1 Introduction

Global nuclear explosion monitoring is outlined in the Comprehensive Nuclear-Test-Ban Treaty (CTBT) and utilizes four verification systems: seismic, hydroacoustic, infrasound, and radionuclide monitoring. While seismic, hydroacoustic, and infrasound technologies can help establish whether an explosion has taken place, radionuclide monitoring is the only technique that can confirm whether an explosion was of nuclear origin.

The current global noble gas detection system was developed from 1995–2005, using metrics and standards from the International Noble Gas Experiment (INGE). The detection network consists of radionuclide stations around the world that collect and analyze air samples and send data to the International Data Center (IDC). The network is populated by three radioxenon detection systems: Automatic Radioanalyzer for Isotopic Xenon (ARIX), Swedish Automatic Unit for Noble Gas Acquisition (SAUNA), and Système de Prélèvement d’Air Automatique en Ligne avec l’Analyse radioXénons atmosphériques (SPALAX) (Haas et al., 2017). The signatures of interest for these systems are 131mXe, 133Xe, 133mXe, and 135Xe. These four radioxenon isotopes are important because they are produced in significant quantities in fission of both 235U and 239Pu, are highly mobile in the environment, and have long detection windows after a suspected explosion (Abdollahnejad et al., 2021; Perkins & Casey, 1996). A target detection limit of 1 mBq/m3 was set for 133Xe, but most systems routinely measure to detection limits below that value. Isotopic ratios of the four isotopes are also used for source discrimination and source attribution (Bowyer et al., 1998; Haas et al., 2017).

To supplement radioxenon monitoring for underground nuclear explosions (UNE), the isotope 37Ar has been considered a signature of interest (Aalseth et al., 2011). 37Ar is primarily produced in a UNE through the neutron activation of 40Ca in soil and rock. The high probability of occurrence of the 40Ca(n,α)37Ar reaction from neutrons released in a nuclear explosion makes it worth consideration as a signature of interest for UNE detection (Haas et al., 2010). 37Ar has a 35-day half-life, which is long enough for meaningful collection and analysis to take place after an explosion.

Using 37Ar as a detection signature in conjunction with the four xenon isotopes in atmospheric measurements could be especially useful in cases where an elevated xenon background results in significantly altered isotopic ratios. Medical isotope production is one of the key sources of the radioxenon background (Saey, 2009). Although some isotopic activity ratios from medical applications can indicate a civilian release, ratios resulting from short irradiation times complicate discrimination between nuclear explosions and civilian sources (Kalinowski et al., 2010). Supplemental measures in the form of an additional radionuclide signature could improve discrimination.

The work described in this paper complements a 2017 study by Haas et al. quantifying the performance of a network of radioxenon detectors. The prior study proposes metrics aside from minimal detectable concentration to assess current radioxenon detection systems in order to evaluate detection systems against their application—detecting nuclear explosions. Two of the metrics are detection probability and discrimination power. Detection probability refers to whether the explosion in question was detectable under the given detection limit and discrimination refers to the probability multiple isotopes were detected in one sample to allow the use of isotopic ratios (Bowyer et al., 1998). This work compares the same metrics that were explored in the original paper but modifies the analysis to be applicable to the detection of 37Ar, with a focus on the detection and discrimination metrics.

Background levels of 37Ar have not yet been thoroughly characterized. Past measurements of atmospheric 37Ar identified frequent nuclear testing as a major source, but dramatic reductions in the frequency of nuclear tests have allowed the background levels to decline (Currie & Lindstrom, 1975). Recent measurements of 37Ar concentrations in the atmosphere have shown that 37Ar background concentration could be somewhere between 1 mBq/m3 and 2 mBq/m3 (Fritz et al., 2021). Existing 37Ar in-field detection systems have detection limits between 500 mBq/m3 and 20 mBq/m3, with laboratory systems having detection limits around 0.2 mBq/m3 (Haas et al., 2010). The lowest detection limit we set for 37Ar detections in this work was 0.1 mBq/m3, which is based on a hypothetical system that:

  • increases sample collection volume by a factor of 10

  • adds many detectors to increase count time to weeks instead of hours

  • adds additional active and passive shielding to reduce background count rates by a factor of 10 compared to current above-ground 37Ar detection systems.

2 Method

A series of HYSPLIT models were used in Haas et al. (2017) to develop the data from representative releases from the Punggye-ri nuclear test site in the Democratic Peoples’ Republic of Korea (DPRK). The set of models consists of 365 10-day dispersion runs, with the simulation start date set as each day in the year 2013. HYSPLIT, an atmospheric transport modeling software, was used to track the spread of radioxenon from the release, assuming a 1 h release period (Stein et al., 2015). Archival weather data from 2013 were used, with a 0.5° global grid. The time step for this archival data was 1 h. The output data from dispersion models in HYSPLIT is a set of concentration files, which can be converted to a set of dilution factors at specified stations, given a series of latitude/longitude coordinates. The map in Fig. 1 shows the relevant radionuclide sampling stations for this study.

