On-Line Analysis of Stable Isotopes of Nitrogen in NH3, NO, and NO2

On-Line Analysis of Stable Isotopes of Nitrogen in NH3, NO, and NO2 at Natural Abundance Levels. Jutta Lauf* ... On-Line Analysis of Nitrogen Stable I...
1 downloads 0 Views 188KB Size
Download PDF
Anal. Chem. 1998, 70, 2750-2756

On-Line Analysis of Stable Isotopes of Nitrogen in NH3, NO, and NO2 at Natural Abundance Levels Jutta Lauf* and Gerhard Gebauer

Lehrstuhl fu¨ r Pflanzeno¨ kologie, Universita¨ t Bayreuth, D-95440 Bayreuth, Germany

Methods were developed for the on-line analysis of stable isotopes of nitrogen (at natural abundance levels) in NH3, NO and NO2 in order to study the contribution of these trace gases to nitrogen cycling in ecosystems. Standard methods for the combustion of organic substances to N2 (for the determination of the 15N/14N ratios by mass spectrometry) failed to satisfactorily oxidize or reduce the respective trace gases. The following oxidation or reduction techniques ensured quantitative conversion of the trace gases to N2 and analytical precision close to the internal precision of the instrument (precision (0.15‰): (1) oxidation of NH3 to N2 at 1150 °C on a NiO surface (which needs a reoxidation before each NH3 analysis; measuring range 11.7-58.8 nmol of NH3, precision (0.30‰); (2) reduction of NO to N2 at 1150 °C on a Ni surface (measuring range 33.3-133.3 nmol of NO, precision (0.28‰); (3) reduction of NO2 to form NO at 420 °C on a Mo surface followed by further treatment as for NO (measuring range 26.0-129.9 nmol of NO2, precision (0.90‰). This last technique was developed due to the poor chromatographic properties of NO2. In the atmosphere, ammonia (NH3), nitrogen monoxide (NO), and nitrogen dioxide (NO2) are trace gases of mostly anthropogenic origin. NH3 is released from animal production,1 from plant surfaces,2 and from a few industrial processes, e.g. removal of nitrogen oxides (NOx) from flue gases.3 NO is produced in combustion processes above 1000 °C4 and during microbiological nitrification5 and denitrification5 and is quickly oxidized to NO2 in the atmosphere.4 NH3, NO, and NO2 are involved in a variety of processes that contribute to environmental pollution. NO and NO2 contribute to the ozone production cycle6 and are greenhouse gases.7 All the nitrogen-containing trace gases mentioned above contribute to atmospheric input of nitrogen to terrestrial ecosys(1) Mannheim, T.; Braschkat, J.; Marschner, H. Z. Pflanzenernaehr. Bodenkd. 1995, 535-542. (2) Mattsson, M.; Schjoerring, J. K. J. Exp. Bot. 1996, 297, 477-484. (3) Meijer, R.; Janssen, F. VGB Kraftwerkstech. 1992, 72, 914-917. (4) Holleman, A. F.; Wiberg, N. Lehrbuch der anorganischen Chemie; de Gruyter: Berlin, 1985; Vol. 33. (5) Stevens, R. J.; Laughlin, R. J.; Burns, L. C.; Arah, J. R. M.; Hood, R. C. Soil Biol. Biochem. 1997, 29 (2), 139-151. (6) Warneck, P. Chemistry of the natural atmosphere; International Geophysics Series 41; Academic Press: New York, 1988; pp 178-194. (7) Lammel, G.; Grassl, H. Environ. Sci. Pollut. Res. 1995, 20, 40-45.

