Seediscussions,stats,andauthorprofilesforthispublicationat:https://www.researchgate.net/publication/27277275
Completeandaccuratemassspectrometric
isotopeanalysisoftroposphericnitrousoxide
ArticleinJournalofGeophysicalResearchAtmospheres·August2003
ImpactFactor:3.43·DOI:10.1029/2003JD003613·Source:OAI
CITATIONS
READS
54
37
3authors:
JanKaiser
ThomasRöckmann
UniversityofEastAnglia
UtrechtUniversity
114PUBLICATIONS1,709CITATIONS
325PUBLICATIONS4,655CITATIONS
SEEPROFILE
SEEPROFILE
CarlBrenninkmeijer
MaxPlanckInstituteforChemistry
381PUBLICATIONS7,488CITATIONS
SEEPROFILE
Allin-textreferencesunderlinedinbluearelinkedtopublicationsonResearchGate,
lettingyouaccessandreadthemimmediately.
Availablefrom:JanKaiser
Retrievedon:10May2016
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 108, NO. D15, 4476, doi:10.1029/2003JD003613, 2003
Complete and accurate mass spectrometric isotope analysis of
tropospheric nitrous oxide
Jan Kaiser1 and Thomas Röckmann
Bereich Atmosphärenphysik, Max-Planck-Institut für Kernphysik, Heidelberg, Germany
Carl A. M. Brenninkmeijer
Abteilung Chemie der Atmosphäre, Max-Planck-Institut für Chemie, Mainz, Germany
Received 10 March 2003; revised 15 May 2003; accepted 22 May 2003; published 13 August 2003.
[1] We describe a manual extraction and purification method for mass spectrometric
isotope analyses of tropospheric N2O. A theoretical framework to correct for
(hydro)fluorocarbon and CO2 interferences is developed and verified experimentally. The
standard deviation for analysis of one sample on a single day is 0.05% for d15N and d18O
and 0.2% for the relative enrichment of the terminal (1d5N) and central (2d5N) nitrogen
atoms. The isotopic composition of N2O in tropospheric background air could thus be
measured with unprecedented precision on samples from six locations. We obtained overall
average values of d15N = (6.72 ± 0.12)% versus air N2 and d18O = (44.62 ± 0.21)%
versus Vienna Standard Mean Ocean Water. Neither a clear spatial pattern from 28N to
79N, nor a temporal trend over the course of 2 years was found. Within the experimental
uncertainties, this is in line with small trends of 0.02 to 0.04%/a derived from analyses
of Antarctic firn air and isotopic budget calculations. Using an independent 2d15N calibration
of our working standard versus air N2, we find large differences in the position-dependent
15 14
N/ N ratios: The mean for all samples is 1d15N = (15.8 ± 0.6)% and 2d15N =
(29.2 ± 0.6)% versus air N2. In light of a new definition for oxygen isotope anomalies, we
reevaluate the existing measurements and derive a 17O excess of 17O = (0.9 ±
0.1)%.
INDEX TERMS: 0322 Atmospheric Composition and Structure: Constituent sources and sinks;
0365 Atmospheric Composition and Structure: Troposphere—composition and chemistry; 0394 Atmospheric
Composition and Structure: Instruments and techniques; 1610 Global Change: Atmosphere (0315, 0325)
Citation: Kaiser, J., T. Röckmann, and C. A. M. Brenninkmeijer, Complete and accurate mass spectrometric isotope analysis of
tropospheric nitrous oxide, J. Geophys. Res., 108(D15), 4476, doi:10.1029/2003JD003613, 2003.
1. Introduction
[2] The global average surface mixing ratio of the atmospheric trace gas nitrous oxide (N2O) was 314 nmol mol1 in
1998 and is currently rising at a rate of (0.25 ± 0.05)%/a
[Prather et al., 2001]. This is of concern, since N2O is both a
greenhouse gas [Yung et al., 1976] and involved in stratospheric ozone depletion [Crutzen, 1970]. Over a time horizon
of 100 years, the radiative forcing per molecule is 296 times
greater than for carbon dioxide [Ramaswamy et al., 2001].
The rate of increase in the global N2O burden is determined
by the imbalance between surface sources (predominantly
microbial nitrogen conversion processes in soils and waters)
and stratospheric sinks (photolysis and reaction with electronically excited oxygen atoms). Because of the long
atmospheric lifetime of 120 years, the variability of tropospheric N2O mixing ratios is very low, which allows one to
1
Now at Department of Geosciences, Princeton University, Princeton,
New Jersey, USA.
Copyright 2003 by the American Geophysical Union.
0148-0227/03/2003JD003613$09.00
ACH
determine accurately the rate of increase [Prinn et al., 2000].
The stratospheric N 2 O sinks are also well quantified
[Minschwaner et al., 1993], so that satisfactory estimates of
the total N2O source flux can be obtained from the sum of the
global increase rate and the sink flux. However, considerable
uncertainties remain in the relative contributions of individual sources to the overall budget [Prather et al., 2001].
[3] Isotope ratio measurements offer additional independent variables that can help to improve the global N2O
budget calculations, provided that fractionation effects by
chemical, physical, and biological processes in the atmosphere are accounted for. N2O becomes enriched in heavy
oxygen (17O and 18O) and nitrogen (15N) isotopes in the
stratosphere because of preferential destruction of isotopically lighter molecules in the sink reactions. Thanks to a
number of recent studies, this phenomenon is now well
understood, both experimentally and theoretically [see
Kaiser et al., 2003; McLinden et al., 2003, and references
therein]. The tropospheric N2O isotope ratios can then be
used to study and quantify N2O emissions. Just as for the
N2O mixing ratio, variations in the isotopic signature are
expected to be small. Because of the stratospheric enrichment
and to achieve mass balance, the global N2O source must be
19 - 1
ACH
19 - 2
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
isotopically lighter than tropospheric N2O [Kim and Craig,
1993]. The imbalance between sources and sinks thus causes
a progressive heavy isotope depletion in N2O over decadal
timescales, similar to the well-known Suess effect for CO2
[Trolier et al., 1996]. On the basis of isotopic budget
calculations, Rahn and Wahlen [2000] predicted a decrease
of the 15N/14N ratio by 0.03% per year but fell short of
deducing a trend from historic N2O isotope measurements
due to the limited analytical precision and a lack of interlaboratory calibration between the various studies. From
analyses of Antarctic firn air, Röckmann et al. [2003a] and
Sowers et al. [2002] could reconstruct small long-term trends
of 0.02 and 0.04% per year for 18O/16O and 15N/14N ratios
which validated the budget calculations by Rahn and Wahlen
[2000].
[4] However, no consistent data set of direct N2O isotope
observations over a multiannual timescale exists up to now.
The present study is an attempt to commence such a data set
by regular monitoring of N2O isotopes at a set of sampling
stations, in a similar fashion to CO2 [Trolier et al., 1996].
Analytical developments undertaken for this purpose are
described in this paper. All sample extraction and purification steps were made off-line, i.e., on vacuum manifolds not
connected to the isotope ratio mass spectrometer. As opposed to an automatic on-line technique [Röckmann et al.,
2003b], this requires large amounts of sample of the order
of hundreds of dm3 of air and standardized reproducible
procedures in manual sample work-up. Supplementing the
established analysis of 18O/16O and average 15N/14N values,
a new technique was applied to measure the individual
isotope ratio of both nitrogen atoms mass spectrometrically,
based upon isotopic analysis of the NO+ fragment produced
in the ion source following electron impact [Brenninkmeijer
and Röckmann, 1999; Toyoda and Yoshida, 1999]. Viton
O-rings had to be excluded from the system as far as possible
to avoid interfering contamination in NO+ fragment analysis
(section 5.2). In combination with techniques to measure
17 16
O/ O ratios [Cliff and Thiemens, 1994; Röckmann et al.,
2001b] a complete and accurate analysis of isotopically
mono-substituted N2O species is now possible. The possibility to use the N2+ fragment for 17O analysis is evaluated,
but the achievable precision is not satisfactory for tropospheric N2O. As already demonstrated by Yoshida and
Toyoda [2000], the central nitrogen atom is enriched in
15
N relative to the terminal nitrogen atom, but our observed
variability is much lower than theirs and reflects an improved analytical precision and possibly more pristine air.
However, despite all efforts, no clearly discernible temporal
or spatial patterns in the tropospheric N2O isotope signature
were detected. We also note that the reported anomalous
deviation of 17O from a mass-dependent relationship to 18O
of 17O = (1.0 ± 0.2)% [Cliff and Thiemens, 1997;
Röckmann et al., 2001b] has to be revised downward by
0.1%, to correct for the 17O anomaly in atmospheric O2
[Luz and Barkan, 2000] which was used as a reference gas in
previous studies.
2. Experimental Methods
2.1. Overview
[5] With the mass spectrometric techniques implemented
so far the isotopic composition of N2O can be measured
either directly on N2O gas [Tanaka et al., 1995], through
preparation of nitrogen (N2) and carbon dioxide (CO2) from
N2O [Kim and Craig, 1990], or through conversion to N2
and oxygen (O2) [Cliff and Thiemens, 1994]. The latter two
techniques have the disadvantage of being very labor
intensive. Therefore the first technique was chosen since it
allows N2O analysis without further chemical preparation
steps and only requires a calibrated N2O working standard
to relate the isotope measurements to international scales
(sections 2.2 and 3.5). Alternatively to mass spectrometry,
infrared spectroscopy can be used for N2O isotope analysis
[Esler et al., 2000; Uehara et al., 2001].
[6] Elemental isotope ratios of N2O are not measured
directly but must be determined from the observed molecular mass spectrum. The molecule ion (N2O+), the NO+
fragment, and the N 2+ fragment with their pertinent
12 (mass-to-charge ratio (m/z) 44 to 48), 6 (m/z 30 to 33),
and 3 (m/z 28 to 30) isotopologues and isotopomers can be
used. However, for N2O samples at natural abundance
isotope levels, only ions with m/z 44 to 46, 30 to 32, and
28 to 29 give sufficient ion currents to achieve the desired
precision within a reasonable measurement time. Mass-tocharge ratio 32 is actually of little use because of O2
interference. Measurements of ion current ratios at m/z 44/
45/46 give the 18O/16O and the average 15N/14N isotope
ratios (section 3), measurements at m/z 30/31 allow one to
deduce position-dependent 15N/14N information (section 5),
and m/z 28/29 measurements enable 17O/16O measurements
of limited precision (section 2.7).
2.2. Units
[7] Isotope ratios are reported relative to an international
standard using d notation
Rsa
d¼
1 1000%o
Rst
ð1Þ
where Rsa is the elemental isotope ratio of the sample (in the
case of N2O, 15R, 17R, and 18R) and Rst the elemental
isotope ratio of the standard. Atmospheric nitrogen has been
established as international standard for 15N abundance
[Mariotti, 1983, 1984]. For oxygen isotopes, Vienna
Standard Mean Ocean Water (VSMOW) and CO2 derived
from Pee Dee Belemnite (VPDB-CO 2 ) are in use
[Gonfiantini, 1978]. As the two N atoms in N2O are not
equivalent, a distinction will be made between the d value of
the terminal N atom in N2O (designated 1d15N), the middle
N (2d15N), and the average of both (d15N). The isotope
ratios 15R1 and 15R2 are defined analogously. Alternative
nomenclatures use d15Nb/d15Na or d546/d456 [Yoshida and
Toyoda, 2000; Yung and Miller, 1997].
