A high volume sampling system for isotope determination of volatile halocarbons and hydrocarbons

Introduction Conclusions References

In particular, stable carbon isotope ratios have been proposed for constraining the origin and fate of atmospheric chloromethane (Harper et al., 2003;Keppler et al., 2005) and bromomethane (Mc Cauly et al., 1999); as the reported carbon isotope ratios of the known sources cover a broad range of δ 13 C values from −40 ‰ for anthropogenic and marine sources to −147 ‰ for chloromethane synthesized from pectin (cf.Keppler et al., 2005).Tremendous progress has recently been made for the carbon isotopic analysis of dissolved halocarbons (Auer et al., 2006), which now allows to extend this approach to short lived halocarbons such as iodomethane (CH 3 I), dibromomethane (CH 2 Br 2 ) and bromoform (CHBr 3 ).Accompanying determination of the carbon isotope ratios of these compounds could provide valuable additional information that can not be derived from the determination of mixing ratios.
For determining carbon isotope ratio of atmospheric VOCs both, air sampling canisters (Rudolph et al., 1997;Tsunogai et al., 1999;Thompson et al., 2002;Archbold et al., 2005;Redeker et al., 2007) and adsorbent tubes (Mead et al., 2008) have successfully been used.Given a minimum carbon amount between 0.25 ng and 1 ng required for carbon isotope analysis, these sampling methods are sufficient Published by Copernicus Publications on behalf of the European Geosciences Union.
To the best of our knowledge, there is only one study reporting carbon isotope ratios for atmospheric bromomethane (Bill et al., 2004) in which a cryogenic sampling system was used.
The main objective of this study was to develop a simple and robust sampling system appropriate for field work enabling the isotope ratio determination of atmospheric halocarbons in the low pptv range, notably bromomethane, bromoform and iodomethane.Here, we describe the configuration and validation of the developed sampling system and report first results from its application within a survey on air samples from an urban (Hamburg, Germany) and a coastal site (Sylt, Germany).

Overview
Briefly, up to 500 l of air were drawn through a cryo trap.The trap was then heated to 125 • C and the analytes were transferred to an adsorption tube for storage and analysis.A direct analysis of the isotopic composition of the target compounds is not possible due to multiple interferences from other compounds.Therefore, the samples were first pre-separated on a GasPro column and the target compounds were recollected on two cryo traps.For carbon isotope determination each fraction was then separated on a PorabondQ column and analysed on a GC-C-IRMS/MS system.

Standards and samples
A Scott EPA TO 15/17 standard containing 65 compounds (1 ppm each in nitrogen) was used as a working standard.For further tests single component standards of chloromethane, bromomethane, iodomethane and dichlorodifluoromethane (each 100 ppm in nitrogen) were used.
Three ambient air samples were taken at the institute building in the center of Hamburg, Northern Germany (53 • 56 86 N, 9 • 97 36 W) from 19 to 24 September 2010.The sampling system inlet was placed at 30 m above the ground.The sampling duration was 2h and the start time varied between 10:00 am and 02:00 p.m. LT (= local time).Possible local sources include traffic exhaust, emissions from the harbor area and volatiles from laboratory activities in the vicinity.
In addition, three marine influenced air samples were taken at the AWI Wadden Sea Station in List/Sylt, Northern Germany (55 • 02 26 N. 8 • 43 96 W) between 29 August and 5 September.The sampling system was placed directly at the coastline 200 m away from the Wadden Sea Station.As for the urban location the sampling duration was 2 h and the start time varied between 10:00 a.m. and 02:00 p.m. LT Potential local sources for halocarbons include emissions from salt marshes, tidal flats and, to a minor degree, harbor activities in List.
72 h back trajectories were calculated by Hysplit 4.8 for an arriving height of 30 m (Hamburg samples) and 2 m (Sylt samples) using NCEP's Global Data Assimilation System (GDAS) data (Draxler and Rolph, 2011).The backward trajectories indicate a prevailing marine and coastal influence for the samples taken at Sylt and a prevailing continental influence for the samples taken at Hamburg.