Fig. 1
figure 1

Map showing release point for the hypothetical release and radionuclide sampling locations (Haas et al., 2017)

Full size image

Once the dilution factor at each station for each point in time is found, it is multiplied by the expected release activity to find the expected concentration of a dispersed pollutant from the HYSPLIT data. Since the “pollutants” studied in both this and the original work are radioactive, a decay correction is also applied to the concentration. Further details about the HYSPLIT models can be found in the 2017 publication (Haas et al., 2017).

The same procedure was used in this study to find the expected concentration of 37Ar at each station for each simulation start date and sample collection date. Radioxenon dilution factors from the HYSPLIT models were converted to 37Ar dilution factors. The dilution factors were then multiplied by the release activity of 37Ar and a decay factor for each point in time to get an expected concentration. The sample collection time used in this study was 6 h because the detection limits utilized in the radioxenon coincident detection portion of the study correspond to the detection limits of Xenon International, which uses a 6 h collection period.

The potential release activity of 37Ar was calculated in a recently published study by Shah et al. (2022). First, an MCNP model with a 235U Watt fission spectrum point source surrounded by granite was used to find the representative neutron flux spectrum resulting from fission neutrons being moderated in the geologic material. The MCNP input consisted of a central sphere of 25 cm of fissioning material surrounded by concentric spheres of geological media. Then, the SCALE modules COUPLE and ORIGEN were used to collapse the neutron cross-section library using the flux spectrum from MCNP, and to generate an isotopic inventory based on the irradiation of the 40Ca in low-calcium granite through the 40Ca(n,α)37Ar reaction. The study revealed that the predicted 37Ar yield in low-calcium granite was 3.57 × 1014 Bq per kiloton. Other geologic materials will produce different activity values, as Shah et al. report. We use the value for low-calcium granite here as a representative point in the distribution of possible values.

The concentration of 37Ar at each sampling location was calculated using the predicted activity multiplied by the release venting fraction (0.1%, 1%, or 10%) and the dilution factor. If the calculated concentration at a given station and time step in the simulation was greater than the detection limit—0.1, 1, 10, 100, or 1000 mBq/m3—then the plume was considered to be detected for that combination of parameters. Each combination of venting fraction and detection limit was analyzed using this method.

A second analysis procedure was done to investigate the possibility of detecting 37Ar in coincidence with any of the four xenon isotopes. The existing xenon decay-corrected dilution factors were multiplied by the release activity of each xenon isotope and the venting fraction—0.1%, 1%, or 10% to recreate the data from the 2017 study. If both the concentration of 37Ar and the concentration of any of the xenon isotopes were above the detection limit, then the release was counted as a coincident detection. The xenon detection limits used for each isotope are outlined in Table 1 and are based on the limits used by the Xenon International detection system.

Table 1 Minimum detectable concentration (mBq/m3) for xenon isotopes for the Xenon International sampling system (Haas et al., 2017)
Full size table

2.1 Assumptions and Limiting Factors

Due to its exploratory nature in investigating the potential capabilities of a hypothetical 37Ar detection system, there are important limitations to this study.

An optimized simulation studying 37Ar, with a 35-day half-life, would use a longer dispersion run. However, since this study was conducted using previously obtained simulation results for nuclides with shorter half-lives, the dispersion run time was limited. Longer dispersion run time may increase the number of stations the plume encounters but would also allow further dilution of the plume itself. The limited dispersion run time is also more appropriate when considering coincident detections with radioxenon.

This study also has limitations due to computational constraints. We only considered one release location. A study with globally distributed release points would yield a more comprehensive look at the effectiveness of a 37Ar system. The start time of the release each day for each simulation was also kept the same for the entire year of simulations. Potential future studies could include additional release times to be more comprehensive.

There is also significant uncertainty on the release fraction for both Xe and Ar. While direct yield xenon is produced in the fireball, the xenon that is born from ingrowth from iodine may not be mobile. Alternatively, the 37Ar is all born in the initial neutron release, but some 37Ar is born in vaporized rock, while some is born in merely fractured or even solid rock. The assumptions used in the paper attempt to provide the reader with the ability to evaluate detection capabilities for various release fractions.

Regarding release timelines, we assume that the gas detected hundreds or thousands of kilometers from the explosion site will most likely result from gas that is mobile at early times and is subject to the least amount of subsurface dilution. While there are examples of releases of gas detected long after the explosion, we do not believe this to contradict the findings of this paper. We acknowledge that future work could focus on the variation in release timelines based on production mechanism.

3 Results and Discussion

3.1 37 Ar Detection

Results are presented to quantify the probability of detecting a release, the number of hypothetical monitoring stations that would detect 37Ar from an explosion, the average number of detections per release, and the possibility of detecting 37Ar in coincidence with the four xenon isotopes of interest.

Figure 2 shows the fractional probability that a release is detected by at least one station. We use the phrase fractional probability to clarify that the values displayed in the figures are not percentages. The results show that a detection system with a detection limit of 0.1 mBq/m3 has a 99% chance of detecting 37Ar from a 10 kt UNE with a 10% vent. As expected, the highest vent fraction and the lowest detection limits led to the highest probabilities of detection.