[8] Since the direct measurement of N2O does not allow
one to compute the final d values versus international standards, it is convenient to set up a N2O working standard as
intermediary between the sample and a fictitious N2O standard with the nitrogen isotope ratio of atmospheric air and the
oxygen isotope ratio of VSMOW. This working standard is
not limited by sample-size restrictions, and several batches of
it can be converted for calibration (section 3.5).
2.3. Air Sampling
[9] Whole air samples were obtained at the locations in
Table 1 with a scheduled sampling interval of 1 to 3 weeks.
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
ACH
19 - 3
Table 1. Sampling Locations for Tropospheric N2O Samples
Station
Country
Location
Altitude
Operating Agency
Spitsbergen
Kollumerwaard
Mainz
Schauinsland
Mount Sonnblick
Izaña, Tenerife
Norway
Netherlands
Germany
Germany
Austria
Spain
79N, 12W
53N, 6E
50N, 8E
48N, 8E
47N, 13E
28N, 16W
474 m
0m
128 m
1205 m
3106 m
2370 m
Norwegian Institute for Air Research
University of Groningen
Max Planck Institute for Chemistry
Umweltbundesamt
Institute for Meteorology, Salzburg
Spanish Meteorological Service
A modified diving compressor (RIX Industries) was used to
fill 5 or 10 dm3 aluminum high-pressure cylinders to up to
120 bar [Mak and Brenninkmeijer, 1994]. Air was drawn
through perfluoroalkoxy (PFA) tubing (outer diameter
12.7 mm) connected to a drying unit filled with Drierite
(CaSO4, with CoCl2 as moistness indicator) at the inlet of
the air compressor. The air cylinders were shipped regularly
to our laboratory in Mainz.
2.4. Extraction of N2O From Bulk Air
[10] An off-line preparation technique was used to extract
N2O quantitatively from bulk air samples. The CO extraction system established in our laboratory served this purpose
[Brenninkmeijer, 1993; Brenninkmeijer et al., 2001]. It
delivered not only CO-derived CO2, but also all other gases
in air with sufficiently low vapor pressures at the boiling
point of liquid nitrogen (77 K). This ‘‘trace gas cocktail’’
consisted mainly of CO2, H2O, and N2O but included also,
among others, nonmethane hydrocarbons (NMHCs), NO2,
chlorofluorocarbons (CFCs), and hydrofluorocarbons
(HFCs). Some of these gases yield isobaric interferences
during the mass spectrometric analysis of N2O isotopes and
consequently had to be removed in two further purification
steps using NaOH to remove CO2 and a preparatory gas
chromatographic separation system.
[11] The extraction line comprised two highly efficient,
Russian-doll-type cooling traps (RDT) [Brenninkmeijer,
1991] prior to the CO oxidation step, which removed the
condensable trace gases from the air stream at a flow rate of
5 dm3 min1. After the designated sample volume had been
processed the first RDT was isolated from the system and
evacuated while it was still frozen. This RDT already
retained more than 99.9% of the CO2 and 99.7% of the
N2O [Röckmann, 1998], so the contents of the second trap
could be ignored. Although the surface area of the glass
fiber thimbles inside the RDT is quite large, the contents
of the first trap could be transferred quantitatively to a
glass U tube by pumping on the first trap for 35 min. From
the U tube the sample was frozen onto Ascarite (NaOHcoated silica, 20– 30 mesh, Sigma) in a glass sample flask.
2.5. Separation From CO2 and H2O
[12] The sample flask with Ascarite consisted of a lower
body and an upper valve part (Louwers Hapert highvacuum stopcock). Both parts were connected by a ball
joint (Rotulex) with a fluorocarbon O-ring seal (Viton). This
design facilitated easy filling and removal of Ascarite from
the bottle.
[13] Each flask was supplied with 2.5 –3.5 g of Ascarite.
Before use it was evacuated until the pressure did not
decrease any more due to water desorption from the alkaline
surface (0.1 mbar). The reaction of CO2 with NaOH
proceeded faster when there was still some water present,
i.e., when the flask was not evacuated further. Usually,
>99.99% of the CO2 reacted within minutes as monitored
by the pressure change in the flask. However, sometimes it
was necessary to heat the flask (<100C) to initiate the
reaction, because otherwise the reaction could take days to
start (probably because the Ascarite was too dry). As soon
as the reaction was initiated, H2O production further accelerated it by dissolution of CO2 and NaOH in the aqueous
phase. After 24 hours on Ascarite the N2O sample contained
<0.1% CO2, which corresponds to a removal efficiency
>99.9999%. Finally, the sample was dried with P2O5 to
remove water from the reaction with Ascarite.
2.6. Gas Chromatographic Purification
[14] After removal of CO2, there were still other gases
present which interfere with isotopic analyses of N2O and
the NO+ fragment. Hydrocarbons in polluted air samples
can pose problems with the analysis of the N2O+ molecular
ion or NO+ fragment ion. Propane (m/z 44, 45, and 46 from
C3H8+ isotopologues), ethanol (m/z 46) or other hydrocarbons (m/z 44, 45, and 46 from C3H8+ fragments) interfere
with N2O+ analysis. Traces of fluorocarbons (m/z 31 from
CF+) or other hydrocarbons (m/z 30 and 31 from C2H6+
fragments) can interfere with NO+ fragment analysis.
[15] The sample with the highest amount of contaminating
gases was from Kollumerwaard. More than 5 mmol mol1
condensable gases other than N2O, H2O, and CO2 must have
been present in the original bulk air sample to account for the
amount of gas present after the CO2/H2O removal step. A
significant fraction was NO2 as suggested by the intense
brown color of the gas, indicating incomplete removal
of NO2 by basic Ascarite. Time permitting, NO2 would
disproportionate completely into NO
2 and NO3 . Most other
samples contained impurities between 1 and 2% of the
amount of N2O.
[16] A gas chromatographic/mass spectrometric (GC/MS)
system was used to remove remaining CO2 and other
impurities (Figure 1). The system was made of glass but
changed later to stainless steel (with Nupro K series valves
and Swagelok fittings), except for the Russian doll trap, the
P2O5 flask, and a calibrated small-volume manometer
(‘‘microvolume,’’ denoted by P2 in Figure 1). This reduced
the number of Viton O-rings which were a source of m/z 31
interference (section 5.2).
[17] The amount of gas was determined prior to and after
gas chromatographic separation in the microvolume with a
piezoresistive sensor. The temperature was recorded at the
time of pressure measurement in the microvolume. Knowing the amount of air processed in the trace gas extraction
process, we thus obtain an independent manometric estimate of the N2O mixing ratio of the air analyzed.
[18] Samples are frozen into an injection loop (35 cm,
3.2 mm OD, stainless steel) and passed over the GC column
ACH
19 - 4
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
Figure 1. Gas chromatographic purification system for N2O. P1 and P2, piezoresistive manometers
(Institute of Geological and Nuclear Sciences, New Zealand); VG, vacuum gauge (Hastings); QMS,
quadrupole mass spectrometer (Balzers).
(Porapak Q, 5 m, 3.2 mm OD, stainless steel, Alltech) using
a stream of purified helium (Messer-Griesheim, grade 5.0;
Supelco High Capacity Carrier Gas Purifier). The gas flow
is split after the GC, and a minute portion is fed via a
1.6 mm OD stainless steel capillary (1 m, 0.5 mm ID) into
a quadrupole MS (Balzers QMS 200). Five to 20 s before
the arrival of the N2O peak at 7.5 min, the four-way
valve is switched over to the collection trap.
[19] As soon as N2O is not detected anymore by the MS,
the four-way valve is switched back. This procedure ensures
complete N2O collection from the gas stream without
alteration of isotope ratios or contamination by other gases
eluting before or after N2O. The timing of the trapping
procedure is crucial since CO2 elutes 1.5 min before N2O
and propane elutes 4.0 min after N2O.
[20] After trapping the purified N2O the collection trap is
isolated from the GC, and a backflush of the column is
initiated. Helium remaining in the collection trap is pumped
off. The trap is thawed, and its contents are transferred to
the P2O5 tube for drying, then quantified manometrically in
the microvolume (P2) and transferred back to the sample
tube for mass spectrometric analysis.
[21] Recovery in blank tests with pure N2O was (99.33 ±
0.04)% with no detectable isotope fractionation. Tests with
the collection trap only gave recovery rates close to 100%.
Closing the split valve did not increase the recovery rate,
possibly indicating a small memory effect. Therefore only
samples with similar d values were purified in succession.
2.7. Mass Spectrometric Analysis
[22] Analyses of N2O+ and the N2+ fragment were carried
out on a Finnigan MAT 252 isotope ratio mass spectrometer; NO+ was analyzed on a Micromass Prism II instrument
with one fixed and two adjustable Faraday cups. Tropo-
spheric samples were measured using conventional dual
inlet systems with cold fingers (microvolumes) to enable
analysis of small samples. Interferences from CO2 in N2O+
analysis and CHF3 in NO+ analysis were corrected for as
described in sections 4 and 5.3.
[23] Assuming a statistical isotope distribution over different isotopologues (while distinguishing 15 N isotopomers), the following relationships hold between
‘‘molecular’’ and elemental isotope ratios:
45
46
R¼
15
R ¼ 15 R1 þ 15 R2 þ 17 R
R1 þ 15 R2
31
29
17
R þ 18 R þ 15 R1 15 R2
ð2Þ
ð3Þ
R ¼15 R2 þ17 R
ð4Þ
R ¼ 15 R1 þ 15 R2 :
ð5Þ
In principle, these four equations can be solved for the four
unknowns, 15R1, 15R2, 17R, and 18R. However, in practice,
the precision in the 17R determination is limited, because it
is calculated from the difference of 45R and 29R, and the
relative intensity of the N2+ fragment ion is only 10% of
the N2O+ molecular ion. The expected uncertainty for 17O
can be calculated from equations (2) and (5) in terms of
d values defining 45d as 45Rsa/45Rst 1, etc., and the average
15 14
N/ N isotope ratio as 15R = (15R1 + 15R2)/2
215 Rst þ17 Rst 215 Rst 15
17
d N
17 R
Rst
st
215 Rst
215 Rst 29
17
d 1 þ 17
d:
Rst
Rst
d17 O ¼45 d
¼45
ð6Þ
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
Since 215Rst/17Rst 19, uncertainties in 45d and 29d are
amplified 20 and 19 times, respectively. Under optimal
experimental conditions a standard deviation of 0.5%
could be achieved for d17O [Kaiser, 2002], still not good
enough to resolve variations in the 17O anomaly of
tropospheric N2O of 0.9%.
[24] Therefore N2O isotope analysis exploits the covariation of 17R and 18R (‘‘mass dependence’’), which can be
cast in a power law
17
R¼A
18
b
R ;
ð7Þ
where b is 0.516 [Kaiser, 2002], corrected from an original
value of 0.515 [Cliff and Thiemens, 1997] and A is found to
be 0.00937035 (exactly), calculated from b = 0.516 and the
isotope ratios of VSMOW 17Rst = 3.799 104 [Li et al.,
1988] and 18Rst = 2.0052 104 [Baertschi, 1976]. The 17O
anomaly in tropospheric N2O can also be taken into
account, and equations (2) – (4) can then be solved for
15
R1, 15R2, and 18R.
