An automated GC-CGC-IRMS setup to measure palaeoatmospheric δ 13 CCH 4 , δ 15 N-N 2 O and δ 18 O-N 2 O in one ice core sample

Abstract. Air bubbles in ice core samples represent the only opportunity to study the mixing ratio and isotopic variability of palaeoatmospheric CH4 and N2O. The highest possible precision in isotope measurements is required to maximize the resolving power for CH4 and N2O sink and source reconstructions. We present a new setup to measure δ13C-CH4, δ15N-N2O and δ18O-N2O isotope ratios in one ice core sample and with one single IRMS instrument, with a precision of 0.09, 0.6 and 0.7‰, respectively, as determined on 0.6–1.6 nmol CH4 and 0.25–0.6 nmol N2O. The isotope ratios are referenced to the VPDB scale (δ13C-CH4), the N2-air scale (δ15N-N2O) and the VSMOW scale (δ18O-N2O). Ice core samples of 200–500 g are melted while the air is constantly extracted to minimize gas dissolution. A helium carrier gas flow transports the sample through the analytical system. We introduce a new gold catalyst to oxidize CO to CO2 in the air sample. CH4 and N2O are then separated from N2, O2, Ar and CO2 before they get pre-concentrated and separated by gas chromatography. A combustion unit is required for δ13C-CH4 analysis, which is equipped with a constant oxygen supply as well as a post-combustion trap and a post-combustion GC column (GC-C-GC-IRMS). The post-combustion trap and the second GC column in the GC-C-GC-IRMS combination prevent Kr and N2O interferences during the isotopic analysis of CH4-derived CO2. These steps increase the time for δ13C-CH4 measurements, which is used to measure δ15N-N2O and δ18O-N2O first and then δ13C-CH4. The analytical time is adjusted to ensure stable conditions in the ion source before each sample gas enters the IRMS, thereby improving the precision achieved for measurements of CH4 and N2O on the same IRMS. The precision of our measurements is comparable to or better than that of recently published systems. Our setup is calibrated by analysing multiple reference gases that were injected over bubble-free ice samples. We show that our measurements of δ13C-CH4 in ice core samples are generally in good agreement with previously published data after the latter have been corrected for krypton interferences.


Introduction
Methane (CH 4 ) and nitrous oxide (N 2 O) are important longlived greenhouse gases that play a significant role in Earth's radiative budget (Solomon et al., 2007).The analysis of ancient air as archived in air bubbles within the polar ice sheets has significantly improved the understanding of Earth's atmospheric and biogeochemical variability.Especially CH 4 records have been used as proxy on the extension of wetlands on the stability of marine clathrates as CH 4 source.On glacial timescales, periods of warmer climate in the Northern Hemisphere correlated with increased concentrations of CH 4 and N 2 O in the atmosphere (Flückiger et al., 2004;Loulergue et al., 2008;Schilt et al., 2010).Moreover, the mixing P. Sperlich et al.: CH 4 & N 2 O isotopes in ice core samples ratios of both gases have significantly increased during industrialization, largely due to human activities (Solomon et al., 2007).Atmospheric mixing ratios of 1774 and 319 ppb have been published in the last IPCC report for CH 4 and N 2 O, respectively (Solomon et al., 2007).With the prospect of even further increasing mixing ratios and their impact on future climate, it is important to thoroughly understand the biogeochemical processes related to both gases.Brenninkmeijer et al. (2003) describe how the isotope fractionation of specific source and sink processes affect the integrated isotopic composition of the respective trace gases in the atmosphere.In an inverse approach, ice core isotope records of CH 4 and N 2 O provide distinct constraints on biogeochemical processes that can be linked to the variability observed in the CH 4 and N 2 O mixing ratios on decadal to millennial timescales (Sowers et al., 2003(Sowers et al., , 2005;;Sowers, 2006Sowers, , 2009;;Ferretti et al., 2005;Schaefer et al., 2006;Fischer at al., 2008;Bock et al., 2010b;Melton et al., 2011a;Sapart et al., 2012).Because ice core records of CH 4 and N 2 O isotopic composition indicate the natural response of specific greenhouse gas sinks and sources to palaeoclimate changes, this information is of great interest to global warming predictions.
Here, we present a method to simultaneously measure δ 13 C isotope ratios of CH 4 as well as δ 15 N and δ 18 O isotope ratios of N 2 O in a single ice core, firn gas or atmospheric sample.By melting ice core samples under vacuum, between 20 and 50 mL STP (standard temperature and pressure) of air can be extracted from 200-500 g of ice for isotopic analysis.Alternatively, atmospheric samples and reference gases can be injected into the system.The system is highly automated and comprises custom made units to separate CH 4 and N 2 O from the main air components (N 2 , O 2 , Ar) and other trace gases (CO 2 , CO) before using a modified gaschromatography combustion unit coupled to a single isotope ratio mass spectrometer (GC-C-IRMS) in continuous-flow mode for isotope ratio determination.The combustion unit converts CH 4 to CO 2 so that the δ 13 C-CH 4 is measured as δ 13 C-CO 2 on the triple collector system of a mass spectrometer (Merritt et al., 1995).Our system includes permanent oxidation, a post-combustion cryo-trap and a second GC column, similar to Melton et al. (2011b).This GC-C-GC-IRMS method ensures a stable oxidation of the combustion reactor with minimized oxygen load into the IRMS, and it excludes interferences of the δ 13 C-CH 4 measurement with Kr inside the IRMS (Schmitt et al., 2013).The system is anchored to the international isotope scales using reference gases that were synthesized after Sperlich et al. (2012) for δ 13 C-CH 4 or calibrated by intercomparison measurements with two external laboratories for δ 15 N-N 2 O and δ 18 O-N 2 O.A novelty of this system is the isotope analysis of two isobaric gas species (CH 4 -derived CO 2 and N 2 O) from one ice core sample using one mass spectrometer in an online measurement mode.Furthermore, we introduce a gold catalyst for quantitative oxidation of CO which -to our knowledge -has hitherto not been used in setups for atmospheric measurements.We suggest the presented system as a powerful tool to provide high-precision isotopic analysis of palaeoatmospheric CH 4 and N 2 O.

