N2O influence on isotopic measurements of atmospheric CO2
-
Upload
independent -
Category
Documents
-
view
1 -
download
0
Transcript of N2O influence on isotopic measurements of atmospheric CO2
RAPID COMMUNICATIONS IN MASS SPECTROMETRY
Rapid Commun. Mass Spectrom. 2004; 18: 1839–1846
Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/rcm.1559
N2O influence on isotopic measurements
of atmospheric CO2
Carmina Sirignano*, Rolf E. M. Neubert and Harro A. J. MeijerCentre for Isotope Research (CIO), Rijksuniversiteit Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands
Received 4 May 2004; Revised 18 June 2004; Accepted 21 June 2004
In spite of extensive efforts, even the most experienced laboratories dealing with isotopic
measurements of atmospheric CO2 still suffer from poor inter-laboratory consistency. One of the
complicating factors of these isotope measurements is the presence of N2O, giving rise to mass
overlap in the isotope ratio mass spectrometer (IRMS). The aim of the experiment reported here
has been twofold: first, the re-establishment of the correction for ‘mechanical’ interference of
N2O in the IRMS, along with its variability and drift, and the best way to quantitatively determine
the correction factors. Second, an investigation into secondary effects, i.e. the influence of N2O
admitted with the CO2 sample on the ‘‘cross contamination’’1 between sample and (pure CO2)
working gas. To make the suspected effects better detectable, isotopically enriched CO2 gas with
different concentrations of N2O has been measured for the first time. No evidence of secondary
effects was observed, from which we conclude that N2O is not a major player in the inter-laboratory
consistency problems. Still, we also found that the determination of the ‘mechanical’ N2O correc-
tion needs to be very carefully determined for each individual IRMS, and should be periodically
re-determined. We show that the determination of the correction should be performed using
CO2/N2O mixtures with concentration ratios around that of the atmosphere, as the extrapolation
from pure gas end member behaviour will give erroneous results due to non-linearities. For our
IRMS, a VG SIRA series II, we find a correction of 0.23% for d45CO2 and 0.30% for d46CO2 of atmo-
spheric samples, (with 0.85% mixing ratio). This implies that the relative ionisation efficiency (E)
value associated with this machine is 0.75. Copyright # 2004 John Wiley & Sons, Ltd.
The spatial and temporal variations of the sources and sinks
of atmospheric CO2 can be distinguished and quantified
applying measurements of stable isotopic ratios of atmo-
spheric CO2,2–7 especially the ratio of the stable carbon
isotopes 12C and 13C. Since the trends in this ratio caused by
anthropogenic emissions are small compared with the natur-
al variations, very high precision and accuracy are needed
in these measurements, as well as long-time stability and
traceability. For d13C an inter-comparability of 0.01%between different laboratories or measurement networks is
demanded, to permit the integration of data from all collec-
tion sites into one global dataset as input for global carbon
cycle models.8 Slightly relaxed requirements apply to d18O.
d13C and d18O are defined in the usual way and given on
the VPDB-CO2 scale,9 where not stated otherwise.
Sources of uncertainty can be found throughout the process
from sampling to data processing. Air samples from remote
places are normally filled into glass flasks for laboratory
analysis manually or applying dedicated devices.5,10–12
Concentrations and isotopic compositions of the sampled
gases might be changed during storage and travel times by
effects such as water interaction13 or leakage and permeation
through the flask valves.14 In the laboratories, different
extraction devices are used to quantitatively freeze out the
CO2 for mass spectrometric isotope analysis.15 Finally,
the calibration of reference gases (derived from calcium
carbonate material), IRMS effects, like reservoir bleed effect
and cross-contamination,1 and the ion correction algorithms
employed9 may vary between laboratories or between mass
spectrometer types. Corrections have to be applied due to:
(i) conversion of the measured d45CO2 to d13C, correcting
for the 12C17O16O contribution to m/z 45;
(ii) conversion of the measured d46CO2 to d18O, correcting
for the 13C17O16O and 13C17O2 contributions to m/z 46;
and
(iii) cryogenic co-extraction of N2O with CO2 (both have
isotopomers with m/z 44, 45, 46).
