Introduction

Beryllium-7 (T1/2 = 53.22 ± 0.06 days, [1]) and the longer-lived 10Be (T1/2 = 1.387 ± 0.012 Ma, [2]) are produced in similar quantities in the Earth’s atmosphere by cosmic-ray-induced spallation reactions, primarily on nitrogen and oxygen [3,4,5,6]. The radionuclides are attached to aerosols, transported to the Earth’s surface via wet or dry precipitation [e.g. 7], and incorporated into ice/water and attached to grain surfaces in sediments or rock.

Hence, 7Be alone, in combination with other short-lived radionuclides such as tritium [8, 9] and 32,33P [10], or the 10Be/7Be ratio can be used as a time-dependent tracer of various transport processes in the Earth’s atmosphere and the environment [3, 11, 12]. Beryllium-7 data can also decipher extraterrestrial processes such as a meteorite’s exposure history [e.g. 13, 14] or variabilities in solar properties such as activity and magnetic field [15, 16].

While the beta-emitter 10Be is routinely and nearly exclusively analysed by accelerator mass spectrometry (AMS) [e.g. 17] after radiochemical separation lasting from a few days [18] to several weeks [19, 20], the gamma-emitter 7Be is usually determined by its 477.6 keV gamma-line with or without radiochemical separation [21, 22]. AMS has been less frequently applied for 7Be quantification. Raisbeck and Yiou [11] used AMS as early as 1988 for analysing 7Be from marine sediments at the now closed French AMS-facility in Gif-sur-Yvette. Nagai et al. [23] performed 7Be-AMS in Japan (MALT) for sea water and rain and Smith et al. [24] in Australia (ANTARES) for snow pit samples, ice, and rain. More recently, 7Be-AMS was developed at the Chinese Xi’an AMS facility [25] and the AMS facility at the Lawrence Livermore National Laboratory (CAMS) for a variety of sample matrices [26]. All AMS facilities extract BeO− from a BeO-AMS-target by Cs-sputtering. Each facility stripped 7Be and its isobar 7Li bare of electrons and separated them by charge using an electrostatic analyser. Only Zhao et al. [27] developed an alternative method of separation, on samples of compressed beryllium powder and lithium metal, whereby 7BeO2+ was selected and measured in their detector. Since 7LiO2+ is unstable, the isobar is suppressed.

In continuation of our earlier approach to make 7Be easily accessible world-wide as a radiotracer for educational purposes [28], where we have successfully developed a radiochemical separation for tens-of-litre-samples of rainwater, we also explored the analytical capabilities of the DREsden Accelerator Mass Spectrometry (DREAMS) facility in Germany [17, 29]. To establish 7Be-AMS at DREAMS, validation measurements by gamma-spectrometry on large rainwater samples were performed at the Institute of Radioecology and Radiation Protection (IRS) in Hannover, Germany [28]. Determination of 7Be and 10Be by AMS in the same sample can be also used as a first quality control. Finally, smaller samples containing much lower 7Be activities were investigated by AMS to demonstrate the full analytical potential.

Experimental

Samples

Two types of rainwater collections were performed on the ground with a minimum distance of 10 m to the next building (see also Table 1):

Table 1 Sample and calibration material details
Full size table
  1. 1.

    Two large (30 L and 118 L) rainwater samples were collected at the IRS in Hannover (52°23′38.5″N; 9°42′07.8″E). The set-up, used for several sampling campaigns, consisted of a rain collector (1 m2) and polypropylene boxes, which resulted in a total collection area of 5.36 m2. For transportation and handling, the rainwater was transferred into 10 L high-density polyethylene (HDPE) storage canisters. No direct acidification in the sampling boxes was performed before transferring the water of the Hann sample (118 L) but it was done for the Hann May sample (30 L).

  2. 2.

    Small rainwater samples (18–51 g, see Table 1) were collected in a single 26 cm × 36 cm box (0.09 m2) at HZDR (51°03′44.9″N 13°57′04.5″E) and Dresden-Loschwitz (51°03′02.8″N 13°49′19.7″E), instantly acidified with a few drops of HCl (7.1 M), transferred to a 50 mL syringe and filtered through a polyvinylidenefluoride filter (pore size: 0.45 µm) into a 50 mL centrifuge tube. The collection box was rinsed with very dilute HCl before the next sample collection.