Extraction system
The extraction unit (Fig. 1) represents the interface for sample and reference gas introduction into the analytical setup and it includes the first step to separate the analytes from the main air components (N 2 , O 2 and Ar).The sample lines of the extraction system are made of 1/4 stainless steel (SST) tubes and Swagelok components (Swagelok, USA) except for the two six-port valves (V1 and V2) that are manufactured by VICI (VICI, USA) and connected to 1/16 SST tubing.All connections are either welded or sealed with metal gaskets to exclude artefacts due to out-gassing of polymer gaskets (Sturm et al., 2004).All analytical lines are either continuously flushed by helium (He, 99.9995 %, Air Liquide, Denmark) or permanently evacuated by the turbo-pump (Pfeiffer, Germany).
The gas manifold enables the injection of gas from up to four different gas tanks or sample flasks into the extraction unit via a mass flow controller (referred to as MFC, manufactured by MKS, model 1179A, specified for N 2 , 200 mL min −1 ).A 100 mL sample volume parallel to the MFC enables aliquotation by expansion of gases from the gas manifold into the volume.The pressure reading of gauge "P1" (Keller, 21Y, max. 2 bar) is used to calculate the amount of air in the sample volume.We use two principle methods to introduce samples: 1. Ice core samples are placed in the 1.25 L melt vessel (SST glass adapter, DN 100, MDC vacuum, UK).Reference gases or air from flask samples can also be introduced into the melt vessel or into the water trap (T1) using manual valves and either the MFC or the sample volume.T1 comprises a 20 cm DN 50 SST tube that is welded to a quick flange connector and sealed with an aluminium gasket.Both 1/4 tubes connecting the water trap are welded onto the opposing quick flange cap.The water trap is filled with glass beads of 3 mm diameter and cooled with a dry-ice/ethanol bath to −78  He).A capillary above T2 maintains a He flush flow of 2 mL min −1 when the valves of T2 towards V1 are closed (Fig. 1).The SST line and the valves that connect the melt vessel with T1 and T2 are thermally insulated and constantly held at 55 • C with a PID-controlled heating wire.
2. Alternatively, the MFC can inject gas from pressurized tanks into the extraction flow directly when V1 and the manual valves are set accordingly (Fig. 1).
For both sample introduction methods, the sample gas is transported by the extraction flow through the CO oxidizer.
Our CO oxidizer is comprised of a 10 cm 1/4 SST tube holding a gold catalyst over a length of 5 cm (Aurolite ™ , Au/TiO 2 , Strem Chemicals) between two glass-wool plugs that are kept in place by SST meshes at both ends.The CO-oxidizer column is held at 60 • C by a PID-controlled heating wire.The CO oxidizer is followed by the separation trap (T3), which is a modification of the principle described by DesMarais (1978).≤ −122 • C, while most N 2 , O 2 and Ar pass through.The effluent of T3 is monitored with a thermal conductivity detector (micro-TCD, VICI) to enable CH 4 and N 2 O mixing ratio analysis from TCD and IRMS peak areas.From T3, the sample is transported through a glass tube of 6 mm outer diameter (OD) and 600 mm length that holds Ascarite ™ (NaOH, Sigma-Aldrich) to remove CO 2 and Mg(ClO 4 ) 2 (Merck) to remove H 2 O.

GC-C-GC-IRMS system
The final purification and analysis of the air samples occurs via a combined GC-C-GC-IRMS setup (Fig. 2) after the ice core and air extraction unit.All valves are twoposition VICI valves, and all tubing either 1/16 SST tubes or fused silica capillaries.Valve V4 provides the transfer flow (10 mL min −1 He) and the analytical flow (1 mL min −1 He).
Coming from T3, the sample is cryo-focussed on T4, a 55 cm section of GC-column (PoraBOND Q, 0.25 mm ID, Agilent) that can be submerged in LN 2 .V4 allows for routing of either the transfer or the analytical flow through the injector valve V3 (2 µL, VICI, USA) for ∼ 1 nmol sized injections of pure CH 4 and N 2 O into T3 or T4.These injections were used to tune the timing of the sequences and to monitor the reproducibility and drift during the test phase of the setup.
The transfer flow from V4 to the ice extraction unit returns to valve V5 of the GC-C-GC-IRMS section (Fig. 2).This helium flow cryo-focuses the purified CH 4 and N 2 O samples in T4.V5 then routes the analytical flow through T4 to introduce the sample gases into GC 1 (PoraBOND Q, 0.25 mm ID, 25 m length, Agilent).GC 1 is submerged in an ethanol bath at −5 • C to increase the chromatographic separation between CH 4 and N 2 O to ∼ 160 s.GC 1 can be heated to 80 • C for desorption of H 2 O.While it is possible to measure δ 15 N and δ 18 O in N 2 O directly using GC-IRMS, CH 4 samples need to be combusted to CO 2 for δ 13 C analysis (Merritt et al., 1995).The CH 4 elutes from GC 1 before N 2 O and is routed to the combustion unit via V6.The combustion reactor contains three Ni wires (99.994 %), three Cu wires (99.9999 %) to store and provide oxygen during combustion and two Pt wires (99.997 %) to catalyse the combustion, (all wires are 0.1 mm OD, Alfa Aesar, UK).A small flow of oxygen is constantly added to guarantee maximum oxidation of the combustion reactor at all times in order to ensure a high CH 4 combustion rate (Cullis and Willatt, 1983).
Water originating from the combustion process is removed by a Nafion membrane (60 cm length, TT-020, Perma Pure, USA) and a helium counter-flow of 7 mL min −1 at −5 • C. The eluting CH 4 -derived CO 2 is trapped on T5 comprising of a fused silica capillary (350 µm ID, 55 cm length) submerged in LN 2 .To increase the trapping efficiency and to prevent CO 2 loss, the capillary trap T5 contains a Ni wire (99.994 % Ni, 100 µm OD, Alfa Aesar), (e.g.Brand, 1995;Behrens et al., 2008).When the CH 4 -derived CO 2 is trapped in T5, V6 redirects the GC flow to analyse the N 2 O first.V6 and V7 are then switched, and T5 is removed from the LN 2 bath to pass the CH 4 -derived CO 2 through GC 2, which is held at 40  The uncertainty estimates are based on the standard deviation (1σ ) apart from a , which is described by Sperlich et al. (2012).The b indicates when the referencing method of NEEM for the isotopic composition of N 2 O is explained by Sapart et al. (2011).