[25] However, the NO+ fragment analysis is slightly
impaired by isotopic scrambling in the ion source which
has to be corrected for; 91.8% of the nitrogen atoms in NO+
come from the central N position but 8.2% are derived from
the terminal N position. This can be deduced from mass
spectrometric analysis of pure 14N15N16O and 15N14N16O
isotopomers [Kaiser, 2002]. A ‘‘scrambling coefficient’’ s is
introduced which is the percentage of nitrogen atoms in
NO+ derived from the terminal N position. The coefficient s
has to be measured in a separate experimental series and its
presence complicates the 31R calculations. With x being the
relative abundance of the specific N2O species, the scrambled ratio 31Rs is
31
Rs ¼
s x
15
NN16 O þ x 14 NN17 O þ ð1 sÞ x N15 N16 O þ x N14 N17 O
:
sx 14 NN16 O þ ð1 sÞx N14 N16 O
ð8Þ
Note that x(15NN16O) = x(15N14N16O) + x(15N216O), etc. To
simplify this, the usual statistical assumption is made, 17R is
extracted, and each term is divided by x(14N216O)
2
31
Rs ¼ s15 R1 þ ð1 sÞ15 R2 þ17 R sð1 sÞð15 R1 15 R2 Þ
15
1 þ s15 R2 þ ð1 sÞ R1
:
ð9Þ
Equation (9) disagrees in the last term from the results of
Yoshida and Toyoda [2000] and Toyoda and Yoshida
[1999]. This term is at its maximum for s = 0.5 and for
large differences in the isotope ratios at the central and
terminal nitrogen positions (15R2 and 15R1). However, s is
only 0.082 in reality and even an extreme assumption of
15
R2 = 1.160 15R1 leads to a correction of 0.006% only for
natural abundance samples. For work with artificially
enriched 15N isotopes this term may become relevant,
though.
[26] A position-dependent analysis of the N isotope ratios
requires a position-dependent calibration of the N2O standard. Otherwise, systematic errors result from the conversion of 31d measurements to 31 R or 31R s. Following
Brenninkmeijer and Röckmann [1999], the position-depen-
ACH
19 - 5
dent isotopic composition of the standard is expressed by a
parameter f
15
f ¼ 15
R1;st
215 R1;st
¼
:
15 R
R1;st þ15 R2;st
st
ð10Þ
Then, the average d15N value relative to the N2O working
standard is
d15 N ¼ f 1 d15 N þ ð1 f Þ2 d15 N
ð11Þ
In terms of d values, f can be expressed by
1 2 d15 Nst 1 d15 Nst
:
f ¼ 2
4 1 þ d15 Nst
ð12Þ
The d values with index ‘‘st’’ are relative to the reference
isotope ratio (air N2). The value 2d15Nst 1d15Nst is the socalled ‘‘site preference’’ [Yoshida and Toyoda, 2000] of the
working standard. Evidently, the d15N value is not simply
the unweighted arithmetic average of 1d15N and 2d15N.
Rather, d15N is biased toward the enrichment at the terminal
or central N position of the sample, depending on the value
of f. This is irrelevant as long as we are dealing with average
d15N values, because 1d15N and 2d15N change correspondingly. However, for position-dependent measurements this
must be considered.
[27] We note that linear combinations of d values (such
as the unweighted average d15N or the site preference) are
to be taken with a grain of salt, since they may not behave
as expected. To this end we define two quantities, d+ =
1/2(2d + 1d) and d = 1/2(2d 1d) (the qualifier 15N was
omitted for clarity), and investigate whether d+ and d obey
the ‘‘addition theorem’’ for d values, i.e., dsa = d + dst +
d dst (indexed d values are relative to the international
standard, and d is the isotope ratio of the sample relative to
the working standard). The result indicates that neither
the unweighted average of 1d and 2d (d+) nor the site
preference (2d) behave according to addition theorem for
d values (however, the weighted average d15N as defined in
equation (11) does). Instead, the following relationships
hold:
þ
þ
þ þ
dþ
sa ¼ d þ dst þ d dst þ d dst
ð13aÞ
þ
þ d
sa ¼ d þ dst þ d dst þ d dst
ð13bÞ
It seems tempting to use equation (13b) to convert the site
preference measured versus the N2O working standard (d)
). However,
to the site preference on international scales (dsa
one should bear in mind that d also depends implicitly on f
15
(or d
st anxd d Nst) in a nonstraightforward way. To
understand this, we define 2d* as the apparent d value of
the central N atom (under the influence of scrambling and a
potentially asymmetric standard)
31
2
d* ¼ 31
Rs 17 R
1:
Rs;st 17 Rst
ð14Þ
ACH
19 - 6
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
Figure 2. Zero-enrichment measurement of N2O working standard gas (Table 2).
Substitution of equations (9) (neglecting the third term),
(10), and (11) into equation (14) yields
2
1
d ¼2 d þ
s 2 d d15 N
d ¼ d* þ
ð1 f Þð1 2sÞ
2
ð1 sÞ 2 d* d15 N
d15 N 2 d 2
¼ d* :
f
f ð1 2sÞ
ð15aÞ
ð15bÞ
Toyoda and Yoshida [1999] obtained essentially the same
result but omitted 17R from their derivation. In conclusion
from the above remarks on linear combinations of d values
we suggest avoiding the use of site preferences and
‘‘average d15N’’ values in quantitative considerations and
relegating the calculation of these properties to illustrative
purposes.
[28] Knowledge of f is obviously important to correct for
the influence of scrambling. J. Kaiser et al. (Mass-spectrometric method for the absolute calibration of the intramolecular nitrogen isotope distribution in nitrous oxide,
submitted to Analytical and Bioanalytical Chemistry,
2003, hereinafter referred to as Kaiser et al., submitted
manuscript, 2003) describe a purely mass spectrometric
procedure to perform a position-dependent calibration of a
sample of N2O. It relies on measurements of mixtures of the
working standard and 15N15N16O against the working
standard. A plot of 31d versus 46d can be used to derive
the ratio 31Rs,st/46Rst which allows computation of 31Rs,st
provided 46Rst has been determined before. Calibration of
46
Rst is possible by conventional techniques as described in
section 3.5. Via this approach, a value of f = 0.4938 ±
0.0003 for our N2O working standard was obtained indicating that the terminal N position is 25% lighter than the
central one. Since the average 15N composition of our
standard was close to air N2 (+1.0%), this means that the
working standard has 1d15N 11% and 2d15N +13%
versus air N2. The sensitivity of 2d15N and 1d15N of a
sample to the accuracy of f is not very large if they are
expressed relative to the working standard [Kaiser, 2002];
however, any asymmetry in the working standard is conferred to 2d15N and 1d15N directly if they are expressed
relative to air N2. Toyoda and Yoshida [1999] have calibrated their own working standard using a chemical approach from which f = (0.5011 ± 0.0002) can be derived.
The average d15N value of their standard is close to ours,
which could lead to the expectation that the 2d15N and 1d15N
values are also similar. However, this is not necessarily the
case, since their d18O value of 23.3% is also very different
from the value of 38.5% for our standard. The differences
in the position-dependent calibration of the working standard lead to significant discrepancies in the position-dependent 15N signature of atmospheric N2O (section 6).
2.8. Sample Size
[29] The mixing ratio (m) of N2O in air samples was
determined by gas chromatography with electron capture
detection (ECD). The nonlinearity of the detector was
accounted for by fitting the response function to a dilution
series. The mixing ratio m was 315 nmol mol1 for the
year 1999, which represented the majority of the tropospheric air samples analyzed. The average amount of
processed air was 400 dm3 (Standard Temperature and
Pressure (STP), 273.15 K, 101,325 Pa), with a range of
300 – 1060 dm3, corresponding to 139 mm3 (Standard
Ambient Temperature and Pressure (SATP), 298.15 K,
105 Pa) of N2O (or 5.6 mmol). In the following sections
all sample amounts given in dm3 or mm3 refer to SATP.
3. Mass Spectrometry for D15N and D18O Values
[30] Before any atmospheric samples were run, several
tests to check the efficiency of the analysis and purification
system were performed.
3.1. Zero-Enrichment Measurement of Standard
Gas Using Bellows and Cold Finger
[31] Working standard gas was admitted to the standard
and sample side bellows of the mass spectrometer, and the
isotope ratios were measured. The standard errors of the
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
Table 2. Precision of 45d and
MAT 252 Mass Spectrometera
46
d Measurements on the Finnigan
45
d/%
No.
No.
No.
No.
1
5
1
5
versus
versus
versus
versus
no.
no.
no.
no.
1,
1,
1,
1,
bellows
bellows
microvolume
microvolume
46
d/%
0.00 ± 0.02 0.01 ± 0.02
0.04 ± 0.02 0.10 ± 0.04
0.05 ± 0.01 0.07 ± 0.03
0.11 ± 0.03 0.14 ± 0.07
Number of
Analyses, n
12
5
5
3
a
No., number.
results shown in Figure 2 and Table 2 (standard number 1
versus standard number 1, bellows) represent the external
precision of the mass spectrometric analysis and were
0.02% for both 45d and 46d and n = 12 measurements. A
similar analysis of a slightly fractionated standard gas
sample (standard number 5 versus standard number 1,
bellows) gives similar results with standard deviations of
0.02% and 0.04% for n = 5.
[32] The same set of measurements was performed for the
microvolume inlet of the isotope mass spectrometer. Aliquots of 300 mm3 or less were admitted to the mass
spectrometer. The standard deviations were similar to the
measurement with bellows. However, the mean values of
45
d and 46d have increased in all cases: The differences
between microvolume and bellow analyses were (0.04 ±
0.03)% for standard number 1 versus number 1 and (0.07 ±
0.05)% for standard number 5 versus number 1 in 45d
whereas the respective changes in 46d were (0.07 ± 0.05)%
and (0.04 ± 0.11)%. This indicates that some fractionation
of the sample occurs when measuring with the microvolume.
3.2. Freezing Standard Gas Onto Ascarite
and Extraction
[33] About 130 mm3 of pure N2O working standard gas
was frozen into a preevacuated Ascarite bottle, and the same
extraction procedure was carried out as for atmospheric air
ACH
19 - 7
samples. However, the gas was not passed over the GC. This
was to check whether N2O isotope ratios are affected by
freezing onto Ascarite. Since the sample size was made
similar to the atmospheric samples, the same corrections for
CO2 contamination had to be applied (section 4). The final
results from eight experiments measured with the microvolume gave 45d = (0.05 ± 0.03)% and 46d = (0.11 ± 0.03)%.
The average yield of N2O was (99.9 ± 0.5)%. The observed
enrichments are not significantly larger for 45d and 46d than
those for the zero-enrichment measurement with the microvolume indicating no fractionation on Ascarite.
3.3. Gas Chromatography of Pure N2O Standard Gas
[34] In two series of experiments, 2 cm3-samples of N2O
standard gas were measured repeatedly with the bellows of
the mass spectrometer, interspersed by GC purification runs.
Each time the samples were frozen back from the bellow.
The aim was to check whether the GC runs fractionated the
samples (Figure 3). Obviously, the d values increase from
analysis to analysis by 0.05% in 45d and by 0.1% in 46d.
