INTRODUCTION
The Centre of Accelerator Mass Spectrometry at the University of Cologne (CologneAMS) recently installed a second ion source (SO-110 B; Stolz et al. Reference Stolz, Dewald, Altenkirch, Herb, Heinze, Schiffer, Feuerstein, Müller-Gatermann, Wotte, Rethemeyer and Dunai2017) from High Voltage Engineering Europa B.V. (HVE, The Netherlands) at the HVE 6 MV Tandetron accelerator, which is used for radiocarbon (14C) analysis of CO2 gas. Until now, all samples were converted to elemental carbon using the automated graphitization equipment AGE-2 (Ionplus AG, Switzerland). Samples smaller than 200 µg C and older than about 25,000 yr BP were not measured so far, because we were not able to produce reliable results (Rethemeyer et al. Reference Rethemeyer, Fülöp, Höfle, Wacker, Heinze, Hajdas, Patt, König, Stapper and Dewald2013). Increasing demands for small-scale 14C analysis of samples such as tiny macrofossils, organic molecules isolated with chromatographic methods, and CO2 samples required the establishment of AMS gas ion source 14C analysis at CologneAMS allowing the analysis of samples as small as 2.5 µg C.
The SO-110 B type ion source was first coupled with the gas injection system (GIS) from Ionplus AG, which introduces CO2 provided in glass ampoules (GIS-AMS). Tests of this system with 14C-free CO2 from gas bottles yielded a constant contamination of about 0.002 µg C. Blank values for samples >20 µg C correspond to 0.002 F14C (Stolz et al. Reference Stolz, Dewald, Heinze, Altenkirch, Hackenberg, Herb, Müller-Gatermann, Schiffer, Zitzer, Wotte, Rethemeyer and Dunai2019). Secondly, an EA3000 elemental analyzer (EA; EuroVector, Italy) was coupled to the GIS (EA-GIS-AMS), in order to reduce pretreatment steps and the time-consuming gas purification required for producing CO2 samples in glass ampoules for GIS-AMS and to make small-scale 14C analysis more efficient. Hence, samples only need to be packed into EA combustion vessels and are consequently combusted into CO2. As soon as the system detects the CO2 peak, the gas is streamed into a zeolite trap with helium (He). After that, the trap is heated and the released CO2 is directly injected into the AMS using the GIS (Stolz et al. Reference Stolz, Dewald, Heinze, Altenkirch, Hackenberg, Herb, Müller-Gatermann, Schiffer, Zitzer, Wotte, Rethemeyer and Dunai2019). First tests of this coupled EA-GIS-AMS setup offered very promising results using IAEA reference standards (Stolz et al. Reference Stolz, Dewald, Heinze, Altenkirch, Hackenberg, Herb, Müller-Gatermann, Schiffer, Zitzer, Wotte, Rethemeyer and Dunai2019).
In this study, we conducted extensive tests to gain detailed information about the contamination introduced during the analysis of CO2 samples using a) the GIS-AMS and b) the EA-GIS-AMS system. The relative importance of exogenous C contribution and its variability are increasing with decreasing sample size (e.g. Rethemeyer et al. Reference Rethemeyer, Fülöp, Höfle, Wacker, Heinze, Hajdas, Patt, König, Stapper and Dewald2013; Ruff et al. Reference Ruff, Szidat, Gäggeler, Suter, Synal and Wacker2010b). Therefore, we determined the lower sample size and age limits of GIS-AMS by measuring multiple series of standards ranging from 2.5–50 µg C. For EA-GIS-AMS analysis, we additionally compared the contamination introduced during sample combustion using different types of vessels made of Sn and Ag.