Sampling system
The sampling system (Fig. 1) was designed with respect to field suitability and integrity of the isotopic composition of the target compounds.To avoid contamination, all system parts were made of stainless steel or silanized glass.
The sampling consists of three consecutive steps: (1) the cryogenic adsorption of the target compounds from ambient air, (2) the transfer of the analytes from the cryo trap to an adsorption tube and (3) the conditioning of the cryo trap.
1.During sampling the valves 1 and 2 were open.300 to 500 l of ambient air were drawn with a membrane pump (KNF Neuberger N86 KNDC, Freiburg, Germany) through the sampling system with a flowrate between 3 and 5 l min −1 .The flowrate and the sampling volume were monitored with a mass flowmeter (Omega FMA-1608A, Deckenpfronn, Germany).The air first passed a particle filter (Sartorius, Teflon membrane filter, diameter: 45 mm, pore size: 0.2 µm) and was then directed through a condenser kept at approximately 5 • C reducing the water vapour of the air.It has previously been shown that a condenser does not affect the recovery of high boiling compounds such as CHBr 3 (Christoph et al., 2002).Finally, the target compounds were enriched in the cryotrap at a temperature of ≤ −110 • C provided by a dry shipper.
2. After sampling, valve 1 and 2 were closed and valve 3 to the adsorption tube was opened.The cryotrap was carefully removed from the dry shipper.Co-trapped argon and traces of O 2 and N 2 were rapidly released into the gas phase.To prevent analyte losses during sample transfer (usually observed at a flowrate >100 ml min −1 ), a restriction capillary (restriction flow 70 ml min −1 at 2 bar) was placed behind valve 3. Without this restriction we observed losses of highly volatile analytes such as chloromethane and dichlorodifluoromethane.After 20 min, valve 4 was opened for 40 min, the cryo trap was heated to 125  with either nitrogen (50 ml min −1 ) or pre cleaned ambient air.Before entering the adsorption tube, water vapour was removed from the gas stream by a Nafion dryer (Perma Pure Inc. NJ, USA) placed in silica gel.During sample transfer the adsorption tube was cooled to −5 • C using a Peltier cooler to prevent breakthrough of the analytes.Under these conditions non-polar compounds with a boiling point as high as 150 • C were completely transferred from the cryo trap to the adsorption tube, while polar water soluble compounds, such as alcohols and aldehydes, were fractionated between the water remaining in the cryo trap and the adsorption tube.After the transfer of samples the adsorption tube was closed and stored at ≤ −80 • C until analysis.
3. After sample transfer, valve 3 was closed and valve 5 was opened for conditioning the cryo trap with a stream of nitrogen (1000 ml min −1 ) for 30 min at a temperature of 125 • C.

Cryofocussing and design of the cryotrap
Classical cryogenic sampling systems use either liquid nitrogen or liquid argon for the extraction of the target compounds from air (Bill et al., 2004;Pupek et al., 2005;Zuiderweg et al., 2011).More recently, a cryostat has been applied in cryogenic sampling systems (Miller et al., 2008).However, both approaches require a well established technical infrastructure limiting the use of such sampling systems in field campaigns, especially for field work in remote areas.In order to overcome the limitations of classical cryogenic sampling systems we employed a dry shipper (Voyageur 12, Air Liquide, Düsseldorf, Germany) as a cooling source.A dry shipper is a dewar that contains an adsorbent taking up liquid nitrogen.The Voyageur 12 can adsorb up to 14 l of liquid nitrogen and provides a temperature of −176 • C for 28 days in the gas phase above the adsorbed liquid nitrogen.It has an approval for air transport allowing cryogenic sampling in remote places where no liquid nitrogen is available.For the sampling system the stopcock of the dry shipper was shortened to 10 cm and a hole of 20 mm diameter was drilled into the stopcock to insert the cryo trap.The effective temperature inside the cryo trap depended on the flushing flow rate and was below −110 • C for flowrates up to 5 l min −1 .The cryo trap consists of an outer stainless steel tube (3/4 inch, 50 cm length) with an air inlet at the side 4 cm below the top and an inner 1/4 inch stainless steel tube that is connected to the sampling pump (Fig. 1).
To achieve complete trapping at −110 • C the space between both tubes is filled with adsorbents separated by plugs of precleaned silanized glass wool.From the top to the bottom the package of the trap was as follows: 0-20 cm: empty; 20-25 cm silanized raschig rings; 25-35 cm silanized glass beads 60-250 µm; 35-41 cm Tenax TA; 41-47 cm Porapak N; 47-50 cm: empty.The empty space at the top and

Adsorption tubes
The adsorption tubes were made of stainless steel 1/4 inch outer diameter, 7 inch length) and filled with 77 mg Carboxen 1016 ® , 215 mg Carbopack X 569 ® , 80 mg Carboxen ® 1003 and 9 mg Tenax ® TA in order of the sampling flow direction.The filling was fixed with stainless steel plugs.Silanized glass wool was used for separating the adsorbents.Before use, the packed tubes were conditioned for 24 h in a tube conditioner (TC 1/2, Gerstel, Mühlheim Germany) at 320 • C and a flow of 100 ml min −1 of nitrogen.