Referencing to the isotope scales
We used four different reference gases in the described setup (Table 1).GIS (glacial isotope standard) refers to a synthetic air mixture that was prepared after Sperlich et al. (2012) with additional N 2 O.An atmospheric air tank was sampled in the year 2008 at a clean-air site of the NEEM camp in northwest Greenland and is hereafter referred to as NEEM.Furthermore, we use two synthetic air mixtures called AL and NOAA that were provided by Air Liquide (Teknisk Luft, Air Liquide, Denmark) and the National Oceanic and Atmospheric Administration (NOAA, Boulder, USA), respectively.The isotope ratios of all gas standards are referenced to the VPDB isotope scale for δ 13 C-CH 4 and to the δ 15 N scale of N 2 air and the δ 18 O VSMOW scale for N 2 O, respectively.The isotopic composition of N 2 O in NEEM is prescribed by extrapolating the atmospheric trend between 1990 and 2002 (Röckmann and Levin, 2005) to July 2008, which is the sampling time of the NEEM cylinder.The integrity of our calibration scale for N 2 O isotopic composition was tested by inter-calibration measurements of NEEM, AL and NOAA by the Institute for Marine and Atmospheric Research in Utrecht (IMAU), the Netherlands (published in Sapart et al., 2011).The isotopic composition of N 2 O in GIS was referenced on the setup described in this paper by calibrating against NEEM and AL.The δ 13 C-CH 4 in NEEM was calibrated by IMAU and the Institute for Climate and Environmental Physics at the University of Bern, Switzerland (Bern), (Jochen Schmitt, personal communication, 2011).GIS was independently referenced for δ 13 C-CH 4 against RM 8563 (Verkouteren, 1999) with the δ 13 C-CO 2 of RM 8563 assigned by Coplen et al. (2006).GIS was previously shown to be in excellent agreement with the externally calibrated NEEM air (Sperlich et al., 2012).AL was referenced for δ 13 C-CH 4 on the described setup based on NEEM and GIS, while NOAA was calibrated for δ 13 C-CH 4 by NEEM only.The mixing ratios of CH 4 and N 2 O in GIS, NEEM and AL were measured at the Max-Planck-Institute for Biogeochemistry (Armin Jordan and Bert Steinberg, personal communication, 2012), while NOAA was provided with certificates.GIS was chosen as working standard for ice core measurements due to the similar mixing ratio of CH 4 .Air samples with higher CH 4 mixing ratios, atmospheric samples such as NOAA, firn air or atmospheric samples, were referenced against NEEM.

Measurement correction and referencing protocol
The ice core measurement routine is shown in Table 2 and the routine for atmospheric gases from air tanks is shown in Table 3. Blocks of three GIS measurements are bracketing the ice core sample and the quality control standard (QCS) measurements.For each reference gas measurement, 40 mL (STP) of GIS were extracted from T1.The offset between the average isotope ratio determined for each block, and the target isotope ratio assigned to GIS is considered to indicate the daily offset of the system to the international isotope scales, including the system drift between the beginning and the end of the measurement day.The isotope ratio measurements of the ice core samples and the quality control standards are corrected for the sample size offset (Sect.3.7) first, and then for the system offset as determined by GIS according to Werner and Brand (2001).

Ice core sample and system preparation
Measurements of ice samples begin with cooling of the H 2 O trap T1 in a dry-ice/ethanol bath.To increase the heat conduction within the glass bead bed, T1 is pressurized with helium to ∼ 1500 mbar for ∼ 2 h.GC 1 is placed in the ethanol bath at −5 • C if it has previously been heated to 80 • C.Meanwhile, ice core samples which are stored at −25 • C are prepared in the cold room at −15 • C. All surfaces of ice core samples are cleaned by removing the top 3-5 mm with a bandsaw.Attached saw dust is then removed with a scalpel or brushed off while the ice sample is carefully checked for anomalous features.To prevent contamination by laboratory air and/or drill-fluid intrusion, cracked parts are removed from each sample.Also, layers with exceptionally high content of dust or soot particles were cut out to prevent artefacts (Rhodes et al., 2013).On average, 30 % is removed from each sample.The decontaminated ice sample and a glass-coated magnet stir bar are then placed inside the melt vessel, which is sealed with a copper gasket and fastened with 16 bolts to 25 Nm torque.Two samples are prepared and stored in a −20 • C chest freezer to be analysed the same day.The analytical system is finally prepared by evacuating T1 as well as all lines and pressure regulators of the atmospheric reference gases connected to the gas manifold (Fig. 1).The measurement routine is started when the pressure inside the extraction system reaches 10 −3 mbar, indicating the system is sufficiently leak-tight and dried after previous analysis.
After every day of ice core measurements, the melt vessel is cleaned with ultra-pure water and detergent soap for laboratory glassware (Alconox ™ , USA), as suggested by Mitchell et al. (2011), before it is dried at 80 • C overnight.To remove trapped water from T1, it is heated to 150 • C with a PID-controlled heating sleeve.The elevated temperature of the extraction line (55 • C, Fig. 1) promotes the drying efficiency of the whole extraction line.Under vacuum, any water is rapidly transferred from the heated parts into the water trap (LN 2 cooled) close to the turbo-pump.This water trap can easily be opened to remove the water.With this method, we avoid opening the connectors in the extraction line.
Table 3. Measurement sequence for air samples from flasks or tanks.The presented sequence is used for triplet measurements of two samples, bracketed by reference gas triplets.Column 2 describes the function of the measured gas, which is described and quantified in column 3 and 4, respectively.Column 5 indicates that all samples are injected through the mass flow controller (MFC) directly into T3 or to be extracted from T1, as indicated in column 6. 2.6 Measurement procedure

