A liquid nitrogen-free preconcentration unit for measurements of ambient N 2 O isotopomers by QCLAS

. Important information about the biogeochemical cycle of nitrous oxide (N 2 O) can be obtained by measuring its three main isotopic species, 14 N 15 N 16 O, 15 N 14 N 16 O, and 14 N 14 N 16 O, and the respective site-speciﬁc relative isotope ratio differences δ 15 N α and δ 15 N β . Absorption laser spectroscopy in the mid-infrared is a direct method for the analysis of the 15 N isotopic composition of N 2 O, yet not sen-sitive enough for atmospheric N 2 O mixing ratios (320 ppb). To enable a fully-automated high precision analysis of N 2 O isotopic species at ambient mixing ratios, we built and optimized a liquid nitrogen-free preconcentration unit to be coupled to a quantum cascade laser (QCL) based spectrometer. During standard operation 10 l of ambient air are preconcentrated on a HayeSep D trap and desorbed in 50 ml of synthetic air. Rigorous tests were conducted, using FTIR, quantum cascade laser absorption spectroscopy (QCLAS), GC-FID and component-speciﬁc ozone and oxygen analysers to investigate recovery rates, conservation of isotopic signatures and spectral interferences after preconcentration. We achieve quantitative N 2 O recovery of > 99% with only minor, statistically not signiﬁcant isotopic fractionation and no relevant spectral interferences from other atmospheric constituents. The developed preconcentration unit also has the potential to be applied to other trace gases and their isotopic composition.


Introduction
Nitrous oxide (N 2 O) is currently the most important anthropogenic ozone depleting substance and exerts the second largest global warming potential weighted (GWP) emission of non-CO 2 greenhouse gases (Ravishankara et al., 2009).
Correspondence to: J. Mohn (joachim.mohn@empa.ch) Since pre-industrial times, N 2 O mixing ratios in the troposphere increased from 270 ppb (parts-per-billion, 10 −9 moles of trace gas per mole of dry air) to the current level of 320 ppb at a rate of 0.26% a −1 . The predominant sources of N 2 O on a global scale are microbial production in agricultural and natural soils as well as rivers, estuaries, coastal areas and oceans. While the major sink, stratospheric destruction, is well quantified, the strengths of individual N 2 O sources remain largely uncertain (Solomon et al., 2007).
Important information about the biogeochemical cycle of N 2 O can be obtained by measuring the intramolecular distribution of 15 N in atmospheric N 2 O (Billings, 2008;Bernard et al., 2006;Ishijima et al., 2007;Rahn and Wahlen, 2000;Röckmann et al., 2003;Röckmann and Levin, 2005;Yoshida and Toyoda, 2000). N 2 O is a linear tri-atomic molecule with one nitrogen atom in the centre (2 or α site) and one at the end (1 or β site) that leads to four possible stable isotopic species with respect to nitrogen isotopes. The most abundant N 2 O isotopic species in the atmosphere are 14 N 14 N 16 O (99%), 15 N 14 N 16 O (0.37%) and 14 N 15 N 16 O (0.37%) Yoshida and Toyoda, 2000). Isotope abundances are usually reported in the δ-notation, where δ 15 N denotes the relative difference in ‰ of the 15 N/ 14 N ratio of the sample versus the reference material (atmospheric N 2 ), analogically δ 15 N β denotes the corresponding ratio for 15 N 14 N 16 O vs. 14 N 14 N 16 O.