Removal of CO 2 and water
Before analysis water and CO 2 must be removed.Chemical traps frequently used in trace gas analysis to remove excessive CO 2 , such as Ascarite or Soda lime (Rudolph et al., 1997;Redeker et al., 2007) reduce the recovery of polyhalogenated methyl halides especially of polybrominated compounds and thus have been ruled out.The adsorbents used in the adsorption tubes have a low affinity for CO 2 and most of the CO 2 from the cryo trap passes the adsorbent tubes during sample transfer.Prior to analysis, the adsorbent tubes were flushed with helium at room temperature (2 min, 20 ml min −1 ) further reducing the CO 2 content of the samples.The remaining CO 2 was removed during the GC preseparation by an initial flushing step.Chemical traps such as magnesium perchlorate or phosphorus pentoxide are not suitable for water removal in highvolume sampling systems due to the large amount of water (up to 30 ml m −3 ) present in ambient air.Water was removed from the sample at three stages during sampling and analysis.First the water vapour pressure in the sampled air was reduced by means of a condenser kept at 5 • C. We found this suitable to prevent clogging of the cryo trap for air samples up to 500 l.The remaining air moisture was frozen out in the cryo trap.During the sample transfer only a small fraction of the water was mobilized from the cryo trap.This water was removed by a Nafion dryer placed in silica gel.Remaining traces of water were finally removed by means of a second Nafion dryer during analysis.
Nafion dryers have the potential for artefact formation.So far, transformation of carbonyl compounds to alcohols and reactions of compounds containing double or triple bonds have been reported (Miller et al., 2008).
During first tests, we attached the Nafion dryer in front of the cryo trap to reduce the water offload to the cryo trap.With this setup we observed elevated levels of bromomethane.Parallel runs with and without the Nafion dryer indicated an artificial formation of bromomethane on the Nafion membrane of up to 10 pptv.Further tests revealed that the sulfonic acid groups of the Nafion membrane can catalyze the nucleophilic substitution of methanol to bromomethane in the presence of bromide.Nevertheless, with the Nafion dryer behind the cryo trap we did not observe any artificial formation of bromomethane.

Carbon isotope ratio determination
A direct analysis of the isotopic composition of the target compounds was not possible due to multiple interferences from other compounds.Therefore, the sample was first pre-separated on a Gas-Pro column and the target compounds were recollected on two cryo traps.For carbon isotope determination each fraction was then separated on a PorabondQ column and analysed on a GC-C-IRMS/MS system.A scheme of the analytical set up is shown in Fig. 2. Two exemplary chromatograms showing the mass 44 trace and the m/z 45/44 ratio from a coastal sample as well as the mass spectra of selected halocarbons are presented in the Supplement.

Pre-separation
The pre-separation was performed on an 6890N/5975B GC-MS (Agilent, Waldbronn, Germany).The analytes were desorbed from the adsorbent tube into a helium gas carrier at 330 • C (15 min., 25 ml min −1 ), directed through a Nafion dryer and refocused on a quartz capillary (0.32 i.d, 60 cm length) immersed in liquid nitrogen.Afterwards the analytes were desorbed at 25 • C and separated on a GasPro column (Agilent, 30 m, 0.32 µm i.d.) with helium as a carrier.The flowrate was set to 5 ml for 4 min to remove the CO 2 and then ramped to 2.7 ml min −1 .The oven temperature program was as follows: 40 • C, hold for 5 min.;6 • C min −1 to 240 • C, hold for 10 min.
About 20 % of the sample were directed into the MS via a split for monitoring the fractionation.The remaining 80 % of the sample were directed to a Valco six port valve (Vici, Schenkon, Swiss) and recollected on two cryo traps (Silica steel 1/16#, 0.32 i.d., 25 cm length), one containing the target compounds and the other containing the remaining components.

Isotope ratio determination
For the isotope ratio determination each fraction was analysed on a Thermo Finnigan GC-C-IRMS/MS system (Trace GC II; combustion Interface III; DeltaV IRMS and DSQ II) system using a CP-PorabondQ column (Varian,25 m,250 µm i.d.) for final separation of the analytes.Again, 20 % of the sample were directed to a Quadrupole mass spectrometer (Thermo Finnigan DSQ II) for monitoring the peak purity.The remaining 80 % were directed to a commercial combustion interface (GC-combustion interface III; Thermo Finnigan) converting the analytes to CO 2 .Water was further removed by a Nafion dryer and about 0.5 ml min −1 . of the Prior to sample analysis, the performance of the IRMS was evaluated by repeated injections of a certified CO 2 reference gas (Air Liquide, Germany, −26.8 ± 0.2 ‰) via an open split.The Scott TOC EPA 15/17 standard was used as a daily working standard to monitor the repeatability.Results are only reported for peaks that met the following quality criteria: (i) peak purity better than 90 % (ii) peak separation better than 90 % valley.
All isotope values are reported relative to the Vienna Pee Dee Belemnite (VPDB) scale and within the lab uncertainties, unless otherwise stated, are reported on the 1 σ level.

Identification, quantification and purity control of VOCs
The analytes were identified by comparison of their retention time and mass spectra with known standards.Further compounds were identified by comparison of the obtained mass spectra with the Nist mass spectral database version 2.0.Primary quantification was done on the Agilent system used for pre-separation against the Scotty TOC EPA 15/17 standard.Compounds quantified against a standard were quantified on the IRMS via the CO 2 intensities against chloromethane and bromomethane as internal standards.The uncertainty of this procedure is estimated to ±15 % on the 1 σ level.This es-timate is based on tests with known standards at different concentration levels.It is further justified from the variability of the mixing ratios determined for long lived compounds such as chlorodifluoromethane at the coastal sites as the variation of the mixing ratios for these compounds is typically less than a few percent in rural and background air masses.

