Supplementary information to : A combustion setup to precisely reference δ 13 C and δ 2 H isotope ratios of pure CH 4 to produce isotope reference gases of δ 13 C − CH 4 in synthetic air

Abstract. Isotope records of atmospheric CH4 can be used to infer changes in the biogeochemistry of CH4. One factor currently limiting the quantitative interpretation of such changes are uncertainties in the isotope measurements stemming from the lack of a unique isotope reference gas, certified for δ13C-CH4 or δ2H-CH4. We present a method to produce isotope reference gases for CH4 in synthetic air that are precisely anchored to the VPDB and VSMOW scales and have δ13C-CH4 values typical for the modern and glacial atmosphere. We quantitatively combusted two pure CH4 gases from fossil and biogenic sources and determined the δ13C and δ2H values of the produced CO2 and H2O relative to the VPDB and VSMOW scales within a very small analytical uncertainty of 0.04‰ and 0.7‰, respectively. We found isotope ratios of −39.56‰ and −56.37‰ for δ13C and −170.1‰ and −317.4‰ for δ2H in the fossil and biogenic CH4, respectively. We used both CH4 types as parental gases from which we mixed two filial CH4 gases. Their δ13C was determined to be −42.21‰ and −47.25‰ representing glacial and present atmospheric δ13C-CH4. The δ2H isotope ratios of the filial CH4 gases were found to be −193.1‰ and −237.1‰, respectively. Next, we mixed aliquots of the filial CH4 gases with ultrapure N2/O2 (CH4 l 2 ppb) producing two isotope reference gases of synthetic air with CH4 mixing ratios near atmospheric values. We show that our method is reproducible and does not introduce isotopic fractionation for δ13C within the uncertainties of our detection limit (we cannot conclude this for δ2H because our system is currently not prepared for δ2H-CH4 measurements in air samples). The general principle of our method can be applied to produce synthetic isotope reference gases targeting δ2H-CH4 or other gas species.

a high purity level of the CH 4 which is achieved by coupling the biogas plant to an industrial purification reactor that increases the CH 4 content to about 95% by removing H 2 S, H 2 O and CO 2 .The remaining 5% comprise mostly N 2 , O 2 , CO 2 and traces of H 2 S. For our purposes, the biogenic 10 CH 4 needed further purification.Non-CH 4 carbon containing molecules that could impact on the isotope measurements were removed.H 2 S, which possibly degrades the analytical systems was furthermore reduced.
To assess the composition and purity of the gases, mass abundance scans were performed by dual-inlet IRMS analysis for high purity fossil CH 4 , biogenic CH 4 as well as purified biogenic CH 4 .The mass abundance scans were then compared to analyze their content of non-CH 4 components.In general, mass abundance scans of a pure CH 4 20 show diverse spectra of masses because certain fractions of CH 4 molecules decompose and/or re-combine to secondary molecules within the ion source, as a result of the electron bombardment (Brunnée and Voshage, 1964).Ions producing these spectra can therefore be an artefact, suggesting the abundance of non existing gas species in the sample gas.
Converting mass abundances to gas compositions therefore introduces an error.The mass abundance scans are thus not the most accurate method to quantify impurities within a CH 4 gas.However, we will show that comparing mass abundance scans of CH 4 with different purity level allows for a sufficient estimate.

Carbon containing impurities
The performed mass abundance scans were evaluated by ISODAT 3.0, the software used to control the mass spectrom- A simple mass balance calculation based on the assumption of a 99.995 % purity level shows, that the average δ 13 C 75 isotope ratio of the impurities would have to be higher than 744 ‰ or lower than -856 ‰ to affect the determined carbon isotope ratios of the biogenic CH 4 by more than the given uncertainty.Because these extreme carbon isotope ratios are highly unlikely in naturally occurring gases, we conclude 80 that our biogenic CH 4 is sufficiently purified from non-CH 4 hydrocarbons.

H 2 S
The low H 2 S content of both purified and unpurified biogenic CH 4 samples could only be detected by the most sensitive 85 Faraday cup detector of the IRMS.Because this detector was saturated for the very abundant CH 4 ions, we related the H 2 S signal on m/z 34 to the O 2 signal on m/z 32 which is constant in both the purified and un-purified biogenic CH 4 .The purification step decreased the H 2 S by 75%.However, this is 90 at a very low concentration levels where the evaluation of the H 2 S peaks is unreliable.Given that H 2 S was reduced to prevent degradation of the analytical system and is not expected to alter the measurements on CH 4 isotope ratios we consider the observed reduction as sufficient.
95 2 Effect of system leakage