The fact that this increase occurs whether or not a GC run
was performed in between indicates that this is due to
fractionation by MS analysis/back-freezing of the sample.
A plot of 46d against 45d gives a slope of 2 which shows
the mass dependence of the fractionation associated with
repeated MS analysis/freeze back cycles. With this fractionation it became clear that any sample analysis should not
involve too many repeated MS analyses. However, if
necessary, it should not be a problem to run a sample
several times over the GC in order to remove impurities
that cannot be removed in a single run.
[35] Next, 12 small samples of N2O working standard gas
(130 mm3) were run over the GC to check whether the GC
purification would introduce any artifacts for smaller atmospheric samples. Again, 45d and 46d only show the expected
enrichment due to microvolume use in the MS measurement, although the standard deviation is slightly higher in
this case. Mean changes in 45d are (0.04 ± 0.04)% and
Figure 3. Repeated MS analyses and gas chromatographic (GC) purification runs of two N2O standard
gas samples. Samples were treated in the same way in series 1 and 2 except for MS analysis 5, where the
N2O sample underwent three GC runs in series 2 but none in series 1. The labels indicate the number of
GC runs that were interspersed with the MS analyses, but for clarity only 46d of series 1 is labeled at each
point.
ACH
19 - 8
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
(0.08 ± 0.07)% in 46d. Both d values are correlated (r2 = 0.74),
indicating a systematic fractionation.
3.4. Extraction of Pure N2O From N2O + CO2
Mixtures With Gas Chromatographic Purification
[36] About 130 mm3 of N2O working standard gas and
70 cm3 of CO2 were frozen into a 120 cm3 valve flask
containing 2.5 to 3.5 g of Ascarite. After CO2 had reacted
with NaOH the flask was processed as for the atmospheric
samples, including gas chromatographic purification. The
average yield of the Ascarite reaction step was (99.5 ± 0.3)%
(n = 13), and that of the GC run was (98.9 ± 0.2)% (n = 13).
Hence the overall yield was (98.5 ± 0.4)%, which is used later
in order to estimate the N2O concentration manometrically.
On average, 45d and 46d values were altered by (0.04 ±
0.06)% and (0.08 ± 0.08)%, respectively. Taken together
with the results from sections 3.1 to 3.3, this means that the
presence of CO2 caused a decrease of 45d by 0.09% and a
decrease of 46d by 0.18%. This may have to do with the
additional production of H2O by reaction of CO2 with
Ascarite. However, from mass balance calculations considering the solubility coefficient of 25 mmol dm3 bar1
for N2O in H2O [Weiss and Price, 1980] and equilibrium
isotope fractionation constants of 15eN2O(g)/N2O(aq) = 0.75%
and 18eN2O(g)/N2O(aq) = 1.06% [Inoue and Mook, 1994]
(all values at 25C), the resulting change in d15N and d18O
would be <0.001%. Furthermore, dissolved N2O would be
released from H2O by drying with P2O5 (section 2.5). The
small isotopic fractionation in the CO2 removal thus remains
unexplained at the moment.
3.5. Calibration of 15N and 18O in the
Working Standard
[37] The N2O working standard (Messer-Griesheim,
99.9999% purity) was calibrated in two different ways:
through conversion to N2 and CO2 by graphite with Pt
mesh as a catalyst and through thermal decomposition to N2
and O2 on a gold surface. Table 3 gives an overview on the
experimental parameters and final results.
[38] N2, O2, and CO2 from N2O decompositions are
measured against working standards of these gases. Those
working standards were calibrated against the international
standards air N2 and NBS-19-CO2 which has a fixed
isotopic composition relative to the international standard,
VPDB-CO2. To this end the O2 working standard was
combusted to CO2 first. The calibration of the N2 and O2
working standards was described by Kaiser [2002].
[39] A quartz tube with a transition piece to Duran glass
was used for the graphite conversions. Only the quartz part
of the reactor was inserted into a tube furnace (Carbolite
MTF 10/15/130) and heated to the designated temperature.
In order to release all residual gases, the reactor, the
graphite, and the catalyst were pumped out under high
vacuum prior to the conversion for at least 1 hour. Furthermore, the first conversion of a day was not included in the
final evaluation. The amount of N2O before conversion and
the combined amount of CO2 and residual N2O after
conversion were determined manometrically and used to
assess the yield of the reaction. Yields higher than 100%
would indicate incomplete conversions.
[40] In case of the Pt/C reaction, the Boudouard reaction
C + CO2 = 2CO (Kp 1 at 700C) produces CO as a by-
Table 3. Overview on Conversion Experiments for Calibration of
the N2O Working Standarda
Experiment
Catalyst
Reactor volume
Sample volume (SATP)a
Temperature
Initial reactor pressure
Duration of conversion
Yield
d18O versus CO2-standard
Number of analyses
d18O versus O2-standard
Number of analyses
d18O versus VSMOWb
Weighted mean
d15N versus N2 standard
Number of analyses
d15N versus air N2
Weighted mean
N2 O + ½ C !
N2 + ½ CO2
N2O ! N2 +
½ O2
Pt
24 cm3
1.2 cm3
(690 – 713)C
180 mbar
1 hour
(99.3 ± 0.7)%
(27.25 ± 0.13)%
17
Au
34 cm3
1.2 cm3
(930 – 940)C
200 mbar
(3 – 12) hours
(100.0 ± 0.1)%
(22.68 ± 0.29)%
9
(38.59 ± 0.20)%
(38.15 ± 0.31)%
(38.45 ± 0.20)%
(14.25 ± 0.09)%
(14.31 ± 0.04)%
11
19
(0.96 ± 0.09)%
(1.02 ± 0.05)%
(1.01 ± 0.03)%
a
SATP, Standard ambient temperature and pressure.
VSMOW, Vienna Standard Mean Ocean Water.
b
product. However, most of the CO reacts back to CO2 when
the reaction is stopped by freezing out CO2, because this
effectively shifts the equilibrium to the C + CO2 side of the
reaction. Any residual CO was removed by Schütze reagent
(acidic I2O5 on silica gel) [Schütze, 1940], because it would
otherwise interfere with the N2 isotope measurement. The
absence of any CO or CO2 contamination in the purified N2
was checked by measurement of the 12C+ ion.
[41] The decomposition to N2 and O2 was performed in a
gold tube (Au purity 99.9%) at 930– 940C. The yield was
controlled by manometric measurement of the residual condensable gas which amounted to 1% of the initial N2O.
However, after passage over a preparatory GC column
(section 2.6), only 1 – 3% of the residual gas turned out to
be N2O, the rest being CO2 that was presumably produced by
carbon containing impurities in the reactor. Thus the overall
N2O conversion was 99.97 – 99.99% and the conversion
products (N2, O2, CO2) quantitatively reflect the initial
isotopic composition of N2O, since kinetic isotope effects
are expected to be small at such high temperatures. Indeed,
measurements on residual N2O from incomplete conversion
indicate fractionation factors of a(15N) = 1.005 – 1.007 and
a(18O) = 1.010 –1.014 at temperatures between 930 and
940C. For the same reason the minor loss of oxygen to CO2
does not influence the isotopic composition of O2 significantly.
On some occasions, the N2O yield was double-checked by
determining the amount of N2 and O2 in a calibrated volume
containing silica gel for freezing in the gas with liquid
nitrogen. It agreed with the stoichiometrically expected
amount indicating no significant leaks of air into the reactor.
[42] N2 from the Pt/C combustion or the N2/O2 mixture
from the decomposition on gold were frozen out on silica
gel and transferred to the isotope ratio mass spectrometer for
nitrogen isotope analysis. N2 could be analyzed directly, but
the N2/O2 mixture (33.3% O2, 66.7% N2) was analyzed
against a mixture of the N2 working standard with the same
amount of O2. The d15N value of the N2 working standard
was (13.10 ± 0.03)% versus air N2 [Kaiser, 2002].
[43] CO2 from the Pt/C conversion could be measured
directly against the CO2 working standard of known d18O =
(29.22 ± 0.14)% versus VPDB-CO2. O2 from the decomposition on gold was first separated from N2 on a preparatory GC system before it could be analyzed against the O2
working standard which has a d18O value of (15.14 ±
0.12)% versus VSMOW.
[44] Both independent determinations of 15N and 18O in
the working standard are in very good agreement. This
gives us high confidence in the quality and accuracy of our
calibration.
4. CO2 Correction for N2O Isotope Analysis
U 0 and V 0 are the actually measured voltages (produced by
the ion currents on the amplifier feedback resistors) for
sample and standard, respectively, at the ion masses 44, 45,
and 46. They are the sum of the true voltages U and V for
N2O only and the voltages from the CO2 contamination, u
and v due to CO2 interference, e.g., 45U 0 = 45U + 45u. The
present treatment is totally symmetric and applies to
possible CO2 contamination of both sample and standard.
No assumption was made on the absence of such a
contamination in the standard. Substituting U by U 0-u and
V by V 0-v and extracting 45U 0/44U 044V 0/45V 0 = 1 + 45d0(N2O)
yields
[45] The CO2 correction for N2O analyses is accomplished in analogy to Tanaka et al. [1995]. So-called
interfering masses are analyzed by peak jumping to
m/z 12 and 46 after measurement of the d values. Tanaka
et al. derived equations of the form
45
46
dðN2 OÞ ¼45 d0 ðN2 OÞ A012 I=44 I
ð16bÞ
where 45d(N2O) is the correct value for the uncontaminated
sample relative to the N2O working standard, 45d0(N2O) is
the measured value for the CO2 contaminated sample, and
12
I and 44I are the ion currents of 12C+ at m/z 12 and of
CO2+ + N2O+ at m/z 44. A0 and B0 are constants. A0 is defined
as 12Ir/44Ir [1 + 45d(CO2)] where 12Ir is the relative ion
intensity of 12C+ to total CO2, 44Ir is the relative ion
intensity of 14N216O to total N2O, and 45d(CO2) corresponds
to CO2 measured against the N2O working standard. The
definition of B0 is analogous to A0.
[46] A dilution series of CO2 standard gas in N2O showed
that the expected linear relationship is reproduced in the
experiment [Tanaka et al., 1995]. However, the derived
expressions for the coefficients appear to be incorrect,
which can be verified when one inserts representative
quantities for 12Ir, 44Ir, and 45d(CO2) (deduced from MS
measurements). Furthermore, the calculated ratio A0/B0 of
0.779 is not reproduced by Tanaka et al. [1995, Figure 3]
where ratios of 0.496 and 0.497 are found.
[47] Therefore the theoretical CO2 correction is recalculated here. Since we measure m/z 12 and m/z 46 for both
standard and sample as interfering masses, the equations
were adapted to this case, but transformations to other
masses are possible without loss of generality.
