Gas chromatography/isotope-ratio mass spectrometry ...

RAPID COMMUNICATIONS IN MASS SPECTROMETRY Rapid Commun. Mass Spectrom. 2003; 17: 1897–1908 Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/rcm.1132

Gas chromatography/isotope-ratio mass spectrometry method for high-precision position-dependent 15N and 18 O measurements of atmospheric nitrous oxide Thomas Ro¨ckmann1*, Jan Kaiser1{, Carl A. M. Brenninkmeijer2 and Willi A. Brand3 1

Max-Planck-Institut fu¨r Kernphysik, Bereich Atmospha¨renphysik, Heidelberg, Germany Max-Planck-Institut fu¨r Chemie, Abteilung Chemie der Atmospha¨re, Mainz, Germany 3 Max-Planck-Institut fu¨r Biogeochemie, Jena, Germany 2

Received 11 May 2003; Revised 20 June 2003; Accepted 21 June 2003

We describe an automated gas chromatography/isotope-ratio mass spectrometry (GC/IRMS) method for the determination of the 18O and position-resolved 15N content of nitrous oxide at natural isotope abundance. The position information is obtained from successive measurement of the isotopic composition of the N2Oþ ion at m/z 44, 45, 46 and the NOþ fragment ion at m/z 30, 31. The fragment ion analysis is complicated by a non-linearity in the mass spectrometer that has to be taken into account. Evaluation of the absolute peak areas allows for a simultaneous determination of the N2O mixing ratio for atmospheric samples. Samples with mixing ratios ranging from a few nmol/mol up to the percent level can be analyzed using different sample inlet systems. The high concentration inlet system provides an easy and quick method to carry out various diagnostic tests, in particular to perform realistic linearity tests. A gas chromatographic set-up with a split column and a backflush possibility improves analytical precision and excludes interferences by substances with long retention times from preceding runs. We also describe a new open split interface that uses only a single transfer capillary to the mass spectrometer for sample and reference gas. Copyright # 2003 John Wiley & Sons, Ltd.

Isotope studies of nitrous oxide (N2O) are conducted to investigate the distribution and cycling of nitrogen through the environment. N2O is a product of microbial nitrification and denitrification processes in soils and in water. To study these processes in more detail, the isotopic composition of the nitrous oxide produced is often measured alongside flux measurements, which can provide independent information on the actual formation processes.1–12 Many studies on nitrification/denitrification mechanisms have been performed using isotopically labeled N2O precursors, but natural abundance methods are now increasingly employed thanks to their steadily increasing precision. In the atmosphere, N2O is an important greenhouse gas and a source of nitrogen oxides in the stratosphere which are involved in ozone destruction. Therefore, the global atmospheric N2O budget is being studied in detail, and in particular the observed increase in the N2O mixing ratio by 0.25% per year over the past decade13 is of concern. Isotope measurements are believed to be an important tool to further constrain the still uncertain global N2O budget, and the number of isotope studies on atmospheric N2O has increased considerably in recent years. This intensified effort has also *Correspondence to: T. Ro¨ckmann, Max-Planck-Institut fu¨r Kernphysik, Bereich Atmospha¨renphysik, Heidelberg, Germany. E-mail: [email protected] { Current address: Department of Geosciences, Princeton University, USA.

fostered new analytical developments. In particular, two additional isotope signatures have become accessible in the recent past. The first was the 17O content of N2O. This signature was tacitly believed to provide no independent information, because 17O changes correlate with the accompanying 18O changes due to the mass dependence of common isotope fractionation mechanisms. In the atmosphere, however, an increasing number of trace gases have become known for which the assumption of mass-dependent fractionation does not hold, among them N2O. N2O has a small but measurable excess of 17O relative to what is expected based on its 18O content, and thus the 17O measurement provides an independent additional isotope signature. Three methods for the determination of d17O in N2O are now available.14– 16 The discovered 17O anomaly17 called for further investigation, and two sources have recently been identified; the strong 17O excess present in atmospheric ozone (O3) can be transferred in a two-step process via NO2 to N2O,16 or via electronically excited atomic oxygen O(1D) in the reaction with N2.18 The second achievement was the development of analytical methods to determine the position-dependent 15N fractionation of N2O. The two nitrogen atoms in the linear NNO molecule are not equivalent, and the 15N abundance may vary independently at the two positions. A major shortcoming of traditional isotope ratio mass spectrometry Copyright # 2003 John Wiley & Sons, Ltd.