METHODS
Pretreatment Methods for GIS-AMS Analysis
For the evaluation of lower size limits and blank values of the GIS-AMS system, different size series of a 14C-free (Pocahontas, POC #3, Argonne Premium Coal) and of a modern standard (Ox-II; NIST SRM 4990C; nominal value 1.3407 F14C) were prepared using sealed tube combustion. Outmost care was taken not to introduce any contamination during the preparation of the standards. The quartz glass ampoules (9 mm OD) for sample combustion and borosilicate glass tubes (4 mm OD), in which purified CO2 samples are collected, were first visually checked for damages and then washed three times with Milli-Q water (Millipore, USA) and three times with acetone (PESTINORM® SUPRA TRACE, grade ≥99.9%, VWR® chemicals, Germany) to remove organic contaminants. The quartz ampoules, cupric oxide (CuO rods, 0.65 × 3 mm, p.a., MERCK KGaA, Germany), and Ag wires (MERCK KGaA, Germany) were pre-combusted at 900°C for 4 hr, while the borosilicate glass tubes were combusted at 450°C for 4 hr. Then the sample was filled into the cleaned quartz tubes using Sn boats for weighing and transferring together with the CuO (CuO:C ratio of about 60:1) and 5 mg Ag wires (Santos and Xu Reference Santos and Xu2017). The ampoules were evacuated (P < 10-4 mbar), flame sealed, and heated in a muffle furnace to combust the standard material to CO2 (900°C, 4 h). The gas was purified on a vacuum rig using a water trap (ethanol dry-ice slurry) and the CO2 was trapped in a borosilicate glass ampoule (4 mm OD) placed in liquid nitrogen and was finally flame sealed (Wotte et al. Reference Wotte, Wordell-Dietrich, Wacker, Don and Rethemeyer2017).
Pretreatment Methods for EA-GIS-AMS Analysis
In order to determine blank values of different EA containers, we measured 14C-free standards (POC #3) in a size range of 3.5–50 µg C. We tested Sn boats (4 × 4 × 11 mm), Sn capsules (3.5 × 9 mm) as well as Ag capsules (5 × 9 mm), all manufactured by Elementar (Germany). Prior to sample combustion, the containers were washed three times with dichloromethane (DCM; SupraSolv®, MERCK KGaA, Germany). The Ag capsules were combusted at 450°C (4 hr) in addition to DCM washing. We omit cleaning in an ultrasonic bath using organic solvents as handling experience showed that vessels become brittle.
System Overview
To enable 14C analysis of CO2 at the HVE AMS system at CologneAMS, operation parameters of the SO-110 B type ion source were first optimized with the objective to maximize a stable C– current output (maximum 12 ± 1 µA, 6 ± 2% C– yield) (Stolz et al. Reference Stolz, Dewald, Altenkirch, Herb, Heinze, Schiffer, Feuerstein, Müller-Gatermann, Wotte, Rethemeyer and Dunai2017). In addition, a new gas-injection-control-software (GICS) controlling data acquisition and hardware was developed that substitutes the software from Ionplus AG (Stolz et al. Reference Stolz, Dewald, Heinze, Altenkirch, Hackenberg, Herb, Müller-Gatermann, Schiffer, Zitzer, Wotte, Rethemeyer and Dunai2019). The GIS sample magazine was also modified to be able to process 16 instead of previously 8 ampoules in one batch, which allows the extension of the autonomous measurement period.
Before starting an AMS measurement, all capillaries are flushed several times with He and a new target is loaded prior to sample CO2 transfer (Stolz et al. Reference Stolz, Dewald, Heinze, Altenkirch, Hackenberg, Herb, Müller-Gatermann, Schiffer, Zitzer, Wotte, Rethemeyer and Dunai2019). Using EA-GIS, the CO2 is transferred from the EA via a 13X zeolite trap into a syringe where it is mixed with He (mixing ratio CO2:He 5%, Stolz et al. Reference Stolz, Dewald, Altenkirch, Herb, Heinze, Schiffer, Feuerstein, Müller-Gatermann, Wotte, Rethemeyer and Dunai2017). For GIS-AMS, CO2 samples in glass ampoules are cracked in the ampoule cracker of the GIS, mixed with He, and injected into the ion source via a syringe (flow rate 1.4 µg C min–1, Stolz et al. Reference Stolz, Dewald, Altenkirch, Herb, Heinze, Schiffer, Feuerstein, Müller-Gatermann, Wotte, Rethemeyer and Dunai2017).