Although all steps have been studied quite extensively,
doubts arise every time an inter-comparison exercise is
conducted, often resulting in a consistency among different
laboratories that is poor compared not only with the precision
requirements, but also the respective laboratories’ internal
precision.
At present, the majority of the groups that perform high
accuracy isotopic measurements on atmospheric CO2 rely on
their suite of ‘‘whole air standards’’, from which they extract
CO2 in the same way they extract CO2 from their samples.
Variability or drifts between the ‘‘whole air CO2’’ and pure
Copyright # 2004 John Wiley & Sons, Ltd.
*Correspondence to: C. Sirignano, Centre for Isotope Research(CIO), Rijksuniversiteit Groningen, Nijenborgh 4, 9747 AG Gro-ningen, The Netherlands.E-mail: [email protected]
CO2 reference materials is quite commonly observed.
Whereas this complicates the calibration of the isotopic
measurement, and—thus—the comparability of one’s pro-
gram with that of other groups, it is generally agreed upon to
be the best way of keeping one’s own measurement series
internally consistent.
Differential drift, i.e. the difference between isotopic
measurements of whole air CO2 and pure CO2, is commonly
attributed to variability in the cryogenic extraction process.
Variability in completeness of extraction, wall exchange
effects, etc., are the type of effects that are usually referred to.
Other possible sources of differential drift, such as N2O
effects, are generally thought to be less of a problem.
The N2O experiments we report on here result from the
unique situation that emerged connected to the isotopic
analysis of atmospheric CO2 from the sampling network of
the Carbon Dioxide Research Group12 of the Scripps
Institution for Oceanography of the University of California,
San Diego (SIO). Our group had started these isotopic
analyses on CO2 already in 1977.16 From 1993 onwards, the
measurements were carried out at SIO itself, with a full year
of overlap between the two laboratories, as well as intensive
exchange of whole air standards and pure CO2 gases. The
unique fact is that from 1977 onwards, until the present day,
the CO2 extraction has been performed at the SIO laboratory
with the same cryogenic device and procedure. The extracted
CO2 samples were transferred, first to CIO, and later ‘‘in-
house’’, in so-called flame-off tubes (Pyrex tubes that were
closed by melting the glass together).
The results of the detailed CIO-SIO inter-comparison
showed a systematic offset in the d13C values of whole air
samples measured at the CIO of (�0.102� 0.005)% compared
with those measured at SIO (i.e. the CIO values being more
negative), while the offset in pure CO2 was found to be
significantly smaller at (�0.049� 0.005)% (determined using
the GS-19 and GS-20 reference materials, which are pure CO2
gases close to atmospheric isotopic CO2 composition; H. A. J.
Meijer et al., in preparation). Obviously, in this case, the
cryogenic extraction procedure cannot be blamed for this
difference.
Therefore, an alternative explanation has been sought.
As the only difference between the two classes of samples is
the presence of N2O, we decided to study the effects of N2O in
detail. Our idea was that there might be an N2O influence
beyond the ‘‘mechanical’’ addition of isotopomers with
masses 44, 45 and 46. For example, the traces of N2O might
influence the ionisation parameters or the cross contamina-
tion1 of the mass spectrometers in different ways. Therefore,
in order to enlarge the suspected effects, isotopically enriched
CO2 gas, mixed with different portions of N2O, has been
applied along with CO2-N2O mixtures of natural isotopic
composition.
Furthermore, we investigated in detail the usual calibra-
tion and correction procedures for N2O with emphasis on the
variability and the possibility of systematic errors.