Sample preparation of rainwater samples

For small samples, about 140 µL of a dedicated low-level (in 10Be) 9Be carrier (2246 ± 11 µg/g) [30] was added to the acidified sample solution (see Table 1). The simplest and fastest chemistry was applied: (a) vortex, (b) addition of ammonia (25%) for Be(OH)2 precipitation, (c) washing one time with 5 mL pH8/9 solution (2 drops NH3aq in 250 mL H2O) to remove NH4+ and Cl−, (d) wet transfer to quartz crucible, (e) drying, (f) ignition to BeO for 2 h at 900 °C in a muffle furnace, (g) mixing with Nb powder (1:4 by weight), (h) pressing in Cu cathodes, sealed from the back with stainless steel pins. Total preparation time was 6–8 h for a batch of seven to ten samples and total costs are around 15 € per sample.

The chemical preparation of the 118 L sample is described in detail by Querfeld et al. [28]. Prior to transferring from the box, the 30 L sample was acidified to pH 1 with HCl, filtered through 2–4 µm Whatmann™ filter paper, concentrated by a rotary evaporator, and evaporated to dryness by an IR-lamp and a hot plate. The residue was exposed three times to 2.5 mL HNO3 (69%) and one time to 1.5 mL HClO4 (70%) to destroy any organics. The residue was taken up in ~ 24 g dilute HCl and 7Be determined by gamma-spectrometry. For inductively coupled plasma optical emission spectrometry (ICP-OES), a small aliquot was sacrificed. The remaining sample (96.8%) was mixed with 2.3 mg of a 9Be carrier (see Table 1). Further chemical separation of Be from other elements followed the approach of Merchel and Herpers [19] used for meteorites and involved repeated hydroxide precipitation, as well as anion and cation exchange. Final sample preparation was performed similar to the method described for small samples above. Before the final Be(OH)2 precipitation, the sample was divided into three parts.

Preparation of AMS calibration material (7Be/9Be)

Beryllium-7 as calibration material for quantification of AMS measurements was produced by proton-activation of LiF (natural isotopic composition; evaporated onto a thin catcher foil) at the Tandetron accelerator [31] of the Institute for Nuclear Research, Debrecen, Hungary, using the nuclear astrophysics beamline [32]. The proton energy was 3.030 MeV, the beam current 300 nA for an irradiation time of 10 min.

This technique is well-established and easy to perform if a cyclotron or accelerator is accessible [33]. The desired activity was pre-calculated such that, by the end of sample preparation of several 7Be-AMS targets, the 7Be/9Be ratio would be roughly similar (same order of magnitude) but slightly higher than the highest expected sample ratio. This is a common practice in AMS measurements since it greatly reduces the potential for cross-contamination in the ion source (see e.g. [34]).

An Al foil (10 µm thick) was accidentally used as the 7Be catcher foil instead of the intended, and simpler to chemically separate, Ni (7.5 µm-thick foil). This contaminated the produced BeO with Al2O3. The Al foil, containing also the freshly produced 7Be, was dissolved in 2 mL HCl evaporated to dryness for gamma-counting to accurately quantify the produced 7Be-activity at HZDR. It was then redissolved in 3 mL HCl (7.1 M), and further treated as small samples described above (but with a 10-times higher amount of a 9Be-carrier). Only about 50% of this material was finally divided into five AMS targets leaving some material as a back-up.

7Be gamma-measurements

In complete sample solutions concentrated by evaporation only and representative aliquots (~ 45%) of large rain samples, 7Be was analysed in a high-purity germanium (HPGe) detector (Ortec: GEM-40200-P; software: Genie2000) [28] at the IRS in Hannover using the characteristic 477.6 keV gamma line [1] (eff: 2% for 1 L Marinelli beaker).