[48] First of all, an exact solution was derived for calculating the correct 45d(N2O) and 46d(N2O) value from the
measured 45d0(N2O) and 46d0(N2O) values. They are defined
as follows:
45
45
dðN2 OÞ ¼
46
U
44 U
45
V
44 V
1;
U0
44 U 0
45 0
d ðN2 OÞ ¼ 45 0 1;
V
44 V 0
45
46
dðN2 OÞ ¼
U
44 U
46
V
44 V
ð17aÞ
46
44
u
v
1 46 0 1 44 0
46
U
V 1;
dðN2 OÞ ¼ 1 þ46 d0 ðN2 OÞ
44
46
u
v
1 44 0 1 46 0
U
V
ð17bÞ
where 45u, 46u, 45v, and 46v can be inferred from the voltage
of interfering mass 12 if the relative intensity of C+ to
12 16 +
C O2 is known. Defining 12r = 12u/44u, we find that 12r
equals (3.9 ± 0.1)% on our instrument. If one assumes that
the contaminating CO2 is always of the same isotopic
composition one can derive 45u and 46u from 44u by
calculating 45u = 44u45R(CO2) 45k and 46u = 44u46R(CO2)
46
k. The ratios 45R(CO2) and 46R(CO2) are the ‘‘molecular’’
isotope ratios of CO2, while 45k and 46k are the ratios of
resistances for the mass 45/46 and mass 44 cups (being 100
and 333 within 1% in our case, but their exact values are of
no importance as will be demonstrated in the next
paragraph). The values 44U 0, 45U 0, 46U 0, 44V 0, 45V 0, and
46 0
V are all measured and thus an exact calculation of the
CO2 correction is possible provided 12r, 45k, 46k, and the
isotopic composition of CO2 are known.
[49] One will mostly find rather a small contamination of
both sample and standard. Then, it is possible to expand
equations (17a) and (17b) and omit terms of second and
higher order
45
45
44
44
45
u
v
u
v
þ
þ
U 0 44 V 0 44 U 0 45 V 0
ð18aÞ
44
44
46
u
v
u
v
þ
þ
:
U 0 44 V 0 44 U 0 46 V 0
ð18bÞ
dðN2 OÞ ¼45 d0 ðN2 OÞ 45
46
46
dðN2 OÞ ¼46 d0 ðN2 OÞ 46
A further approximation is viable for a small contamination
of the standard (v V 0), since then 44v/44V 0 = 44v/(44V +
44
v) 44v/44V. Furthermore, 45V and 46V can be expressed in
terms of 44V as follows: 45V = 44V 45Rst(N2O)45k and 46V =
44 46
V Rst(N2O)46k. One obtains
1;
U0
44 U 0
46 0
d ðN2 OÞ ¼ 46 0 1:
V
44 V 0
46
45
44
u
v
1 45 0 1 44 0
45
45 0
U
V
dðN2 OÞ ¼ 1 þ d ðN2 OÞ
1
44
45
u
v
1 44 0 1 45 0
U
V
ð16aÞ
dðN2 OÞ ¼46 d0 ðN2 OÞ B012 I=44 I;
19 - 9
ACH
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
45
46
Rst ðN2 OÞ
dðN2 OÞ ¼45 d0ðN2 OÞ 12 r
46 0
45
RðCO2 Þ 1 þ d ðN2 OÞ
45
Rst ðN2 OÞ 1 þ45 d0 ðN2 OÞ
12
12
u
v
46 U 0
46 V 0
12
12 u
v
1 þ46 d0 ðN2 OÞ 46 0 46 0
U
V
ð19aÞ
ACH
46
19 - 10
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
dðN2 OÞ ¼46 d0ðN2 OÞ 46
12
Rst ðN2 OÞ 46 RðCO2 Þ 12 u
v
:
12 r
46 R ðN OÞ 46 U 0
46 V 0
st
2
12
12 u
v
1 þ46 d0 ðN2 OÞ 46 0 46 0 :
U
V
ð19bÞ
Now, the final approximation is the assumption 45d0(N2O) 1
and 46d0(N2O) 1. Actually, both values are 6% in the
present case. This gives
45
46
dðN2 OÞ ¼45 d0 ðN2 OÞ 46
dðN2 OÞ ¼
46
Rst ðN2 OÞ
12 r
45
12
12
u
v
dðCO2 Þ 46 0 46 0
U
V
ð20aÞ
12
12
Rst ðN2 OÞ 46
u
v
:
d ðN2 OÞ d
ð
CO
Þ
2
12 r
46 U 0
46 V 0
ð20bÞ
0
46
Hence we obtain two correction factors, A and B, being the
factors before the terms in parentheses in equations (20a) and
(20b). Furthermore, their ratio, A/B, is given by
45
d(CO2)/46d(CO2), which is 0.5 in our case, in accordance
with the findings of Tanaka et al. [1995, Figure 3].
[50] In an actual mass spectrometric analysis, 12u, 12v,
46 0
U , and 46V 0 were measured after the measurement of
45 0
d (N2O) and 46d0(N2O), and the corrections were calculated
using equations (20a) and (20b). The value 46Rst(N2O) is
known from the working standard calibration (section 3.5),
and 12r is derived from the CO2 mass spectrum (see above).
Note that if both ion beams are well balanced, the equation
44 0
U = 44V 0 holds, and with the above approximations for
small enrichments and contamination 46U 0 = 46V 0 is valid too.
[51] The values 45d(CO2) and 46d(CO2) for a sample of
our CO2 working standard (d13C = 42.504% and d18O =
29.217% versus VPDB-CO2) were measured relative to
the N2O working standard and found to be 492% and
939%, respectively. Hence the ratio of correction factors
(A/B) is 45d(CO2)/46d(CO2) = 0.524, which is in agreement
with results from mixing experiments (Figure 4). Here one
obtains a ratio of 0.526 ± 0.001. For convenience, the terms
in parentheses in equations (20a) and (20b) are abbreviated
as 12b. Similar plots of d against 22b (from CO22+) or against
the mixing ratio of CO2 in the artificial N2O + CO2 mixture
give the same ratios but higher standard deviations of
±0.005, because CO22+ is 5 times less abundant than
12 +
C in the CO2 mass spectrum.
[52] Atmospheric CO2 has a slightly different composition than the CO2 standard gas used for calibration of the
CO2 correction curve. Therefore the actual coefficients for
CO2 correction are A = 29.5%/100% and B = 54.7%/100%
(which are the adjusted values from Figure 4).
+
5. Mass Spectrometry of NO Fragment Ions
5.1. Zero-Enrichment Measurements of 31D Values
[53] Similar tests as for d15N and d18O analysis were
performed in case of the NO+ fragment. Zero-enrichment
measurements (section 3.1) with the microvolume of the
Micromass Prism II mass spectrometer gave essentially
unchanged mean 31d values of (0.03 ± 0.02)% from a set
Figure 4. Mixtures of CO2 in N2O:
46 0
d (N2O) versus 12C+ interference.
45 0
d (N2O) and
of five analyses. However, problems appeared in early
analyses of atmospheric samples (section 5.2) and in further
blank tests of the analytical system (section 5.4).
5.2. Contamination Problems in 31D Analysis
[54] Whereas initial analyses of atmospheric samples
gave similar 45d and 46d values, the 31d values varied by
several per mill absolutely. Furthermore, repetitive GC
purification steps of the same sample revealed that 31d
generally decreased. However, the quadrupole mass spectrometer connected to the preparatory GC did not reveal any
compounds that eluted before or after the N2O peak.
Varying the time frame during which gas was frozen out
from the GC effluent showed that the contaminant must
elute after N2O. Upon close inspection of mass spectra of
severely contaminated samples taken on the MAT 252
instrument, peaks were found at m/z 69 and sometimes also
at m/z 50 and 51 that did not exist in the reference gas
spectrum. With the help of a mass spectral database (Wiley/
National Institute of Standards and Technology, 1990),
trifluoromethane (CHF3, Freon F-23) was identified as the
most likely candidate for this kind of contamination (distinct peaks at m/z 31, 51, and 69 due to CF+, CHF2+, and
CF3+). Therefore tests of CHF3 and other fluorine containing
gaseous compounds were run on the GC. Retention times
were 8.8 min for CHF3, 2.5 min for CF4, 14.0 min for
CH2F2, and 14.8 min for CF3Cl. The retention time of N2O
in the setup as described above is 8.0 min. For an amount
of 130 mm3 that is representative of a real atmospheric
sample, we have to sample the effluent of the GC from
7.5 to 9.0 min, which overlaps with the CHF3 peak. Thus
CHF3 is the contaminant responsible for the observed
variation of 31d.
5.3. CHF3 Correction
[55] To account for CHF3 contamination, interfering
masses were measured at m/z 51 and m/z 69, but the
correction via m/z 69 is preferred, because of an unidentified
broad background at m/z 50 to m/z 52 in the MS. This
correction can be derived along the lines of the CO2
correction (section 4). One has to consider the measured
ionization efficiencies of 14N16O+ relative to 14N216O+ (30r =
30 44
I/ I 0.32) and of CHF2+/CF3+ to CF+ (51r = 51I/31I 1.7,
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
ACH
19 - 11
there was no correlation between measured values 31d0 and
b either, rendering the CHF3 correction established in
section 5.3 apparently useless. Possibly, there are other
contaminants present in the analytical system that produce
ions of m/z 31 (such as CF+) and cause a greater variability
of 31d. It was hoped that the gas chromatographic purification step would solve this problem.
[57] However, after the GC purification step the raw,
uncorrected 31d0 values turned out to be significantly
enriched relative to the standard gas and still displayed a
higher standard deviation than the zero enrichment tests
(section 5.1): 31d0 = (0.39 ± 0.17)% (n = 12). The value 31d0
was not correlated to 45d (Figure 6), making isotopic
fractionation an unlikely cause for the enrichment. Moreover, 31d0 was not correlated to 69b, either. A tentative CHF3
correction of the data as established in section 5.3 gave a
corrected value of 31d = (0.05 ± 0.29)%. Although the mean
31
d value is closer to zero now, the standard deviation even
increased.
[58] To investigate this phenomenon further, the samples
were analyzed a second time on the MAT 252 after 31d
analysis on the Prism II. 45d and 46d showed an increase in
line with expectations from section 3.3 (+0.09% and
+0.15%), but 69b and 51b clearly increased too. This must
be attributed either to impurities that were frozen out from
the MS or to contamination released from Viton O-rings in
the sample flask valves. To estimate the amount of m/z 69
impurity at the time of 31d measurement, the average of 69b
for both analyses on the MAT 252 was taken and 31d0
corrected accordingly. The corrected mean 31d now had a
satisfactory value of (0.09 ± 0.08)% (section 5.1). A plot of
31
d versus 45d showed much less scattered data compared to
the plot of uncorrected 31d0 versus 45d (Figure 6).
[59] By repeated MS analysis/freeze back cycles on the
Prism II we could rule out that impurities were present in
the MS itself which caused continuous increases in 31d.
However, we suspected that out-gassing O-rings could be
the reason. Therefore, in addition to changing the vacuum
manifold for GC purification from glass (with Viton O-rings
in the valves) to an all-stainless-steel system (section 2.6),
all Viton O-rings including those in the valves of the
Ascarite flasks were replaced by butyl rubber O-rings, and
the CFC-based vacuum grease (Fomblin, BOC Edwards)
used until then for the O-rings was replaced by silicone
grease (Dow Corning). The glass sample flasks were
replaced by stainless steel flasks with Nupro valves for
the most recent analyses of atmospheric N2O. The 45d and
46
d values are not affected by the described artifacts and
were therefore retained for the older analyses (the analysis
sequence does not correspond to the sampling sequence).
These modifications to the vacuum system prevented
any further significant CF+ interference in NO+ fragment
analysis.