Trapping and desorption efficiency of the sampling system
The trapping and desorption efficiency was tested for the entire sampling procedure on a 0.2 nmole level by injecting 5 ml of the Scott TOC EPA 15/17 and 50 µl of the iodomethane standard into a stream of nitrogen and preconcentrating the analytes on the cryo trap at a flow of 5 l min −1 for 100 min (total volume 500 l).Extraction was done as described above and the recovery rates were calculated against direct injections of the standard mixture onto the Agilent GC-MS system used for pre-separation and quantification.
The results of these tests are given in Table 1.The recovery rates for the entire sampling procedure averaged 97.8 % (range from 87 % to 110 %) and the reproducibility ranged from 1 % to 7 %.recovery rates with increasing boiling points for compounds with boiling points above 80 • C.However, there are no statistically significant differences between the recovery rates for direct injection on the adsorption tubes and those for the entire sampling process.Therefore, we conclude that no significant losses of the target analytes occur during sampling.

Reproducibility versus concentration
Due to the low mixing ratios of some target compounds such as CH 3 Br, CH 3 I and CHBr 3 (usually <10 pptv), we determined the linearity and reproducibility of the carbon isotope ratios for low levels of carbon in the sub-nmole range.This was performed by repeated injections of chloromethane and bromomethane into the GC-IRMS.The injected carbon amounts ranged from 0.02 to 20 nmol (n ≥ 6 for each concentration level) corresponding to mixing ratios between 1 and 1000 pptv for a 500 l sample.The results of the reproducibility tests are depicted in Fig. 3.The carbon isotope ratio determination was free of a systematic bias for carbon amounts down to 0.02 nmole.The overall standard deviation for carbon amounts between 20 and 1 nmole were 0.25 ‰ for chloromethane (n = 30) and bromomethane (n = 12) respectively.The standard deviation for each concentration level (n = 6) varied between 0.12 ‰ and 0.30 ‰.However, below 1 nmole the standard deviation for both compounds increased and was 2.6 ‰ for chloromethane and 1.8 ‰ for bromomethane at carbon levels below 0.04 nmole.This corroborates the study of Rudolph et al. (1997) and Redeker et al. ( 2007) who observed similar deviations for comparable amounts of carbon.

Reproducibility of the analytical system
Since the sampling procedure showed an excellent recovery of 98 ± 5 %, it is expected to be free of analytical artefacts.
Blanks were checked on a routine base at regular intervals and the blanks contributed less than 0.5 % to the overall signal.Thus, we can rule out blank contribution as a significant source of bias.Adsorptive losses or incomplete desorption of the target compounds remain as the most important source of bias during sampling.The kinetic isotope effects for such physical processes are generally small.For instance, adsorption of various aromatic hydrocarbons does not appear to cause significant carbon isotope fractionations (Goldstein and Shaw, 2003 and references therein).Even adsorptive losses of 10 % that are associated with a hypothetical kinetic isotope effect of 5 ‰ would lead to a bias of less than 0.6 ‰.Thus, the reproducibility of the carbon isotope ratio determination was only tested for the analytical system.The effect of the pre-separation was tested on the 0.2 nmole level with the Scott TOC EPA 15/17 standard.5 mL of the standard were injected into the GC-MS system, pre-separated and analysed for stable carbon isotope ratios as described in Sects.3.41 and 3.42.Results were compared to the isotope ratios obtained without pre-separation.Without pre-seperation the isotopic composition could only be determined for 7 compounds because of coelutions and peak overlaps.After pre-separation, isotope ratios for 31 compounds Standard deviations (n = 6) ranged from 0.5 ‰ for propene and benzene to 2.8 ‰ for bromoform and generally decreased with increasing numbers of carbon atoms.The average standard deviation for C 1 -compounds, C 2 -compounds and C 3 -compounds were 1.5 ‰, 1.1 ‰ and 0.9 ‰, respectively.A Mann-Whitney-U-test (p < 0.05) revealed usually no significant influence of the pre-separation on the carbon isotope ratios.However, a significant difference between both procedures has been observed for bromoform (direct: −38.6 ± 3.1 ‰; n = 6; pre-separation: −33.8 ± 2.8, n = 6; p = 0.025) and for 1,1-dichloroethane (direct: −24.0 ± 0.4 ‰; pre-separation: −22.3 ± 0.6, n = 6; p = 0.004).

Ambient air samples
The carbon isotope ratios and mixing ratios of the ambient air samples as well as results from previous studies are presented in Table 2.The range of the carbon isotope ratios of selected halocarbons is depicted in Fig. 4. Data are only presented for compounds which could be clearly identified either by comparison with standards or by their mass spectra and which met the quality criteria outlined above.In total, we could determine carbon isotope ratios of 37 compounds with mixing ratios between 0.3 pptv for chlorobenzene and up to 1600 pptv for propane.The high amounts of hydrocarbons in urban air samples caused strongly tailing peaks in the IRMS and thus prevented the carbon isotope determination of several organohalogens.This applied for chloroethane, iodomethane, chloroform, and bromoform in all and for carbon tetrachloride and 1,1,2-trichloro-1,2,2-triflouorethane in some of the urban air samples.