For the calibration material, the dissolved Al foil (2 mL HClc) evaporated to dryness, was also analysed by gamma-spectrometry, but instead at HZDR. The evaporated sample of ~ 0.5 cm diameter was situated in a plastic vial atop the detector end cap, centrally in the symmetry axis of the detector resulting in an effective distance of the sample material to the detector end cap of ~ 1.5 cm. The n-type HPGe detector (60% relative efficiency) was placed in a graded shield setup including 15 cm thick lead shielding [35]. The gamma-ray detection efficiency in the measurement geometry was (7.76 ± 0.23)%, determined by a calibrated point-type 7Be source (earlier measured in far geometry (21.5 cm) using 137Cs and 133Ba activity standards from the Physikalisch-Technische Bundesanstalt (PTB)), which was placed in a plastic vial equal to the one used for the sample. The 477.6 keV peak contained (161,400 ± 400) counts after 20.2 h. By using the half-life of (53.22 ± 0.06) days and a branching ratio of (10.44 ± 0.04) % for the 7Be decay to the first excited state of 7Li [1], the number of 7Be nuclei was determined to be (1.84 ± 0.05) × 109 (equivalent to 277 Bq), referred to 1 June 2017 12:36 h CEDT.

AMS measurements (7Be and 10Be)

AMS measurements were performed at the 6 MV Tandem accelerator at DREAMS [17, 29]. The AMS targets (BeO) are loaded into a carousel where a piston injects them automatically into the ion source. In the ion source, the sample material is bombarded by Cs+ ions, sputtered, and a beam of negative BeO− ions is extracted. The sputtering process is destructive, thus, care must be taken to ensure sufficient material is used for each radioisotope of interest. The negative ion beam reaches the bouncer magnet, which sequentially selects the isotope to be injected into the tandem by mass.

The 10Be/9Be ratios were measured first using Fast Sequential Injection (FSI) detailed in [17, 29, 36]. The tandem voltage was chosen so as to maximize the 2+ charge state yield of Be isotopes. Due to the available positions of the offset Faraday cups on the high-energy (HE) side, simultaneously measuring the stable 9Be (at the Faraday cup) and 7Be through the reference path, at the fixed terminal voltage, is not possible. To circumvent this, the isotopes were measured using software called Slow Sequential Injection (SSI) obtained from High Voltage Engineering Europa B.V. and used e.g. at the AMS facility in Seville, Spain [37].

In SSI mode, 7BeO− ions were injected by the low-energy bouncing magnet into the tandem accelerator at a terminal voltage of 4.728 MV. At this voltage, the production of 2+ charge state is dominant (~ 70%) [38]. The stripped 7Be2+ ions were bent by the HE-magnet down the reference path and stripped to the 4+ charge state by a 3.1 µg/cm2 silicon nitride foil. At constant mass of 7 amu, the charge state of 4+ is selected using the electrostatic analyser (ESA) in order to suppress the isobar and decay product of 7Be, 7Li, since the charge state of 4+ does not exist for 7Li.

For the measurement of 9Be, the bouncer magnet voltage was changed to allow 9BeO− into the tandem accelerator, which was set to terminal voltage of 3.2 MV leading to a charge state yield of ~ 68% for 9Be2+. The tandem voltage was changed to allow the bending of the 9Be2+ into the movable Faraday cup following the HE-magnet while keeping the magnetic field fixed.

Each SSI sequence involved the injection of 7BeO− (7Be16O−) for 40 s and 9BeO− (9Be16O−) for 2 s. Unfortunately, due to a timing flaw in the setup of the measurement, the bouncer and tandem voltages were changed in such a way that they briefly scattered high currents of particles from the tandem into the gas ionization chamber (GIC) at each transition. As a quick fix to preserve the foils and the GIC itself, two additional steps (2 s each) were set up to offset the changing of the bouncer voltage from the changing of the terminal voltage. Finally, the software added a default pause of 5 s between each step. Thus, the steps of a single sequence follow: 2 s 9Be, 5 s pause, 2 s “dummy1”, 5 s pause, 40 s 7Be, 5 s pause, 2 s “dummy2”, 5 s pause.

With this approach, the cathode was being sputtered, consuming the BeO material, for 36% of the time with no data recorded. In the end, the radioisotope of interest reaches the GIC as 7Be4+ and is counted atom by atom.