Ice core samples
In order to minimize the time between the measurements, samples are introduced into the extraction unit while the previous sample is analysed in the GC-C-GC-IRMS section of the setup.This results in a sample interval of only 42 min.
Ice core samples and reference gases are treated as similar as possible in order to balance analytical effects during analysis, following the principle of identical treatment (Werner and Brand, 2001).However, one difference exists in the application and extraction of ice core and reference gas sample: the MFC is used to inject reference gas into T1, from where it is extracted into T2.In contrast, ice core samples release the air during melting inside the melt vessel from where the gas is continuously cryo-pumped through T1 into T2.This is the only gas-handling difference between ice core sample and reference gas.We tested for isotope fractionation effects based on the different gas handling but found no significant isotope fractionation in our setup (see Sect. 3.7).The continuous extraction technique minimizes the pressure and the time the gas is in contact with the melt water as to reduce gas dissolution.
Every analysis begins with the evacuation of the extracting unit.The thoroughly cooled T1 represents the cavity for reference gas extractions, and is evacuated for 6 min prior to reference gas introduction.To measure an ice core sample, the melt vessel is attached during the previous reference gas measurement so it can be evacuated for 20-30 min without delaying the measurement routine.During this time, the melt vessel is cooled in an ice water bath to prevent the sample from melting ahead of schedule.The measurement routine then continues with cooling T2 in LN 2 for 8 min in preparation for the extraction of ice core air or reference gases.Towards the end of cooling T2, reference gases are injected into T1, or the ice bath under the melt vessel is replaced by a hot water bath (60 • C) to start melting the ice.When T2 has cooled for 8 min, the air extraction is started by opening the valve between T1 and T2 and the melt vessel for ice extractions.After 40 mL (STP) of reference gas have been extracted for 13 min, the pressure within the extraction unit (P2) decreased from 14 mbar to 0.07 mbar, indicating 99.5 % extraction efficiency.The pressure decrease during extractions is very reproducible, and only varies with the amount of air and the variable melting time of ice core samples.By ensuring that reference gas and ice core samples are extracted to similar pressures, isotope fractionation effects based on variable completeness of extraction cancel out.The IRMS operating software (ISODAT) is started as soon as the pressure in the extraction system is lower than 0.11 mbar (P2), thereby ensuring identical extraction efficiency for samples.
When the extraction has ended, reference gases and samples are subject to identical analytical procedures, precisely timed by the ISODAT script.First, T3 is cooled to −154 • C within 1 min and is held constant by the PID controller.After T3 has cooled for 4 min, the extraction is stopped.T2 gets heated from −196 to 50 • C in less than 1 min, thereby desorbing the sample within the extraction flow and transferring the sample from T2 through the CO oxidizer to T3.T3 separates N 2 , O 2 and Ar from the sample gas mixture and retains CH 4 , N 2 O, CO 2 and Kr on the Hayesep D. At the same time, the temperature of T3 is increased to −122 • C to improve the separation performance of T3 (Umezawa et al., 2009).Residual N 2 , O 2 and Ar will be separated later in the GC-C-GC-IRMS section.The TCD signal to measure the amount of N 2 , O 2 and Ar is recorded using a LabView script.The separation in T3 is completed after 400 s.V2 is then switched and T3 heated to −10 • C so the transfer flow can carry the desorbing trace gases through the chemical trap into the capillary trap T4, which was submerged in LN 2 10 s before.CH 4 and N 2 O are focussed in T4 after 360 s, V2 is then switched and T3 heated to 50 • C to vent residual water.Next, V5 is switched and the analytical flow transports both sample gases from T4 through GC 1 as T4 gets lifted from the LN 2 bath.It takes about 220 s for the CH 4 to pass through GC 1 and to enter the combustion oven.After 200 s, the post-combustion trap T5 is lowered into a LN 2 bath to trap the CH 4 -derived CO 2 .
After 80 s of post-combustion trapping, all CH 4 -derived CO 2 is trapped in T5, and V6 gets switched to route the analytical flow directly to the open split and the IRMS for the isotopic measurement of N 2 O.Meanwhile, V7 gets switched to a pure helium flow of 1 mL min −1 to remove the oxygen from T5.When the N 2 O measurement is completed, V6 gets switched to route the CH 4 -derived CO 2 from T5 through GC 2 into the open split for isotopic analysis in the IRMS.In the mean time, T1 and T2 are evacuated in preparation for the following sample.

Air samples
The analysis of air samples is similar to the analysis of ice core samples, and differs only in the fully automated introduction of air samples if the samples are applied from pressurized vessels.Air tanks can be connected to the gas manifold where the MFC introduces the samples through V1 into the extraction flow (Fig. 1).A sample flow time of 90 s allows for the sample and helium flows to equilibrate before V2 is switched to direct the sample into the pre-cooled T3.After this point, the analysis is exactly the same as described in Sect.2.6.1;however, only one sample is introduced and analysed at a time.If the air samples are at sub-ambient pressures, they are introduced to T2 through the sample volume to be extracted like the reference gases during ice core measurements.

Sampling for system blank
System blank tests are determined after every ice core sample extraction (Table 2).For this purpose, the valves of the extraction line get closed when the ice extraction is completed.The sample vessel with the extracted sample is thereby kept under vacuum, but it can collect air from potential leakage into the extraction unit.The extraction volume is then extracted for 5 min and processed as a regular sample.

CO oxidizer
Using the post-combustion trap strictly requires the elimination of any potential interference of non-CH 4 gases (CO and CO 2 ) with the CH 4 -derived CO 2 in the mass spectrometer.To exclude any spurious CO contribution, CO is quantitatively oxidized to CO 2 and then removed by the Ascarite trap (Fig. 1).Schütze reagent is often used in analytical setups to oxidize CO to CO 2 at room temperatures when the CO is to be analysed for its isotopic composition of δ 13 C and δ 18 O (e.g.Brenninkmeijer, 1993).Alternatively, Sofnocat or hopcalite can be applied as oxidation catalyst (e.g.Kato et al., 1999).However, the CO-oxidation efficiency of both catalysts decreases with increasing moisture content (McPherson et al., 2009).To our awareness, this is the first system described for atmospheric measurements using Aurolite as catalyst to oxidize CO to CO 2 .The efficiency of Aurolite is reported to even improve with increasing moisture content in the sample gas stream (e.g.Date et al., 2002;McPherson et al., 2009), making it particular suitable for CO elimination in air extracted from melted ice core samples.Our CO oxidizer is heated to 60 • C to increase the CO oxidation  (Date et al., 2002), while the oxidation of CH 4 can be ruled out at temperatures below 130 • C (Walther et al., 2009).We chose Aurolite on TiO 2 support because of its superior conversion efficiency and gold particle stability compared to Al 2 O 3 and ZnO (Walther et al., 2009).
To test the system, synthetic CH 4 -and CO 2 -free air with a CO mixing ratio of 350 ppb was mixed in the setup described by Sperlich et al. (2012).This air mixture was applied to the system with and without the CO oxidizer as well as with a trap holding Schütze reagent (iodine pentoxide, 99.99 %, Sigma-Aldrich) inside a 6 mm OD glass tube over a length of 150 mm.The timing of the GC-C-GC-IRMS system was adjusted to fully combust, trap and analyse the residual CO that eluted from GC 1.Note that the CO that was oxidized in the CO oxidizer or in the Schütze reagent was chemically trapped on the Ascarite trap (Fig. 1) and therefore not detected.A series of experiments with Aurolite showed no CO-derived CO 2 peak that exceeded the normal blank peak, while up to 90 % of the CO passed through the Schütze reagent unoxidized.Our tests suggest that Aurolite quantitatively oxidizes CO in atmospheric samples and is significantly more reliable to oxidize CO than commercial Schütze reagent, and it also has the benefit of being immune to moisture.Based on the gold nanoparticle size stability reported by Date et al. (2002) and Walther et al. (2009), and the fact that our CO oxidizer is in contact with far less than 1 L of air per measurement day, we expect the CO oxidizer to be stable for many years.

