Technical Note: Inexpensive modification of Exetainers for the reliable storage of trace-level hydrogen and carbon monoxide gas samples

. Atmospheric trace gases such as dihydrogen (H 2 ), carbon monoxide (CO) and methane (CH 4 ) play important roles in microbial metabolism and biogeochemical cycles. Analysis of these gases at trace levels requires reliable storage of discrete samples of low volume. While commercial sampling vials such as Exetainers® have been tested for CH 4 and other greenhouse 10 gases, no information on reliable storage is available for H 2 and CO. We show that vials sealed with butyl rubber stoppers are not suitable for storing H 2 and CO due to release of these gases from rubber material. Treating butyl septa with NaOH reduced trace gas release, but contamination was still substantial, with H 2 and CO concentrations in air samples increasing by a factor of 3 and 10 after 30 days of storage in conventional 12 mL Exetainers. Among the rubber materials tested, silicone showed the lowest potential for H 2 and CO release. We thus propose to modify Exetainers by closing them with a silicone plug, and sealing 15 them with a stainless steel bolt and O-ring for long-term storage. Such modified Exetainers exhibited stable concentrations of H 2 and CH 4 exceeding 60 days of storage at atmospheric and elevated (10 ppm) concentrations. The increase of CO was still measurable, but nine times lower than in conventional Exetainers with treated septa, and can be corrected for due to its linearity by storing a standard gas alongside the samples. The proposed modification is inexpensive, scalable and robust, and thus enables reliable storage of large numbers of low-volume gas samples from remote field locations.

Investigating the turnover of environmental H 2 and CO at trace concentrations generally requires the collection of discrete 30 samples for analysis with a sensitive gas chromatography system (GC). Ideally, the GC is field-deployable and measurements can be conducted in situ, which eliminates the problem of gas storage altogether (King and Weber, 2008;Meredith et al., 2017). Alternatively, a field-laboratory or shipboard setup may allow for short-term (minutes to hours) storage of samples in syringes or gas-tight sampling bags (Conrad and Seiler, 1988). However, such arrangements are not always feasible (e.g. at remote or inaccessible field sites), and samples have to be stored for longer periods (days to months). 35 Glass flasks or bulbs of 0.5 -1 L volume sealed with hoses or O-rings have been found to provide stable, long-term storage of a variety of atmospheric trace gases and their stable isotope ratios (Rothe et al., 2005;Thrun et al., 1979), and are routinely used for sampling campaigns involving isotopic ratios of H 2 (Chen et al., 2015;Schmitt et al., 2009). Other similar designs include stainless steel flasks and cylinders Sulyok et al., 2001). However, these systems have two major downsides: they require the collection of large sample volumes of >1 L, and they are prohibitively expensive for large-scale 40 field campaigns with hundreds or thousands of samples.
For these reasons, glass vials of a few mL volume closed with butyl rubber septa (either crimp-capped or screw-capped, e.g. Exetainers®) are among the most widely used containers to store gas samples in various scientific fields. They are relatively inexpensive and fit many GC autosamplers, thus making them ideally suitable for large measurement campaigns. In particular, Exetainers have shown good stability (<5% deviation from a reference gas) of trace gases such as methane (CH 4 ), nitrous 45 oxide (N 2 O) and carbon dioxide (CO 2 ) of up to one month, regardless of temperature (Faust and Liebig, 2018;Glatzel and Well, 2008;Rochette and Bertrand, 2003). Longer storage times induced deviations of up to 30% from the reference gas, but could be corrected for by storing a standard gas together with the samples (Faust and Liebig, 2018;Laughlin and Stevens, 2003).
To our knowledge, storage of H 2 and CO at trace levels in Exetainers or crimp-cap vials have not been investigated. However, 50 anecdotal reports of rapid sample contamination with H 2 and CO from butyl stoppers has been an ongoing concern. Here we demonstrate that various butyl rubber septa commonly used for sealing glass vials release significant amounts of hydrogen and carbon monoxide over short periods of time. We then propose a simple modification to seal Exetainers with silicone sealant and a stainless steel bolt, and confirm the long-term stability of H 2 , CH 4 and, to a lesser extent, CO concentrations.