THEORETICAL BACKGROUND
For mass spectrometric stable isotope analysis of carbon
dioxide from atmospheric samples, CO2 has to be separated
from the other air constituents. The most widely used proce-
dure for extracting CO2 is by quantitatively freezing out the
CO2 fraction in a cold trap at liquid nitrogen temperature
(�1968C), combined with a separation step from water
vapour, kept at �808C. Unfortunately, nitrous oxide (N2O,
an atmospheric greenhouse gas with concentrations around
320 ppb) has very similar physical properties and thus cannot
be separated cryogenically from CO2. Moreover, the masses
of the N2O isotopomers equal those of CO2 with the main
mass numbers 44, 45 and 46 amu. Thus atmospheric N2O
interferes with the mass spectrometric isotope analysis of
CO2, turning the apparent d13C and d18O values of CO2
too negative by (0.21� 0.01)% and (0.30� 0.01)%,17–20
respectively.
Until now any alternative procedure of removing the N2O
fraction proved to be inadequate because of insufficient
precision (e.g. separation by gas chromatography21) or
because it destroys the oxygen isotope ratio (reducing
N2O in a copper oven19). A correction for the N2O contri-
bution after measurements of its fragments in the same
time (following Friedli and Siegenthaler18 or Mook and
Jongsma19), or after N2O concentration measurements, is still
the most common practice. This, however, depends strongly
on the mass spectrometer type and even on the actual
operating conditions.
As reported by Friedli and Siegenthaler,18 the observed
isotopic ratios R45m for masses 45:44 and R46
m for 46:44 of the
CO2 and N2O mixtures can be expressed as:
R�m ¼ R�
C þ �ER�N
1 þ �E� ð1Þ
Here � ¼ ½N2O�=½CO2� is the concentration ratio of the two
species, which can also be expressed as the ratio of the
respective partial pressures � ¼ pN2O=pCO2, further on re-
ported as mol N2O/mol CO2 (sometimes omitted) or in %units. E, termed ‘ionisation efficiency ratio’,20 is the ratio of
the ionisation yields of N2O and CO2. With ionisation yield
we mean the ion current of m/z¼ 44 (I44) produced by the
CO2 or N2O gas, with a certain partial pressure p.
E ¼I44ðN2OÞ
pN2O
I44ðCO2ÞpCO2
ð2Þ
where the subscripts c, n and m refer to CO2, N2O and mea-
sured, respectively, and the symbol * stands for either 45 or
46.
The observed raw delta value measured on a mass
spectrometer when a CO2/N2O mixture has been admitted
can be expressed as:
��m ¼ R�m
R�reference
� 1
¼ �E
1 þ �E� R�
n � R�reference
R�reference
þ 1
1 þ �E� R�
c � R�reference
R�reference
ð3Þ
Writing the delta values that one would obtain measuring
N2O as if it were CO2, i.e. versus a CO2 standard, as:
��n ¼ R�n
R�reference
� 1 ð4Þ
Copyright # 2004 John Wiley & Sons, Ltd. Rapid Commun. Mass Spectrom. 2004; 18: 1839–1846
1840 C. Sirignano, R. E. M. Neubert and H. A. J. Meijer
and substituting in Eqn. (3), the relation between the mea-
sured and the corrected delta values can be given as:
��m ¼ �E
1 þ �E��n þ 1
1 þ �E��c ð5Þ
Finally, from the last equation the corrected value for CO2
samples, contaminated by co-extracted N2O, can be derived:
��c ¼ ��m þ Eð��m � ��nÞ� ð6Þ
This equation shows that the deviation introduced by the
N2O contribution is with a good approximation linearly
dependent on the mixing and ionisation efficiency ratios, the
variation introduced into the brackets by dm being negligible
as compared with dn and the measurement uncertainties.