2750 Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

tems, and plants are able to use the nitrogen for growth.8,9 Input of nitrogen from the atmosphere may induce nutritional imbalances in plants, especially in poor soils, and excessive deposition of atmospheric nitrogen is closely linked to forest decline across large areas in Europe.10 A wide range of studies have suggested that natural variation in the ratio of stable isotopes of nitrogen in soil, in plants, in dry and wet nitrogen deposition, and in nitrogen-containing trace gases is a powerful tool in studying the mechanisms and pathways of NH3, NO, and NO2 in ecosystems, including natural ecosystems. Efforts to characterize the 15N/14N variation patterns of these gases in the atmosphere6 and of the processes responsible for their production require fast and high-precision analysis of hundreds of samples. Such studies have been hindered by analytical limitations associated with natural concentrations which, in the atmosphere, are in the lower ppb range. Until now there have been few measurements of the abundance of isotopes of nitrogen in ammonia, nitrogen monoxide, and nitrogen dioxide.11-14 In all of these studies, samples were prepared for isotopic mass spectroscopy using off-line techniques. The gases were trapped chemically as salts from a specified volume of air. These salts were then converted to N2 using Cu/ CuO in evacuated sealed glass ampules and subsequently measured via a viscous-flow inlet system or by use of an elemental analyzer coupled to an isotope-ratio mass spectrometer. Recent advances in isotope-ratio gas chromatography/mass spectrometry15 have initiated new studies of the on-line analysis of isotope abundance in trace gases. Applications for on-line analysis of N2O are already available because it is present in higher concentrations in the atmosphere and is inert under tropospheric conditions.16,17 The purpose of this report is to present new techniques for the on-line analysis of stable isotopes of nitrogen in NH3, NO, and NO2 at concentrations found in the atmosphere (natural abundance levels). EXPERIMENTAL SECTION Glassware. All glassware used in this study was deactivated with a solution of 5% (CH3)2Cl2Si (dimethyldichlorosilane) in (8) Sommer, S. G.; Jensen, E. S. J. Environ. Qual. 1991, 20, 153-156. (9) Segschneider, H.-J.; Wildt, J.; Fo¨rstel, H. New Phytol. 1995, 131, 109-119. (10) Schulze, E.-D. Science 1989, 244, 776-783. (11) Freyer, H. D. Pure Appl. Geophys. 1978, 116, 393-404. (12) Moore, H. Atmos. Environ. 1977, 11, 1239-1243. (13) Heaton, T. H. E. Atmos. Environ. 1978, 21, 843-852. (14) Heaton, T. H. E. Tellus 1990, 42B, 304-307. (15) Mathews, D. E.; Hayes, J. M. Anal. Chem. 1978, 50, 1465-1473. (16) Yoshida, N.; Matsuo, S. Geochem. J. 1983, 17, 231-239. (17) Brand, W. A. Isotopes Environ. Health. Stud. 1995, 31, 277-284. S0003-2700(98)00005-5 CCC: $15.00

© 1998 American Chemical Society Published on Web 06/02/1998

Table 1. Summary of the GC Columns and the Oxidation or Reduction Techniques Used for Experiments 1-6a GC column experiment 1 29.4 nmol of NH3 50.0 nmol of NO experiment 2 29.4 nmol of NH3 50.0 nmol of NO experiment 3 29.4 nmol of NH3 50.0 nmol of NO experiment 4 5.9-64.7 nmol of NH3 16.7-133.3 nmol of NO experiment 5 81.5 nmol of NO2 experiment 6 81.5 nmol of NO2

furnace II

furnace III

PoraPlot Q (Figure 1) PoraPlot S (Figure 1)

NiO, 900-1300 °C Ni, 900-1300 °C

Cu, 640 °C Cu, 640 °C

PoraPlot Q (Figure 1) PoraPlot S (Figure 1)

NiO, 900-1300 °C Ni, 900-1300 °C

PoraPlot Q (Figure 1) PoraPlot S (Figure 1)

furnace I

NiO, 1150 °C Ni, 1150 °C

NiO, 1150 °C

PoraPlot Q (Figure 1) PoraPlot S (Figure 1)

NiO, 1150 °C Ni, 1150 °C

PoraPlot S (Figure 2)

Mo, 320-460 °C

Ni, 1150 °C

PoraPlot S (Figure 2)

Mo, 420 °C

Ni, 1150 °C

a The GC and combustion interface flow pathways are shown in Figures 1 and 2. NH was separated from other compounds on a PoraPlot Q 3 column at 140 °C (0.32 mm i.d., 30 m long, 1.5 mL/min He carrier gas flow at 140 °C; Chrompack, Middelburg, The Netherlands). NO was separated from other compounds on a PoraPlot S column at -77 °C (ID 0.32 mm, 15 m long, 5 mL/min He carrier gas flow at -77 °C; Chrompack).