69
Figure 5. Dilution series showing the influence of CHF3
contamination on the measured 31d0 value. The value 69b =
2.3% corresponds to a molar fraction of 0.01% CHF3 in
N2O.
69
r = 69I/31I 2.9). The exact approach is described here
(see section 4 for an explanation of the symbols)
31
u
1 31 0
31
31 0
U
d¼ 1þ d
1;
31
v
1 31 0
V
ð21Þ
which results with the same approximations as for the CO2
correction in
69
69 u
Rst ðN2 OÞ 1 1 46 0
v
d
ð
N
O
Þ
1
þ
2
46 U 0
46 V 0
Rs;st ðN2 OÞ 30 r 69 r
46
31
d ¼31 d0 31
ð22aÞ
51
51 u
Rst ðN2 OÞ 1 1 46 0
v
d
ð
N
O
Þ
1
þ
2
46 U 0
46 V 0
Rs;st ðN2 OÞ 30 r 51 r
46
31
d ¼31 d0 31
ð22bÞ
The factors before the bracketed term are denoted 69C and
51
C, the bracket terms 69b and 51b, respectively. They
amount to 69C = 9.70%/% and 51C = 5.70%/%, which
compares favorably to the results of a dilution series of
CHF3 in N2O (Figure 5). The pertinent ratios 69C/51C are
1.703 (theory) and 1.665 ± 0.003 (experiment).
5.4. Further Blank Tests and Modifications to the
Analytical System
[56] Checks of the 31d integrity upon freezing pure N2O
on Ascarite (section 3.2) gave similar zero enrichments as
for 45d and 46d analyses but with a higher standard deviation: (0.10 ± 0.33)% (n = 8). Almost no correlation could
be found between 45d and 31d (r2 = 0.12). Furthermore,
6. Tropospheric N2O Samples
6.1. Analyses of Four Mainz Air Samples Taken on a
Single Day
[60] Having established accuracy and precision of the
overall method for standard gas samples, it was possible to
apply it to atmospheric samples. The problems due to
CHF3 contamination shall be illustrated on a set of atmo-
ACH
19 - 12
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
Figure 6. Plot of uncorrected 31d0 (solid squares) and corrected 31d (open squares) values versus 45d for
standard gas runs over the GC. The value 31d0 is barely correlated to 45d (r2 = 0.07).
spheric air samples taken from the roof (height above
ground: 12 m) of the Max Planck Institute for Chemistry
in Mainz on a single day in November 1999. The prevailing moderate westerly winds on that day excluded contamination by nearby point sources, corroborated by normal
urban CO mixing ratios of 200 nmol mol1. Samples
were drawn through PFA tubing (perfluoroalkoxy copolymer, Dupont) and directly processed over the CO extraction
line.
[61] Figure 7 shows that there is still a large CHF3
contamination after only one GC purification (indicated
by high 69b values). Especially the first sample contained
22% impurities (determined manometrically) that had to
be removed by the preparatory GC. A contribution of more
than 10% to these impurities came from CHF3. We believe
that this was caused by the PFA sampling line, possibly by
gaseous decomposition products of PFA or by gases dissolved in or adsorbed to PFA. The correlation of 31d0 with
69
b for the first GC run shows the expected slope of 10%/%
(Figure 5), but the results for the second GC run deviate
significantly (5%/%) although the slope is less well
defined. It is assumed that other impurities than CHF3 are
responsible for this. After the raw 31d0 values were corrected
for CF+ interference by the latter slope the final results after
two GC runs show a good agreement in the 2d15N values for
the four samples (Table 4). Analyses of these four air
samples were still impaired by the use of glass flasks with
Viton O-rings (section 5.4). However, typical 69b values for
the most recent analyses were 0.01% or less, proving the
successful removal of major contaminants by eliminating all
potentially interfering Viton O-rings from the vacuum
manifold and the sample flasks. In the absence of a clear
cause for the remaining minor CF+ interference, we assumed that it was due to CHF3 and used the better
understood slope of 10%/% to correct the raw 31d0 values
but note that the required correction is small in any case.
[62] The values d15N and d18O bear similar errors as the
blank runs. Therefore the methodology is considered to be
valid for atmospheric samples too, and we assume that the
standard deviation of the mean from the four samples
represents the total uncertainty of the analytical method.
The values have been corrected for the small interference of
residual CO2 impurities (typically by 0.05/0.1% or less for
d15N/d18O) but not for the 17O isotope anomaly of atmospheric N2O and the little fractionation encountered in the
check runs of standard gas samples (section 3.4). Correcting
for the O isotope anomaly requires downward revisions of
d15N by 0.05% and 0.09% for 2d15N [Kaiser, 2002;
Toyoda and Yoshida, 1999].
[63] The mixing ratio determined by GC-ECD (section 2.8)
was (315 ± 0.5) nmol mol1. The results of the manometric
determination in the microvolume of the vacuum line
gave on average (311 ± 2) nmol mol1 (Table 4),
in reasonable agreement with the gas chromatographic
analysis.
6.2. Time Series of N2O Measurements at
Various Stations
[64] An overview of the results of N2O isotope measurements in tropospheric air from six sampling stations (Table 1)
Figure 7. Correlation of 31d0 with 69b (section 5.3) for two
GC runs of Mainz air samples.
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
ACH
19 - 13
Table 4. Analyses of Four Air Samples Taken in Mainz on 10 November 1999
Analysis
d15N/%
1
2
3
4
Mean
Rel. to air N2/VSMOW
5.53
5.62
5.53
5.52
5.55 ± 0.05
6.56 ± 0.08
1 15
d N/%
4.40
4.42
4.82
4.49
4.53 ± 0.19
15.8 ± 0.9
2 15
d N/%
d18O/%
15.22
15.41
15.62
15.28
15.38 ± 0.18
29.0 ± 0.9
5.69
5.78
5.74
5.69
5.73 ± 0.04
44.40 ± 0.23
Mixing Ratio/
nmol mol1
312
313
308
309
311 ± 2
d values are generally relative to 18O/16O and average 15N/14N isotope ratios of the working standard. d values in the last row are relative to
international standards and include the uncertainty of the working standard calibration (Table 3).
over 2 years of measurements is given in Figure 8 and
Table 5. Obviously, the variability of the isotopic composition of N2O is rather low compared to the strong gradients
observed in the stratosphere [Griffith et al., 2000;
Röckmann et al., 2001a; Yoshida and Toyoda, 2000] but
larger than the analytical precision estimated from the
analysis of four tropospheric air samples obtained in Mainz
on a single day (Table 4). Local variations of source fluxes
might account for this but should influence N2O mixing
ratios as well. However, N2O mixing ratios were virtually
invariant (315.8 ± 1.8 nmol mol1) and did not show
any correlation to d values. Only one sample obtained
at Mount Sonnblick on 12 August 1999 was found to
be significantly depleted in heavy isotopes (d15N = 5.3%,
d18O = 42.2%; not shown in Figure 8) and a high mixing
ratio of 328 nmol mol1. Although it was attempted to
exclude pollution by local point sources during sampling,
these excursions must be attributed to such a source.
[65] No clearly discernible temporal or spatial trends are
present in the tropospheric data in Figure 8. Nevertheless,
we note that ln(1 + d15N) and ln(1 + d18O) are correlated to
each other (Figure 9), just as in the stratosphere. However,
the slope of 0.54 ± 0.04 differs from the values of yapp =
1.16 – 1.26, derived from the most recent studies of stratospheric N2O [Kaiser, 2002]. The value yapp is defined as the
ratio of the apparent 15N and 18O fractionation constants,
i.e., yapp = 15eapp/18eapp. It is also different from the slope of
2 found in isotope measurements of soil emissions [Gros
et al., 2003], the largest single N2O source [Prather et al.,
2001]. This seems to indicate that the correlation between
d15N and d18O is an experimental artifact and neither of
stratospheric nor of surface origin. For example, kinetic
isotope effects due to diffusion should induce a slope close
to 0.5 (because of the mass differences of 18O to 16O and
15
N to 14N). In contrast to this mass-dependent relationship,
an anticorrelation between 1d15N and 2d15N is found (slope
0.88 ± 0.05; r2 = 0.82) which reflects the way by which
1 15
d N is calculated (1d15N = 2d15N 2d15N) and the absence
of large variations in d15N.
[66] Our results for the mean oxygen and average nitrogen isotope ratios of tropospheric N2O (last row in Table 5)
agree within errors with most previous studies (see Rahn
and Wahlen [2000] for an overview) and the values reported
by Yoshida and Toyoda [2000] (d15N/% = 7.0 ± 0.6 and
d18O/% = 43.7 ± 0.9) which were obtained by on-line gas
chromatography-isotope ratio mass spectrometry. Although
the mean mixing ratio of our samples is identical to the
value of Yoshida and Toyoda (315.7 ± 2.4 nmol mol1), we
observe much smaller variations of d15N and d18O. This
indicates that it is possible to obtain a higher precision with
the off-line technique used here, although at the expense of
a more laborious sample preparation.
[67] The so-called ‘‘site preference’’ (= 2d15N 1d15N)
is clearly different in both studies; (45.8 ± 1.4)% were
found in the present case as opposed to (18.7 ± 2.2)%
reported by Yoshida and Toyoda [2000]. The large deviation is a consequence of the position-dependent 15N calibration of the working standard. For the same reason, the
individual 1d15N and 2d15N values differ for the two studies.
Obviously, the two calibrations are in disagreement, because the isotopic composition of tropospheric N2O must
be nearly constant. Assuming hypothetically that our N2O
Figure 8. Time series of d15N, d18O, 2d15N, and 1d15N in
tropospheric N2O.
ACH
19 - 14
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
Table 5. Overview of Tropospheric N2O Analyses From Six Sampling Locationsa
Location
Latitude
Spitsbergen
Kollumerwaard
Mainz
Schauinsland
Mount Sonnblick
Izaña
Mean of all
79N
53N
50N
48N
47N
28N
d15N
6.80
6.58
6.61
6.69
6.69
6.78
6.72
±
±
±
±
±
±
±
0.14 n
0.03 n
0.07 n
0.09 n
0.09 n
0.11 n
0.12 n
1 15
2 15
d N
= 23
=3
=8
= 11
= 36
=9
= 90
15.8
15.4
15.7
15.6
15.9
15.9
15.8
±
±
±
±
±
±
±
0.4
0.3
0.4
0.5
0.7
0.6
0.6
d18O
d N
n
n
n
n
n
n
n
=
=
=
=
=
=
=
16
2
7
7
27
8
67
29.4
28.6
28.9
29.0
29.2
29.4
29.2
±
±
±
±
±
±
±
0.4
0.2
0.4
0.4
0.7
0.5
0.6
n
n
n
n
n
n
n
=
=
=
=
=
=
=
16
2
7
7
27
8
67
44.72
44.39
44.49
44.61
44.55
44.63
44.60
±
±
±
±
±
±
±
0.25
0.04
0.15
0.15
0.18
0.13
0.21
n
n
n
n
n
n
n
=
=
=
=
=
=
=
23
3
8
11
35
9
89
a 15
d N values are relative to air N2, and d18O is relative to VSMOW. Standard deviations represent variation at individual sampling locations and do not
include uncertainty of working standard calibrations.
working standard gas had the same intramolecular 15N
distribution as Yoshida and Toyoda’s ( f = 0.5011), we
would obtain a value of (15.7 ± 1.1)%, closer to the value
measured by Yoshida and Toyoda. Since the site preference
is not expected to vary much in atmospheric background
air, this emphasizes the need for more work regarding
position-dependent 15N calibrations. We have tried to
account for the uncertainty of all relevant parameters in
the position-dependent calibration of our working standard
gas [Kaiser et al., submitted manuscript, 2003], but if there
are unknown systematic errors in, for example,
18
R(SMOW), the results may nevertheless be incorrect.
