 Activities of environmental samples measured gravimetrically (deposited activity), and using TDCR and imaging techniques
Full size table

In the final phase, the radioluminescence imaging method was applied to the examination of nuclear materials of various types. The first inspection was performed on a set of 3 depleted uranium samples: 2 industrial radiography collimators with the bullet in the center (see Fig. 6a). The produced radioluminescence image highlighted an alpha-active edge of the left sample; upon further investigation of the black collimator piece, the alpha-active area belonged to a polished rim section. The increased activity in the polished rim is due to damage to the protective coating, which then allows alpha particles to leak through. The total surface activity was quantified at (2.2 ± 0.3) kBq.

Fig. 6
figure 6

Processed radioluminescence images of nuclear materials samples measured in air. Raw UV images are smoothed and denoised through thresholding before overlapping with the corresponding color images

Full size image

The second inspection features a fresh natural uranium pellet. Figure 6b shows the radioluminescence profile of the pellet measured in two orientations. As with the previous source, the luminous glow is not centered with the sample but originates from the part of the pellet with a chipped surface. The calculated total surface activity of the pellet is (0.61 ± 0.13) kBq.

The anisotropic nature of the fresh UO2 pellet radioluminescence suggests that alpha particles escape through the defects in the surface. The pellet is an uncoated sintered uranium dioxide powder ceramic, which can retain noble gases in its microporous structure. Given the 238U decay chain, these factors hint at the enhanced radon emanation from defects in a sample as opposed to a normal diffusion through the uranium dioxide lattice [20].

Conclusions

In this measurement campaign, the radioluminescence mapping system developed within the RemoteALPHA project was successfully applied to the characterization of environmental samples and inspection of nuclear materials, showing the potential of remote unsupervised monitoring of alpha contaminations. The activities deduced from the radioluminescence measurement in air compare well with those measured using the triple-to-double coincidence ratio (TDCR) technique. For higher activity leaf (Leaves 1) and soil samples (Soil 1), the results agree to within 7%. In the N2 + NO atmosphere, the disagreement is larger (about 25%) predominantly due to an uneven gas flow through the chamber. Very low activity (< 164 Bq) environmental samples (Sand 1 and Leaves 2) have been quantified only in UV-C under an N2 + NO atmosphere with an agreement better than 13%.

The analysis of uranium samples has illustrated the potential application of the system towards item (source, container) integrity checks during the transfer and storage of nuclear materials. With the optical detection method, the hotspot localization and quantification can be done on a sample of arbitrary shape and size without limiting the inspection to flat external surfaces.

Radioluminescence imaging has matured into a robust method for remote localization of alpha sources that would benefit a handful of radiation monitoring applications. The satisfactory accuracy of the imaging technique with an established traceability chain brings a novel tool for the alpha metrology community, while the unsupervised and remote nature of measurement allows automating procedures in the fields of nuclear safeguards, decommissioning of nuclear facilities, and nuclear forensics. The environmental application of the radioluminescence imaging technique can be used to efficiently localize and inspect alpha contaminations in the aftermath of a nuclear fallout or a dirty bomb explosion.