Demonstration of trace-gas release from rubber materials
A selection of typical rubber materials used for sealing gas-sampling containers were tested for release of H 2 , CO and CH 4 by incubating them for several days in closed containers. Exetainers of 12 mL volume were chosen for containers as they fitted the autosampler of the GC, thus minimising variability due to manual sample injection. Twelve different treatments were prepared in quadruplicates; details of the treatments and rubber materials are summarised in Table 1. Some treatments involved 60 pre-treating butyl rubber by i) boiling for 2 h in 0.1 M NaOH and then twice in Milli-Q water (Lin et al., 2012); and ii) washing with a surfactant (Tween 20) followed by rinsing and autoclaving twice (Ruth Henneberger, personal communication). After pre-treatments, rubber materials were weighed to 2 g or 4 g depending on treatment, then cut to smaller pieces to fit into Exetainers. Five control Exetainers were prepared empty. All samples were closed with a NaOH pre-treated Exetainer septum (typical weight 0.6 g); thus, all treatments including controls were equally exposed to this septum in the lid. In addition, empty 65 3 mL control Exetainers were prepared in triplicates for each timepoint to investigate the effect of a smaller sample volume.
At the start of the experiment, all samples were flushed with a reference gas for at least 5 min at high flow (> 1 L min -1 ) with pressure regulated to 1.8 bar at closure. The reference gas consisted of pre-calibrated industrial-grade pressurized air with concentrations of 0.568 ppm H 2 , 0.665 ppm CO and 1.96 ppm CH 4 . Samples were then measured 5 times over the course of 9 days; samples were measured repeatedly, except for 3 mL control Exetainers which were discarded after measurements due to 70 insufficient volume for repeated sampling. Measurements of H 2 , CO and CH 4 were performed on a VICI TGA 6k equipped with an autosampler and a sample-injection loop of 1 mL, using a pulse-discharge helium ionisation detector as described earlier . The detection limits for the three gases were 61 ppb for H 2 , 10 ppb for CO and 170 ppb for CH 4 as determined by replicate samples (n=7) of zero-grade air zero (BOC Australia, North Ryde NSW, Australia).

Exetainer modifications
Simple modifications of Exetainers (silicone-sealed Exetainers, SEs) are proposed to minimise contamination of gas samples containing H 2 and CO at trace levels. The modifications consist of i) a small, permanent plug of silicone sealant that can be pierced with needles during short-term sample handling, and ii) a stainless-steel bolt and O-ring instead of a butyl septum to 80 provide a long-term seal blocking diffusive gas exchange. Selleys Wet Area silicone sealant was used to fill the inside top 5 -10 mm of the Exetainer glass vials ( Fig. 1a and 1b). The sealant was administered with a 10 mL plastic syringe for finer handling. First, a thin layer was applied to close the bottom end of the plug, then the plug was filled upwards, to avoid slow expulsion of sealant due to overpressure in the Exetainer. A surplus of about the same amount of sealant was administered at the top to account for potential contraction during curing (Fig. 1c). After curing for 1 -2 weeks, the surplus was cut at the rim 85 of the glass thread with a scalpel, while also scraping off any residual silicone from the rim. Newly prepared SEs were then flushed with high-purity N 2 at slight overpressure and placed in a vacuum chamber for 2 weeks to extract residual gases entrapped in the silicone. However, this vacuum step is not strictly necessary, as good results were also achieved with silicone Exetainers flushed with N 2 and left in ambient air for several weeks.
Well-prepared SEs should be gas-tight in the short term (minutes to hours) and hold overpressure for several days or weeks. 90 However, gas diffusion through silicone is rapid compared to butyl rubber (refs), thus SEs require a second seal for long-term storage. For this purpose, commercially available buttonhead stainless-steel bolts (M6 x 10 mm, grade 304) were inserted into empty Exetainer plastic lid, and a Viton O-ring (9.25 mm ID, 12.7 mm OD) was added to seal the bolt against the Exetainer glass rim (Fig. 1d). For sample access or analysis, the bolt and O-ring can simply be removed, with the silicone plug containing the sample. For analysis times of several hours, e.g. when using an autosampler, SEs can be closed with a conventional treated 95 septum to minimise diffusive exchange.