The bulk nitrogen δ value (δ 15 N bulk =(δ 15 N α +δ 15 N β )/2) tropospheric N 2 O was found to be enriched by 6.3±0.3‰ to 6.72 ± 0.12‰ relative to air-N 2 (Yoshida and Toyoda, 2000;Toyoda et al., 2004;Kaiser et al., 2003;Park et al., 2004), with a strong site preference (SP=δ 15 N α −δ 15 N β ) of 18.7±2.2‰ for the central nitrogen atom (Yoshida and Toyoda, 2000). Tropospheric N 2 O mixing ratios and isotopic composition are largely controlled by back injection of stratospheric N 2 O as well as N 2 O emissions from microbial activity and combustion processes. In the stratosphere, the abundance of the heavier isotopologues and the SP increases with altitude to values as high as 140‰ for δ 15 N bulk  and 100‰ for SP (Toyoda et al., 2001), mainly due to isotopic fractionation during UV-induced photolysis and oxidation. Isotopic signatures of different N 2 O emission sources were found to vary largely between approx. −50‰ (soil) to 10‰ (coal combustion) for δ 15 N bulk and −5‰ (soil) to 35‰ (ocean) for the SP (Toyoda et al., 2009, and references therein). An overview on atmospheric applications of N 2 O isotopic analysis can be found in a review publication of Brenninkmeijer et al. (2003). The required analytical precision for δ 15 N α and δ 15 N β depends strongly on the application and may vary between < 0.1‰ for firn air analysis to 1‰ for subsoil N 2 O studies.
Currently, the only technique for N 2 O isotopic measurements at ambient mixing ratios is laboratory-based isotoperatio mass-spectrometry (IRMS) in combination with flasksampling and cryogenic preconcentration. However, isotopomers such as 14 N 15 N 16 O and 15 N 14 N 16 O cannot be distinguished directly by IRMS, as they have the same mass.
To determine the site-selective isotopic composition, the 15 N content of the NO + and N 2 O + fragment and molecule ions of N 2 O has to be analyzed (Brenninkmeijer and Röckmann, 1999;Toyoda and Yoshida, 1999). Therefore, the masses m/z 44, 45, and 46 (for N 2 O) and m/z 30 and 31 (for NO) are monitored either simultaneously (on last generation IRMS) or after two separate sample injections. Although IRMS is a well-known method with excellent precision (0.05‰-0.9‰) and accuracy (Röckmann and Levin, 2005;Well et al., 2008;Yoshida and Toyoda, 2000), it has some disadvantages such as the large size of the instrument, which hinders field measurements. The required field sampling and sample storage strongly limits its use for the characterization of variations in N 2 O isotopic species at time-scales relevant to many environmental processes.
Laser spectroscopy is a valuable alternative, because compact and continuously-operating instruments can be developed that allow for extended field measurements. This offers the possibility for process studies and opens a completely new field of applications. Furthermore, the method is inherently selective, even for molecules with the same mass, because it is based on the highly characteristic rotationalvibrational transitions of the different N 2 O isotopic species. The absorption lines of the various species can be easily resolved at low cell pressure. However, there are only a few spectroscopic measurements on isotope ratios of N 2 O, and these were performed on pure N 2 O samples (Gagliardi et al., 2005;Nakayama et al., 2007;Uehara et al., 2001Uehara et al., , 2003 with the exception of two recent studies (Wächter et al., 2008;Wächter and Sigrist, 2007). The reported precisions range between 0.2‰ and 9‰.
Recent advances in laser technology provided a new class of mid-infrared (IR) semiconductor lasers: the unipolar intersubband quantum cascade lasers (QCL). They have the advantage of operating at non-cryogenic temperatures (quasi room-temperature) with high spectral purity and relatively high power output (Faist et al., 1994). The application of these light sources for high-precision isotope ratio measurement has been recently demonstrated for CO 2 by Nelson et al. (2008) and Tuzson et al. (2008a, b). Spectroscopy in the mid-IR region is particularly effective for molecules that exhibit strong fundamental vibrations, such as CO 2 and N 2 O, with line intensities about two orders of magnitude higher than in the near-IR. We recently demonstrated a precision of 0.5‰ for δ 15 N at N 2 O mixing ratios of 90 ppm (Wächter et al., 2008) using a pulsed QCL emitting at 4.6 µm. Atmospheric N 2 O mixing ratios (320 ppb), however, are more than two orders of magnitude lower and are therefore currently not accessible for direct isotopic analysis by laser spectroscopy.