Preparing the IRMS for measurement of two different gases
While only a small fraction of the analyte molecules that enter the ion source of an IRMS get ionized and reach the detector, other sample molecules can be subject not only to unintended physical and chemical interactions with the analytical system itself but also with molecules of other gas species present in the analyser.Potential effects include system memory, adsorption and desorption of gases altering the background levels, ion reactions (Anicich, 1993), reactions enabled by the hot filament and a combination of all (Brand, 2004).Measuring two isobaric gases such as N 2 O and CO 2 in the same sample on the same IRMS can reduce the precision significantly if the conditions inside the IRMS vary throughout the measurements (Carter and Barwick, 2011).To increase the stability of our analyser, we inject a large pulse of pure reference gas into the ion source for 1 min before the start of each measurement sequence.The following measurement sequences are comprised of 10 or 12 flat-topped peaks of pure CO 2 or N 2 O reference gas for CH 4 -derived CO 2 and the N 2 O measurement, respectively (Fig. 3).This enables simultaneous improvement and monitoring of the reproducibility before the sample peak enters the IRMS.The standard deviation of the three flat-topped reference gas peaks preceding the sample is 0.029 ‰ for (marked with star) includes a very small CO 2 remainder (not visible) stemming from system blank and incomplete CO 2 removal by the Ascarite.We use its isotopic composition to monitor the performance of the Ascarite trap (Fig. 1).
δ 13 C-CO 2 , 0.061 ‰ for δ 15 N-N 2 O and 0.086 ‰ for δ 18 O-N 2 O, respectively (average of 66 measurement sequences).These values refer to reference gas peak 8, 9 and 10 in the CH 4 -derived CO 2 sequence and 9, 10 and 12 in the N 2 O sequence (peak number 11 in the N 2 O sequence is not considered due to the small contribution of CO 2 (caption of Fig. 3).The traces show baseline separation for ≥ 5 s between reference gas and sample peaks in all chromatograms, which is the time interval used for baseline determination as part of the peak integration parameters.Measurements of atmospheric and synthetic air mixtures with precisely referenced δ 13 C-CH 4 (Sperlich et al., 2012) suggest this conditioning method of the IRMS showed no significant η effect (Brand, 2004) over δ 13 C-CH 4 range of 5 ‰, reflecting the magnitude of δ 13 C-CH 4 variability in firn and ice core samples (e.g.Fischer et al., 2008).

Excluding krypton interference
Krypton -which is abundant in the atmosphere at ∼ 1 ppm (Aoki and Makide, 2005) -was found to co-elute from the GC column with CH 4 and interfere with the analysis of CH 4derived CO 2 (Schmitt et al., 2013).The system described here deploys a post-combustion cryo-trap followed by a second GC column (Fig. 2).This combination ensures cryogenic  4).This test showed a crippled peak from 86 Kr ++ on m/z 43 which precedes the δ 13 C-CH 4 measurement of atmospheric air with baseline separation of 10 s between Kr-and the CH 4 -derived CO 2 peak on m/z 43 (Fig. 4).In this case, the sample peak produced a small signal on m/z 43, as can also be seen for the three reference gas peaks (pure CO 2 ) at the beginning of both sequences.
In comparison, injections of pure CH 4 via V3 into T4 underwent exactly the same analytical steps in the GC-C-GC-IRMS section but did not show the crippled 86 Kr ++ peak on m/z 43 at all (Fig. 4).This test identifies the crippled peak from 86 Kr ++ in the air sample and proves that our δ 13 C-CH 4 measurements are not affected by Kr interferences.

Shot noise
We apply the approach of Merritt et al. (1994) to calculate the shot noise for the different analytes and the average peak areas of reference gases and samples.We calculated a shot-noise range on the 1-σ level of 0.04-0.05‰ for δ 13 C-CH 4 , 0.11-0.17  the variability of the sample preparation.Generally, the shotnoise ratio is larger for N 2 O than for CO 2 due to the lower abundances of the rare isotope (1.1 % 13 C, 0.37 % 15 N and 0.201 % 18 O, e.g.Sessions, 2006) and the signal intensity of N 2 O as compared to CH 4 -induced CO 2 , due to lower mixing ratios (e.g.Schilt et al., 2010) and ionization efficiency (Friedli and Siegenthaler, 1988;Ghosh and Brand, 2004).

Sample recovery of injected air standards
Figure 5 shows the sample recovery as the response of the system to sample size variations between 0.5 and 3.2 nmol CH 4 and 0.3 and 1.1 nmol N 2 O.The measurements are based on the extractions of air samples between 20 and 70 mL, respectively (Sect.3.7).Since the valves to route the MFC injection of reference gases into the extraction line were manually controlled, a small variability of the applied volume and hence in IRMS peak area of CH 4 -induced CO 2 and N 2 O could not be avoided.The linear regression of the average IRMS peak area over the injected sample size shows an R 2 of 0.998 and 0.987 for CH 4 and N 2 O, respectively.We therefore conclude the extraction unit and the components of the GC-C-GC-IRMS system responded linearly to sample size variability, indicating appropriate system design and timing steps enabling quantitative analysis within the expected sample size range.

System blank tests
The detected system blanks represent an integrated signal of system leakage and contamination within the analytical gas streams, but more importantly, they represent ice-samplespecific leakage through the melt vessel gasket as well as sample remnants from incomplete extractions.Another theoretical source of N 2 O blank peaks is the microbial production of N 2 O from NO 3 and NH 4 .This N 2 O production might take place after the ice core extraction, while the melted sample remains in the melt vessel (Jochen Schmitt, personal communication, 2013).However, we have no evidence for this process and cannot distinguish between all contributing sources.We compared the IRMS peak areas of each blank test to the IRMS peak areas of the preceding ice core sample for 35 ice core measurements, and found blank peak areas with an average size of 1.5 and 3.0 % of the preceding CH 4 and N 2 O sample, respectively.The N 2 O blank size neither correlates with the CH 4 blank nor could any anomalous signal be detected on the TCD, precluding leakage from laboratory air.We speculate that the higher N 2 O blank values resulted from ice core air remnants due to higher solubility as compared to CH 4 or in situ production during melting.In fact, 12 out of 35 blank tests showed a peak area that exceeded 5 % of the peak area of the preceding N 2 O sample.The blank tests of the other 23 ice core samples averaged to 1.1 % of the sample peak area.These tests might indicate limitations in the extraction efficiency for N 2 O during the melt extraction as compared to the extraction efficiency of CH 4 .Because the isotope ratios of such small peaks are not unambiguously detectable, we do not apply blank corrections to our isotope measurements.
During the test phase of the setup, we installed a helium flow of 200 mL min −1 to strip dissolved gases from the melted ice core sample after the majority of the gases have been extracted as described in Sect.2.6.1.Unlike Behrens et al. (2008) and Bock et al. (2010a), we pumped on the extraction line at the downstream end of T2 to absorb the sample gases on T2 while removing the helium for 5-10 min.Measurements of ice core samples that were extracted with the stripping method did not show a significant blank reduction or a reduced variability of isotopic analysis for CH 4 and N 2 O, and did not conclude a more efficient extraction method.The stripping technique was thus excluded from the setup.