Hydrocarbons and oxygenated VOCs
The carbon isotope ratios obtained for the hydrocarbons agree with the results of previous studies (Rudolph et al., 1997;Tsunogai et al., 1999;Redeker et al., 2007).δ 13 Cvalues of the alkanes propane, butane, isobutane, pentane and isopentane ranged from −31.8 ‰ to −25.0 ‰ with individual standard deviations between 0.8 ‰ to 2.2 ‰.Except for butane that co-eluted with 1,2-dichloro-1,1,2,2tetrafluoroethane, the compounds were well separated.The portion of 1.2-dichloro-1.1.2.2-tetrafluoroethane to the butane carbon was less than 2 % in urban samples and therewith negligible.In samples from the coastal site this portion amounted to 5 to 20 % and thus may have affected the carbon isotope ratio determination of butane.However, the carbon isotope ratios from both sites showed no significant differences.Even though the average mixing ratios between both sites differed by roughly one order of magnitude (urban site: 304-1620 pptv; coastal site: 11-193 pptv), we observed no significant differences in δ 13 C-values neither between the alkanes nor between the two sites.This is in line with the small fractionation factors (1.41-3.44 ‰) reported for the reaction of alkanes with OH-radicals (Rudolph et al., 2000).Furthermore, the alkanes most likely stem from local traffic related sources at both sites and therefore obtain similar isotopic signatures.On average the alkenes, propene and 2-butene were enriched in 13 C by 3.0 ‰ relative to the alkanes, which is conform to previous studies (Redeker et al., 2007).Propadiene and 2-butyne were even more enriched with δ 13 C-values of −18.1 ± 0.3 ‰ and −13.6 ± 8.2 ‰, respectively.Propene, the only unsaturated hydrocarbon that could be determined at both sites, showed no site specific differences in the δ 13 C values.As the atmospheric degradation of propene by OHradicals is assigned with a considerable fractionation factor of 11.7 ‰ (Rudolph et al., 2000), the lack of a site specific difference in δ 13 C thus points towards local sources rather than towards propene transported over a long distance from urban to coastal areas.
We also could determine carbon isotope ratios of several oxygenated compounds including dimethylether (DME), furane, propanal and propenal, which were identified based on their mass-spectra.The carbon isotope ratios of the aldehydes (−25.4 ± 1.6 ‰) were mainly in the same range as those of the alkenes with the exception of propenal that was strongly enriched in the urban air samples (−17.2 ± 2.3 ‰).Furane showed δ 13 C values of −29.0 ± 1.3 ‰.DME was depleted relative to the saturated hydrocarbons showing a mean δ 13 C-value of −34.1 ± 0.8 ‰ in the urban samples and of −37.2 ‰ at the coastal site.

Organohalogens
The mixing ratios of the long-lived CFCs and chloromethane generally fell within ±30 % of the atmospheric background levels.The average carbon isotope ratios of the organohalogens covered a broad range of δ 13 C-values.Vinylchloride was strongly enriched in 13 C with an average δ 13 C value of 0.5 ‰.In contrast, hexafluoropropene was strongly depleted in 13 C with average δ 13 C value of −57.1 ‰.