toluene at room temperature for 24 h and then washed in ethanol at room temperature for 24 h. The furnaces of the combustion interface were exclusively connected with deactivated fused silica capillaries, 0.32 mm i.d. Sample Preparation. Pure NH3 (99.98%), NO (99.5%), and NO2 (98%) purchased in lecture bottles (Linde, Munich, Germany) were used for all experiments. Because of the small volume of the lecture bottles (0.38 L) δ15N increased by about 4‰ as the bottle was emptied. This drift was corrected to achieve comparability of the data obtained from different experiments. To avoid contamination of the pure gases with atmospheric N2 during the removal process, one arm of a T-shaped glass tube was connected via a PTFE tube to the lecture bottle with the directly opposite arm vented to the hood. The third arm was sealed with a Teflon rubber septum in an Ultra-Torr connector (Cajon, Solon, OH). With a 100 µL gastight syringe (Hamilton, Bonaduz, Switzerland) the septum was penetrated and pure gas was removed at a constant gas flow. The T-shaped glass tube was constructed in such a way that the end of the syringe needle reached into the constant gas flow. The first gas sample was removed after a minimum of 5 min of constant flow. The amount removed of each pure compound ranged between 5.9 and 133.3 nmol. The volume to mass relationship was corrected based on the daily air pressure and temperature using gas law. Mass Spectrometer. The 15N/14N ratios of N2 gas generated from the respective trace gases were measured using a Finnigan MAT Delta S (Finnigan MAT, Bremen, Germany) isotope ratio mass spectrometer (IRMS) coupled to a Hewlett-Packard gas chromatograph (GC) 5890 Series II (Wilmington, DE) via a Finnigan MAT combustion interface II (Finnigan MAT, Bremen, Germany). Isotope ratios were compared to those of a N2 reference gas (99.9995%, Linde), which was previously calibrated against the reference substances N1 and N2 provided by the IAEA (Vienna, Austria). Results are reported in δ notation, where δ15Ν [‰] ) (RSA/RST - 1) × 1000. RSA and RST refer to the ratio 15N/ 14N in the sample and in air N .18 The internal reproducibility of 2 the mass spectrometer measurement of N2 is typically (0.15‰

or better. The ion current at m/z 30 (14N16O or 15N15N) provided a quantitative measure of completeness of the reduction of NO and NO2 to N2. Detection of any ion currents at m/z 30 above the natural abundance of 15N15N indicates that N2 generation conditions were insufficient and that regeneration or replacement of furnace material was needed. The unit of the intensity of the ion current is [V]. An integration above time results in [V s]. This unit is used in the following to report the recovery of N2. GC and Combustion Interface. Figure 1 illustrates the GC combustion interface coupling as modified for subsequent experiments. After injection, gas samples passed through furnace I and were subsequently separated on a GC column (Table 1). The samples then passed through furnaces II and III. The furnaces were equipped with ceramic tubes (320 mm long, ID 0.8 mm, OD 1.5 mm; Friatec AG, Mannheim, Germany) filled with metal wires (125 mg).19 The furnaces catalyzed specific oxidation or reduction reactions as follows: (a) NiO (99.98% Ni, OD 0.25 mm, oxidized with O2 at 1000 °C for 30 min to NiO; Goodfellow, Cambridge,

(18) Mariotti, A. Nature 1983, 303, 685-687.

(19) Lauf, J.; Gebauer, G. Deutsche Patentanmeldung 196 50 444.9, 1996.

Figure 1. Schematic diagram of the GC combustion interface IRMS coupling as used for the experiments 1-4.

Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

2751

Figure 2. Schematic diagram of the GC combustion interface IRMS coupling as used for the experiments 5 and 6.