System calibration with bubble-free ice measurements
We follow the approach of Sowers and Jubenville (2000), Bock et al. (2010a) and Sapart et al. (2011) to calibrate our ice core extraction systems using artificial, bubble-free ice (BFI).We measured a total number of 35 air standard extractions over eight BFI samples.The first air standard was injected over the frozen BFI sample in the evacuated melt vessel.The BFI was then melted completely before the extraction was started.After the first extraction, the melted BFI sample was kept in the melt vessel under vacuum.Air standards were then injected over the melted BFI sample and the extraction started after 2 min.The air standards were taken from four different air mixtures (GIS, AL, NEEM, NOAA, Table 1), each with different isotopic composition and different mixing ratios of CH 4 and N 2 O, respectively.By analysing between 20 and 70 mL (STP) of the air standards, the calibration experiments covered a large range of sample sizes including the amount of sample air to be expected from the extraction of 200 to 500 g of glacial and interglacial ice core samples (Fig. 6).Two BFI samples that were produced in the same BFI batch contained dissolved gases that only affected their first BFI measurement that included their melt process.This observation is in line with the description of Bock et al. (2010a), who found dissolved gases in BFI samples even though the BFI contained no visible gas inclusions.The stability of the consecutive extractions over melted BFI showed that a melted sample was completely degassed by the previous air extraction.Changing the air standards of two consecutive injections showed no memory effect in the isotope ratios of CH 4 or N 2 O, highlighting quantitative extraction.A significantly increased variability in both N 2 O isotope ratios occurred during two days of BFI measurements.The data of these measurements were excluded, resulting in less N 2 O data of BFI measurements as compared to CH 4 (Fig. 6).We speculate that this effect is based on disequilibration effects inside the Hayesep D in trap T2 after longer standby times.Regular use of T2 or pre-conditioning it with adsorption-desorption cycles in case T2 has not been used for a while stabilized the N 2 O isotope ratios.However, CH 4 measurements were unaffected.The system returned to stable analysis the following day.
The isotope ratio of the BFI measurements was referenced via reference gas injections of 40 mL (GIS) to the respective international isotope scales.The linear regression of the isotope ratio offset versus the sample peak size is used to correct for differences in analyte amount of ice-core-air extractions from the melt vessel and reference gas extractions from T1 (Fig. 6).We find no significant isotope fractionation between the extractions from T1 and from the melt vessel for experiments with identical amounts of analytes.We therefore conclude that there is no significant offset between the extraction of reference gases from T1 and ice core samples from the melt vessel, and that the occurring fractionation is only based on the analyte amount.The correction according to the linear regression (Fig. 6) therefore covers the necessary corrections that relate all samples to the isotope scales as defined by 40 mL of the reference gas (Sect.2.3).The sample-sizecorrected BFI measurements show a standard deviation (1σ ) of 0.09 ‰ for δ 13 C-CH 4 , 0.5 ‰ for δ 15 N-N 2 O and 0.7 ‰ for δ 18 O-N 2 O.

Quality control standard measurements
One QCS measurement was included in each routine for ice core measurements (Table 2).The QCS were varied in size and evaluated as an unknown sample with corrections for size-and system variability (Sect.3.7).The results of 17 QCS measurements for δ 13 C-CH 4 , δ 15 N-N 2 O and δ 18 O-N 2 O of the QCS from an ice core measurement campaign are plotted in the performance chart (Fig. 7).Because the BFI tests showed no detectable isotope fractionation between the BFI extractions from the melt vessel and the reference gas extractions from T1, we consider the QCS measurements representative to indicate the magnitude of the measurement uncertainty that is inherent to the analysis of ice core samples.The standard deviation of the QCS measurements is therefore an important measure for the precision of the system.The estimated measurement uncertainty for δ 13 C-CH 4 is 0.08 ‰ and 0.6 ‰ for both δ 15 N-N 2 O and δ 18 O-N 2 O, based on the 1σ standard deviation of the QCS measurements.

Reproducibility of ice core measurements
We measured 13 ice core samples (200-500 g) from the EU-ROCORE and NEEM ice core (gas age 657-1766 AD) to compare the performance of the setup with published δ 13 C-CH 4 data (Fig. 8).Eleven of these samples are measured as replicates that are divided into four groups of 2-4 samples with a maximum age difference within each group of less than 14 yr.These sample groups can be assumed to contain air of similar composition (Buizert et al., 2012).The pooled standard deviation for δ 13 C-CH 4 in these 11 samples of 0.07 ‰ is a representative measure for the reproducibility of ice core measurements.Unfortunately, we cannot provide the corresponding N 2 O data of these samples as the measurement routine for N 2 O measurements was not fully developed at the time these ice core measurements were done.