Quantification of Contamination
The amount of exogenous C introduced during sample preparation and analysis can be quantified using the model of constant contamination as given by Hanke et al. (Reference Hanke, Wacker, Haghipour, Schmidt, Eglinton and McIntyre2017). It is based on the assumption that the measured F14C (FM) and its mass (mM) are composed of the values of the standard itself (FStd, mStd) and of contributions from a constant exogenous contamination (FC, mC). As mC cannot be determined directly, it is commonly assumed that the contamination is composed of two end-members, namely a 14C-free (F14C = 0) and a modern (F14C = 1) proportion (e.g. Santos et al. Reference Santos, Southon, Griffin, Beaupre and Druffel2007; Lang et al. Reference Lang, Früh-Green, Bernasconi and Wacker2013; Gierga et al. Reference Gierga, Schneider, Wiedemeier, Lang, Smittenberg, Hajdas, Bernasconi and Schmidt2014). The resulting F14Cc therefore depends on the contribution of each endmember between 0 and 1. In order to determine the respective proportions, size series of 14C-free (POC #3) and modern (Ox-II) standards were prepared and measured. For technical reasons, we only apply this model to POC#3 and Ox-II series analyzed with the GIS-AMS, but not to EA-GIS-AMS data consisting of only 14C-free standard material. The amount of exogenous C (mC) was determined by fitting the two independent parameters (FM and mM) using the Matlab script published by Haghipour et al. (Reference Haghipour, Ausin, Usman, Ishikawa, Wacker, Welte, Ueda and Eglinton2019).
RESULTS AND DISCUSSION
GIS-AMS Measurements
In order to identify and quantify contamination introduced during sample preparation for CO2 analysis with GIS-AMS, we prepared and measured four size series of 14C-free (POC #3, Figure 1A, D) and two series of modern standard material (Ox-II, Figure 1B). Three of these size series of POC #3 (total n = 44) and two of the Ox-II series (total n = 27) were prepared by weighing defined amounts of standard material into individual quartz ampoules (Figure 1A and B). For the fourth 14C-free series (n = 14), standard material equivalent to ~1 mg C was combusted in one tube and split into aliquots (Figure 1D) with sizes between 50 µg C and 2.5 µg C.
F14C vs. mass C [µg] measured by GIS-AMS: A) for 14C-free (POC#3) and B) modern (Ox-II) standard material (red circles). Fitting results are displayed as blue lines with 1-σ uncertainty (dashed blue lines). The fit assumes an error margin of 25%, outliers are displayed in black. C) shows the modelled solution space of χ² (indicated by colors) that depends on both contamination pools. The small circle indicates one possible solution for the fitting procedure as displayed by F14Cc and mc. D) displays the series of aliquots produced from one large sample, which has not been assessed using the χ² procedure, because only 14C-free standards have been prepared as aliquots.
The 14C results of the modern standard (Ox-II) do not show a distinct size dependency but rather scatter within 1-σ uncertainty around the consensus value (Figure 1B). Contamination assessment using the Matlab script from Haghipour et al. (Reference Haghipour, Ausin, Usman, Ishikawa, Wacker, Welte, Ueda and Eglinton2019) for the Ox-II standards revealed a best fit when assuming minuscule amounts (0.02 µg C) of fossil exogenous C. Based on the used mixing model, we assume that mostly modern contamination is introduced during sample preparation (Figure 1C).
The 14C results of the three blank series weighed directly into quartz ampoules are in a similar range with an average of 0.0050 ± 0.0004 F14C (n = 4) for 50 µg C and increase in 14C concentration at 20 µg C (0.0163 ± 0.0015 F14C, n = 3, Figure 1a), as smaller samples are increasingly affected by the constant contamination than larger ones. In contrast, the series of aliquots prepared from one large sample gives much better blank values for smaller sample sizes of 2.5–20 µg C between 0.0029 ± 0.0007 F14C (20 µg C) and 0.0178 ± 0.0079 F14C (3.5 µg C; Figure 1D), which indicates that the contamination is mainly introduced during the pretreatment including the weighing and transfer of samples/standards into the quartz ampoules and the subsequent combustion as also observed by Santos et al. (Reference Santos, Southon, Drenzek, Ziolkowski, Druffel, Xu, Zhang, Trumbore, Eglinton and Hughen2010) and Santos and Xu (Reference Santos and Xu2017). The constant contamination for the total data set (n = 44 POC #3 and n = 27 Ox-II standards) consists of 0.30 ± 0.08 µg C (Figure 1C). Based on our model, the corresponding isotopic ratio is F14Cc = 0.93 ± 0.23 of which 0.28 µg C is modern and 0.02 µg C fossil contamination.
Figure 2 shows the result of the total GIS-AMS data set (A: POC#3 blanks, n = 44; B: Ox-II, n = 27) that was corrected for the calculated contamination given in Figure 1C. The data scatter close to the respective consensus values after application of the correction for contamination.