For these reasons we suggest to combine laser spectroscopy with preconcentration techniques to increase N 2 O mixing ratios from ambient levels to ≈60 ppm. The developed custom-built preconcentration unit enables us to make use of the precision and site specificity of QCLAS at a temporal resolution of approx. 30 min. The preconcentration technique we employ is fully automated and is based on a commercial closed-loop cryogenic unit and thereby free of consumables like liquid nitrogen. Here we describe the most important properties that were obtained through systematic optimization and characterization: (a) excellent recovery of N 2 O and other trace gases; (b) conservation of the N 2 O isotopic composition, and (c) negligible impact of other atmospheric constituents (e.g. CO 2 , H 2 O, CO, O 3 , CH 4 and C 2 H 2 ) on the analyzed N 2 O isotope ratios. Our results also indicate that the coupling of preconcentration and laser spectroscopy may also be feasible for the measurement of mixing ratios and isotopic composition of other trace gases (e.g. many VOCs, CH 4 , CO).

Preconcentration unit
The technology of our preconcentration unit is based on a previously developed system called "Medusa" which was originally designed for halogenated compounds measurements by gas-chromatography mass-spectrometry (Miller et al., 2008). We re-designed and optimized the system for the preconcentration of N 2 O isotopic species and their subsequent quantification by laser spectroscopy. One main reason to choose this system is the combination of a relatively mild adsorbent trap with a low-temperature refrigeration unit reaching <−150 • C. Our preconcentration unit differs from "Medusa" by a number of key elements, including: an optimized setup with two multiport valves and a low temperature adsorbent trap. Furthermore, all gas flows are significantly increased and regulated by mass flow controllers (MFC, Redy Smart Series, Vögtlin Instruments, Switzerland) and the complete preconcentration unit is controlled and monitored by LabVIEW (National Instruments Corp., USA).
The instrument flow scheme is illustrated in Fig. 1, showing one particular position of the valves (Valco Instruments Inc, Switzerland), which are used to isolate or combine various sections of the sample pathways. The 6-port valve is equipped with an elongated-groove rotor and driven by a multi-position actuator which enables more than 2 valve positions. Rotating the 4-port valve will change the flow direction through the HayeSep D trap. The preconcentration trap consists of stainless steel tubing (ID 1.6 mm, OD 2.1 mm) filled with 200 mg HayeSep D (100-120 mesh, Hayes Separations Inc., USA) and coiled onto an aluminium tube standoff. The standoff is attached to a copper base-plate cooled by the refrigeration unit (PCC: Polycold Compact Cooler, Brooks Automation, USA), and located in a vacuum chamber. The trap reaches temperatures of ≈ −150 • C and can be heated resistively to > 100 • C, using a custom-built trap heating unit coupled with an Omega PID controller (CN77354-C2, Omega Engineering Inc., USA).