1898

T. Ro¨ckmann et al.

(IRMS) techniques for the determination of 15N and 18O in N2O is that it is impossible to distinguish the two molecules 15 14 16 N N O and 14N15N16O, which both contribute to the m/z 45 ion beam. Thus, past measurements of 15N in N2O only yielded the average of the 15N fractionation at both positions. Note that this signature is a mathematical average that cannot be related directly to a physical or chemical process. One way to resolve both individual 15N fractionations is high-resolution measurement of the optical absorption spectra of the different N2O isotopomers. Methods based on Fourier transform infrared (FTIR) spectroscopy19 and tunable diode laser absorption spectroscopy (TDLAS)20 have been successfully developed recently. In addition, IRMS can be employed to distinguish between the two positions. The method, developed simultaneously by two groups,21,22 is based on mass spectrometry of the NOþ fragment which is formed from N2O following electron impact in the ion source of a mass spectrometer. If the fragment retains the nitrogen atom that was originally attached to the oxygen atom (which is true apart from a certain degree of scrambling, see below), determination of the 15N content of the fragment yields the 15 N fractionation at the central position of the parent N2O molecule. The isotopic composition at the terminal position can then be calculated from the central position (NOþ fragment) and the average (overall N2O) measurement. The principle of this method was already used in 1950 by Friedman and Bigeleisen23 for nitrous oxide produced from isotopically enriched NH4NO3. Apparently, this method was never adapted for natural abundance measurements. Resolution of the atomic fractionation at the two individual positions also requires a new notation. Isotope ratios are commonly reported in d-notation, where d is the relative difference of the ratio of a rare isotope to the corresponding abundant isotope (e.g., 18R : w(18O)/w(16O) where w is the abundance of the respective isotopomers) in a sample relative to an international standard ratio, reported in per mill: 18 O ¼ 18 RSA =18 RST 1 1000 % : In the case of 15N, the corresponding d15N value has been used for the average 15N fractionation. Unfortunately, for the individual positions, three different notations are presently used. We have numbered the nitrogen isotopes from the terminal position as numbers 1 and 2 (Fig. 1), and, correspondingly, 1d15N represents the 15N fractionation at the terminal position and 2d15N at the central position. Yoshida and Toyoda24 have designated the positions with Greek letters a and b starting with the central position, thus d15Nb represents the 15N fractionation at the terminal position and d15Na at the central position. Yung and Miller25 used abbreviations for the N2O isotopomers 15N14N16O (546) and 14 15 16 N N O (456).

Figure 1. Mesomeric forms of N2O in resonance with each other. Locants 1 and 2 indicate the numbering of the nitrogen atom positions for the position-dependent 15N determinations. Copyright # 2003 John Wiley & Sons, Ltd.

The newly developed methods have considerable impact on the field and yield important new insights into the global N2O isotope budget. In particular, the enrichment in the stratosphere that was confirmed in 199726 was investigated immediately. Both FTIR and IRMS techniques27–29 have confirmed that the stratospheric 15N enrichments are position-dependent, as postulated in theoretical calculations.25 Furthermore, laboratory studies have shown that the main cause of the observed enrichments is the (position-dependent) 15N fractionation in the UV photolysis of N2O.28,30–35 Moreover, the 15N fractionation in the second important N2O sink, reaction with O(1D), is qualitatively and quantitatively different at the two positions compared with photolysis.36 Whereas 15N14NO is photolyzed faster than 14N15NO, 14 15 N NO is removed faster in the reaction with O(1D). This difference provides the possibility to distinguish between the two processes in the atmosphere based on the isotope fractionation pattern they leave in N2O, and a corresponding signal has indeed been detected in the lower stratosphere, where reaction with O(1D) is most important.28 These examples show that there is much information to be gained from position-dependent 15N measurements on N2O. The new techniques may also be a valuable tool in the study of nitrification and denitrification pathways and the quantification of N2O source fluxes to the atmosphere, and first measurements on N2O sources are indeed presently becoming available.1– 3,37,38 The mass spectrometric method had been initially developed and described for offline N2O extraction and measurements of the N2O isotopomers using a dual-inlet system.21,22 In follow-up studies the technique was soon realized in continuous-flow mode,1–3,24,28,29,38,39 but this method has not been described in detail to date. Here we describe a continuous-flow method to determine the position-dependent 15N fractionation as well as the 18O fractionation in atmospheric nitrous oxide. Several improvements over the original method have been made, and different modes of sample admission allow an easy measurement at N2O ratios ranging from mixing ratios as low as few nmol/mol up to the percent level. When peak areas are not constant, the largest source of error is the non-linearity of the IRMS source for the NOþ measurement, and a convenient method is described to determine and monitor this nonlinearity that may be instrument-dependent.

EXPERIMENTAL Mass spectrometer configuration A prerequisite for the mass spectrometric determination of the position-dependent 15N fractionation in N2O is an IRMS collector configuration that allows simultaneous monitoring of NOþ ions at m/z 30 and 31, in addition to the conventional set-up with collectors at m/z 44, 45 and 46. Note that it is not necessary to monitor all five masses simultaneously. One possible configuration is the common universal triple collector assembly40 with one narrow central collector and two wider collectors for the other masses, which facilitates measurements of the N2Oþ precursor ion and the NOþ fragment ion in consecutive runs. This is the set-up that has been used for the work presented in this paper. Alternatively, three Rapid Commun. Mass Spectrom. 2003; 17: 1897–1908