Precision of the setup
Multiple gases which differ in the isotopic composition as well as in the mixing ratio of CH 4 and N 2 O were injected over BFI to calibrate the system.One measurement of AL in variable amounts was included as QCS measurement during every day of ice core measurements to monitor the performance of the analytical setup including the data processing.Finally, we showed the reproducibility for the δ 13 C-CH 4 measurements of the setup by the pooled standard deviation of 11 pre-industrial ice core samples between 200 and 500 g.We find the uncertainty estimate based on the analysis of ice www.atmos-meas-tech.net/6/2027/2013/core samples most representative.However, its comparability is restricted because each ice core sample represents a unique air mixture that may be affected by atmospheric variability and the stochastic nature of bubble trapping.Therefore, even two adjacent ice core samples are not necessarily 100 % identical.This problem of system calibration can be circumvented by repeated analysis of reference gases from pressurized tanks.For the δ 13 C-CH 4 measurement, the uncertainty estimates as derived from the BFI measurements, the QCS measurements and the pooled standard deviation of ice core sample measurements are in good agreement, suggesting that a realistic measurement uncertainty is estimated by all three methods.For the isotopic analysis of N 2 O, the uncertainty estimate of the BFI and the QCS measurements also agree well.Based on the good match of the uncertainty estimates discussed for δ 13 C-CH 4 , we suggest that the uncertainty of N 2 O measurements can reliably be estimated from the BFI and QCS measurements.For all measured parameters, we chose to state the uncertainty with the highest value, independent of the method from which it was derived.We therefore conclude a measurement uncertainty of 0.09 ‰ for δ 13 C-CH 4 (BFI), 0.6 ‰ for δ 15 N-N 2 O (QCS) and 0.7 ‰ for δ 18 O-N 2 O (BFI), which is comparable to or better than those of Sowers et al. (2003), Ferretti et al. (2005), Schaefer and Whiticar (2007), Behrens et al. (2008), Sowers (2009), Sapart et al. (2011) and Melton et al. (2011b).We measured a total of 13 EUROCORE and NEEM ice core samples for δ 13 C-CH 4 and compared our results with NEEM ice core data that were recently published by Sapart et al.(2012) as shown in Fig. 8.The data from Sapart et al. (2012) were corrected for the Kr effect, which was not known at the time of publication.The Kr-corrected IMAU data are between 0.15 and 0.88 ‰ more depleted in δ 13 C-CH 4 , depending on the CH 4 mixing ratio, where the highest correction was applied to the samples from the industrial period.Our data and the Kr-corrected data from Sapart et al. (2012) will be referred to as CIC and IMAU Kr , respectively.We selected six data points of the IMAU Kr dataset between 677 and 1757 AD and measured between one and four samples per IMAU Kr data point on our setup.The mean gas age of the CIC and the respective IMAU Kr samples differed between 1 and 24 yr.Greenland ice core samples have been shown to integrate the atmospheric variability over a period of about 35 yr (Buizert et al., 2012).We therefore consider the compared samples to represent similar air samples.CIC and IMAU Kr data show excellent agreement during five time intervals that cover a δ 13 C-CH 4 range of nearly 2 ‰ (Fig. 8).However, we found a disagreement between IMAU Kr and the mean of four CIC samples (three Eurocore, one NEEM) during the δ 13 C-CH 4 excursion reported by Sapart et al. (2012) at 964 AD, where the IMAU Kr sample is more enriched in δ 13 C-CH 4 by ∼ 1 ‰.Between 900 and 1000 AD, three IMAU Kr samples agree well, which rules out that the disagreement is based on one outlier.Since we can exclude Kr interference as the reason for the offset, we speculate that the differences between CIC and IMAU Kr are due to differences in the ice core samples, the sample preparation or the ice extraction technique.
At about 950 and 1000 AD, Rhodes et al. (2013) show several spikes of excess CH 4 , measured on a shallow NEEM ice core.The strongest of these CH 4 artefacts (+78 ppb) was detected in a depth of 272.1 m, which is close to the samples of Sapart et al. (2012) from 964 and 1009 AD (269.5 m and 278.3 m, respectively) but even closer to the depth of the one NEEM sample that was measured at CIC for this intercomparison (272.8 m).Note that the samples of Rhodes et al. (2013) were taken from a shallow core at NEEM, while the discussed NEEM samples are taken from the main core.The age-depth relation may vary slightly between the two cores, which hampers drawing unambiguous conclusions.
One hypothesis is that impurities in the ice are related to excess CH 4 (Rhodes et al., 2013) and thus possibly caused the disagreement between CIC and IMAU Kr .While ice core layers that contained visible particles were removed during sample preparation at CIC, small particles were observed and retained within one IMAU Kr sample that shows the disagreement.However, adjacent IMAU Kr samples with and without visible particles agree well in δ 13 C-CH 4 , as is confirmed from other depth ranges, thereby questioning the hypothesis that particles in the ice necessarily create artefacts in δ 13 C-CH 4 .Thus, the link between particles and δ 13 C-CH 4 offset remains speculative.
One analytical difference between CIC and IMAU Kr is the technique to extract air from ice core samples.While a continuous melt-extraction technique is used at CIC, IMAU Kr data are based on a dry-extraction technique (Sapart et al., 2011).Both techniques are optimized for high extraction efficiency, and both techniques are shown to produce very similar results.It is therefore unlikely that conceptual differences of the extraction techniques alone caused the observed disagreement in δ 13 C-CH 4 .
Because the data comparison between CIC and IMAU Kr shows excellent agreement during all other time intervals, we find our setup suitable for δ 13 C-CH 4 measurements in ice core samples.A δ 13 C-CH 4 record of firn air samples was measured on our setup and also shows very good agreement with the results from other systems as published by Sapart et al. (2013).

δ 15 N-N 2 O and δ 18 O-N 2 O
A flask intercomparison study for N 2 O isotope ratios in three different gas mixtures was conducted to compare our setup to the setup described by Sapart et al. (2011).The air mixtures used for this test varied by ∼ 7 and ∼ 6 ‰ in δ 15 N and δ 18 O, respectively (Table 1).The results for all gases showed excellent agreement within the uncertainty of the measurements and were published earlier (Sapart et al., 2011).
So far, intercomparison measurements on isotope ratios of N 2 O ice core samples have not been made in the ice core community, and data for the time period of our δ 13 C-CH 4 comparison measurements (657-1766 AD) are lacking.We therefore cannot give an ice core intercomparison at this stage but emphasize the need for such a study in the future.

Summary and conclusions
We introduced Aurolite as reliable catalyst for the oxidation of CO.Based on our test we conclude that Aurolite is useful to produce CO-free air for reference gas mixing purposes.We expect that Aurolite could also be used to analyse δ 14 C-CO; however, more tests would be required to prove this hypothesis.We described our setup to measure δ 13 C-CH 4 , δ 15 N-N 2 O and δ 18 O-N 2 O isotope ratios in air-and ice core samples and thoroughly discussed its performance and measurement uncertainty based on bubble-free ice and quality control standard measurements.We proved the reproducibility of the analytical system and the suggested data-processing method based on detailed experiments with sample amounts that can be expected in 200-500 g of glacial and interglacial ice core samples.We furthermore compared our ice core measurement for δ 13 C-CH 4 with published data which prove that our setup is capable of highlighting small yet significant isotope variations with excellent precision.A previously published intercomparison of isotope measurements of N 2 O in reference gases proved the described setup to be in very good agreement with established systems.A high-resolution dataset of δ 15 N-N 2 O and δ 18 O-N 2 O measurements from ice core samples between 657 and 1766 AD is not available for intercomparison.However, we discussed measurement control strategies analogue to our δ 13 C-CH 4 measurements that suggest our setup is suitable to reliably measure the isotopic composition of N 2 O in ice core samples.We conclude that the precision of our setup for δ 13 C-CH 4 , δ 15 N-N 2 O and δ 18 O-N 2 O measurements is 0.09, 0.6 and 0.7 ‰, respectively.Especially the excellent precision of our setup for δ 13 C-CH 4 and its independence of Kr interference make our setup suitable to analyse the variability of the interhemispheric gradient of CH 4 and of small changes of δ 13 C-CH 4 in high temporal resolution.The described setup enables measurement of two samples per day, which allows production of large datasets in future.Upcoming work should also include the harmonization of similar analytical systems and sample preparation protocols to minimize interlaboratory offsets as well as the investigation of measurement artefacts.