For many studies, these discrepancies are of little importance, since tropospheric N2O can be used as a reference,
e.g., for the interpretation of stratospheric isotope ratios or
for the variation of d values in firn air.
[68] Summarizing, we note, as expected, little variation in
the stable isotope composition of tropospheric N2O between
28N and 79N and over timescales of 2 years. The
variability is slightly larger than the total uncertainty of
the off-line analysis system. Judging from the observed
correlation between d15N and d18O, this may be caused by
residual sampling artifacts which have not yet been
accounted for in the error analysis. However, no such
correlation is observed between 1d15N and 2d15N. Small
+0.05
+0.06
%/a in d15N and 0.0250.04
%/a
trends of 0.0410.03
were derived by Röckmann et al. [2003a] from Antarctic
firn air analyses. Changes of 0.08% and 0.05% are
therefore expected over the course of two years which can
clearly be reconciled with the direct measurements obtained
in the present study.
where b is the three-isotope exponent for mass dependently
fractionated N2O and equals 0.516 [Cliff and Thiemens,
1997; Kaiser, 2002]. Values of d are relative to VSMOW. If
d17O is expressed relative to another standard that does not
lie on the same mass-dependent fractionation line as N2O
and VSMOW (e.g., air O2), a correction term has to be
included in equation (23)
1 þ d17 OðN2 O versus air O2 Þ
17 O ¼ b
1 þ d18 OðN2 O versus air O2 Þ
1 þ d17 Oðair O2 versus VSMOWÞ
b 1
1 þ d18 Oðair O2 versus VSMOWÞ
ð24Þ
For air O2, d17O = 11.92% and d18O = 23.50% versus
VSMOW [Luz et al., 1999], and the required correction of
17O is approximately 0.14%. We note that more recent
measurements of the d18O value of air O2 versus VSMOW
gave slightly higher values between 23.7 and 23.8%
[Coplen et al., 2002; Kaiser, 2002], but accompanying
measurements of d17O are wanting, so the consequences for
the 17O correction are uncertain. Until now, studies of
7. Oxygen Isotope Anomaly of Tropospheric N2O
7.1. Definitions
[69] ‘‘Mass-independent’’ oxygen isotope effects were
initially quantified in an operational way by 17Oy d17O g d18O. The factor g has a representative value
close to the three-isotope exponent b (equation (7)). This
definition was sufficient as long as the measurement precision was limiting but leads to inconsistencies if accurate
results are desired [Miller, 2002; Young et al., 2002]. The
definition in equation (23) is therefore preferred, because it
is derived, without approximations, from the mass-dependent fractionation law (equation (7)) and uses standard
addition theorems for d values
17 O ¼
1 þ d17 O
b 1;
1 þ d18 O
ð23Þ
Figure 9. Correlation between ln(1 + d18O) and ln(1 +
d15N) of N2O in tropospheric samples. A linear least squares
fit through all data weighted by assumed errors of 0.05% in
both ln(1 + d15N) and ln(1 + d18O) is shown with 95%
confidence intervals as follows: ln(1 + d15N) = (17.0 ±
1.6)% + (0.54 ± 0.04) ln(1 + d18O) (r2 = 0.60).
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
oxygen isotope anomalies used 17Oy and expressed d
values relative to air [Cliff et al., 1999; Cliff and Thiemens,
1997; Röckmann et al., 2001b]. The new definition and the
discovery that a correction to 17O is required warrants a
reevaluation of previous results.
7.2. Atmospheric Measurements
[70] Oxygen isotopes of N2O were measured on molecular O2 derived from the remainder of the tropospheric N2O
after N2O+ and NO+ analysis. N2O was decomposed to N2
and CO2 (section 3.5). CO2 was subsequently converted to
CH4 and H2O which is finally treated with F2 to give O2 and
HF. In the form of O2, d17O and d18O can be determined
without interference from other isotopes [Brenninkmeijer
and Röckmann, 1998]. An alternative approach by Cliff and
Thiemens [1994] obtains O2 by cryogenic separation from
N2, following N2O decomposition on a gold surface.
[71] All samples exhibit a clear 17O excess [Cliff et al.,
1999; Cliff and Thiemens, 1997; Röckmann et al., 2001b].
The results of our workgroup give 17O = (0.93 ± 0.08)%
(n = 10), which agrees well with the value of (0.85 ±
0.19)% (n = 124) derived from Cliff and Thiemens’s [1997]
larger data set. Had we used the outdated definition for the
17
O excess and air O2 as a standard, values of 17Oy = (1.01 ±
0.08)% and (0.98 ± 0.19)% would have been obtained. The
difference between these two values is actually smaller than
between the values for 17O, because our O2 working
standard is 29% lighter in 18O than tropospheric N2O.
[72] As noted by Röckmann et al. [2001b], the most
prominent difference between the two data sets is the range
of observed d18O values. Whereas our data show d18O
values between 44.4 and 45.2% (section 6.2), the measurements from Cliff and Thiemens [1997] show a large spread
from 38 to 43% and are generally lower. This variability of
d18O is only conceivable in the vicinity of strong N2O
sources, because N2O is generally well mixed in the
troposphere owing to its long lifetime of 120 years.
However, variations of sample size were deemed to be
absent by Cliff and Thiemens within their error of 1% and
are unlikely at the remote sites where part of their samples
were obtained which may point to a lack of precision in
sampling, but not analysis, since the precision of the N2O
decomposition method was estimated to be better than
0.1% for d17O and d18O [Cliff and Thiemens, 1994]. The
differences in the magnitude of d18O may be a calibration
problem. The results from Röckmann et al. [2001b] are in
good agreement with other reports of d18O values for
tropospheric N2O, but the results from Cliff and Thiemens
[1997] are the lowest of all [Rahn and Wahlen, 2000;
Yoshida and Toyoda, 2000]. In conclusion, we find a
well-defined 17O excess of 17O = (0.9 ± 0.1)% (assuming
b = 0.516).
8. Conclusions
[73] The present work establishes the first long-term
data set of 18O/16O and position-dependent 15N/14N ratios
of tropospheric N2O over the course of 2 years at six
sampling stations between 28N and 79N (section 6.2).
The average d15N and d18O values of all samples were
(6.72 ± 0.12)% versus air N2 and (44.62 ± 0.21)% versus
ACH
19 - 15
VSMOW, respectively. No clear temporal trends or meridional variations could be identified. Given the experimental uncertainties, this does not represent a contradiction
to trends of 0.02 to 0.04%/a reconstructed from firn air
analyses and expected from isotope budget calculations.
Large differences in the position-dependent 15N abundance
were found with mean values of 1d15N = (15.8 ± 0.6)%
(terminal N atom) and 2d15N = (29.2 ± 0.6)% (central N atom).
A previous study found a smaller difference of 19 %
between 2d15N and 1d15N [Toyoda and Yoshida, 1999].
However, this is most likely not an indication of true tropospheric variability (which is expected to be very low), but
rather a reflection of deviating position-dependent calibrations of the working standard. Both calibrations involved
entirely different techniques, and a future direct standard
intercalibration is certainly warranted. Concerning the average d15N and d18O values, we strove for highest accuracy and
used two independent calibration methods (section 3.5 and
Table 3) which gave good agreement.
[74] Our manual (off-line) extraction, purification, and
mass spectrometric isotope analysis technique is more
laborious than online coupling to a gas chromatograph but
has higher precision and allows the analysis of 17O/16O
ratios on the same sample as used for d18O and positiondependent d15N analysis. About 100 dm3 of tropospheric air
give sufficient N2O for a traditional dual inlet system with a
cold finger for gas admission, but under standard conditions, 400 dm3 are used (sections 2.3 and 2.8). Gases
condensable at 77 K are separated from dry air samples
(section 2.4), CO2 is removed chemically from the condensable trace gas mixture (section 2.5), and N2O is further
purified on a preparatory gas chromatograph (section 2.6)
before mass spectrometric analysis (section 2.7). In principle, all elemental isotope ratios (18O/16O, 17O/16O, and
15 14
N/ N at both positions of the molecule) can be obtained
in a single run of the mass spectrometer, drawing on NO+
and N2+ fragment analysis in addition to conventional
measurements of the N2O+ molecule ion. A last-generation
isotope ratio-mass spectrometer with a suitable multiple
collector configuration may even allow the simultaneous
measurement of all ions (at m/z 28, 29, 30, 31, 44, 45, and 46).
However, the low relative abundance of the N2+ fragment
and a nonlinearity effect in its formation restrict the
precision for d17O to 0.5%. To achieve the desired
precision and accuracy, we scrutinized the analytical
technique for its key steps (sections 3.1 – 3.4) and present
a systematic treatment of the theoretical basis for data
reduction (section 2.7). Minor errors in previous descriptions of necessary calculations in data analysis are
identified and the concept of site preference is given a
critical appraisal. An unexpected contamination due to CF+
interference complicates NO+ fragment analysis (section 5).
Also, residual CO2 interferes with N2O (section 4). A
theoretical framework to correct for these interferences is
developed and verified experimentally. Modifications to
the analytical systems that eliminate potential interferences
from (hydro)fluorocarbons (such as outgassing of Viton
O-rings) are described (section 5.3).
[75] In section 7 we have reevaluated previous measurements of oxygen isotope anomalies in tropospheric N2O. A
downward revision of 17O by 0.1% to (0.9 ± 0.1)% is
required, primarily due to the fact that the previously used
ACH
19 - 16
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
reference, atmospheric O2, has a small negative anomaly
relative to VSMOW.
[ 76 ] Acknowledgments. We would like to thank Wolfgang
Hanewacker for careful sample processing, Rolf Hofmann for gas-chromatographic analyses and managing the sampling network, as well as our
collaborators for sample collection at the various stations. Dave Griffith
inspired deeper thoughts on linear combinations of d values.
References
Baertschi, P., Absolute 18O content of standard mean ocean water, Earth
Planet. Sci. Lett., 31, 341 – 344, 1976.
Brenninkmeijer, C. A. M., Robust, high-efficiency, high-capacity cryogenic
trap, Anal. Chem., 63(11), 1182 – 1184, 1991.
Brenninkmeijer, C. A. M., Measurement of the abundance of 14CO in the
atmosphere and the 13C/12C and 18O/16O ratio of atmospheric CO with
applications in New Zealand and Antarctica, J. Geophys. Res., 98(D6),
10,595 – 10,614, 1993.
Brenninkmeijer, C. A. M., and T. Röckmann, A rapid method for the
preparation of O2 from CO2 for mass spectrometric measurement of
17O/16O ratios, Rapid Commun. Mass Spectrom., 12(8), 479 – 483, 1998.
Brenninkmeijer, C. A. M., and T. Röckmann, Mass spectrometry of the
intramolecular nitrogen isotope distribution of environmental nitrous
oxide using fragment-ion analysis, Rapid Commun. Mass Spectrom.,
13(20), 2028 – 2033, 1999.