Craig and Keeling22 were in 1963 the first to correct for the
N2O contribution in the isotopic analysis of atmospheric
carbon dioxide, suggesting a correction of þ0.31% for d13C
and þ0.44% for d18O. Later on Mook and van der Hoek,20
for samples of natural isotopic composition, approximated
Eqn. (6) to:
��c ¼ ��m � E��n� ð7Þ
and suggested a correction as follows:
�13C � �13Cm ¼ ð343 � 6Þ�E ð% Þ�18O � �18Om ¼ ð497 � 6Þ�E ð% Þ ð8Þ
They also took into account the different ionisation yield of
the two gas species, that they determined by comparing the
behaviour of the pure gases, stating that the efficiency ratio
depends neither on the inlet pressure, nor on the concentration ratio.
The literature values of E given for different mass spectro-
meters range from 0.68 to 0.75,18–20,23 leading to reported
corrections that vary from 0.20 to 0.22% for d13C and from
0.29 to 0.32% for d18O with r¼ 0.85%.
EXPERIMENTAL
Three series of four different artificial mixtures each, i.e.
CO2 with N2O traces, were prepared. Series AN and BN
were prepared using CO2 with a natural isotopic composition
and the series BE using CO2 enriched both in 13C and in 18O.
Series AN and series BN/BE were prepared and measured in
different time periods with an 18 month time lag. Table 1
gives a summary of the sample descriptions. Doubly labelled
CO2 was used in order to highlight possible secondary N2O
effects, such as an enhanced cross contamination of the
machine working standard due to N2O admitted to the
source with the sample. The enriched gas was prepared at
the CIO. For the oxygen signature, enriched water, usually
employed as tracer in biological applications, was brought
into isotopic equilibrium with the CO2 sample. For the carbon
signature, CO2 was developed from highly enriched sodium
carbonate (98% Na213CO3). The CO2 gas, enriched or not, was
transferred into 250 mL stainless steel cylinders, and sealed
with two manually operated Nupro valves series H (Nupro
Company-Willoughby, OH, USA). The same cylinders
were used later also to contain the final samples. After their
filling with CO2, the cylinders of each series were connected
to a common manifold (pumped to vacuum beforehand) and
left open and connected for about a week to allow the gas to be
isotopically the same in all the cylinders of a certain batch.
This appeared to be sufficient for the natural series only.
Repeated measurements of all four cylinders of series BN
showed a standard deviation of 0.015% for delta 45 and
0.011% for delta 46 as measured with respect to the machine
reference gas. No differences in the composition of the gas
could be found in the different containers. Therefore, just
one cylinder was used as a test cylinder to directly compare
the pure CO2 gas with the N2O-spiked ones. The enriched
samples, however, stayed significantly different from each
other even after 1 week of diffusive exchange. Thus they
had to be characterised individually by repeated indepen-
dent measurements before the addition of N2O. Furthermore,
the oxygen measurements of the enriched samples in parti-
cular were influenced by daily variations in the machine
behaviour. This could be confirmed by the variations in the
measurements of an independent enriched CO2 gas standard
(GS-45), routinely measured in three flasks at the end of every
measurement day. This part of the test was conducted in early
summer, when, according to our experience, the increase
in temperature and thus the variability of the laboratory
Table 1. The isotopic characteristics of the pure gases (all values are given in%with respect to our machine reference gas) and
mixing ratios of the artificial mixtures used
Samplename
r assigned (molN2O/mol CO2)
r calculated*(mol N2O/mol CO2)
d45CO2 raw(w.r.t. mach.
Ref.) pure gases SD
d46CO2 raw(w.r.t. mach.
Ref.) pure gases SD
CO2 Series AN AN0 0 0 7.11 0.02 �4.22 0.02AN1 (4.21� 0.04)10�4 4.58 10�4
AN2 (8.83� 0.09)10�4 9.17 10�4
AN3 (1.68� 0.02)10�3 1.82 10�3
N2O �349.16 0.35 �491.60 0.30CO2 Series BN BN0 0 0 6.78 0.02 �0.37 0.05
BN1 (3.59� 0.04) 10�4 3.08 10�4
BN2 (7.25� 0.07)10�4 6.70 10�4
BN3 (1.45� 0.01)10�3 1.37 10�3
CO2 Series BE BE0 0 0 446.09 0.02 510.77 0.08BE1 (3.61� 0.04)10�4 3.38 10�4 445.39 0.02 508.12 0.14BE2 (7.23� 0.07)10�4 6.58 10�4 445.89 0.04 509.86 0.28BE3 (1.44� 0.01)10�4 1.35 10�3 446.01 0.04 509.98 0.23
* Calculated from the ion beam of m/z 30, according to Eqn. (9).