UK) was used for the oxidation of NH3 to N2. NiO (mp 1990 °C)4 was favored over CuO (146 kJ + 2CuO f Cu2O + 1/2O2, 900 °C)4 as an oxidation reagent because of greater potential reaction temperatures. (b) Ni (Goodfellow, 99.98%, OD 0.25 mm) was used for the reduction of NO to N2. Ni (mp 1452 °C)4 was favored over Cu (mp 1083 °C)4 because of the greater potential reaction temperatures. (c) Cu (Goodfellow, 99.95%, OD 0.25 mm) was tested as an additional reduction reagent for NOx. Cu is commonly used in elemental analzyers as a reduction reagent for NOx at 600640 °C after oxidation reactions. As a common reaction step it was used for testing the completeness of the NO reduction. (d) Mo (Goodfellow, 99.95%, OD 0.25 mm) was used for the reduction of NO2 to NO. NO2 is a reactive gas and has poor chromatographic performance.4,20,21 Mo was chosen for the precolumn reduction of NO2 to NO because it is already commonly used for detection of NOx by chemiluminescence, a robust and routine method in environmental analysis.22 The NO produced was further reduced as described previously. All furnaces could be equipped with an empty ceramic tube which, at 30 °C, had no chemical function and served only as a capillary to cross the furnace. All furnaces could be heated to 1300 °C. H2O generated during the combustion process was removed from the carrier gas stream with a Nafion tube. CO2, delivered with the He carrier, was removed with a liquid nitrogen trap to prevent interference with δ15N determination. The generated N2 was transported to the IRMS with the He carrier gas stream. The material in furnace II could be regenerated on-line with O2 (99.995% Linde) or H2/N2 (30/70 vol %, Linde) depending on the material used. In principle, oxidation or reduction reactions can also be completed using the same processes before the sample mixture is separated on the GC column. Figure 2 shows the GC column between furnace II and furnace III after the H2/O2 and the back-flush valve. This arrangement was used for on-line precolumn reduction of NO2 in furnace II. Measurements. Six different types of experiments were performed to optimize the recovery of N2 from NH3, NO, and NO2. The general features of these experiments were as follows: experiments 1 and 2 were performed to optimize the reaction temperature of the NiO or the Ni reactor with (experiment 1) and without an additional Cu reduction reactor (experiment 2). These experiments were performed for NH3 and NO. In experi(20) Mitra, G. D.; Ghosh, S. K. J. High Resolut. Chromatogr. Chromatogr. Commun. 1985, 8, 150-151. (21) Wilhite, W. F.; Hollis, O. L. J. Gas Chromatogr. 1968, 6, 84-89. (22) VDI Richtlinien 2453, Blatt 6, 1980.

2752 Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

ment 3 the completeness of the conversion of NH3 and NO to N2 was tested. A dependency of the δ15N values on the NH3 or NO sample size was tested in experiment 4. In experiment 5 the temperature of the Mo reactor for reduction of NO2 to NO was optimized. The Mo amount required for optimum reduction of NO2 to NO was tested in experiment 6. Table 1 shows the GC and combustion interface conditions and the flow diagrams as well as the sample amounts used in these experiments as subsequently described in detail for the respective gases. NH3 Measurements. In experiment 1 we tested the recovery of N2 using the NiO oxidation reactor at a variety of temperatures while holding the Cu reduction reactor at a fixed temperature. In experiment 2 we tested the recovery of N2 again using the NiO oxidation reactor at a variety of temperatures but without a Cu reduction reactor. In both experiments 1 and 2, the NiO wire was oxidized with O2 for 30 min at 1000 °C, prior to use each day. In experiment 3 we tested the completeness of the oxidation of NH3 to N2. In principle, NH3 may be oxidized either before or after the GC column. If the precolumn oxidation is incomplete, excess NH3 is oxidized in the second furnace. In this case, two N2 peaks are generated from the NH3 sample. In experiment 4 we varied the amount of NH3 injected to test the dependency of the δ15N recovery on sample size. In addition, two different methods for the preconditioning of the NiO in furnace II were compared: (a) daily oxidation of NiO with O2 in furnace II at 1000 °C for 30 min; (b) 10 or 20 s oxidation with O2 at 1150 °C before each sample injection. NO Measurements. By analogy with the NH3 experiments, in experiment 1 the recovery of N2 was tested at various temperatures of the Ni reduction reactor while the Cu reduction reactor temperature was held fixed. In experiment 2 we tested the recovery of N2 using the Ni reduction reactor at various temperatures but without a Cu reduction reactor. In experiment 3 we tested the completeness of the NO reduction to N2. In principle, NO may be reduced before the GC column with excess NO, generating a second peak that can be easily measured at m/z 30 (14N16O). In experiment 4 the amount of NO injected was varied to test the dependency of the δ15N recovery on the sample size. In addition, two different methods for the preconditioning of the Ni in furnace II were compared: (a) no preconditioning of Ni and (b) 30 s of Ni reduction with H2/N2 at 1150 °C before sample injection. NO2 Measurements. In experiment 5 we tested the recovery of N2 and measured δ15N-NO2 values using the Mo reduction reactor at various temperatures while holding the Ni reduction reactor at a fixed temperature. In experiment 6 we tested the recovery of N2 and measured δ15N-NO2 values using the Mo reduction reactor with different quantities of Mo in furnace II: (a) 1 wire ) 125 mg of Mo, (b) 2 wires ) 250 mg of Mo, (c) 3 wires ) 375 mg of Mo. Statistics. Means were calculated from three measurements. Error bars represent (1 standard deviation. Differences between means within each experiment were tested for significance with a one-way ANOVA after checking for homogeneity of variances (Levens test). When treatment effects were significant, the LSD test with a significance level p < 0.05 was used to distinguish the means. Mean values between two experiments were tested for significant differences with Student’s t test for paired samples.