High-precision position-dependent

narrow collectors for m/z 44, 45 and 46 and a separate set of collectors for m/z 30 and 31 can be employed, also in consecutive runs. Users of the Micromass Prism II isotope-ratio mass spectrometer may move the collectors between runs to switch between the two configurations (which is of course problematic in routine use). Much of our work in the past using offline mass spectrometry for samples from laboratory experiments was even conducted using the oxygen cup configuration (m/z 32, 33, 34) on a Finnigan MAT 252 mass spectrometer. Here, m/z 30 and 31 do not fit precisely on the oxygen isotope collectors for m/z 32 and 33, so that the measurement is performed on the flanks of the mass peaks rather than on the flat peak top. Even in this undesirable mode, reproducible measurements can be made although the mass spectrometric precision is reduced.15,36 A new generation of IRMS machines is now available which allow simultaneous monitoring of more than three masses. This enables the determination of the positionresolved 15N fractionation and the 18O fractionation in one single measurement, when the collectors are configured to monitor all masses of interest simultaneously. The fact that all isotope ratios can then be determined on the same aliquot of sample will eliminate one source of statistical error from the measurements, namely the instrumental drift between measurements. In the present instrument-determined setup of running two aliquots consecutively, the individual position-dependent signatures are calculated from two different measurements performed at different times, which introduces statistical errors. A second obvious advantage of performing the entire measurement in one single run is the reduction in measurement time. At least for high-precision atmospheric measurements, still relatively large samples have to be extracted because of the low mixing ratio. Finally, when sample size is limited, using the entire sample for both measurements results in a large peak area and gain of precision. Nevertheless, we show below that high-precision isotope measurements on small N2O samples are still possible with the three-collector mass spectrometers that are commonly used throughout the world today.

Sample extraction and preconcentration To prepare a suitable N2O sample for subsequent gas chromatography, N2O has to be extracted from the bulk air (or other bath gas) first. For air samples, where CO2 is usually present in a more than 1000-fold excess relative to N2O, the majority of this CO2 has to be removed, as a CO2 peak of this magnitude would saturate the chromatographic column. The online extraction and preconcentration procedure follows the principle of the ThermoFinnigan PreCon device41 and can easily be operated using the PreCon software. Therefore, it will only be described briefly. Figure 2 shows the experimental set-up. The preconcentration system is generally separated into a sample carrier flow and an analytical flow, which are provided by two separately controllable flows of a pure He carrier gas (purity 99.999%, further purified using a gas purifier, Supelco). Samples can be injected into the first He carrier gas stream in several ways for the different applications, as detailed in the next section. The gas then passes through a chemical trap (12 mm o.d., 20 cm Copyright # 2003 John Wiley & Sons, Ltd.

15

N and

18

O measurements of N2O 1899

length) filled to 70% with Ascarite (NaOH on silica, Aldrich, 20–30 mesh) to remove CO2 (if present) from the sample, and 30% MgClO4 (Merck) at the end to remove the water. For atmospheric samples, the Ascarite trap reduces CO2 by 5 to 6 orders of magnitude, i.e., the initial CO2/N2O ratio of >1000 is reduced to 99% pure 15N14N16O (ICON Inc., USA) contains 8–9% of the 15N16O fragment (mass 31) rather than the mass 30 fragment 14N16O only; vice versa, NO produced from pure 14N15N16O contains 8– 9% of the mass 30 fragment. The purity of each labeled isotopomer was independently checked using an FTIR spectrometer. No interferences from the respective other isotopomers could be detected, indicating that the purity is even higher than the 99% specified. To determine the scrambling correction precisely, several mixtures of normal and enriched N2O were prepared. Small quantities of the >99% pure 15N14N16O, 14N15N16O and 15N15N16O compounds were mixed with reference N2O, using calibrated small volumes with capacitance manometers that were tested for linearity by comparison with a laboratory pressure standard (Digiquartz, Paroscientific). The mixing was double-checked using isotope dilution, i.e., the measured enrichment of mass 45 (average 15N enrichment) of the mixtures relative to the reference N2O was used to calculate the amount of enriched N2O added to the reference gas. The very good agreement between the two measurements, as shown in Fig. 4, indicates the absence of significant systematic errors. Without isotope scrambling, admixture of 15N14N16O should produce no observable 15N enrichment in the NOþ fragment, whereas, for admixture of 14N15N16O, the entire enrichment that is observed in m/z 45 of the parent N2O should show up in the NOþ fragment. In Fig. 5, the observed 15N enrichments in the NO fragment are plotted versus the enrichments in the precursor N2Oþ ion. For addition of 14N15N16O, the enrichment in the NOþ fragment is only 91.1% of that observed for m/z 45, i.e., in the fragmentation process 91.1% of the 14N15N16O molecules added to the original gas produce 15N16O (m/z 31) in the ion Rapid Commun. Mass Spectrom. 2003; 17: 1897–1908


Figure 4. Mixtures of labeled and non-labeled N2O were prepared barometrically and double-checked using isotope dilution on a dual-inlet mass spectrometer (ThermoFinnigan MAT 252). The slope of 1 shows the absence of systematic errors, whereas the remaining scatter is mainly statistical error in the pressure measurements. source. When 15N14N16O is added, an enrichment of 8.3% of the corresponding enrichment in m/z 45 is still observed in the NOþ fragment, showing that in this case 15N16Oþ is produced. These measurements show that an isotope scrambling of about 8.5% occurs in the ion source. This is slightly higher than, but still in good agreement with, the value of 8.1% determined by Toyoda and Yoshida,21 and slightly lower than the value of 9% found in the early study by Friedman and Bigeleisen.23 We note that in a recent publication a scrambling of 19% was found using a different mass spectrometer.2 To consider the scrambling of NOþ quantitatively, a ‘scrambling coefficient’ s is introduced which is defined as the share of nitrogen atoms in NOþ derived from the terminal N atom. Then, the scrambled ratio 31Rs reads: 31