Long-term storage test of silicone-sealed Exetainers
Long-term storage tests were conducted with 3 mL and 12 mL SEs containing different reference gases, to test sample stability and tightness of SEs. Quadruplicate SEs of both volumes were prepared for each of the 5 timepoints by flushing with either pre-calibrated pressurised air (0.643 ppm H 2 and 1.99 ppm CH 4 ; CO not detected) to test for gas release from materials, or a calibration gas (10.2 ppm H 2 , 9.90 ppm CO and 10.1 ppm CH 4 ) to test for tightness of the seal. The SEs were prepared eight 105 months prior to flushing and set to 1.8 bar final pressure. For comparison, a batch of 12 mL SEs was left uncapped and flushed with calibration gas, and a batch of 12 mL conventional Exetainers was capped with NaOH treated septa and flushed with pressurised air. Samples were analysed alongside the reference gases after 3, 10, 30, 60 and 92 days of storage. https://doi.org/10.5194/bg-2020-282 Preprint. Discussion started: 21 September 2020 c Author(s) 2020. CC BY 4.0 License.

Hydrogen and carbon monoxide release from rubber materials 110
A short-term storage test using pre-calibrated pressurised air as reference gas was conducted with various sealing rubbers. All tested rubber materials released H 2 and CO during short storage times of a few hours or days, with orders of magnitude differences between materials (Figs. 2 and 3). Untreated Exetainer septa released the highest amounts of both H 2 and CO; control-corrected mean concentrations in 12 mL exetainers reached 16 ppm and 33 ppm from 2 g of material after 9 days (UT2, Fig. 2). This was 30 and 50 times higher than the initial reference gas, and close to 20 times higher than empty control 115 Exetainers. The concentration increase was linear and roughly proportional to mass, with double the material (4 g) approximately doubling the concentrations of H 2 and CO (UT4, Fig. 2). Rates of H 2 and CO release where thus nearly identical for UT2 and UT4 treatments when normalised to mass (Fig. 3).
Treated materials performed significantly better. Boiling Exetainer septa with 0.1 M NaOH reduced H 2 and CO release by a factor of 4 (Tex; Fig. 2); however, corrected concentrations still reached 3 ppm and 8 ppm at the end of the experiment, with 120 an increase of 3 pmol H 2 and 10 pmol CO every h per g material (Fig. 3). Empty control Exetainers of 12 mL volume closed with a treated septum showed mean H 2 and CO concentrations of 0.89 ppm and 1.4 ppm after 9 days, 1.6 and 2.1 times higher than initial air concentrations. Smaller 3 mL Exetainers showed an even higher contamination with 1.4 ppm H 2 and 1.7 ppm CO after 7 days.
Grey crimp-cap stoppers treated in the same way performed slightly better, with little H 2 release and lower CO release than 125 treated Exetainer septa (Tgr, Figs. 2 and 3). Washed and autoclaved black crimp-cap stoppers were comparable to grey stoppers for CO, but released slightly more H 2 (Tbl; Figs. 2

and 3). Untreated O-ring materials performed better than all butyl rubbers;
Viton O-rings showed lower H 2 and CO release, and nitrile O-rings lower H 2 release than butyl rubbers (FKM and NBR; Figs. 2 and 3). Silicone sealants showed the lowest potential for H 2 and CO release (SPA, SPB and SSW; Fig. 2), with SSW performing best. Interestingly, H 2 and CO release rates were similar for both gases for silicones, while nearly twice as high for 130 CO than H 2 for all butyl and O-ring rubbers (Figs. 2 and 3).
None of the tested rubber materials caused increasing CH 4 concentrations (Fig. 2), except Viton O-rings with a small release of approximately 0.3 pmol g -1 h -1 (Fig. 3). Most treatments showed similar CH 4 concentrations than controls, which occasionally led to negative concentrations when corrected (Fig. 2). https://doi.org/10.5194/bg-2020-282 Preprint. Discussion started: 21 September 2020 c Author(s) 2020. CC BY 4.0 License. Fig. 2: Hydrogen (H2), carbon monoxide (CO) and methane (CH4) release from various sealing rubbers, corrected for release in empty control vials. Replicate samples (n=4) of 2 g or 4 g of rubber material were prepared in 12 mL Exetainers sealed with a treated Exetainer septum, and flushed with pressurised air as reference gas. Controls consisted of empty Exetainers (n=6) sealed with a treated Exetainer septum. Samples were measured repeatedly except for Con3 samples where n=3 replicates were sacrificed for each measurement, due to low gas volume. 140 https://doi.org/10.5194/bg-2020-282 Preprint. Discussion started: 21 September 2020 c Author(s) 2020. CC BY 4.0 License. Fig. 2 for treatments legend.