N2O influence on isotopic measurements of atmospheric CO2 1841
Copyright # 2004 John Wiley & Sons, Ltd. Rapid Commun. Mass Spectrom. 2004; 18: 1839–1846
conditions influence the measurement performance. We sus-
pect that traces of water vapour, entering the inlet system
with every sample change, affect the cross contamination of
the mass spectrometer. Enriched samples with high precision
demands are the most susceptible to this.
The addition of N2O was done by freezing N2O gas, kept at
a known pressure in a calibrated volume, into the sample
cylinders cooled down to liquid nitrogen temperature. The
mixing ratios assigned volumetrically are accurate to��1%,
taking into account the errors in the pressure, temperature
and volume determinations.
The N2O/CO2 mixing ratios (see Table 1) vary largely
around the atmospheric mixing ratio. We roughly aimed at
one ‘‘blank’’ sample, containing pure CO2, one sample at half
of the present atmospheric mixing ratio (0.4%), one sample at
the present atmospheric mixing ratio (0.8%), and one sample
with double the present atmospheric mixing ratio (1.6%).
The assigned mixing ratios were checked by inference
based on the mass spectrometric determination of the ion
current of m/z¼ 30 (NOþ) and after subtraction of the back-
ground signal representing fragments, mainly 12C18O,
produced from CO2 (ðI30ÞCO2):19
½N2O�½CO2�
� �SAMPLE
¼
ðI30ÞSAMPLE � ðI30ÞðCO2ÞSAMPLE
ðI30ÞSTANDARD � ðI30ÞðCO2ÞSTANDARD
½N2O�½CO2�
� �STANDARD
ð9Þ
The standard for this ‘mass 30’ N2O determination was
prepared volumetrically, in a similar way as described above
for the mixtures; it had a N2O/CO2 ratio of 0.0089� 0.0005,
i.e. ten times natural, and was measured along with the B
series, that means at a different time than series A. The ‘mass
30’ results are also reported in Table 1, but are not used in
further calculations. Comparison shows only fair agreement,
which implies that the ‘mass 30’ correction method appears to
be vulnerable for systematic errors.
The actual experiment consisted of two separate parts; first,
verifying the ionisation efficiency of the pure gases CO2 and
N2O, admitted into the mass spectrometer at different inlet
pressures, and second, measuring the artificial mixtures for
their isotopic composition and comparing the results with
those obtained applying the theoretical correction described
in the previous paragraph.
For the first part of the experiment, aliquots of pure gases
were expanded from the cylinders into small glass flasks and
then admitted into the VG SIRA series II IRMS, of which the
electron energy was 58 eV, as for the routine measurements.
For the second part, the test cylinders were directly con-
nected to the manifold of the machine. Each time one aliquot
of the gas mixtures, trapped in the volume between the two
sealing valves, was expanded into the inlet system. After-
wards the pure N2O gas, used for the spiking, was measured
for its delta values ‘‘as if it were CO2’’, with respect to the
regular machine reference gas.
RESULTS AND DISCUSSION
The mass spectrometric response to different inlet pressures
of the pure gases used for the mixtures is shown in Fig. 1(a).