s

15

N and

18


Figure 5. Observed 31R plotted versus observed 45R for mixtures of natural and labeled N2O. Starting with the reference N2O in the bottom left corner, two dilution series were performed using 14N15N16O (solid circles) and 15 14 16 N N O (solid squares). Several of the already mixed materials were used for further tests by adding the respective other compound (open symbols). In the absence of isotope scrambling, a slope close to 1 is expected for the addition of 14 15 16 N N O, as indicated by the dashed line, whereas a constant is expected for the addition of 15N14N16O (with the unlabeled position having the typical ambient 15N content). For the actual mixtures, the slope for the addition of 14 15 16 N N O is 0.911 on average, and for the addition of 15 14 16 N N O we find 0.083. Small corrections can be applied to account for the influence of 15N15N16O and 14N14N16O.15 Measurements were made using conventional dual-inlet mass spectrometers (ThermoFinnigan MAT 252 and Micromass PRISM II).

N14 N16 O þ 15 N15 N16 O þ 14 N14 N17 O þ 14 N15 N17 O Rs ¼ 14 14 16 s N N O þ 14 N15 N16 O þ ð1 sÞ 14 N14 N16 O þ 15 N14 N16 O ð1 sÞ 14 N15 N16 O þ 15 N15 N16 O þ 14 N14 N17 O þ 15 N14 N17 O þ 14 14 16 s N N O þ 14 N15 N16 O þ ð1 sÞ 14 N14 N16 O þ 15 N14 N16 O 15

Assuming a statistical isotope distribution, this can be rearranged to: Rs ¼ s15 R1 þ ð1 sÞ15 R2 þ 17 R

31

sð1 sÞ

15

R1 15 R2

abundance samples. For work with artificially enriched 15N isotopes, this term may become relevant though.

2

1 þ s15 R2 þ ð1 sÞ15 R1 ð6Þ

Note that the last term in the above equation is missing in the formula from Yoshida and Toyoda.21,24 This term arises from the fact that also the double-substituted isotopomer 15 15 16 N N O contributes to the scrambled as well as the unscrambled fragment at m/z 31. It has a maximum value for s ¼ 0.5 and for large differences in the isotope ratios at the central and terminal nitrogen positions (15R2 and 15R1). However, since s is only 8–9% in reality, even an extreme enrichment of the central over the terminal N position of 160% leads to a correction of only 0.006% for natural Copyright # 2003 John Wiley & Sons, Ltd.

ð5Þ

Dependence on the reference gas As mentioned above, the isotopic composition at the terminal position cannot be determined independently due to interferences.22 Therefore, the individual isotope signatures are derived from the measurement of the NOþ fragment and the average fractionation. However, the mass spectrometric measurement of the fragment ion compares the 15N content of sample and reference gas at the central position (including the scrambling as discussed above) whereas the measurement of the average fractionation compares the average 15N content of sample and reference gas. A problem arises if 15N is not distributed evenly between the two positions in the reference gas. Then the average measurement does not yield the Rapid Commun. Mass Spectrom. 2003; 17: 1897–1908


1904

arithmetical average of the individual fractionations at the two positions, but the weighted average of the intramolecular distribution of 15N in the reference gas, represented by the parameter f, where: ð7Þ f ¼ 15 R1;st = 15 R1;st þ 15 R2;st is the fraction of 15N at the terminal position in the reference gas. Then, the average 15N fractionation can be calculated from the individual fractionations as: 15 N ¼ f 1 15 N þ ð1 f Þ 2 15 N

ð8Þ

The simplest case occurs for f ¼ ½, since then 15 Rst ¼ 15R1,st ¼ 15R2,st and thus both 31Rst and 31Rs,st are equal to 15Rst þ 17Rst [Eqns. (3) and (6)]. The average d15N value is equal to the arithmetic average of 1d15N and 2d15N. f > ½ means that the average d15N value is biased towards the enrichment at the terminal N position of the N2O molecule; f < ½ means that d15N is biased towards the enrichment at the central N position. This is independent of any scrambling and serves to illustrate how a potential asymmetry in the intramolecular 15N distribution in the working standard may influence the average d15N. As long as we are dealing with average d15N values only, this is irrelevant. However, for position-dependent measurements, it has to be taken into account. We define 2d* as the apparent d value of the central N atom under the influence of scrambling and a potentially asymmetric standard: 2

31

¼ 31

Rs 17 R 1 Rs;st 17 Rst

ð9Þ

Substitution of Eqns. (6) (neglecting the third term), (7) and (8) into Eqn. (9) yields: s 2 2 15 N 2 ¼ 2 þ ð10Þ ð1 f Þð1 2sÞ 1

¼ 2 þ

15 N 2 2 ð1 sÞ 2 15 N ¼ f f ð1 2sÞ

ð11Þ

Like the magnitude of the scrambling, the parameter f has to be quantified separately. The absolute determination of the intramolecular distribution of 15N within N2O is difficult. From the ratio 31R 0.39% observed in the mass spectrometer, Brenninkmeijer and Ro¨ckmann concluded that f is at least roughly in agreement with 0.5 for the reference gas. Two groups have attempted an absolute position-dependent calibration of their working standard gases. When translated to tropospheric N2O, Toyada and Yoshida,21 using a chemical conversion method, found differences in the enrichments at the central and terminal nitrogen position of 18.7 2.2%, whereas Kaiser et al.48 find a value of 45.8 1.4% using a purely mass spectrometric calibration. At present, this large discrepancy cannot be resolved, but we note that as long as d values are reported relative to a N2O standard, 1d15N and 2 15 d N are not too sensitive to the precise value of f, so that the assumption of f ¼ 0.5 for tropospheric N2O does not lead to large errors. However, this is clearly different if 1d15N and 2 15 d N are reported relative to atmospheric N2. Here, the precise knowledge of f is essential. Copyright # 2003 John Wiley & Sons, Ltd.