We are aware that in general the SSI mode is inferior to the FSI mode with respect to data precision. The SSI mode is susceptible to larger uncertainties resulting from potential drifts in the Be current resulting from variations in ion output from the sample itself and transmission through the machine set-up. However, this uncertainty is yet outweighed by the one from counting statistics.

In small samples and aliquots of large samples (~ 40%), 7Be and 10Be were quantified versus the aforementioned 7Be calibration material and the in-house 10Be standard SMD-Be-12 [17], where the latter is traceable to NIST4325 [39].

AMS measurement of 7Li to test chemical and AMS isobar suppression

Because we reduced the chemical sample separation to a minimum for all small volume samples and also for the calibration material (containing macroscopic amounts of LiF), we wished to ensure that the isobar 7Li was sufficiently depleted. To compare 7Li concentrations in BeO prepared from all sample materials we allowed both 7Li3+ and 7Be3+ into the GIC to separate them via their difference in stopping power. At the same time, we checked whether separation of 7Li in the GIC is already sufficient or whether full separation of 7Li from 7Be using additional stripping to the 4+ charge state is required. As for the case described above, 7 amu and 2+ charge state ions from the accelerator were bent into the reference path by the HE-magnet. Both passed through the degrader foil acquiring a distribution of charge states. The ESA and final 30° magnet were tuned to accept the expected energy of 7Li3+ into the GIC. Each of these steps were performed with most of the apertures in—transmitting only near-axis components—to reduce the abundant Li signal. A few BeO samples from each sample type (some small Dresden and both large Hannover rainwater samples), as well as the LiF calibration material (Table 1), some BeO blanks, and graphite pressed in Al and Cu sample holders, were studied in this manner.

Results and discussion

Chemical and AMS separation of the isobar 7Li

The results of the relative isobaric abundances in BeO from various samples (small and large rainwater samples, LiF calibration material, blanks) together with graphite pressed in Al and Cu, respectively, are presented in Fig. 1.

Fig. 1
figure 1

Relative abundance of 7Li in BeO from different materials, carbon pressed in Al and Cu, respectively, and BeO blanks

Full size image

The objectives of these investigations were: (1) proving whether or not 7Be can be separated from its isobar 7Li in the detector, and (2) to test the rapid chemical separation protocol for its performance.

Firstly, we found that separation by stopping power in the gas ionization detector was clearly inferior to separation by charge state since the ion beam had to be severely attenuated (with most apertures in) to be at a level where the 7Li could be introduced into the detector without damaging it. Since separation by charge states was proven and superior with respect to source-to-detector efficiency, there was no further investigation on this modality.

Secondly, the small volume samples and the blanks processed with fast and cheap chemical processing had roughly the same amount of 7Li as the larger samples that were treated with more lengthy procedures. Thus, for the purpose of reducing the concentration of the isobar present in each target to a level not harming the silicon nitride foil, the simple chemistry on small samples is sufficient.

AMS results (7Be and 10Be) for small and large rainwater samples

The AMS results are presented in Table 2 and Figs. 2 and 3. All 7Be data was corrected to the collection time. The 7Be/9Be ratios are more a confirmation of sample preparation procedure than a meaningful result since their value depends mostly on the volume of sample collected and the amount of 9Be carrier added. The 7Be/9Be ratios confirm that the activity of the reference material was well-prepared to suit the purpose having similar (slightly higher) ratios for calibration materials than samples to be analysed.

Table 2 AMS isotopic ratios and 7Be concentration at time of collection
Full size table
Fig. 2
figure 2

Concentrations of 7Be (left) and 10Be (right) of rainwater water samples. Note samples Drs 05_05, Drs 05_06, Drs 2, and Drs 3 were collected at the start of rainfall containing a larger amount of dust. Drs 5 was also collected at the start of its rainfall but it had rained the night before, likely depleting the air of particulate matter. *Both Hannover samples contain excess 10Be and are depleted in 7Be

Full size image
Fig. 3
figure 3

10Be/7Be isotope ratios of all rainwater samples (left) and of small samples from Dresden only (right). *Both Hannover samples contain excess 10Be, and Hann was also depleted in 7Be

Full size image

The isotopic ratios of the AMS measurements and the activities provided by the gamma-counting measurements were converted to 7Be atoms g−1 at the respective collection times. Hann was measured to be (4.14 ± 0.17) × 103 atoms g−1 and (4.37 ± 0.23) × 103 atoms g−1 and Hann May was measured to be (24.68 ± 0.95) × 103 atoms g−1 and (23.44 ± 0.76) × 103 atoms g−1 by gamma-counting and AMS, respectively. Thus, the AMS results are in excellent agreement with the gamma-counting validation measurements.