Fig. 1 .
Fig. 1.Extraction unit design and flow scheme.Valve symbols with a handle symbolize manually operated valves, while symbols without a handle indicate automatically actuated valves.Filled valve symbols indicate closed valves; open symbols represent open valves.The figure shows the valve positioning of the extraction line during the melt extraction of an ice core sample.The red lines mark the section of the extraction unit which is thermally insulated and heated to 55 • C. The letters A and B identify the ports of V2 that are connected to the respective port A at V5 and port B at V4 of the GC-C-GC-IRMS setup (Fig. 2).Port A transfers the sample to the GC-C-GC-IRMS unit, while port B receives the carrier-gas flow for A.

Fig. 2 .
Fig. 2. Flow scheme of the GC-C-GC-IRMS unit.The letters A and B show the valve ports that connect to the ports A and B of the ice extraction unit, respectively.The CH 4 and N 2 O sample is delivered to port A of the GC-C-GC-IRMS unit by the helium stream that flows out of the GC-C-GC-IRMS unit at port B.
• C (PoraBOND Q, 0.25 mm ID, 25 m length, Agilent) prior to measuring the CH 4 -derived CO 2 .Heliumflushed purge housings are used on valves V3-V6.The sample gases enter the IRMS (Delta V Plus, Thermo Fisher, Germany) through the open-split interface (ConFlo IV, Thermo Fisher, Germany).Before the N 2 O-and CH 4 -derived CO 2 samples are introduced into the IRMS, multiple "on/off" peaks of pure N 2 O and CO 2 reference gases are applied to stabilize the IRMS (Sect.3.2).

P.
Sperlich et al.: CH 4 & N 2 O isotopes in ice core samples efficiency

Fig. 3 .
Fig. 3. Chromatograms with the three m/z 44, 45 and 46 traces for CH 4 (bottom panel) and of N 2 O (top panel) measurements.The enlargement in the top panel shows the background before the N 2 Osample peak in detail (units in mV).The flat-topped peaks in the main chromatograms are "on/off" peaks from pure N 2 O and CO 2 reference gases.The chromatograms show the measurement of an ice core sample with low N 2 O and CH 4 mixing ratios, typical for glacial periods.Reference gas peak 11 of the N 2 O measurement (marked with star) includes a very small CO 2 remainder (not visible) stemming from system blank and incomplete CO 2 removal by the Ascarite.We use its isotopic composition to monitor the performance of the Ascarite trap (Fig.1).

Fig. 4 .
Fig. 4. A small krypton signal is visible when comparing the m/z 43 traces of δ 13 C-CH 4 measurements in an atmospheric air mixture (dark-blue line, right axis) and in a pure CH 4 gas (red line, left axis), where the latter contains no Kr.Both peaks of CH 4 occur on m/z 43, while the baseline separated peak induced by remnant Kr occurs only in the atmospheric sample.To avoid an overlap of the displayed peaks, the chromatogram of the pure CH 4 is shifted by 15 s.
‰ for δ 15 N-N 2 O and 0.15-0.22‰ for δ 18 O-N 2 O.The calculated shot-noise ratio may therefore explain between 40 and 50 % of the measurement uncertainty for δ 13 C-CH 4 and between 20 and 29 % and 21-31 % of the uncertainty of the δ 15 N and δ 18 O measurements of N 2 O, respectively.The remaining uncertainty is most likely based on

Fig. 5 .
Fig. 5. Sample recovery.Green circles show the injected CH 4 amount [nmol] and the resulting IRMS peak area [Vs] as detected during bubble-free ice measurements with the injection of four different air mixtures in variable amounts.Blue circles display analogue information for N 2 O.The two blue crosses show outliers of N 2 O that suffered from a loss of IRMS peak area for unknown reasons.Both outliers are therefore not included in the calculation of the regression, which are shown with 95 % confidence interval.

P.
Sperlich et al.: CH 4 & N 2 O isotopes in ice core samples

Fig. 6 .
Fig. 6.Impact of sample size variability on the isotopic composition of δ 15 N-N 2 O (A), δ 18 O-N 2 O (B) and δ 13 C-CH 4 (C).The offsets in isotopic composition of CH 4 and N 2 O were determined in 33 and 26 BFI extractions, respectively.The dark-grey lines indicate the IRMS peak area of the reference gas measurements (GIS), which were used to reference the BFI samples to the respective isotope scales.The shaded zones indicate the IRMS peak area range of ice core sample measurements.Coloured lines display the linear regressions that are used to correct the isotopic composition of sample measurements for the difference in IRMS peak area between sample and reference gas.

Fig. 7 .
Fig. 7. Results of QCS measurements.The circles of the performance chart display the isotope ratio measurements of δ 15 N-N 2 O (A), δ 18 O-N 2 O (B) and δ 13 C-CH 4 (C) as determined during the QCS measurements by injecting variable amounts of AL.The crosses relate to the y axis on the right-hand side and indicate the IRMS peak area of the respective measurement.The lines show the average δ values of the measurements.

Figure 8 :Fig. 8 .
Figure 8: Comparison of CIC and IMAU Kr δ 13 C-CH 4 ice core data (the i 943 have been corrected for the Kr-effect).Blue circles indicate data com 944 good agreement was found.The orange circle indicates the comparis 945 disagreement of ~1 ‰.Error bars indicate the measurement uncerta 946 measurements used for the laboratory comparison.Bracketed numb 947 samples measured at CIC for the respective comparison while numbe 948 time A.D. of the IMAU Kr samples.949 950 951

Table 1 .
Applied gases.Reference gas names are listed in column 1. Column 2, 3 and 4 show the calibration path for δ 13 C-CH 4 , the determined δ 13 C and the mixing ratio of CH 4 , respectively.The referencing path for N 2 O is mentioned in column 5, while column 6, 7, and 8 indicate the δ 15 N, the δ 18 O and the mixing ratio of N 2 O, respectively.The standard deviation for the isotopic composition of N 2 O in GIS is larger compared to the other gases due to the calibration with two gases during measurements that were optimized to measure δ 13 C-CH 4 .The larger scatter results from the measurement of small N 2 O amounts.Note that the mean value of the QCS measurements for the isotopic composition of N 2 O proves very good accuracy for the calibration of GIS (Fig.7).

Table 2 .
Measurement sequence for ice core samples.A sequence for two ice core sample measurements comprises 11 extraction measurements.Column 2 describes the function of each extraction.Column 3 and 4 show the name and amount of the extracted sample, while column 5 specifies the way the samples are introduced and column 6 shows from which analytical component the respective sample is extracted, where MV stands for melt vessel.