All gases were admitted to the source through the sample
side capillary and the variation in pressure was obtained by
expanding different aliquots of the same gas, filling different
flasks with different pressures, as well as controlling the inlet
volume by adjusting the sample bellow. On the vertical axis
the major beam currents are given as registered on the
m/z¼ 44 Faraday cup of the detector. The gases admitted to
the source of the IRMS were: CO2, both enriched and natural
(higher ionisation efficiency), and pure N2O (lower ionisation
efficiency). The inlet pressure was varied from a few tenths of
a hPa up to 39.5 hPa, resulting in ion beams between 0.1 and
7 nA for CO2 with natural isotopic composition and 5.41 nA
for N2O. The inlet pressure of enriched CO2 was varied
from 0.8 up to 31.85 hPa, producing a current from 0.79 up
to 5.59 nA. As expected the enriched CO2 curve lies below
the natural CO2 one, due to the lower 44CO2 abundance that
accompanies the enrichment in 45CO2 and 46CO2. From mass
balance considerations, the ionisation yield of enriched CO2
(d45CO2¼ 446%, d46CO2¼ 509%) would be expected to be
just 99.3% of that of CO2 of natural isotopic composition.
This results also in a higher relative ionisation efficiency of
N2O when compared with enriched CO2 than when com-
pared with natural CO2. Shown in Fig. 1(b) are the E values,
computed by comparing the current of the N2Om/z¼ 44 with
that of CO2 at the same inlet pressure. The fit curves are
computed from the cubic splines fitting the observation in
Fig. 1(a). These results show a clear picture of the non-linear
response of the ionisation yield at different inlet pressures of
a gas. The assumption that the relative yield of the two species
is independent of the gas pressure in the ion source, com-
monly adopted to estimate E for a given mass spectrometer,
is clearly not even applicable in the case of pure gases. Con-
sidering a pressure domain between 15 and 32 hPa, E is esti-
mated to vary from 0.80 to 0.87, for the N2O yield compared
with that of enriched CO2, and from 0.78 to 0.83, when the
N2O yield is compared with that of natural CO2. Approxi-
mately at the centre of this range, 24 hPa for enriched CO2
and 23 hPa for natural CO2, are the inlet pressures that give
a m/z¼ 44 current of about 4 nA, on which reference and
sample beam are normally adjusted for the measurement.
At these inlet pressures N2O produces signals of 3.4 and
3.2 nA, that give E values of 0.85 and 0.80, respectively,
with a difference too large to be explained simply by isotope
mass balance considerations. Besides this, both these estima-
tions differ from the best guess of E (0.75) obtained from
measurements on the artificial mixtures as described below.
The plot in Fig. 1(c) compares the behaviour of pure
enriched CO2, the same as used for the mixture, pure N2O
and a mixture of the two in equal parts. The behaviour of the
mixture does not follow quantitatively the expected response
estimated from the behaviour of the pure gases (solid line in
the plot).
A summary of the measurements on the artificial mixtures
is given in the plots in Figs. 2–4. The averages of the mea-
surements (5 and 10 repetitions, respectively, on AN and BN)
on the ‘natural CO2’/N2O mixtures have been reported as
completely raw data in terms of delta 45 and delta 46 on the
machine working gas scale. The linear relationship expressed
by Eqn. (6) is completely fulfilled. Regression analysis
between the isotopic results and the assigned mixing ratios
shows values comparable to each other and to those reported
1842 C. Sirignano, R. E. M. Neubert and H. A. J. Meijer
Copyright # 2004 John Wiley & Sons, Ltd. Rapid Commun. Mass Spectrom. 2004; 18: 1839–1846
in the literature: for the coefficients we find E �45m � �45
n
� �of
�(268� 10)% and �(267.7� 6.0)%, and E �46m � �46
n
� �of �(378� 19)% and �(320� 20)%, for the AN and BN
series, respectively. If we assume an atmospheric N2O/CO2
mixing ratio of 0.85%, computed from the most recent
concentrations and trends given for both species by
Houghton et al.,24 then the corrections to be applied would
be 0.23 and 0.30% for d45CO2 and d46CO2, respectively. Some
d46CO2 values of the BN series have suffered from the
consequences of traces of water in the mass spectrometer
inlet. This is even more so for the values of the enriched series
(BE), measured during the same period. For obvious reasons
the samples of the BE series could not be measured with the
same accuracy as the natural series. However, the isotope
ratios show a linear dependence of the mixing ratio (r) as
described in Eqn. (6), with a coefficient E ��m � ��n� �
estimated
from a least-squares fit to be (583� 33)% for �45c � �45
m
� �and
(910� 90)% for E �46c � �46
m
� �. These coefficients depend on ��m,
so they increase as the CO2 gas gets enriched. As mentioned
above, in this case, the deviation from the CO2 gas present in
each cylinder before the addition of N2O had to be regarded,
instead of the absolute d-value. Comparing the theoretical
corrections with those measured, no peculiar effects or evident
deviations from the expectations can be shown, indicating that
no evident secondary effects due to N2O ‘contamination’ can
be proven.