Sample admission Pressurized air samples In the past we have analyzed samples from various pressurized air cylinders and from flasks at atmospheric pressure and devised various convenient admission systems. For pressurized samples, the air sample is injected into the He flow via a prime/purge valve V1, which is part of the commercial PreCon system (Fig. 2(a)). To prevent pressures above atmospheric pressure in the preconcentration trap (which may lead to condensation of oxygen), the entire PreCon system is operated below atmospheric pressure by pumping with a membrane pump at the vent through a suitable capillary. This reduces the pressure at the injection point to roughly 400 mbar for the pure He carrier gas and to slightly below atmospheric pressure when an air sample is admitted. A constant sample flow rate of 25 mL/min is set by a mass flow controller (MFC). Before the MFC, a Valco multiport valve allows the simultaneous connection of presently up to four sample cylinders. All valves can be operated automatically using the mass spectrometer software. Depending on the N2O mixing ratio of the sample and on the kind of measurement (N2Oþ or NOþ fragment), different amounts of air need to be processed to obtain the desired peak areas and thus precision in the mass spectrometer. With the present set-up, the amount of air admitted to the system can be adjusted easily by opening V1 for the adequate time. For example, at atmospheric N2O mixing ratios (315 nmol/mol), we usually process 125 or 167 mL of air (admission time 300 or 400 s). For one stratospheric air sample with a mixing ratio of 19 nmol/mol we processed 625 mL (1500 s admission time) and still got 30% of the peak area; for the NOþ measurement, we even processed 2.5 L (6000 s admission time).

Ambient pressure air samples The samples at atmospheric pressure we have processed thus far came in flasks equipped with a septum. Two capillary tubes, one for the He carrier gas flow and one for the outlet, were injected with needles through the septum and connected to the prime/purge vales V2 and V1, respectively (Fig. 2(b)). By closing a third valve V3 between V1 and V2, and opening V1 and V2, the carrier gas was directed through the sample flasks and the sample was admitted. Again, the admission time can be adjusted to yield the desired peak area. Also, the membrane pump at the vent keeps the pressure in the system below atmospheric pressure.

High concentration samples High concentration samples, either from the condensable fraction of air samples (‘trace-gas cocktails’) or from synthetic mixtures prepared for testing purposes, were injected using a further 6-port valve (Fig. 2(c)) with a sample loop of the desired size (we used typically 50–1000 mL sample loops for mixing ratios up to 1%; in principle, one should even be able to inject pure N2O this way using a sufficiently small sample loop). Again, the samples were prepared at approximately atmospheric pressure in glass flasks equipped with a septum, but now the sample was purged by a very small He flow (0.1 mL/min) through the sample loop using a doublehole needle. After injection, the contents of the sample loop Rapid Commun. Mass Spectrom. 2003; 17: 1897–1908


were flushed by a flow of He (4 mL/min) through a glass trap filled with Ascarite (6.35 mm, 20 cm, to remove CO2 if present) into the preconcentration trap and processed as described above. The advantage of this system is the considerably reduced admission time, the possibility to inject various diagnostic trace gas mixtures (e.g., N2O in He, N2O in pure CO2) and the possibility to directly inject the abovementioned ‘trace-gas cocktails’ without dilution into a bath gas. When the sample flask is large enough (e.g., a 1-L flask), its contents are virtually not changed by the small dilution of the carrier gas He, and peaks of similar size can be injected at a high rate for long times, which is useful for adjusting system parameters and for reproducibility tests. When the sample flask is small, its contents get diluted exponentially with time and a series of continuously decreasing peaks is admitted, which provides for a useful and easy linearity test (see below). At the same time, this decrease in peak size is also a drawback for the measurement of actual samples from the concentrates. Clearly, it is not practical to obtain 1 L of condensable trace gases out of atmospheric air, which is equivalent to about 2.8 m3 of air. For our experiments, we filled 15-mL glass flasks to atmospheric pressure with the trace gas concentrate (equivalent to 40 L of air). For these samples, the peak size continuously decreases which is problematic for the NOþ measurement due to the nonlinearity. However, this non-linearity can be corrected for by measuring mixtures of the reference gas in a flask of the same size. The isotope ratio can then be determined by comparing a series of reference gas peaks of varying size with a series of sample peaks.