The operation sequence and parameter settings (e.g. flow rates, temperatures and timing) of the preconcentration unit were optimized to obtain a sharp N 2 O concentration profile during desorption. The principle of operation is described in the following paragraphs (see Figs. 1 and 2): after an initialization phase that sets all parameters to their starting values, the HayeSep D trap is purged by back-flushing with 50 sccm (standard cubic centimetre per minute) of high purity synthetic air and heated to approx. 100 • C (phase I). This phase assures the absence of residual trace gases on the preconcentration trap and reproducible starting conditions. After cooling the trap down to ≈ −150 • C, the loading (phase II) is initialized by setting the sample gas flow to 500 sccm, flushing the lines and rotating the 6-port valve to the loading position. H 2 O and CO 2 are removed from the measuring gas by a Nafion TM drier (MD-050-72S-1, PermaPure Inc., USA) and a chemical trap filled with Ascarite (3 g, 10-35 mesh, Fluka, Switzerland) bracketed by Mg(ClO 4 ) 2 (2·3 g, Fluka, Switzerland). To purge residual gas molecules onto the trap, a short post-flush with synthetic air is applied at the end of phase II. Rotating the 4-port valve to position 2 (backward flow) and adjusting the flow of synthetic air to 50 sccm initiates desorption (phase III). After setting the trap temperature to −35 • C, an immediate release of N 2 O (boiling point −88.5 • C) can be observed by FTIR spectroscopy (see Fig. 2). Finally, all parameters and conditions are set back to their initial values.The duration of trap loading (phase II) and desorption (phase III) are adjusted based on the desired/available sample volume (standard procedure: 20 min at 500 sccm sampling flow results in 10 l sample volume) and the desired N 2 O output mixing ratio, usually around 60 ppm for high precision N 2 O isotopic analysis by QCLAS. The minimal desorption time for quantitative N 2 O recovery depends on the molar N 2 O input and was determined by online FTIR spectroscopy to be between 60 s for 140 nmol and 180 s for 4000 nmol N 2 O. Typically a 10 l ambient air sample is focused into 50 ml desorption volume, corresponding to a concentration factor of 200.

On-line FTIR analysis of mixing ratios
We employed a Nicolet Avatar 370 MCT spectrometer (Thermo Nicolet Corp., USA) equipped with a heated (40 • C) low-volume (50 ml) flow-through gas cell with 1 m pathlength (Model LFT-210, Axiom Analytical Inc., USA). An infrared spectrum with 0.5 cm −1 optical resolution was recorded every three seconds, and trace gas (N 2 O, CO, CH 4 , CO 2 and H 2 O) mixing ratios were retrieved by classical least square fitting in selected wavelength regions (Mohn et al., 2004. Calibration spectra were recorded under identical instrumental and spectroscopic conditions from certified and diluted calibration gases or by continuous injection of liquids into nitrogen. The detection limits (VDI, 1995) for the trace gases in the undiluted gas were: 100 ppb N 2 O, 200 ppb CH 4 , 1 ppm CO 2 , 200 ppb CO, 10 ppm H 2 O. The expanded standard uncertainty for these components is around 10% (95% confidence level). The response time (t 95 ) for the experimental setup (sampling line, FTIR gas cell) was 55 s at the applied flow rate of 50 sccm and 13 s employing a fed-in flow of 450 sccm synthetic air in front of the gas cell to determine the minimal desorption time of the preconcentration unit.

Analysis of the 15 N isotopic composition of N 2 O by QCLAS
The QCL based spectrometer used in the present project is an improved version of the instrument used by Wächter et al. (2008). The experimental setup consists of a single-mode, pulsed quantum cascade laser emitting at 4.6 µm, a multipass absorption cell (optical path length 56 m, volume 500 ml; Aerodyne Research Inc., USA) and a detection system with pulse normalization. Laser control, data acquisition and realtime signal processing are performed by the TDLWintel software (Aerodyne Research Inc., USA). The main upgrades include: (a) a new RT-QCL (Alpes Lasers SA, Switzerland) with about 10 times more power and a considerably narrower line width of 0.0068 cm −1 ; (b) a new generation thermoelectrically cooled detector (PVI-3TE-5, Vigo System, PL) with a specific detectivity (D*) > 10 11 cm Hz 1/2 W −1 at 4.5-  4.6 µm, coupled to a miniaturized preamplifier and TECcontroller (MTCC-01, Vigo System, PL). The detector assembly is mounted on a water-cooled cold-plate, which maintains a stable heat-sink temperature of 15±0.1 • C. Further improvements were achieved by redesigning the laser current and temperature control unit, optimizing the laser pulser electronics and the 3-D-alignment stage of the collimating optics for a better alignment stability. The selected absorption lines for this study are shown in Fig. 3. They are located within the scanning range of the laser and have sufficiently strong line strength with similar absorbance for each isotopic species. Mixing ratios for the different species are obtained by simultaneously fitting a low-order polynomial to the spectral baseline and a Gaussian-convoluted Voigt profile to the observed absorption lines, taking into account the measured path length, gas temperature (∼307 K), pressure (8 kPa) and laser line width. For the site-specific relative isotope ratio differences δ 15 N α and δ 15 N β , a precision of 0.2-0.3‰ was obtained at 90 ppm N 2 O using the Allan variance approach (Werle et al., 1993, data not shown n.a.: not analyzed (a) gas mixtures prepared from 9 or 90 ppm N 2 O reference gas and synthetic air; line 1: additionally with pressurized air and mixing ratios determined by FTIR.