Stability of trace gas samples in silicone-sealed Exetainers
Gas samples of two reference gases (pressurised air and a 10 ppm calibration mix) were stored up to three months in Exetainers 145 with different seals, and compared against fresh reference gas ( Fig. 4 and Fig. S1). Conventional 12 mL Exetainers with treated septa showed a significant increase in H 2 and CO concentrations during storage of pressurised air, with deviations from fresh gas larger than 5 % after 3 days of storage (Fig. 4); CH 4 concentrations were within 5 % of fresh gas for at least 30 days, and slightly higher afterwards. In contrast, pressurised air stored in 12 mL SEs sealed with a stainless steel bolt remained stable (deviation < 5 %) for more than 60 days for both CH 4 and H 2 . Concentrations of CO increased from below detection limit (~70 150 ppb) to 0.18 ppb in 3 days and around 0.6 ppm in 92 days of storage; however, the rate of increase was approximately linear and nearly an order of magnitude lower compared to conventional Exetainers (Fig. S1). The 10 ppm calibration gas remained stable in bolt-sealed 12 mL SEs for all three gases for at least 60 days, and deviations were less than 10 % after 92 days of storage (Fig. 4). The smaller 3mL SEs showed higher variability in both fresh and stored gases, with CH 4 and H 2 stable for up to 10 days (Fig. 4). Both CH 4 and H 2 concentrations increased with storage in pressurised air, and decreased in the calibration 155 mixture; for CO we observed a clear increase in both reference gases. We also tested the calibration mix in SEs with the silicone left open to lab air, and observed the expected exponential decrease for all three gases as concentrations equilibrated with air. The blue fill indicates the difference between freshly flushed and stored gas is within 5 % of fresh gas. Note that CO was not detected in fresh pressurised air, and concentration is plotted as zero.