RESULTS Open split interface The benchmark test for the new open split system has been the successive injection of reference gas peaks. During the entire run, the He dilution capillary is kept in ‘split in’ position and the reference gas is injected for 20 s and retracted for 30 s. In routine operation a reproducibility of 0.03% for d15N and 0.04% for d18O is obtained for ten N2O reference peaks, which is comparable to the commercial inlet systems. The reproducibility is worse for the NOþ fragment ion (0.1% for 31d), and this value is sometimes only reached after a few runs. The reduced precision is likely due to the comparatively small fraction of NOþ ions (17–35% of the N2Oþ ions on different mass spectrometers) and the correspondingly smaller peaks. Additional production of NO2 or NO on the filament as observed previously45 may play a role.

15

N and

18


þ

Non-linearity for NO fragment analysis As mentioned above, a serious limitation for mass spectrometry of the NOþ fragment is the non-linearity. Linearity tests are usually performed on continuous-flow machines by running series of reference peak injections, successively increasing the concentration of the reference gas and thus the peak amplitude. Using this procedure, we have repeatedly found that the non-linearity usually amounts to approximately 5%/ V (a feedback resistor of 3 108 is used to convert the major ion beam current into a voltage, i.e., 1 V corresponds roughly to 3 nA), and it can be much worse when the source is not well tuned. The general problem with this kind of linearity test is that it is carried out with almost rectangular peak shapes (reference gas injections), whereas sample peaks usually involve transient signals. A series of peak injections from a sample loop filled from a small volume, where the sample is slowly diluted by the reference gas He as described above for the admission of high concentration samples, offers an easy way to carry out a linearity test using ‘real peak shapes’. Figure 6 displays results of a typical measurement series with peak areas decreasing from 3.2 to 1.4 Vs and a corresponding decrease in the 31d values. We show here the change in the d value with changing peak area rather than amplitude, because we generally found a better correlation. The nearly perfect exponential decay of the peak areas illustrates how the sample is slowly diluted by the carrier gas He. The isotope values are also described well by an exponential fit so that a measurement series can easily be corrected for varying peak areas.

Air samples To test and monitor the performance of the entire analytical procedure for air samples with atmospheric N2O mixing ratio or less, reproducibility and linearity tests have been performed regularly over many months using a reference atmospheric air cylinder filled at the Schauinsland Station in the Black Forest, Germany. The results in Fig. 7 show the standard deviation of a single measurement within one series of 5 up to 20 measurements as a function of peak area. The peak

Blanks Using the new sample admission procedures, the preconcentration system is operated below atmospheric pressure. This bears the risk that atmospheric air entering the system via leaks may contaminate the samples. Therefore, blank runs are routinely carried out (e.g., after the Ascarite trap is changed) without injection of an air sample. In these blank runs, no detectable N2O peaks could be found, which shows the absence of significant air leaks. Copyright # 2003 John Wiley & Sons, Ltd.

Figure 6. Series of sample injections from a small volume using the double-hole needle. The gas in the sample vessel is continuously diluted by the carrier gas He. Thus, the areas decrease from run to run. A simultaneous decrease in the 31d value is observed, due to the non-linearity of the mass spectrometer. Rapid Commun. Mass Spectrom. 2003; 17: 1897–1908

1906


Table 1. Reproducibility test using several stratospheric air samples. The values listed are raw data versus the laboratory reference N2O gas obtained on different days N2O (ppb)

d15N

d18O

11

279.6

4

34.1

8.636 8.931 8.935 70.489 70.405

8.096 8.464 8.406 60.593 59.871

Sample #

31

Figure 7. The achievable precision of an isotope ratio measurement depends on the peak area. Below approximately 3 Vs, the errors increase. These data were obtained from routine reproducibility and linearity tests of the system using a real air sample. Each point represents the standard deviation of a series of 5 to 20 measurements.

area is usually varied by varying the time the sample is admitted to the preconcentration unit, sometimes also by varying the flow rate of the sample into the system. It is evident that for peak areas above approximately 3 Vs, the error of a single measurement is about 0.1% for d15N and about 0.2% for d18O. Below 3 Vs, the standard deviations increase, but they remain approximately half as large for d15N compared with d18O. Due to the relatively small yield of NOþ fragment ions (see above), and to avoid too long admission times and limit sample consumption, fragment analyses are usually carried out at peak areas at or slightly below 2 Vs, giving a standard deviation of 0.3%. With decreasing peak area, this error also increases. The measured d values of these reproducibility and linearity tests give a good indication of the linearity of the system. Figure 8 shows that average 15N and 18O results only display a very small degree of non-linearity even down to an

12

293.9

1

39.8

d

10.451 10.506 89.317 89.060

area of only 1 Vs. The non-linearity for the fragment ion is also clearly apparent here, illustrating that a linearity correction is required if samples do not have a constant peak area. Clearly, such extensive measurement series cannot be performed for all samples. In particular, samples from the stratosphere are rather limited in size. Nevertheless, for several stratospheric samples multiple analyses were carried out and the results are displayed in Table 1. Even for these low concentration samples with high enrichments, measurements of high precision are possible.

Measurement of the N2O mixing ratio In addition to the d values (isotope information), the mass spectrometer report also includes the peak areas recorded during the measurement. This information can be used to derive the total N2O mixing ratio of an air sample. By scaling the MS peak areas to the peak area of the laboratory standard gas that was always measured between the stratospheric samples (air sample with nominally 318 nmol/mol N2O), we obtained an independent measure of the mixing ratio. Note that the MS measurements actually provide two independent measurements, one on the N2Oþ and one on the NOþ fragment, which can again be averaged to reduce statistical error.