Validation of N 2 O preconcentration unit
To validate the N 2 O preconcentration unit, tests were performed on the quantitative recovery of N 2 O and other trace gases (Sect. 2.2.1), on the conservation of the N 2 O δ values ( δ 15 N=δ 15 N precon −δ 15 N original ) (Sect. 2.2.2), and on the composition of preconcentrated ambient air and its potential interferences with N 2 O isotopic analysis (Sect. 2.2.3).
To investigate the trace gas recoveries and isotopic fractionation effects, gas mixtures with known mixing ratios and/or isotopic composition were prepared from standard reference gases by dynamic dilution with high purity synthetic air (79.5% N 2 purity 5.0, 20.5% O 2 purity 5.0, N 2 O < 0.3 ppb) and/or pressurized air using a pair of MFCs. Standard reference gases applied include 9.0 ppm and 90 ppm N 2 O in synthetic air (Messer Schweiz, Switzerland) to determine the trace gas recoveries and gravimetrically prepared 90.8 ppm N 2 O in synthetic air (Wächter et al., 2008) with a distinct isotopic composition of δ 15 N α =25.9±0.4‰ and δ 15 N β =25.2±1.0‰ as analysed by IRMS (Toyoda and Yoshida, 1999) to investigate the conservation of the N 2 O δ values. Table 1 summarizes the gas mixtures used during the various validation experiments. For the study of spectral interference we relied on ambient air sampled with a membrane pump (PTFE coated membrane, N 035 AT.18, KNF Neuberger, Switzerland).

Trace gas recovery
For the determination of the trace gas recovery, 1 to 10 l of a gas mixture were supplied at a flow rate of 100 sccm, containing N 2 O, CH 4 and CO at ambient and H 2 O and CO 2 at sub-ambient mixing ratios (Table 1). While these tests were based on the standard preconcentration procedure, no chemical trap was employed to investigate the influence of H 2 O and CO 2 traces on the N 2 O adsorption. On-line FTIR measurements were performed to obtain time-resolved information on the trace gas mixing ratios desorbed from the HayeSep D trap. The amount of desorbed trace gases was calculated as the integral over time of the output trace gas mixing ratio and the flow rate of synthetic air applied for desorption. The chosen approach has the advantage of analyzing the recovery for different trace gases under standard operation conditions but does not allow a highly precise budgeting because it relies on a fast integration over a wide range of mixing ratios and on MFCs and gas flows that are different during adsorption and desorption.
To achieve an even better precision for the determination of the N 2 O recovery, a second approach was used. These further experiments were designed to yield the same mixing ratio before and after the preconcentration step due to the use of the same MFC (MFC2) and flow parameters (2400 s and 100 sccm) during adsorption and desorption, and by collecting the desorbed sample in a 10 l Tedlar bag (Product No. 245-08, SKC Inc., USA). The sample gas was dynamically diluted from a tank, containing either 9.0 or 90 ppm N 2 O in synthetic air, with synthetic air to ambient (350 ppb), intermediate (900 ppb) or process (9.0 ppm) mixing ratios (Table 1). N 2 O mixing ratios in the sample gas and in the desorbed sample (Tedlar bag) were compared based on QCLAS at 2188.19 cm −1 . Experiments at every mixing ratio were replicated thrice. Small differences (<1%) between the adsorbed and desorbed gas volumes, as recorded from the MFC output, were corrected.