Combined data set of A) 14C-free (POC#3) and B) modern (Ox-II) standard size series prepared in quartz ampoules for GIS-AMS analysis. Both sets have been corrected by the calculated contamination given in Figure1C and scatter around consensus values (solid black lines).
The constant contamination derived from the GIS system only, which was determined by directly injecting CO2 from gas bottles into the AMS ion source, is 0.012 µg modern C (1–23 µg C, Stolz et al. Reference Stolz, Dewald, Heinze, Altenkirch, Hackenberg, Herb, Müller-Gatermann, Schiffer, Zitzer, Wotte, Rethemeyer and Dunai2019). Thus, sample preparation and combustion introduce most, about 0.30 ± 0.08 µg C, exogenous C.
EA-GIS-AMS Measurements
For 14C analysis using EA-GIS-AMS, samples were weighed into containers for combustion in the EA. We investigated the contamination derived from different EA containers in three series using 14C-free material (POC #3) combusted in Sn boats (n = 7), Sn capsules (n = 7), and Ag capsules (n = 6) (Figure 3).
Overview of EA-GIS-AMS results for size series of blank material (POC#3) combusted in Sn boats (circle), Sn capsules (triangle), and in Ag capsules (square).
So far, only 14C-free standards have been tested using the EA-GIS-AMS method due to a pending replacement of the EA and new installation of an isotope ratio mass spectrometer (IRMS). In order to quantify the amount of exogenous C using the contamination mixing model for this approach, we use a similar F14Cc of 0.93 like for GIS-AMS. This assumption is made to make results qualitatively comparable to other EA-GIS-AMS studies. We use this as a conservative estimate, because GIS-AMS sample handling involves more handling steps compared to EA-GIS-AMS, including the weighing of samples in Sn boats, the transfer of sample material from Sn boats into quartz ampoules as well as the handling and re-sealing on the vacuum line into a glass ampoule. More detailed investigations are planned once the new systems are operating.
In addition, we tested different EA sample containers including Sn boats and capsules cleaned with DCM, which removes contamination more efficient than Acetone or Milli-Q water (Welte et al. Reference Welte, Hendriks, Wacker, Haghipour, Eglinton, Günther and Synal2018; Ruff et al. Reference Ruff, Fahrni, Gäggeler, Hajdas, Suter, Synal, Szidat and Wacker2010a), and Ag capsules cleaned by heating at 450°C. To directly compare the results for the Sn boats and Sn capsules, we calculated the amount of exogenous C for each series using the parameters obtained from the combined ampoule data set (F14Cc = 0.93 ± 0.23) (Table 1). Blank values for Sn boats are 0.0127 ± 0.0012 F14C for 50 µg C and up to 0.0877 ± 0.0172 F14C for 3.5 µg C (mc = 0.36 ± 0.09 µg C). Sn capsules have slightly better blank values of 0.0090 ± 0.0010 F14C (50 µg C) and 0.0593 ± 0.0093 F14C (3.5 µg C; mc = 0.37 ± 0.09 µg C). Sn boats and capsules thus introduce similar amounts of contamination. Our results determined with blank material combusted in Sn containers are in a similar range to results of other studies, which however combusted empty Sn containers including Welte et al. (Reference Welte, Hendriks, Wacker, Haghipour, Eglinton, Günther and Synal2018; mc = 0.58 ± 0.04 µg C for DCM-washed Sn boats and mc = 0.30 ± 0.03 µg C for DCM-washed Sn capsules) and Haghipour et al. (Reference Haghipour, Ausin, Usman, Ishikawa, Wacker, Welte, Ueda and Eglinton2019; mc = 0.28 ± 0.08 µg C, F14C 0.6 ± 0.2).
Summary of blank assessment for 50 µg and 3.5 µg C large samples measured with GIS-AMS and EA-GIS-AMS analysis. The quantification is based on 14C-free standard material (POC#3) combusted in quartz ampoules (GIS) or in EA Sn containers using the Matlab script by Haghipour et al. (Reference Haghipour, Ausin, Usman, Ishikawa, Wacker, Welte, Ueda and Eglinton2019). We assume a similar F14C for GIS-AMS as determined for EA-GIS-AMS.
*1(n = 4); *2(n = 3), *3based on GIS-AMS data evaluation.