‘Trace-gas cocktails’

Figure 8. Linearity test of the analytical procedure of atmospheric air samples. d values are the raw data versus the reference N2O gas. Note that the variations here also include day-to-day variability of the instrument and several small changes in the analytical set-up. Copyright # 2003 John Wiley & Sons, Ltd.

Injections of the extracts of condensable gas are a particularly good example for samples with variable peak areas. The sample is continuously diluted with helium and, as shown in Fig. 6, the d value for the NOþ fragment is strongly dependent on peak area. Similar measurements can easily be carried out using N2O reference gas mixtures filled into the same flasks. Figure 9 shows results for two different reference gases and one air sample. To mimic the atmospheric trace gas mixtures as closely as possible, 1:1000 mixtures of the reference gases in CO2 were prepared for these experiments. The two different reference gases have a similar average 15N content, but a different distribution between the two positions. One reference gas is produced from ammonium nitrate, the other one from adipic acid production. Figure 9 shows that the non-linearity for the two reference gas mixtures and the air sample are identical within the errors. Thus, a calibration curve can Rapid Commun. Mass Spectrom. 2003; 17: 1897–1908


15

N and

18


and reference gas does not reduce the analytical precision, but greatly increases flexibility, when a mass spectrometer is used for various applications.

Acknowledgements We thank Bernd Knape for his help with the construction of the analytical system and for carrying out many of the measurements presented in this paper. John Crowley checked the purity of the isotopically labeled N2O gases by FTIR spectroscopy.

Figure 9. Linearity tests for two different N2O reference gas mixtures (1% N2O in CO2) with different 15N enrichment at the two positions, and one condensable ‘trace gas cocktail’ from a tropospheric air sample. 15-mL volumes were filled with the gas mixtures and processed as described above. For the reference gases, data are from various measurement series during the measurement period. The points for the air sample represent one single measurement series like the one shown in Fig. 6. For the three gases, the slopes of the fits are identical within the errors. This means that the d value of a sample versus the reference gas can be obtained from the vertical distance of a measurement point from the fit line through the reference gas data [Eqn. (12)].

be established using the reference gas mixtures, and the d value of a sample versus the reference gas can be calculated from the vertical distance at the respective peak area according to: 31 ¼

31 31 SA REF 31 1 þ REF

ð12Þ

CONCLUSIONS 18

O and position-dependent 15N measurements on atmospheric N2O are possible with high precision using an online technique. The reproducibility for a 125-mL sample at ambient N2O concentration is 0.1% for d15N, and 0.2% for d18O; for a 420-mL sample we achieve 0.3% for the d31 fragment measurement. The new method can in principle be realized on any gas isotope ratio mass spectrometer that has a collector configuration to simultaneously monitor m/z 30 and 31, in addition to the conventional collectors for ions at m/z 44, 45 and 46. Using different sample inlet devices, samples with mixing ratios from few nmol/mol up to the percent level can be analyzed. The implementation of a split-column set-up for the gas chromatographic separation is a significant improvement over existing N2O isotope systems, particularly important for fragment ion analysis. A non-linearity for the NOþ fragment ion of approximately 1.5% per Vs peak area has to be taken into account. Using the high concentration sample inlet, the non-linearity can be determined easily and quickly. The new open split interface using only one transfer capillary to the mass spectrometer for sample Copyright # 2003 John Wiley & Sons, Ltd.