Monitoring Argon in the sample
Mass abundance scans using dual-inlet IRMS were performed on each combusted CH 4 sample.Any leakage of laboratory air into the offline combustion setup would in-100 crease the relative abundance of Argon (Ar) compared to the mass abundance of CO 2 .We interpreted the ratio of the peak heights of Ar (m/z 40) to CO 2 (m/z 44) as indicator for laboratory air leakage into the offline combustion setup during a combustion process.To increase the sensitivity of this com-105 parison, we used the Ar signal of a Faraday cup with a 100 times stronger.The resulting ratio is 0.006 +/-0.0003 for the pure CO 2 -40339.We found the same ratio in the combusted samples stemming from fossil CH 4 .We found a ratio of 0.03 in the biogenic CH 4 which is higher due to the containing 110 5% of atmospheric air.We found these ratios to be stable for the respective CH 4 type.These measurements give us confidence that the measurements were not affected by leakage into the system during the aliquotation, the combustion or the sample transfer into the dual-inlet IRMS.

Sensitivity test for laboratory-air leakage
The effect of an undetected CO 2 blank can be estimated with the following sensitivity test.The offline combustion system was evacuated, closed off and tested for leaks in over-night test prior to each measurement.A sample would only be in-120 troduced if the pressure increase overnight stayed below the detection limit of our piezoelectric pressure gauge (1 mbar precision).We give a conservative estimate of the maximum effect an undetected leak would have as blank contribution.We assume a pressure increase of 1 mbar in the system of ∼ 125 600 ml volume, which corresponds to a leakage of 0.6 ml.If we conservatively assume 1 part per thousand of the leakage is CO 2 with a δ 13 C isotope ratio of -7 ‰ a small biogenic CH 4 sample with a δ 13 C of -56 ‰ would be most affected.A mass balance calculation shows that the maximum possi-130 ble blank contribution would affect a 120 ml sample of biogenic CH 4 with -56 ‰ by 0.0002 ‰ which is by two orders of magnitude lower than the precision of our method.We conclude that any undetected potential blank has no significant effect on our results.The blank contribution is most 135 likely even lower because most of the combusted samples were about 50% larger than anticipated in the calculation.Also, the lab air contains more depleted CO 2 from human breath which is more depleted in 13 C. Any additional CH 4 leakage would furthermore reduce this blank effect due to its 140 lower isotopic leverage.

References
Table 1.Table S1: Mass abundances in the purified biogenic CH4 as evaluated by ISODAT 3.0.The evaluation parameters were extremely tuned to be able to capture very small abundances and therefore compromising the accuracy of the quantification.Hence, peak numbers 8 and 9 show opposing sizes in peak area and height.The CH4-derived ions were identified following Brunnée and Voshage (1964) 115

Fig. 1 .
Fig. 1.Figure S1: Shown is a scan of mass abundances between m/z 5 and m/z 60 when purified bio-methane is introduced into the ion source of the dual inlet IRMS.The y-axis is set to logarithmic scale to better visualize the low abundances.Chemical formulas and m/z ratios identify the respective peaks, indicated by the arrows.
Fig. 1.Figure S1: Shown is a scan of mass abundances between m/z 5 and m/z 60 when purified bio-methane is introduced into the ion source of the dual inlet IRMS.The y-axis is set to logarithmic scale to better visualize the low abundances.Chemical formulas and m/z ratios identify the respective peaks, indicated by the arrows.
to the sum of peak areas from mass abundances that possibly results of non-CH 4 hydrocarbons (m/z 26, 29, 30, 44) which were weighted for the maximum number of carbon atoms.This is based on the assumption that the entire signal on m/z 26, 29, 30 and 44 is derived 50 from hydrocarbons and that all ions causing the peak on m/z 44 comprised of C 3 H + 8 with three carbon atoms instead of CO + 2 .This is important because one molecule of C 3 H 8 produces 3 molecules of CO 2 in the combusted sample.Therefore, one molecule of C 3 H + 8 contributes three times stronger Based on the comparison of the mass abundance scans from purified biogenic CH 4 and fossil CH 4 , we can assume a similar purity level.
45molecules was compared 55 to the signal on m/z 44 than one molecule of CO 2 .We also assume that non-CH 4 hydrocarbons are stable in the 4 reveals that about 0.67% of the detected ions were derived from ions containing at least two carbon atoms.The company pro-65 viding the high purity fossil CH 4 specifies the purity level of its CH 4 N45 with 99.995% with a mixing ratio of non-CH 4 hydrocarbons ≤ 20 ppmv.Therefore, we conclude that inside the ion source. .