Isotopic fractionation
To validate the conservation of isotopic signatures, a N 2 O reference gas (90.8 ppm N 2 O) analysed by IRMS (δ 15 N α = 25.9±0.4‰ and δ 15 N β =25.2±1.0‰) was diluted using highpurity synthetic air prior to preconcentration, followed by Tedlar bag sampling and QCLAS analysis (Table 1). Ambient (350 ppb), intermediate (900 ppb) and process (9.0 ppm) N 2 O mixing ratios were supplied to the preconcentration unit. The standard preconcentration procedure was applied with an adsorption flow rate of 500 sccm during 20 min and 50 sccm (60-150 s) desorption flow. At elevated N 2 O input mixing ratios (9.0 ppm) the desorbed sample gas was dynamically diluted by 1:10 during desorption by a fed-in flow, resulting in mixing ratios of ≈60 ppm for all experiments. A series (n=5-70) of preconcentration cycles were pooled in one Tedlar bag to obtain a gas sample of 3 to 5 l. Experiments at each mixing ratio level were replicated three to five times.

Spectral interference
To evaluate potential spectral interferences from ambient air, samples were taken through a 5 m, 6 mm outer diameter PTFE tube from outside the laboratory building. The corresponding trace gas mixing ratios were taken from the routine measurements at the ≈300 m distant station Dübendorf of the Swiss National Air Pollution Monitoring Network (NABEL). Where no NABEL data were available (N 2 O and CO 2 ), we assumed typical ambient mixing ratios as shown in Table 1. The standard preconcentration procedure (Sect. 2.1.1) was followed, and the desorbed gaseous species were analyzed either continuously (O 3 , N 2 O, CO 2 , H 2 O, CH 4 , CO) or after Tedlar bag sampling (C 2 H 2 , O 2 ). Continuous data after preconcentration were obtained by FTIR (Sect. 2.1.2) for N 2 O, CO 2 , H 2 O, CH 4 , and CO. Ozone was measured using an UV absorption O 3 analyzer (TEI 49C, Thermo Fisher Scientific Inc., USA) after dynamic dilution with synthetic air. Oxygen was determined with an O 2 analyzer (Servomex 570A, Spectris plc, UK), and acetylene mixing ratios were obtained using a GC-FID (Agilent 6890 with Plot Al 2 O 3 (KCl) column) after preconcentration with a Peltier cooled HayeSep D trap. The mixing ratios and corresponding preconcentration factors were then used to simulate absorption spectra and evaluate possible spectral interferences.

Trace gas recovery
Quantitative recovery of the adsorbed trace gases is a prerequisite to assure the representativeness of the analyzed gas sample and to avoid fractionation effects. Two different approaches were chosen to tackle this subject: (i) determination of the trace gas recovery with increasing adsorption volume to examine the trap capacity (Fig. 4); (ii) high-precision recovery of N 2 O at different levels of mixing ratios ( Table 2).
The capacity of the adsorbent trap for different trace gases was evaluated based on a sequence of experiments where 1 to 10 l of a gas mixture with constant composition (see Table 1) were first adsorbed on the preconcentration trap and then desorbed with a flow of 50 sccm of synthetic air and analyzed by on-line FTIR spectroscopy. For every trace gas the molar amounts in the input and output gas volume were calculated, integrating over time the respective mixing ratios and gas flows. The CO mixing ratio in the output flow was below the FTIR detection limit of 200 ppb and the amount of desorbed CH 4 was nearly constant and did not depend on gas sample volume (Fig. 4). This indicates that the mild adsorbent HayeSep D does not quantitatively retain CH 4 and CO at a trap temperature of approx. −150 • C which is above their respective boiling points of −161 • C and −191.5 • C.
For high precision N 2 O isotopic analysis by QCLAS the loss of CO is an advantage because of its absorption in the selected spectral region. The application of the preconcentration unit for CO and CH 4 may still be envisaged by employing a stronger adsorbent (i.e. Molsieve 5Å) or an alternative refrigerant reaching a lower adsorption temperature.