Discussion
Our results clearly demonstrate that commercially available gas-sample vials sealed with butyl rubber are not suited for the 165 storage of gas samples with trace levels of H 2 and CO. Among tested rubber materials, butyl rubbers exhibited the highest rates of H 2 and particularly CO release, regardless of any pre-treatment. Untreated Exetainer septa performed particularly https://doi.org/10.5194/bg-2020-282 Preprint. Discussion started: 21 September 2020 c Author(s) 2020. CC BY 4.0 License. poorly; measured rates indicate that concentrations of H 2 and CO in air samples stored in conventional 12 mL Exetainers could essentially double and quadruple every 24h. Treatment of septa with NaOH greatly reduced potential contamination, but concentrations in air samples could still double approximately every two weeks for H 2 and every week for CO. The level of 170 contamination was proportional to the volume of the vial and the mass of rubber (a proxy for the specific surface area in contact with the enclosed gas phase); therefore, form factor and volume of vials have to be considered. For example, serum vials with an increased volume of >100 mL and thus lower exposed surface area relative to volume may still be suitable for gas storage or incubation experiments lasting a few weeks, if treated stoppers are used (Islam et al., , 2020Kessler et al., 2019).
Viton and nitrile rubbers performed much better than butyl septa for H 2 , though CO release was still substantial. In principle, 175 butyl septa could be replaced with Viton or nitrile; some commercial products are available for crimp-cap vials and may be an alternative for laboratories with an existing stock of such vials. We have not tested septa made from these materials however, and tightness, possible treatments and resistance to multiple piercings will need to be evaluated alongside potential contamination. Particularly for CO, Viton and nitrile may not be much better than treated butyl rubber.
Silicone sealants showed the lowest potential for H 2 and CO release. Assuming a second barrier to control diffusion (discussed 180 below) and a typical septum weight of 0.5 g, contamination after one month of storage would theoretically accumulate to 0.1 ppm H 2 and 0.2 ppm CO in vials sealed with silicone. This is close to the contamination introduced by piercing the septum with a needle, which inevitably introduces a small amount of ambient air (Lin et al., 2012). Such a rate can be acceptable for many projects, particularly when investigating relative changes. However, diffusive exchange through silicone rubber is rapid and requires an impermeable second barrier to prevent this, as proposed with our SEs. 185 Specific mechanisms for H 2 and CO release from different rubbers have not been investigated here. We can however highlight some common patterns among the different rubbers tested. First, a near-linear increase in H 2 and CO in all rubber materials, regardless of pre-treatment and concentration, indicates a zero-order reaction and thus points towards degradation of material as underlying cause, rather than release of entrapped or dissolved gas. Entrapped or dissolved gases would slowly equilibrate with the gas phase, thus rates would show an inverse correlation with concentrations. Silicones showed slightly reduced rates 190 towards the end of the release experiment (Fig. 2), but the trend was weak. We cannot exclude the possibility of entrapped or dissolved gases at higher concentrations in the materials, but this appears less realistic when considering negligible release of CH 4 . Chemical degradation of rubbers can occur via multiple and complex reactions and strongly depend on its composition and polymer structure (Dubey et al., 1995). Likely, silicones degrade differently to butyl and O-ring rubbers, as indicated by the ratio of H 2 to CO release of 1:1 for silicones and at least 1:2 for other rubbers. Regardless of the cause of H 2 and CO release 195 however, among the rubbers tested here, only silicones and possibly Viton seem to be suitable for storing small gas samples with H 2 and CO at trace levels for longer than a few days.
Our modified SEs sealed with a stainless steel bolt and O-ring provided stable storage for gas samples with H 2 and CH 4 at ambient levels, and with all three gases at elevated levels, for up to two months (Fig. 4). The SEs thus allow long-term storage of small volumes of trace-level H 2 samples, and extend reliable storage times for CH 4 samples by one month compared to 200 conventional Exetainers (Faust and Liebig, 2018). There was still significant contamination of gas samples with CO at ambient https://doi.org/10.5194/bg-2020-282 Preprint. Discussion started: 21 September 2020 c Author(s) 2020. CC BY 4.0 License.
levels; yet, the increase was nine times lower than in conventional Exetainers with treated septa. Most importantly, the increase appeared to be linear, and thus offers the potential to correct for this bias by storing a reference gas alongside the samples (Laughlin and Stevens, 2003). The 3 mL SEs were stable for 3 to 10 days for H 2 and CH 4 , and showed deviations thereafter.
Yet, for these gases concentrations stabilised within 10 to 30 days to a slightly higher level with pressurised air, and a slightly 205 lower level with calibration mix. We suspect this to be a result of equilibrium partitioning between silicone and gas phase, which is much more pronounced in the 3 mL vials with a higher silicone-gas ratio. This may lead to a "memory" effect when gases from previous samples can partition into the silicone, then back into the gas phase in the new sample. In our case, 3 mL SEs were stored in lab air with slightly higher H 2 and CH 4 concentrations than pressurised air, and lower concentrations than the calibration mix. Therefore, partitioning between gas and silicone could explain the observed pattern in 3 mL SEs, but 210 seemed largely negligible in 12 mL SEs. To minimise this effect, we suggest purging SEs with high-purity N 2 after measurement, and storing them in an atmosphere of known composition, ideally similar to future samples. Surprisingly, concentrations of H 2 were much less affected than CO during storage in SEs, unlike in emission experiments where silicones showed similar rates of H 2 and CO release ( Fig. 2 and 3). However, we also observed high variability in H 2 emissions from silicones, with some replicates showing H 2 increase comparable to CH 4 , others to CO. We suspect this was an 215 effect of heterogeneous curing of the large silicone plugs we prepared for the emission experiment, which cured for three weeks before cutting to smaller pieces. In contrast, the SEs used in the long-term storage test were made 8 months prior to the test and were fully cured. We therefore speculate that silicones emit H 2 only during curing, while CO appears to be released at a consistent rate.
We have now employed 12 mL SEs for several projects with 400 to 600 samples and storage times of up to one month. Once 220 prepared, SEs could be pierced numerous times without compromising tightness. Although not explicitly tested, we observed less contamination when piercing the silicone plug with needles compared to thinner butyl septa, particularly when the needle was withdrawn slowly and the silicone was allowed to seal before the needle was fully out. Stainless steel bolts, O-rings and Exetainer plastic caps have been used numerous times without compromise. Overall, the proposed system is simple to implement, inexpensive, robust and reliable, and can be integrated in many existing sampling and processing pipelines. 225

Author contributions
PAN, EC and PLMC designed the experiments, and PAN, EC and TJ carried them out. CG and PLMC provided logistical and technical support. PAN prepared the manuscript with contributions from all co-authors.

Data availiability
All data can be made available by the authors upon request.