Fig. 2
figure 2

Probability that a release of 37Ar resulting from a 10 kt explosion in low-calcium granite is detected in one or more samples

Full size image

These results can also be used to help inform what detection limits a potential 37Ar system should have to get a desired detection probability. Any system with a detection limit of 1 mBq/m3 or better has at least a ~ 57% chance of detecting 37Ar from a 10 kt UNE in granite, given a 1% or 10% vent.

Figure 3 shows the probability that a specific station detects a release. Most stations had a non-zero probability of 37Ar detection with a 10% vent, meaning that the concentration of 37Ar at that station was higher than the 0.1 mBq/m3 detection limit. The stations most likely to detect 37Ar at any venting fraction for this specific release were JPP38 and RUP58, which are both close to the release point. The only station that did not have any detections of 37Ar during any of the simulations was CNP21. Lack of a measurable 37Ar concentration at this station might have been due to its location near a mountain range in northern China or due to weather conditions, mainly wind speed and direction at the time of the hypothetical release. Either of these could have stopped the plume from reaching the station and could explain the lack of detection at that location.

Fig. 3
figure 3

Probability by sampling location that a release is detected in one or more samples. This graph assumes a detection limit of 0.1 mBq/m3

Full size image

Prevailing wind patterns would explain why the detection probability at CNP20 is lower than RUP60 even though it is much closer to the test site. Plume width explains why the probability of detection for RUP58 and RUP60 are similar, even though RUP60 is much further away from the test site. It is more likely for the plume to have passed by RUP58 but spread in width as it traveled to hit RUP60.

Figure 4 shows the average number of stations detecting each release, represented by a box plot. The boxes indicate the upper and lower quartiles, the solid black line shows the median, and the small square within the box shows the average. Most combinations of detection limits and venting fractions could detect a release, with exceptions for detection limits at or above 100 mBq/m3. As expected, the median number of stations detecting a release decreased with decreasing vent fraction and increasing detection limit.

Fig. 4
figure 4

Average number of stations detecting each release

Full size image

Figure 5 shows the average number of samples with a detectable concentration. As in Fig. 4, the boxes indicate the upper and lower quartiles, the solid black line shows the median, and the small square within the box shows the average. Most combinations of venting fraction and detection limit detected a release, with the exceptions being 100 mBq/m3 or higher detection limit. The average was above the median for all detections, which indicates a positively skewed distribution of the data points. One reason for this skew could be the substantial number of outliers shown.

Fig. 5
figure 5

Average number of samples with a detectable concentration

Full size image

In addition to determining the capabilities of a 37Ar only system, the capabilities of coincident detections – detecting both 37Ar and xenon isotopes – were explored. Coincident presence of 37Ar and radioxenon could help determine whether a nuclear explosion actually took place, given that it is often difficult to discriminate with only xenon samples. The detection probabilities for 37Ar being detected along with any of the four xenon isotopes of interest are nearly identical to the detection probabilities for an 37Ar-only system, shown earlier in Fig. 2, indicating that detection of argon is the limiting factor given these release activities and detection limits. While detecting 37Ar is the more limiting factor in coincident detection, it can be useful in mitigating background issues with current xenon detection systems.

4 Conclusions

This study demonstrates the potential of 37Ar as an additional signature for nuclear explosion monitoring. With a detection limit of 1 mBq/m3, a network of 37Ar detectors would have a greater than ~ 57% probability of detecting 37Ar produced from a 10 kt UNE with the characteristics assigned in this study. Additionally, 37Ar can be detected in coincidence with radioxenon isotopes, helping to increase confidence in signature analysis. Recommendations:

  1. 1.

    A monitoring system for 37Ar should have a detection limit of 1 mBq/m3 or lower to be a complimentary signature to any of the xenon systems currently in use.

  2. 2.

    Attempts should be made to lower the detection limit to a 0.1 mBq/m3 system, which could increase the detection probability. Figure 2 shows that improving detection limits of Ar-37 detection systems from 1 mBq/m3 to 0.1 mBq/m3 would improve detection probability by almost a factor of four for well-contained tests with 0.1% noble gas vent fractions. This is assuming that background levels are below these values.

  3. 3.

    More robust measurements of current 37Ar background and good characterization of 37Ar sources are essential to be able to set a more accurate detection limit for a 37Ar system.

Improvements in Ar purification systems, particularly more efficient and quicker separation of Ar from collected gas, could increase the number of air samples purified per day, which in turn could help improve detection limits. A new system developed by Riedmann and Purstschert based on cryogenic cooling and counter-heating takes between 3–4.5 h for complete separation of argon from 80 L of atmospheric air, which means 2 samples per day could be purified (Riedmann & Purtschert, 2016). Additionally, although existing 37Ar detection capabilities are designed for On-Site Inspections, improved separation chemistry and reduced backgrounds are both possible in a fixed system.

In the future, the possibility of using 39Ar as an additional signature will be explored in a similar study. 39Ar is another isotope of interest because it can be used as a long-lived indicator of an underground nuclear explosion since it has a half-life of 269 years. 39Ar is produced through the neutron activation of 39 K in soil and rock; this is a threshold reaction with a non-zero cross section above 1 MeV.