Figure 3. Temperature dependency of the NH3 oxidation on a NiO surface in furnace II with respect to the N2 recovery and the δ15N of the recovered N2 under two different conditions: (a) furnace I with an empty ceramic tube at 30 °C and furnace III with a Cu reduction reactor at 640 °C; (b) furnace I and III with an empty ceramic tube at 30 °C. The data points are means of replicate analyses (n ) 3). The error bars are standard deviations, which are in some cases smaller than the symbols. Significance levels of differences between the two experiments are noted by stars: *, p < 0.05; **, p < 0.01; ***, p < 0.001. Different letters denote significantly different groups within one experiment.

All statistics were computed using the statistical package SPSS.23 RESULTS AND DISCUSSION NH3 Measurements. Figure 3 illustrates the relationship between temperature of the NiO surface in furnace II and (a) recovery of N2 and (b) the δ15N of the N2 recovered from oxidation of NH3. In the first experiment, furnace I was fitted with an empty ceramic tube at 30 °C and furnace III with a Cu reduction reactor at 640 °C, whereas in the second experiment furnaces I and III were fitted with empty ceramic tubes at 30 °C. The recovery of N2 reached a plateau at temperatures greater than 950 °C in both experiments and was not significantly different between the two experiments. Thus, oxidation of NH3 to NO in the NiO reactor was negligible. The δ15N-NH3 values rose with temperature and reached a plateau at temperatures between 1150 and 1200 °C. The temperature dependency of the δ15N-NH3 values was clear in both experiments. δ15N-NH3 values were consistently ∼1.5‰ greater (more negative) in experiment 1 (Cu reactor in furnace III) than in experiment 2 (empty ceramic tube in furnace III). The difference in δ15N-NH3 values between (23) Norusis, M. J. SPSS/PC+; SPSS Inc.: Chicago, IL, 1986.

experiments was significant throughout the tested range of temperatures and was most significant at 1150 °C. It is assumed that a partially oxidized Cu reactor was responsible for 15N fractionation in experiment 1. CuO tends to bleed oxygen at temperatures above 250 °C in the presence of reducing substances.4,24 We assumed that the liberated oxygen reacted with the generated N2 to form small amounts of NOx. For this reason, we rejected the use of a Cu reduction furnace. The optimum temperature range for the NiO reactor was between 1150 (δ15NNH3: -4.14‰ ( 0.04‰) and 1200 °C (δ15N-NH3: -3.94‰ ( 0.18‰); over this range, δ15N-NH3 values were not significantly different. The reproducibility of the δ15N-NH3 values in this temperature range was close to the internal precision of the IRMS. In the following experiments we used a NiO reactor at 1150 °C. When we tested the conversion of NH3 to N2 before GC separation using a freshly oxidized NiO reactor (experiment 3), there was no N2 peak (data not shown). NH3 oxidation in the NiO reactor was thus complete since a second N2 peak would be generated by any NH3 remaining after passing through the first NiO reactor (furnace I) which would then be chromatographed as NH3 and oxidized to N2 in the second NiO reactor (furnace II). Figure 4 illustrates the dependency of recovery of N2 and the δ15N-NH3 values on the quantity of NH3 injected (experiment 4). The efficiency of oxidation of NH3 to N2 decreased with increasing sample size and was generally improved by a 20 s reoxidation of the NiO reactor at 1150 °C before each NH3 analysis when compared to that achieved using daily reoxidation for 30 min at 1000 °C. For small sample sizes (up to 14.7 nmol of NH3) there was little difference between the methods of reoxidation (as indicated by the overlap of confidence intervals). The incomplete recovery of N2 using a NiO reactor which was reoxidized once each day, produced a marked inverse relationship between δ15N-NH3 and sample size. In contrast, δ15N-NH3 was largely independent of sample size when the reactor was reoxidized immediately before each analysis. There was little difference between the two methods of oxidation when sample sizes were small (up to 22.9 nmol). δ15N-NH3 measurements could be reproduced with a standard deviation of (0.30‰ over a range from 11.7 to 58.8 nmol of NH3 with a 20 s reoxidation of the NiO reactor before each analysis. The selection of sample sizes was based on the lower and upper detection limits of the GC/IRMS system. A reoxidation time of only 10 s at 1150 °C before each analysis was insufficient to reduce the dependency of the δ15N-NH3 on the NH3 sample size (data not shown). Optimum conditions for δ15N-NH3 measurements include a NiO reactor at 1150 °C together with a reoxidation for 20 s before each NH3 analysis. NO Measurements. The recovery of N2 from NO using a Ni surface was dependent on temperature and was improved by omitting the Cu reduction reactor (Figure 5). N2 recovery was significantly greater when the Cu reduction reactor was omitted at Ni reactor temperatures of 950, 1050, and 1100 °C. δ15N-NO values were not significantly affected by omission of the Cu reactor and generally decreased with increasing temperature (Figure 5). NO was not fully reduced using a Cu reduction reactor at 640 °C under the conditions prevailing in the (24) Meritt, D.; Freeman, K. H.; Ricci, M. P.; Studly, S. A.; Hayes, J. M. Anal. Chem. 1995, 67, 2461-2473.

Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

2753

Figure 4. Dependency of the N2 recovery and the δ15N values of the recovered N2 on the NH3 sample size under two oxidation conditions of the NiO reactor: (a) daily reoxidation for 30 min at 1000 °C and (b) 20 s of reoxidation at 1150 °C before each NH3 analysis. Data points are means of replicate analyses (n ) 3). Error bars are standard deviations, which are in some cases smaller than the symbols. Solid lines are regressions, and dotted lines indicate the 95% confidence interval. Regressions for the N2 recoveries: (a) daily oxidation, y ) 1.28x - 4.78, r2 ) 0.999; (b) 20 s of reoxidation before each analysis, y ) 1.47x - 3.18, r2 ) 0.999. Regressions for the δ15N-NH3 values: (a) daily oxidation, y ) (-5.3 × 10-2)x - 4.20, r2 ) 0.833; (b) 20 s of reoxidation before each analysis, y ) (-8.0 × 10-3)x - 4.11, r2 ) 0.375.

furnaces of the combustion interface II.24 We assume that the Cu quantity in the furnace is insufficient to prevent a constant stream of oxygen escaping from previously formed CuO. The high temperatures and the presence of Cu as a catalyst promote the formation of NO from formerly created N2.24 This side reaction leads to N2 losses with isotopic fractionation. In the following experiments, we adopted a reaction temperature of 1150 °C for the Ni reactor (δ15N-NO: -50.23‰ ( 0.28‰) because this temperature is in the range where recovery of N2 was greatest with the smallest standard deviations. Reproducibility of the δ15N-NO measurements at 1150 °C was roughly twice that of the internal precision of the IRMS. When we tested the conversion of NO to N2 before GC separation (experiment 3) in the Ni reactor, no evidence of a second peak (due to NO which might have escaped reduction in the Ni reactor in furnace I and was reduced after GC separation in furnace II) at m/z 30 was given (data not shown). The quantity of NO injected was linearly related to the recovery of N2 over the range set by the lower and upper detection limits of the GC/IRMS system (16.7-133.3 nmol of NO, Figure 6). 2754 Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

Figure 5. Temperature dependency of NO reduction on the Ni surface in furnace II with respect to the N2 recovery and the δ15N of the recovered N2 under two different conditions: (a) furnace I with an empty ceramic tube at 30 °C and furnace III with a Cu reduction reactor at 640 °C; (b) furnaces I and III with an empty ceramic tube at 30 °C. The data points are means of replicate analyses (n ) 3). The error bars are standard deviations, which are in some cases smaller than the symbols. Significance levels of differences between the two experiments are noted by stars: *, p < 0.05; **, p < 0.01; ***, p < 0.001. Different letters denote significantly different groups within one experiment.