We have so far not tested heat treatment (550°C) of DCM-washed Sn vessels, which can further reduce contamination levels (Welte et al. Reference Welte, Hendriks, Wacker, Haghipour, Eglinton, Günther and Synal2018). The size series combusted in Ag capsules gave 14C concentrations, which are about one order of magnitude higher, between 0.1921 ± 0.0293 – 0.5492 ± 0.0322 F14C, than for the Sn containers and the data show no size dependency but scatter strongly (Figure 3). This result indicates a much higher and more variable contamination originating from the Ag capsules, which was likely introduced to the activated surface of the capsule after heating, as suggested by Welte et al. (Reference Welte, Hendriks, Wacker, Haghipour, Eglinton, Günther and Synal2018) and Haghipour et al. (Reference Haghipour, Ausin, Usman, Ishikawa, Wacker, Welte, Ueda and Eglinton2019).
Based on the results of our tests, we now chose for our laboratory routine Sn boats cleaned three-times with DCM. However, we plan to perform further tests with Ag capsules combusted at higher temperatures (800°C) as suggested by Fewlass et al. (Reference Fewlass, Talamo, Tuna, Fagault, Kromer, Hoffmann, Pangrazzi, Hublin and Bard2018) and Al containers that can be combusted at temperatures of up to 550°C, which may be an alternative to the usage of solvent-cleaned Sn suggested also by Welte et al. (Reference Welte, Hendriks, Wacker, Haghipour, Eglinton, Günther and Synal2018).
CONCLUSION
In this study, we investigated the contamination introduced during sample preparation for gas ion source 14C analysis using our HVE 6 MV Tandetron accelerator coupled with the Ionplus AG GIS and an EuroVector EA3000. We prepared various size series of standards ranging from 2.5–50 µg C in glass ampoules for 14C analysis using GIS-AMS and in different EA combustion containers for EA-GIS-AMS in order to quantify exogenous C contributions introduced, applying the model of constant contamination. The constant contamination introduced during sample preparation for GIS-AMS analysis is 0.30 ± 0.08 µg C (F14Cc = 0.93 ± 0.23; 0.28 µg modern C, 0.02 µg fossil C). Blank values increase from 0.0050 ± 0.0004 F14C for 50 µg C to 0.0163 ± 0.0015 F14C for 20 µg C.
For EA-GIS-AMS analysis, the samples need to be weighed into EA combustion vessels, which introduce different amounts of contamination depending on the material and type of the vessel itself and on the cleaning procedure. Lower blank values were obtained for 14C-free material combusted in DCM-cleaned Sn boats (0.0127 ± 0.0012 F14C for 50 µg C and up to 0.0877 ± 0.0172 F14C for 3.5 µg C; mc = 0.36 ± 0.09 µg C) compared to DCM-cleaned Sn capsules (0.0090 ± 0.0010 F14C for 50 µg C and 0.0593 ± 0.0093 F14C for 3.5 µg C; mc = 0.37 ± 0.09 µg C). DCM-cleaned and heated Ag capsules produced the highest blank values indicating that either combustion at 450°C is not removing contamination sufficiently or that the surface of the containers is activated by heating and adsorbs contamination. F14C results of EA-GIS-AMS are in similar ranges compared to our GIS-AMS results, except for the Ag capsule series, which have a higher and variable 14C content (0.1921 ± 0.0293 – 0.5492 ± 0.0322 F14C).
With our study we demonstrate that small-scale 14C analysis of gaseous samples using the HVE SO-110 B ion source and a modified GIS from Ionplus AG is now operational at CologneAMS. The amounts of exogenous C introduced during sample processing can be accounted for using adequate correction methods. As the sample preparation for EA-GIS-AMS is more time efficient and less expensive, we are planning further tests with a new set-up including an IRMS system. However, GIS-AMS analysis still remains the technique of choice for the 14C analysis of small gas samples, e.g. from incubation studies.
ACKNOWLEDGMENTS
This work was financially supported by German Science Foundation (DFG) within SFB 806 and FOR 1806 as well as by funds from the German Ministry of Science and Education (BMBF) within the project “KoPf”. We would like to thank Ulrike Patt for guidance and troubleshooting during sample pretreatment as well as our student assistants (Thorsten Domann, Reaz Hossain, Elisabeth Krewer, and Vera Schmitt) who helped during sample processing.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/RDC.2019.143