Figure 2 shows both the concentrations of 7Be and 10Be in atoms per gram of rainwater. The 7Be values varied relatively little for all of the samples that were collected late in a rainfall; e.g. 8–10% deviation from their mean 7Be value for Drs 0, Drs 1, Drs 4, Drs 6. However, five samples (Drs 05_05, Drs 05_06, Drs 2, Drs 3, Drs 5) were all collected at the very beginning of their respective rainfall and, with the exception of Drs 5, show an elevated amount of 7Be and 10Be. This effect is more prominently seen in the 7Be, which would result almost exclusively from “recent” atmospheric production whereas some 10Be may be present as old, re-suspended dust causing an elevated background level of 10Be. Both radionuclides suggest scavenging and enrichment in dust for rain collected at the beginning of a rainfall. Drs 5 does not follow this trend, likely because it had rained the previous night, which washed out most of the dust in the air before this rainfall collection.

Beryllium attaches to (container) surfaces at pH levels ≥ 5. Because the Hannover sample containers were used for repeated rain collections without the (crucial) acidification, the consequences for the Hann samples were: (1) Some Be remaining in the boxes, i.e. non-quantitative transfer, resulting in 10Be and 7Be concentrations that were too low (but correct 10Be/7Be), and (2) Short-lived 7Be decaying in the boxes while waiting for the next rain sample (up to several weeks, see collection time in Table 1). The 7Be-depleted Be will be partially leached from the boxes for the next sample. This results in too low 7Be relative to 10Be, i.e. too high 10Be/7Be. It has to be mentioned that sample Hann was collected in several portions (also in many boxes, each of them having absorbed Be), thus, the first consequence adds for Hann. The second consequence affects both, the Hann sample and the next sample, Hann May. Both Hannover samples contain excess 10Be and Hann was additionally depleted in 7Be (Table 2).

An important lesson to be learned from these observations is that acidification (pH < 5) of the rainwater before transfer is highly recommended in order to overcome the above mentioned consequences. In addition, cleaning of the sampling containers with slightly acidic water after each sampling campaign is crucial. Beryllium will also be adsorbed on dust and organic matter (e.g. leaves, pollen, insects) in the sample, thus, acidification before any filtration is extremely important. Since the sampling boxes had not been cleaned with acidic water before both Hann and Hann May were collected, and the acidification for Hann was also omitted before filtration, the 7Be and 10Be values of both were found to be unreliable for the purpose of this study and cannot be used for further interpretation. It should be noted, though, that the original intention of the initial project, detailed in Querfeld et al. [28], was to collect 7Be only. Thus, no attention had been paid to remove residual 10Be or 7Be-depleted Be between (fractions of the Hann) samples. Cleaning steps to remove previous beryllium from the sample collection container used in Dresden were performed as the study goal developed further.

The 10Be/7Be ratios (Table 2; Fig. 3) of rainwater samples collected in Dresden are in agreement with each other, range from 1.2 to 2.5, and are consistent with values presented by other labs. For rainwater samples, others report 10Be/7Be ranges of 1.5–2.9 [11], 1.4–2.9 [12], 1.3 [25], and 1.7–4.1 [26]. In air filter samples, Zanis et al. [40] and Yamagata et al. [41] report ranges of 1.5–2.7 and 1–2, respectively. The effects of the previously discussed deviations in the collection procedure in Hannover are obvious in this plot and illustrate the importance of acidifying the rainwater to prevent adhesion to the collector walls as well as rinsing the collectors with acidified water between sampling campaigns for accurate 7Be and 10Be concentrations and 10Be/7Be ratios.