-Dichlorodifluoromethane (CFC-12):
The average mixing ratios of dichlorodifluoromethane (614 ± 60 pptv for the urban site and 554 ± 87 pptv for the coastal site) showed no significant difference between the two sites.The average δ 13 C value was −39.6 ‰ with slightly enriched δ 13 C values at the urban site (coastal site: −41.2 ± 0.2 ‰; urban site: −37.9 ± 1.1 ‰).Our data are in the range reported by Redeker et al. (2007), who gave an average δ 13 C value of −37.1 ± 3.9 ‰ from a coastal and an urban area in Ireland, but are depleted in comparison to those reported by Mead et al. (2008) from Bristol, UK (−33.5 ± 0.8 ‰).
-Chloromethane: Chloromethane mixing ratios were 524 ± 36 pptv for the urban and 620 ± 30 pptv for the coastal site.The δ 13 C value average over both sites was −38.0 ± 4.1 ‰ (n = 7) with a slight difference between the urban (−39.9 ± 1.9 ‰) and the coastal site (−36.2± 0.7 ‰).The values mirror previously published results.Tsunogai (1999) and co-workers reported an average δ 13 C of −36 ‰ for the marine background in the subtropical Pacific.Thompson et al. (2002) determined an average δ 13 C value of −36.4 ± 1.6 ‰ from a remote site in the arctic (Alert, Canada).Slightly more depleted values of −39 ± 2.3 ‰ were reported by Redeker et al. (2007) from Belfast, Ireland. -Bromomethane: The δ 13 C-values of bromomethane in the urban samples were −41.5 ± 3.3 ‰ (n = 4) being in excellent agreement with the only previous reported values of −43.0 ± 1.7 ‰ from a suburban site in Berkely, USA (Bill et al., 2004).At the coastal site the δ 13 C-values averaged −31.0 ± 0.3 ‰.This isotopic enrichment by 10 ‰ is accompanied by a decrease of the average mixing ratios from 10 ± 3 pptv at the urban site to 7 ± 1 pptv at the coastal site.However this difference in the mixing ratios is mainly driven by one urban sample showing an elevated mixing ratio and thus we cannot state a systematic relation between the carbon isotope ratios and the mixing ratios.
-Iodomethane: Iodomethane coeluted with carbon disulfide, dichloromethane and propenal.Although the mixing ratios in all samples were generally sufficient for the carbon isotope ratio determination we were only able to determine carbon isotope ratios of iodomethane in the coastal samples after a careful adjustment of the time intervals for the fractionation of the samples.
Iodomethane belongs to the few compounds revealing a strong within site variation of the carbon isotope ratios.For the three coastal samples determined, δ 13 C values were −79.8 ‰, −40.4 ‰, and −40.8 ‰.Given the respective mixing ratios of 1.9, 1.0 and 4.9 pptv no clear relation between variations of the δ 13 C values and of mixing ratios becomes evident.Iodomethane has a relative short atmospheric lifetime of only a few days (Harper and Hamilton, 2003).Therefore both, the mixing ratios and the carbon isotope values, more likely provide a snapshot than an integrated signal, which may explain the lack of any correlation.There are no literature data on carbon isotope ratios of atmospheric iodomethane available to compare with.But our own unpublished data from incubation experiments and greenhouse experiments revealed δ 13 C values of −47 ± 11 ‰ for iodomethane emitted from different halophytes which correspond with the more enriched atmospheric values in our study.The strongly depleted δ 13 C value was determined in air masses coming from the open North Sea while the more enriched δ 13 C values were determined in more coastal influenced air masses.A potential important source in open oceans currently under discussion is the photolytic formation of iodomethane in the sea surface layer (Moore and Zafirou, 1994).Nevertheless, we can currently not substantiate the reasons for the observed large within site variation of the carbon isotope ratios. -Bromoform: A reliable determination of bromoform δ 13 C values was only possible in the coastal samples.In the urban air samples the determination was hampered by co-eluting C8 hydrocarbons.The average δ 13 C values of bromoform in the coastal samples was −18.3 ± 4.6 ‰ (−22.9 ‰, −13.8 ‰ and −18.2 ‰).
The isotope ratios reported here were corrected for the isotopic shift observed for the standard (see Sect. 4.2.2).
Due to this correction we estimated the overall reproducibility for the δ 13 C determination of bromoform to ±4.1 ‰ on the 1 σ level.This is close to the natural variability in these samples and it thus remains unresolved weather the variability of the δ 13 C values for bromoform reflects the natural variability or simply the analytical uncertainty.As for iodomethane, there are no literature data available on the isotopic composition of atmospheric bromoform.However, similar isotope ratios were reported for bromoform produced in incubation experiments by the brown algae Fucus serratus (δ 13 C of −15 ‰) and the planktonic algae Dunaliella tertiolecta (δ 13 C of −24 ‰) (Auer et al., 2006).In the same study the δ 13 C value of dissolved bromoform in a sea water sample from the Baltic Sea was determined to −28 ‰.
-Chloroethane: δ 13 C values of chloroethane could only be determined in the samples from the coastal site.In the urban air samples the tailing of the butane peak impeded a reliable carbon isotope ratio determination of this compound.Average mixing ratios were 2.0 ± 0.7 pptv and the δ 13 C values (−36.5 ± 1.7 ‰) were comparable to those of chloromethane.
-Vinyl chloride: Vinyl chloride was only detectable in the urban air samples with mixing ratios of 5.3 ± 3.5 pptv.With an average δ 13 C value of −0.5 ± 2.1 ‰ it was strongly enriched in 13 C as compared to the other chlorinated compounds.To the best of our knowledge, no isotopic data for atmospheric vinylchloride are so far published.Based on our reference standard one might assume a δ 13 C value of −30.1 ‰ for industrially produced vinyl chloride.In the atmosphere, vinylchloride is rapidly degraded mainly by OH-radicals.The fractionation factor (ε) of this reaction has not been determined yet.
As the OH-radical attacks the double bound it may be in the same order as the ε of 11.4 ‰ reported for the atmospheric degradation of propene by OH-radicals (Rudolph et al., 2000).In addition, incubation experiments with different soils indicate a large ε of 21.5 to 26.0 ‰ for the microbial degradation of vinylchloride in soils (Bloom et al., 2000).Thus, this extraordinary high δ 13 C value of vinylchloride may result from its rapid atmospheric degradation and/or evasion of isotopically enriched vinylchloride from sources such as landfills.