As shown in Fig. 4, for N 2 O and CO 2 a linear relation in the molar amounts between the trap output and the sample volumes was obtained, with expanded standard uncertainties for the regression lines of less than 1%. The linearity holds for the full investigated range, while the slight deviation from unity of 2% to 6%, is well below the uncertainty due to potential systematic offsets in the analysis of input and output mixing ratios by FTIR or in the input and output gas flows.
The efficiency of N 2 O recovery is critical, because the adsorption/desorption processes may lead to significant isotopic fractionation (Bertolini et al., 2005;Brand, 1995;. Therefore, a second, more precise approach was applied which allows the direct comparison of the sample gas before and after preconcentration by precise QCLAS measurements (Sect. 2.2.1). Using this experimental setup, the input and output mixing ratios agreed within 0.2% at ambient (350 ppb), intermediate (900 ppb) and process levels (9.0 ppm) ( Table 2). Apart from the standard deviation for repetitive measurements, systematic deviations from e.g. subtle differences between the adsorbed and desorbed gas volumes cannot be excluded. However, such effects are considerably smaller than 1%, corresponding to a >99% recovery and thus, we expect only minimal, if any, isotopic fractionation.

Isotopic fractionation
To further confirm the conservation of isotopic signatures the isotopic composition of samples at ambient (350 ppb), intermediate (900 ppb) or process mixing ratios (9 ppm) were compared before and after preconcentration as described in Sect. 2.2.2. All experiments were done with gases that were enriched in δ 15 N α by 25.9±0.4‰ and δ 15 N β by 25.2±1.0‰.
Agreement of the N 2 O isotopic composition with/without preconcentration was in general very good (see Table 3). The δ 15 N α values of N 2 O after preconcentration are within 0.1‰ compared to the unprocessed calibration gas. For δ 15 N β , the data seems to indicate a slight increase in the values after preconcentration, particularly at lower N 2 O input mixing ratios. The uncertainty of the mean values is however rather large, and in a one-sided student's t-test the effect is not significant at the 0.05 significance level. We, nevertheless, plan to repeat the experiments in the future with the next generation of QCLAS and a directly coupled preconcentration unit. It should also be noted that the calibration routine will include the preconcentration step and can thus be expected to largely account for small fractionation effects if they really occur.

Spectral interference
At the precision levels needed for N 2 O isotope ratio analysis, even very weak spectral interferences of other compounds may be detrimental. To analyze spectral interferences of ambient air constituents the following approach was applied: (i) identification of molecular species (based on HITRAN database) that may cause spectral interferences through their absorption features; (ii) quantitative analysis of these components after preconcentration of ambient air, and (iii) assessment of spectral interferences based on spectral simulations. Although the chosen absorption lines are free of any strong or obvious interference, we have considered all molecules in the HITRAN database (Rothman et al., 2005) for lines in the relevant wavelength region. The line strengths of the relevant molecules are plotted in Fig. 5b, weighted for their ambient molecular abundance (Table 1). Based on this abundance weighted representation, the most relevant interfering compounds are H 2 O and CO 2 , followed by CO, O 3 and C 2 H 2 . Apart from its spectral interference, the water content in the measuring gas has to be reduced considerably to avoid clogging of the adsorption trap. This is accomplished by permeation drying and Mg(ClO 4 ) 2 , while CO 2 is largely removed by the Ascarite trap. For the other gas compounds, the preconcentration procedure is quite selective, as will be shown below.