Potential for lower 7Be AMS uncertainties and lower detection limits

The total uncertainty of 7Be AMS data of all rain samples is 6–7% with the exception of Drs 5 being at 22%. In contrast to all other samples (chemical yield for BeO ≥ 94%), the chemical yield for BeO for Drs 5 was inexplicably very low (21%), hence, it produced low ion currents and was quickly consumed. Drs 05_05 and Drs 05_06 were measured for a short time during the first feasibility test runs yielding higher uncertainties. Due to insufficiently low count rates in a later beam time, they were not re-measured. In all cases, the uncertainties were dominated by counting statistics. An improvement in the switching time of the SSI-mode measurements can only improve this. We are working with the developers of the proprietary SSI-mode software to eliminate the need for the aforementioned “dummy” steps. This as well as optimizing all other switching times will maximize the amount of counting time, and therefore number of counts detected on a given sample, reducing the relative uncertainty. The 3% uncertainty of the measurement of the calibration material by decay counting (for 0.8 day) provided an additional but small systematic contribution. If necessary, it is expected that this can be reduced to 1% by measuring longer and in an underground facility [42]. The counting background, mainly due to cosmic ray-induced effects, can in principle be further suppressed by two or three orders of magnitude, using e.g. the new HZDR underground facility Felsenkeller [43]. Last but not least, reducing the time between sample collections, chemical processing, and AMS measurements will also improve statistics as more short-lived 7Be has not yet decayed and is still available for the actual measurement.

Our detection efficiency for 7Be (including the negative ion yield) from the target to the detector is 2.4 × 10−4. Our blank value for 7Be/9Be (5 × 10−16) is comparable to AMS blank values elsewhere [e.g. 23, 24, 26]. Decay counting has been reported for 7Be samples with activities as low as 25 mBq [44], however such low activities require counting times of many days (e.g. 16 days). For activities lower than this, a considerable amount of time must be invested to measure with reasonable levels of uncertainty [23]. A detection limit of ~ 7 mBq can be reached within unreasonable counting times of 100 days at the Felsenkeller [43]. A value of about 60 mBq, enriched by filters collecting for 24 h air volumes of 500–1000 m3/h, is documented from investigations of 7Be concentrations in air as measured by the CTBTO (Comprehensive Nuclear-Test-Ban Treaty Organisation) global monitoring system [45]. Our AMS measurements push this threshold to 0.6 mBq, which is at least one-to-two orders of magnitude more sensitive and requires only very simple and fast chemical preparation as well as much smaller samples. The efficiency can improve by removing the aforementioned “dummy” steps and by optimizing the tuning/machine setup (could improve by a factor of three), which is also expected to reduce our uncertainty.

Conclusions and outlook

Measurements of 7Be and 10Be have proven feasible at DREAMS on the same BeO prepared from rainwater samples. Simple, cheap, and fast chemistry to prepare BeO-AMS targets from small rainwater samples worked very well when special attention was paid to acidifying rainwater before transfer from collection boxes and filtration takes place. The overall costs (human resources and chemical products) and speed for chemical separation had been reduced making 7Be-AMS very compatible to e.g. counting techniques (without or including radiochemical enrichment) for routine analyses. Especially, for projects with high sample through-put AMS can be superior over long-lasting counting.

Total uncertainties of sample measurements were usually 6–7%. Validation measurements by gamma-counting of the two large rainwater samples were in excellent agreement with our AMS results. The detection limit for 7Be measurement at DREAMS is 0.6 mBq, which is one-to-two orders of magnitude better than “standard/ordinary” and “sophisticated” decay counting (e.g. in an underground laboratory). Both the limit and uncertainty can be improved by more precise decay counting measurements of the calibration material, the removal of so-called “dummy” steps currently required by the switching software, and better tuning conditions.

Our 7Be and 10Be data showed that the very first rain (< 5 min) collected was enriched in particulate matter. Thus, AMS optimised for small samples has the potential for time evolution studies of rain using 7Be and 10Be as a natural tracer. Additionally, the low detection limit and the high sample throughput will enable future studies where high-precision measurements of small timescale phenomena, thus, small sample volumes are of interest.