-Chloroform: Chloroform stable isotope ratios could only be determined in the coastal samples due to strong chromatographic interferences from 2-methyl-1-butene in the urban samples.δ 13 C values analysed were −33.8 ‰, −33.1 ‰ and −44.8 ‰.The relative enriched isotope ratios were observed in air masses from the North and North East respectively while the depleted 13C value of −44.8 ‰ was observed in westerly air masses that passed along the Dutch and the German coast.It is noteworthy that this depletion in 13 C for chloroform occurred not in the same sample as the depletion in 13 C for iodomethane.This depletion in 13 C is surprising as it was not accompanied by significant differences in the mixing ratios (10.0 ± 2.1 pptv, n = 3).Anyhow the mixing ratios found here for chloroform were by almost two orders of magnitude lower than those reported by Redeker et al. (2007) but the carbon isotope ratios found here fell into the range of −37.4 ± 6.4 ‰ given for chloroform in that study.In contrast, the carbon isotope ratios reported for chloroform by Mead et al. (2008) were on average more than 10 ‰ enriched in 13 C. -Carbontetrachloride: The average found δ 13 C value of carbontetrachloride was −28.9 ± 1.9 ‰ (n = 3) and average mixing ratios were 104 ± 8 for the coastal site.This agrees well with the results of Mead et al. (2008) who reported an average δ 13 C value of −27.1 ± 1.2 ‰.The average mixing ratio of carbontetrachloride in the urban samples was 92 ± 25 pptv with a mean δ 13 C value of −36.5 ± 5.3 ‰.The isotopic signal in the urban air samples might be influenced by incomplete recovery of carbontetrachloride in the target fraction (85 and 90 %) and thus has to be taken with great care. -Chlorodifluoromethane: In contrast to dichlorodifluoromethane, chlorodifluoromethane was significantly depleted in 13 C at the urban site (δ 13 C of −53.2 ± 4.6 ‰) as compared to the coastal site (−44.2± 3.3 ‰).Our values from the coastal site resemble those reported in the study of Redeker et al. (2007) who provided an average δ 13 C value of −42.9 ± 5.6 ‰ for Belfast (Ireland).More enriched δ 13 C values of −33.9 ± 1.0 ‰, have been reported from Bristol, UK (Mead et al. 2009).Interestingly, Redeker et al. (2007) observed a slight although statistically not significant enrichment of 13 C for chlorodifluoromethane in northerly air masses as compared to westerly air masses and air masses from Europe. -Trichlorofluoromethane: The isotope ratios of trichlorodifluoromethane were −29.5 ± 5.3 ‰ for the urban site and −31.5 ± 2.6 ‰ for the coastal site.As for dichlorodifluoromethane, no significant differences in δ 13 C between the sites were observed.Our values corroborate the results of Redecker et al. (2007) who gave an average δ 13 C value of −27.3 ± 4.4 ‰.
-Pentafluoroethane, norflurane and bromotrifluoromethane: Unusually 13 C enriched carbon isotope ratios were observed for pentafluoroethane (16.5 ± 5.3 ‰) and norflurane (4.3 ± 3.3 ‰).A thorough reanalysis of these data revealed interferences on the m/z 45 and m/z 46 signal.As both compounds elute shortly after carbonylsulfide in an interval where the m/z 46/44 ratio is still affected by sulphur, we presume these interferences to result from the formation of fluorosulfur-compounds in the combustion interface or in the ion source.This is further supported by the carbon isotope ratio measurements of bromotrifluoromethane (CBrF 3 ).Interestingly, we observed the opposite effect for bromotrifluoromethane, eluting 15.3 s after carbonylsulfide.In the urban air samples it was recollected in the same fraction as carbonylsulfide and yielded an average δ 13 C value of −86.0 ± 0.4 ‰.Again a reanalysis of our data revealed substantial interferences on the m/z 45 and m/z 46 signal For the coastal samples, where CBrF 3 and carbonylsulfide were recollected in different fractions, we obtained an average δ 13 C value of −39.1 ± 10.0 ‰. and found no indication for interferences.As CBrF 3 has an average atmospheric lifetime of 65 years and the mixing ratios for both sites were comparable (3.7 ± 0.2 pptv) this huge discrepancy is rather due to the analytical interferences than to atmospheric degradation or source-related processes.
For chloromethane and bromomethane these differences between both sites are in line with our current understanding of the atmospheric cycling of these compounds.But they are surprising for the long lived CFC's and HCFC's with atmospheric lifetime of several decades along with the lack of any significant actual sources implies a natural variability close to the experimental uncertainty.With exception of the dichlorodifluoromethane measurements at the coastal site, both the within site variability and the in between site variability exceed what one might expect for such inert tropospheric trace gases.If we compare our data with those of Redecker et al. (2007) and Mead et al. (2008), it becomes obvious that the repeatability (reproducibility within a lab) reported in all three studies is comparable.But the average δ 13 C values reported for several (H)CFC's show substantially larger variations pointing towards a poor reproducibility.For instance the average δ 13 C values for dichlorodifluoromethane vary from −33.5 ‰ to −41.2 ‰ and those for chlorodifluoromethane vary from −33.9 ‰ to −53.2 ‰.If the assumption that these trace gases are inert in the troposphere is valid we can currently state only a poor reproducibility for these compounds.
On other hand we can currently not rule out that CFC's and other long-lived organohalogens are degraded in ocean surface waters (Yvon and Butler, 1996;Yvon-Lewis and Butler, 2002).For carbon tetrachloride the oceanic lifetime driven by hydrolysis has been estimated to 2599 days and might be reduced to 94 days due to (micro)biological activity as suggested by Butler et al. (1997).The estimates of oceanic lifetimes of chlorofluorocarbons that showed an isotopic enrichment in our study range from 1100 days to more than 120 000 days without considering an additional biological sink in the oceans.As already pointed out by Yvon-Lewis and Butler (2002), there is substantial evidence for a microbial degradation of chlorofluorocarbons from different environmental settings .If such degradation processes also occur in the surface ocean and are assigned with a substantial large fractionation factor they may imprint the isotopic composition of these trace gases in the atmosphere.Nevertheless, any justification of the isotopic variability of the long lived CFC's and HCFC's require a careful evaluation of potential sources of errors and bias for each compound.
The enrichments in 13 C of chloromethane and bromomethane at the coastal site were not accompanied by a significant decrease of the mixing ratios which would point towards an enhanced degradation in the marine boundary layer e.g.due to reactions with chlorine radicals.Therefore, we suppose atmospheric degradation processes not to be the decisive factor for these differences.We assume these isotopic differences are caused by the air sea exchange of these compounds in concert with partial degradation in surface oceans altering the isotopic signatures.Bromomethane from both, intrinsic sources and from the atmosphere, is known to be rapidly degraded in marine surface waters by biotic and abiotic processes with overall degradation rates of up to 20 % per day (King and Saltzman, 1997 and references therein).The abiotic degradation due to hydrolysis and transhalogenation is assigned with a large ε of 69 ± 8 ‰ (King and Saltzman, 1997) and the degradation of bromomethane by methylotrophic bacteria is assigned with an ε of 45 ± 10 ‰ (Miller et al., 2001).Simultaneously, bromomethane is produced in the surface water.Although the isotopic composition of the intrinsic bromomethane is unknown, it appears reasonable to presume the bromomethane emitted back into the atmosphere to be isotopically enriched considering its rapid degradation and the exceptional strong isotopic fractionation of this process.