δ15N-NO over the smaller range from 33.3 to 133.3 nmol of NO were independent of sample size but were about 4‰ more depleted in 15N at lower sample sizes. We assumed that trace contaminations of N2 and/or CO2 interfere with the δ15N determination at lower sample sizes. Reduction of the Ni reactor for 30 s with N2/H2 before each NO analysis led to a loss in precision of the N2 recovery and the δ15N-NO determination (data not shown). It seems likely that the system became contaminated with N2 from the H2/N2 reduction mixture. NO2 Measurements. The recovery of N2 increased sharply between temperatures of about 360 and 420 °C using a Mo reactor for NO generation from NO2 and a Ni reactor for N2 generation from NO after chromatography (Figure 7). Recovery was not significantly different between 400 and 440 °C. The δ15N-NO2 values also showed a strong dependency on temperature in this range and were significantly least negative at 420 °C, coinciding with the temperature of the greatest N2 recovery. We selected an optimum temperature of 420 °C (δ15N-NO2: -47.64 ( 0.43‰) for the Mo reactor in subsequent work. The Mo reactor showed significant conditioning phenomena with respect to N2 recovery and δ15N-NO2 (Figure 8). The

Figure 6. Dependency of the N2 recovery and the δ15N values of the recovered N2 on the NO sample size. The data points are means of replicate analyses (n ) 3). The error bars are standard deviations, which are in some cases smaller than the symbols. The solid line is a regression for the N2 recovery, and the dotted lines indicate the 95% confidence interval: y ) 0.38x - 3.24, r2 ) 0.979.

Figure 7. Temperature dependency of the NO2 reduction to NO on a Mo surface in furnace II with respect to the N2 recovery and the δ15N values of the N2 recovered from NO. The data points are means of replicate analyses (n ) 3). The error bars are standard deviations, which are in some cases smaller than the symbols. Different letters denote significantly different groups within the experiment.

recovery of N2 reached a plateau after about 600 nmol of NO2 had been reduced. Recovery was then stable. The δ15N-NO2 showed a slight drift which was related to the cumulative amount of NO2 reduced (cumulative amount of NO2 ) ∑ of all NO2 injections ever reduced). The drift in δ15N-NO2 could be corrected using regression equations. Before the plateau is reached (phase I) and after the reduction capability of the Mo is exhausted (phase III), the N2 recoveries and the δ15N-NO2 values were poorly reproducible. The drift in δ15N-NO2 can be minimized by increasing the Mo amount in the reactor. Figure 9 and Table 2 summarize important characteristics for the dependency of the δ15N-NO2 values on the cumulative amount of NO2 reduced in a Mo reactor for three different conditions: (a) 125 mg of Mo, (b) 250 mg of Mo, and (c) 375 mg of Mo per reactor. We defined an acceptable range in the cumulative amount of NO2 reduced in a Mo reactor as being that in which we could obtain reproducible δ15N-NO2 values and we used this definition to obtain the data shown in Figure 9. The recovery of N2 from NO2 did not change significantly with the Mo content of the reactor. In contrast, the Mo content determined the amounts of NO2 necessary for conditioning which increase with increasing quantities of Mo in the reactor. The drift of δ15N-NO2 over the defined acceptable range (Figure 9) was smallest with the greatest quantity of Mo in the reactor. Similarly, standard deviations of the δ15N-NO2 measurements were also smallest with the greatest Mo content (Table 2).

On the other hand, reduction capacity was not increased beyond 250 mg of Mo. While the reasons for this observation are uncertain, it might be assumed that the NO2 remains in the 375 mg Mo reactor for a longer period of time due to a greater resistance to gas flow. Ignoring the δ15N-NO2 drift over the acceptable range, the standard deviation for the δ15N-NO2 determination for the 375 mg Mo reactor was (1.08‰. By using a linear regression to correct for the drift, the precision could be improved to (0.15‰. Our suggestion for the optimum conditions for the NO2 reduction to NO was to use a 375 mg Mo reactor at 420 °C. As for NH3 and NO, the sample size for the δ15N-NO2 measurement is limited by the lower and the upper detection limit of the GC/IRMS system. Recovery of N2 was linear in the range of 13.0-129.9 nmol NO2 (y ) 8.05x - 3.48; r2 ) 0.9812) with a standard deviation for δ15N-NO2 values of (3.15‰. A reduction in the range from 26.0 to 129.9 nmol of NO2 improved the standard deviation for the δ15N-NO2 determination to (1.15‰ (data not shown). The known relationship for drift in δ15N-NO2 with cumulative amount of NO2 reduced could be applied to produce a standard deviation of (2.87‰ for sample sizes in the range 13.0129.9 nmol of NO2. If the range is reduced to 26.0-129.9 nmol of NO2, the standard deviation for δ15N-NO2 improves to (0.90‰. We assumed that trace contaminations of N2 and/or CO2 interfere with the δ15N-NO2 determination in the lower sample size range. The novel methods of on-line analysis of 15N/14N ratios in nitrogen-containing trace gas samples described here will help Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