REFERENCES 1. Toyoda S, Yoshida N, Miwa T, Matsui Y, Yamagishi H, Tsunogai U. Geophys. Res. Lett. 2002; 29: 10.1029/ 2001GL014311. 2. Sutka RL, Ostrom NE, Ostrom PH, Gandhi H, Breznak JA. Rapid Commun. Mass Spectrom. 2003; 17: 738. 3. Popp BN, et al. Glob. Biogeochem. Cycle. 2002; 16: 10.1029/ 2001GB001806. 4. Ostrom NE, Russ ME, Popp B, Rust TM, Karl DM. Chemosphere: Global Change Sci. 2000; 2: 281. 5. Barford CC, Montoya JP, Altabet MA, Mitchell R. Appl. Environ. Microbiol. 1999; 65: 989. 6. Mariotti A, Germon JC, Hubert P, Kaiser P, Letolle R, Tardieux A, Tardieux P. Plant Soil 1981; 62: 413. 7. Mandernack KW, Rahn T, Kinney C, Wahlen M. J. Geophys. Res. 2000; 105: 17709. 8. Yoshida N, Hattori A, Saino T, Matsuo S, Wada E. Nature 1984; 307: 442. 9. Yoshida N. Nature 1988; 1988: 528. 10. Yoshinari T, Wahlen M. Nature 1985; 317: 349. 11. Webster EA, Hopkins DW. Biol. Fertility Soils 1996; 22: 326. 12. Wada E, Yoshida N, Yoshioka T, Yoh M, Kabaya Y. Mitt. Int. Verein. Limnol. 1996; 25: 115. 13. Prather M, et al. In Climate Change 2001; The Scientific Basis: Contribution of Working Group I to the Third Assessment Report of the Intergovernmental Panel on Climate Change, Houghton JT, et al. (eds). Cambridge University Press: Cambridge, 2001; 239–287. 14. Cliff SS, Thiemens MH. Anal. Chem. 1994; 66: 2791. 15. Kaiser J. PhD thesis, University of Mainz, 2002. 16. Ro¨ckmann T, Kaiser J, Crowley JN, Brenninkmeijer CAM, Crutzen PJ. Geophys. Res. Lett. 2001; 28: 503. 17. Cliff SS, Thiemens MH. Science 1997; 278: 1774. 18. Estupinan EG, Nicovich JM, Li J, Cunnold DM, Wine PH. J. Phys. Chem. 2002; 106: 5880. 19. Esler MB, Griffith DWT, Turatti F, Wilson SR, Rahn T, Zhang H. Chemosphere Global Change Sci. 2000; 2: 445. 20. Uehara K, Yamamoto K, Kikugawa T, Yoshida N. Sensor Actuat B-Chem. 2001; 74: 173. 21. Toyoda S, Yoshida N. Anal. Chem. 1999; 71: 4711. 22. Brenninkmeijer CAM, Ro¨ckmann T. Rapid Commun. Mass Spectrom. 1999; 13: 2028. 23. Friedman L, Bigeleisen J. J. Chem. Phys. 1950; 18: 1325. 24. Yoshida N, Toyoda S. Nature 2000; 405: 330. 25. Yung YL, Miller CE. Science 1997; 278: 1778. 26. Rahn T, Wahlen M. Science 1997; 278: 1776. 27. Griffith DWT, Toon GC, Sen B, Blavier J-F, Toth RA. Geophys. Res. Lett. 2000; 27: 2485. 28. Ro¨ckmann T, Kaiser J, Brenninkmeijer CAM, Crowley JN, Borchers R, Brand WA, Crutzen PJ. J. Geophys. Res. 2001; 106: 10403. 29. Toyoda S, Yoshida N, Urabe T, Aoki S, Nakazawa T, Sugawara S, Honda H. J. Geophys. Res. 2001; 106: 7515. 30. Toyoda S, Yoshida N, Suzuki T, Tsuji K, Shibuya K. In Int. Conf. Study of Environmental Change Using Isotope Techniques, Vienna, Austria, International Atomic Energy Agency: Vienna, 2001. 31. Kaiser J, Brenninkmeijer CAM, Ro¨ckmann T. Phys. Chem. Chem. Phys. 2002; 4: 10.1039/b204837j. 32. Kaiser J, Brenninkmeijer CAM, Ro¨ckmann T, Crutzen P. Atmos. Chem. Phys. Disc. 2002; 2: 1735. Rapid Commun. Mass Spectrom. 2003; 17: 1897–1908

1908


33. Ro¨ckmann T, Brenninkmeijer CAM, Wollenhaupt M, Crowley JN, Crutzen PJ. Geophys. Res. Lett. 2000; 27: 1399. 34. Rahn T, Zhang H, Wahlen M, Blake GA. Geophys. Res. Lett. 1998; 25: 4489. 35. Zhang H, Wennberg PO, Wu VH, Blake GA. Geophys. Res. Lett. 2000; 27: 2481. 36. Kaiser J, Brenninkmeijer CAM, Ro¨ckmann T. J. Geophys. Res. 2002; 107: 10.1029/2001JD001506. 37. Yamulki S, Toyoda S, Yoshida N, Veldkamp E, Grant B, Bol R. Rapid Commun. Mass Spectrom. 2001; 15: 1263. 38. Pe´rez T, Trumbore SE, Tyler SC, Matson PA, OrtizMonasterio I, Rahn T, Griffith DWT. J. Geophys. Res. 2001; 106: 9869. 39. Ro¨ckmann T, Kaiser J, Brenninkmeijer CAM. Atmos. Chem. Phys. Disc. 2002; 2: 2021.

Copyright # 2003 John Wiley & Sons, Ltd.

40. Habfast K. In Modern Isotope Ratio Mass Spectrometry, Platzner IT (ed). John Wiley: Chichester, 1997; 11–82. 41. Brand WA. Isotopes Environ. Health Stud. 1995; 31: 277. 42. Kaiser J, Brenninkmeijer CAM, Ro¨ckmann T. J. Geophys. Res. 2003; in press. 43. Leckrone KJ, Hayes JM. Anal. Chem. 1997; 69: 911. 44. Ferretti DF, Lowe DC, Martin RJ, Brailsford GW. J. Geophys. Res. 2000; 105: 6709. 45. Tanaka N, Rye DM, Rye R, Avak H, Yoshinari T. Int. J. Mass Spectrom. Ion Processes 1995; 142: 163. 46. Kaiser J, Brenninkmeijer CAM, Ro¨ckmann T. Geophys. Res. Lett. 2003; 30: 10.1029/2002GL016073. 47. Assonov SS, Brenninkmeijer CAM. Rapid Commun. Mass Spectrom. 2001; 15: 2426. 48. Kaiser J, Park S, Boering KA, Brenninkmeijer CAM, Hilkert A, Ro¨ckmann T. Anal. Bioanal. Chem. 2003; submitted.

Rapid Commun. Mass Spectrom. 2003; 17: 1897–1908