Conclusions
In this study a simple and field suitable cryogenic sampling system for subsequent determination of carbon isotope ratios of a wide range of VOCs was developed.The dry shipper has shown to be a suitable and easy-to-use cooling source that can replace conventional dewars or cryostats in various applications.Up to 30 samples can be taken over a period of two weeks without the need of any technical infrastructure allowing for sampling campaigns in remote areas.Recovery rates for the entire sampling procedure ranged from 92 to 105 % with low standard deviations.The analytical repeatability (1 σ ) for carbon isotope determination on the 0.2 nmole level ranged from 0.5 ‰ to 2.8 ‰.With a sampling volume of 500 l, carbon isotope ratios of compounds with typical mixing ratios between 1 and 10 pptv can be determined with a precision better than 2 ‰.Nevertheless, the determination of these compounds in urban air masses is often hampered by high loads of hydrocarbons.
We reported isotope ratios for a broad range of VOCs in urban and marine air in Northern Germany.Several compounds have not yet been analyzed for their isotopic composition.For the organohalogens having been measured in previous studies the carbon isotope ratios found here are consistent with those previously reported.
The δ 13 C values for bromomethane from the urban site of −41.5 ± 3.3 ‰ are in excellent agreement with those reported by Bill et al. (2004).In contrast, bromomethane δ 13 C values were enriched in 13 C by about 10 ‰ in the coastal samples.A similar but less pronounced trend was observed chloromethane.We hypothesize that these differences are related to atmosphere-ocean exchange with fractionating biotic and abiotic degradation processes in the surface ocean and suggest carbon isotope ratio determination as a promising tool for better constraining the role of the ocean in the global cycle of these compounds.

Figure 3 : 29 Fig. 3 .
Figure 3: Reproducibility of the δ 13 C measurements versus the carbon amount for CH 3 Cl and CH 3 Br.Results are given as deviation from the mean.The respective means were calculated from all measurements with carbon amounts from 1.0 to 20 nmol.The mean δ 13 C values of the single component standards were -32.5 ‰ and -40.7 ‰ for CH 3 Cl and CH 3 Br, respectively.

Fig. 4 :Fig. 4 .
Fig. 4: Variability of the δ 13 C values of selected halocarbons in urban and coastal air sam Urban air samples are marked with red diamonds and coastal samples are marked with dots.

Table 1 .
Comparison of the carbon isotope ratios obtained with and without pre-separation for the Scott Speciality Gases TOC 15/17 standard and recovery rates for a sampling volume of 500 l.Results are only given for compounds that were free of interferences from other compounds.

Table 2 .
Averaged concentrations and isotopic values for all compounds reported in this paper from the coastal and the urban sampling site.