Brenninkmeijer, C. A. M., C. Koeppel, T. Röckmann, D. S. Scharffe,
M. Bräunlich, and V. Gros, Absolute measurement of the abundance
of atmospheric carbon monoxide, J. Geophys. Res., 106(D9), 10,003 –
10,010, 2001.
Cliff, S. S., and M. H. Thiemens, High-precision isotopic determination of
the 18O/16O and 17O/16O ratios in nitrous oxide, Anal. Chem., 66(17),
2791 – 2793, 1994.
Cliff, S. S., and M. H. Thiemens, The 18O/16O and 17O/16O ratios in atmospheric nitrous oxide: A mass-independent anomaly, Science, 278(5344),
1774 – 1776, 1997.
Cliff, S. S., C. A. M. Brenninkmeijer, and M. H. Thiemens, First measurement of the 18O/16O and 17O/16O ratios in stratospheric nitrous oxide: A
mass-independent anomaly, J. Geophys. Res., 104(D13), 16,171 – 16,175,
1999.
Coplen, T. B., et al., Compilation of minimum and maximum isotope
ratios of selected elements in naturally occurring materials and reagents,
Water Res. Invest. Rep. 01-4222, 131 pp., U.S. Geol. Surv., Reston, Va.,
2002.
Crutzen, P. J., The influence of nitrogen oxides on the atmospheric ozone
content, Q. J. R. Meteorol. Soc., 96, 320 – 325, 1970.
Esler, M. B., D. W. T. Griffith, F. Turatti, S. R. Wilson, T. Rahn, and
H. Zhang, N2O concentration and flux measurements and complete
isotopic analysis by FTIR spectroscopy, Chemosphere: Global Change
Sci., 2, 445 – 454, 2000.
Gonfiantini, R., Standards for stable isotope measurements in natural compounds, Nature, 271, 534 – 536, 1978.
Griffith, D. W. T., G. C. Toon, B. Sen, J.-F. Blavier, and R. A. Toth, Vertical
profiles of nitrous oxide isotopomer fractionation measured in the stratosphere, Geophys. Res. Lett., 27(16), 2485 – 2488, 2000.
Gros, V., C. A. M. Brenninkmeijer, P. Jöckel, J. Kaiser, D. Lowry, E. G.
Nisbet, P. O’Brien, T. Röckmann, and N. Warwick, Use of isotopes, in
Emissions of Chemical Species and Aerosols Into the Atmosphere, edited
by C. Granier, Kluwer Acad., Norwell, Mass., 2003.
Inoue, H. Y., and W. G. Mook, Equilibrium and kinetic nitrogen and oxygen isotope fractionations between dissolved and gaseous N2O, Chem.
Geol., 113, 135 – 148, 1994.
Kaiser, J., Stable isotope investigations of atmospheric nitrous oxide, Ph.D.
thesis, Johannes Gutenberg Univ., Mainz, Germany, 2002.
Kaiser, J., T. Röckmann, C. A. M. Brenninkmeijer, and P. J. Crutzen,
Wavelength dependence of isotope fractionation in N2O photolysis, Atmos. Chem. Phys., 3, 303 – 313, 2003.
Kim, K.-R., and H. Craig, Two-isotope characterization of nitrous oxide in
the Pacific Ocean and constraints on its origin in deep water, Nature, 347,
58 – 61, 1990.
Kim, K.-R., and H. Craig, Nitrogen-15 and oxygen-18 characteristics of
nitrous oxide: A global perspective, Science, 262(5141), 1855 – 1857,
1993.
Li, W.-J., B. Ni, D. Jin, and T.-L. Chang, Measurement of the absolute
abundance of oxygen-17 in V-SMOW, Chin. Sci. Bull., 33(19), 1610 –
1613, 1988.
Luz, B., and E. Barkan, Assessment of oceanic productivity with the tripleisotope composition of dissolved oxygen, Science, 288(5473), 2028 –
2031, 2000.
Luz, B., E. Barkan, M. L. Bender, M. H. Thiemens, and K. A. Boering,
Triple-isotope composition of atmospheric oxygen as a tracer of biosphere productivity, Nature, 400, 547 – 550, 1999.
Mak, J. E., and C. A. M. Brenninkmeijer, Compressed-air sample technology for isotopic analysis of atmospheric carbon monoxide, J. Atmos.
Ocean. Technol., 11(2), 425 – 431, 1994.
Mariotti, A., Atmospheric nitrogen is a reliable standard for natural 15N
abundance measurements, Nature, 303, 685 – 687, 1983.
Mariotti, A., Natural 15N abundance measurements and atmospheric nitrogen standard calibration, Nature, 311, 251 – 252, 1984.
McLinden, C. A., M. J. Prather, and M. S. Johnson, Global modeling of the
isotopic analogues of N2O: Stratospheric distributions, budgets, and the
17O-18O mass-independent anomaly, J. Geophys. Res., 108(D8), 4233,
doi:10.1029/2002JD002560, 2003.
Miller, M. F., Isotopic fractionation and quantification of 17O anomalies in the oxygen three-isotope system: An appraisal and geochemical significance, Geochim. Cosmochim. Acta, 66(11), 1881 – 1889,
2002.
Minschwaner, K., R. J. Salawitch, and M. B. McElroy, Absorption of solar
radiation by O2: Implications for O3 and lifetimes of N2O, CFCl3, and
CF2Cl2, J. Geophys. Res., 98(D6), 10,543 – 10,561, 1993.
Prather, M., et al., Atmospheric chemistry and greenhouse gases, in Climate
Change 2001: The Scientific Basis: Contribution of Working Group I to
the Third Assessment Report of the Intergovernmental Panel on Climate
Change, edited by J. T. Houghton et al., pp. 239 – 287, Cambridge Univ.
Press, New York, 2001.
Prinn, R. G., et al., A history of chemically and radiatively important gases
in air deduced from ALE/GAGE/AGAGE, J. Geophys. Res., 105(D14),
17,751 – 17,792, 2000.
Rahn, T., and M. Wahlen, A reassessment of the global isotopic budget of
atmospheric nitrous oxide, Global Biogeochem. Cycles, 14(2), 537 – 543,
2000.
Ramaswamy, V., O. Boucher, J. Haigh, D. Hauglustaine, J. Haywood,
G. Myhre, T. Nakajima, G. Y. Shi, and S. Solomon, Radiative forcing
of climate change, in Climate Change 2001: The Scientific Basis:
Contribution of Working Group I to the Third Assessment Report of the
Intergovernmental Panel of Climate Change, edited by J. T. Houghton et
al., pp. 349 – 416, Cambridge Univ. Press, New York, 2001.
Röckmann, T., Measurement and interpretation of 13C, 14C, 17O and 18O
variation in atmospheric carbon monoxide, Ph.D. thesis, Ruprecht-KarlsUniv., Heidelberg, Heidelberg, Germany, 1998.
Röckmann, T., J. Kaiser, C. A. M. Brenninkmeijer, J. N. Crowley,
R. Borchers, W. A. Brand, and P. J. Crutzen, Isotopic enrichment of
nitrous oxide (15N14NO, 14N15NO, 14N14N18O) in the stratosphere and
in the laboratory, J. Geophys. Res., 106(D10), 10,403 – 10,410, 2001a.
Röckmann, T., J. Kaiser, J. N. Crowley, C. A. M. Brenninkmeijer, and P. J.
Crutzen, The origin of the anomalous or ‘‘mass-independent’’ oxygen
isotope fractionation in atmospheric N2O, Geophys. Res. Lett., 28(3),
503 – 506, 2001b.
Röckmann, T., J. Kaiser, and C. A. M. Brenninkmeijer, The isotopic
fingerprint of the pre-industrial and the anthropogenic N2O source,
Atmos. Chem. Phys., 3, 315 – 323, 2003a.
Röckmann, T., J. Kaiser, C. A. M. Brenninkmeijer, and W. Brand, Gaschromatography, isotope-ratio mass spectrometry method for high-precision position dependent 15N and 18O measurements of atmospheric
nitrous oxide, Rapid Comm. Mass Spectrom., 17(16), 1897 – 1908,
2003b.
Schütze, M., Ein neues Verfahren zur direkten Sauerstoffbestimmung
in organischen Substanzen, Z. Anal. Chem., 118(7/8), 245 – 258,
1940.
Sowers, T., A. Rodebaugh, N. Yoshida, and S. Toyoda, Extending records
of the isotopic composition of atmospheric N2O back to 1800 A.D. from
air trapped in snow at the South Pole and the Greenland Ice Sheet Project
II ice core, Global Biogeochem. Cycles, 16(4), 1129, doi:10.1029/
2002GB001911, 2002.
Tanaka, N., D. M. Rye, R. Rye, H. Avak, and T. Yoshinari, High precision
mass spectrometric analysis of isotopic abundance ratios in nitrous oxide
by direct injection of N2O, Int. J. Mass Spectrom. Ion Processes, 142(3),
163 – 175, 1995.
Toyoda, S., and N. Yoshida, Determination of nitrogen isotopomers of
nitrous oxide on a modified isotope ratio mass spectrometer, Anal.
Chem., 71(20), 4711 – 4718, 1999.
Trolier, M., J. W. C. White, P. P. Tans, K. A. Masarie, and P. A. Gemery,
Monitoring the isotopic composition of atmospheric CO2: Measurements
from the NOAA Global Air Sampling Network, J. Geophys. Res.,
101(D20), 25,897 – 25,916, 1996.
Uehara, K., K. Yamamoto, T. Kikugawa, and N. Yoshida, Isotope analysis
of environmental substances by a new laser-spectroscopic method utilizing different pathlengths, Sens. Actuator B-Chem., 74(1 – 3), 173 – 178,
2001.
KAISER ET AL.: SPECTROMETRIC ISOTOPE ANALYSIS OF TROPOSPHERIC NITROUS OXIDE
Weiss, R. F., and B. A. Price, Nitrous oxide solubility in water and seawater, Mar. Chem., 8, 347 – 359, 1980.
Yoshida, N., and S. Toyoda, Constraining the atmospheric N2O budget
from intramolecular site preference in N2O isotopomers, Nature, 405,
330 – 334, 2000.
Young, E. D., A. Galy, and H. Nagahara, Kinetic and equilibrium massdependent isotope fractionation laws in nature and their geochemical and
cosmochemical significance, Geochim. Cosmochim. Acta, 66(6), 1095 –
1104, 2002.
Yung, Y. L., and C. E. Miller, Isotopic fractionation of stratospheric nitrous
oxide, Science, 278(5344), 1778 – 1780, 1997.
ACH
19 - 17
Yung, Y. L., W. C. Wang, and A. A. Lacis, Greenhouse effect due to
atmospheric nitrous oxide, Geophys. Res. Lett., 3(10), 619 – 621, 1976.
C. A. M. Brenninkmeijer, Abteilung Chemie der Atmosphäre, MaxPlanck-Institut für Chemie, D-55020 Mainz, Germany.
J. Kaiser, Department of Geosciences, Princeton University, Princeton,
NJ 08540, USA. ([email protected])
T. Röckmann, Bereich Atmosphärenphysik, Max-Planck-Institut für
Kernphysik, D-69117 Heidelberg, Germany.