Mixing ratios of H 2 O and CO 2 were efficiently reduced during preconcentration, below their FTIR detection limits of 10 and 1 ppm, respectively. Both H 2 O and CO 2 were shown to be efficiently removed by chemical trapping with  Mg(ClO 4 ) 2 and Ascarite to ppm or even ppb levels, respectively (Röckmann and Levin, 2005;. The average ozone output mixing ratio was only about six times the ambient mixing ratio which corresponds to a 3% recovery, assuming a concentration factor of 200. This can be expected because ozone is subject to rapid decomposition during desorption at hot metal surfaces. CH 4 and CO are not efficiently trapped by the preconcentration unit, due to their low boiling points, as reflected by a breakthrough of >98%. The C 2 H 2 output mixing ratio was around 125 ppb, which would correspond to an input mixing ratio of 0.62 ppb assuming a quantitative recovery. This value is in-between the annual average C 2 H 2 mixing ratios observed at the remote NABEL-station Rigi-Seebodenalp and the urban station Zurich with 0.4 and 1.0 ppb (FOEN and EMPA, 2009), respectively, which is in accordance with the suburban character of the sampling station Dübendorf. O 2 and N 2 are to a large extent lost during trapping. The remaining bulk gases increased the O 2 /N 2 ratio by 2-3% in the synthetic air employed for desorption, most probably due to the larger N 2 loss during trapping as a result of the difference in boiling points of −183.0 • C and −195.8 • C of O 2 and N 2 , respectively. This slight change in the gas matrix should, however, not have any significant effect on the isotope ratio measurements since the reference gases will also be supplied in synthetic air and treated similarly.
A hypothetical absorption spectrum of spectroscopically interfering components was generated, based on the analyzed output mixing ratios and typical conditions employed for QCLAS measurements (Sect. 2.1.3). To obtain a meaningful representation, the spectrum of interfering components after preconcentration of ambient air was multiplied by a factor of 1000 (Fig. 5), reflecting the required precision for the N 2 O isotope ratios in the sub-per mille range. The only relevant signals might be due to very weak O 3 , CO and C 2 H 2 absorption. The resulting effect on the mixing ratios of the N 2 O isotopic species is, however, below 1% (even with a 1000×magnification) in a typical retrieval of mixing ratios from measured spectra, which translates into a shift of less than 0.01‰ on the δ-scale.

Conclusions
We built, optimized and validated a liquid nitrogen-free fully-automated preconcentration unit to increase N 2 O mixing ratios by a factor of 200 from ambient levels to about 60 ppm. The key element is the combination of a relatively mild adsorbent with a low temperature refrigeration unit, achieving trapping temperatures of approximately −150 • C. During standard operation, 140 nmol of N 2 O are adsorbed from 10 l of ambient air within 20 min and desorbed in 50 ml of high purity synthetic air. The validation experiments showed quantitative N 2 O recovery with only minor, statistically not significant isotopic fractionation, and negligible spectroscopic interferences from other atmospheric constituents on the analyzed N 2 O isotope ratios.
Based on these encouraging results we intend to directly couple the preconcentration unit to a QCL spectrometer to permit fully-automated high precision (<0.5‰ for δ 15 N) N 2 O isotopic analysis at ambient mixing ratios (≈320 ppb) with a temporal resolution of less than 30 min. Expected spectroscopic improvements, e.g. using continuous-wave (cw) QCL and/or spectral ratio method will provide enhanced precision without significant modifications in the instrumental setup. Additionally, δ 18 O-N 2 O measurements would be feasible by selecting a QC laser with an appropriate scanning range. Future modifications of this basic setup may permit to analyze other trace gases that are not directly accessible by currently available spectroscopic techniques (i.e. many VOCs) and to quantify isotopologues of substances at relatively low mixing ratios, such as CH 4 , CO and organic compounds by employing a stronger adsorbent (i.e. Molsieve 5Å) or an alternative refrigerant reaching a lower adsorption temperature. of the original "Medusa" for GC-MS. Wilhelm Gutjahr helped to conduct the numerous experiments needed to optimize the experimental conditions. Davide Ferri provided the FTIR gas cell, and Erwin Pieper and Willy Knecht contributed with their competence in mechanical design and electronics.
Edited by: F. Keppler