2755

Table 2. Dependency of the δ15N-NO2 Values of N2 Obtained from NO2 on the Cumulative NO2 Amount That Had Reacted on the Mo Surface in Furnace II for Three Different Conditions: (a) 125 mg of Mo, (b) 250 mg of Mo, and (c) 375 mg of Mo Mo quantity δ15N-NO2 [‰] range mean ( sd n N2 recovery [V s] mean ( sd acceptable range cumulative amount of NO2 reduced [nmol] reduction capacity within the acceptable range [nmol NO2] δ15N-NO2 drift/100 nmol of NO2 [‰] δ15N-NO2 drift within the reduction capacity [‰]

125 mg

250 mg

375 mg

-34.55 to -39.63 -36.47 ( 3.41 10

-34.53 to -37.78 -36.45 ( 1.53 14

-37.04 to -38.12 -37.61 ( 0.32 12

35.320 ( 5.78

43.033 ( 8.24

39.757 ( 1.78

163.0-978.3 815.3 0.52 5.08

163.0-1223.9 1060.9 0.31 3.25

652.2-1550.0 897.8 0.12 1.08

Figure 9. Dependency of the δ15N-NO2 values of N2 obtained from NO2 on the cumulative NO2 amount that had reacted on a Mo surface in furnace II for three different conditions: (a) 125 mg of Mo, (b) 250 mg of Mo, (c) 375 mg of Mo. Solid lines are regressions for the respective amounts of Mo in furnace II: (a) y ) (-1.1 × 10-2)x 30.34, r2 ) 0.803; (b) y ) (-4.1 × 10-3)x - 33.62, r2 ) 0.827; (c) y ) (-9.8 × 10-4)x - 36.53, r2 ) 0.809. The regressions refer only to the acceptable range for δ15N analyses (phase II).

Figure 8. Dependency of the N2 recovery and the δ15N values of N2 obtained from NO2 on the cumulative NO2 amount that had reacted on the Mo surface in furnace II (375 mg of Mo). Three phases are differentiated. Phase I: conditioning of the Mo reactor. Phase II: acceptable range for the δ15N-NO2 analyses. Phase III: exhausted Mo reactor. Solid line indicates the regression for phase II.

greatly in studies where rapid analysis of large numbers of gas samples is required. Previous methods which utilized stand-alone methods of sample preparation are labor intensive and thus expensive. The precision of the techniques for the measurement of stable isotopes of nitrogen in NH3, NO, and NO2 were close to the internal precision of modern IRMS and thus applicable for studies utilizing the natural abundance of nitrogen isotopes as well as for those in which the abundance of 15N is artificially enriched (i.e. tracer studies). About 5 ppm of NH3, NO, or NO2 in an air volume of 100 mL is required for analysis. On-line cryofocusing systems for trace 2756 Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

gases for air volumes of 100 mL are already commercially available.17 Given that concentrations of NH3, NO, and NO2 in ambient air are in the lower ppb range,6 about 100 L of ambient air is needed to produce the minimum sample size for nitrogen isotope analysis. Sample sizes of 100 L of ambient air, however, require new trapping techniques (e.g. those based on molecular sieves and cryofocusing), and such techniques are currently under development. ACKNOWLEDGMENT Mark Adams gave many valuable suggestions for the improvement of the English of this manuscript. Financial support by the Bayreuther Institut fu¨r Terrestrische O ¨ kosystemforschung (BMBF PT BEO 51-0339476B) is gratefully acknowledged. The paper contributes to the EU project COGANOG (FAIR 3-CT96-1920). Received for review January 5, 1998. Accepted April 2, 1998. AC9800053