As the conclusive step of this experiment, we assigned the
best estimate for the ionisation efficiency ratio E according to
Figure 1. Mass spectrometric response to different inlet pressures of the pure gases used for the mixtures. (a) m/z¼ 44
currents versus the pressure registered by the inlet transducer of the VG SIRA series II mass spectrometer. (b) Relative ionisation
efficiency (E) of N2O compared with CO2 natural and enriched, calculated from pure gases. Symbols represent E values
computed as the ratio of the N2O and CO2m/z¼ 44 at the same inlet pressure. The lines are computed from the cubic spline fits in
(a). (c) Representation of the behaviour of a 1:1 mixture compared with that of its pure gas constituents. The solid line in the plot is
the calculated mean response of those of the pure gases.
N2O influence on isotopic measurements of atmospheric CO2 1843
Copyright # 2004 John Wiley & Sons, Ltd. Rapid Commun. Mass Spectrom. 2004; 18: 1839–1846
the measurements conducted on the mixtures belonging to
the AN, BN and BE series. d46CO2 values have been excluded
from the best estimation of the E parameter, because some of
them, belonging to the BN and BE series, suffered from the
consequences of traces of water in the inlet of the mass
spectrometer. The deviation of the true d45CO2 values from
those computed applying Eqn. (6) is:
��45CO2 ¼ �45c � �45
m þ E �45m � �45
n
� ��
� �ð10Þ
Dd45CO2 must be equal to zero when the correct parameters
are applied. As ‘true value’ for (d45c) we took the original CO2
composition of the samples before the N2O spiking. This is a
single value per series for the natural mixtures AN and BN, as
measured on the pure CO2 samples along with the mixtures.
For the enriched series individual ‘true values’ were assigned
to each different mixture, according to the initial isotopic
composition of the pure CO2 gas. The results of this exercise
are reported in Figs. 5(a) (series AN), 5(b) (series BN), and 5(c)
(series BE), respectively, in which we plotted the results for E
values from 0.6 to 0.9, in steps of 0.02. The closer the arbitrary
E values get to the best E value, the more the corrected values
approach the true pure CO2 value and the deviations
Dd45CO2 tend to be constant at 0. A summary of this simu-
lation can be obtained by plotting versus E the slope or the
correlation coefficient of the regression line through the
different data sets obtained by each value of E (shown in
Fig. 6). The definition of the best E value to be assigned to the
VG SIRA series II instrument is given by the intercept of the
slope or the correlation coefficient functions with the x-axis,
associated to a horizontal line (slope¼ 0) fitting perfectly
uncorrelated values.
As shown in Fig. 6, the best value for E is 0.749� 0.003,
showing no correlation at all between the final CO2 isotopic
values with changing CO2/N2O ratios. The estimation
derived from the analysis of the mixtures containing enriched
CO2 is just slightly higher (0.750) than the estimations based
Figure 2. Isotope measurements on series AN N2O/CO2
mixtures. Mass ratios 45/44 (left axis) and 46/44 (right axis)
are expressed as d45CO2 and d46CO2, respectively. [N2O]/
[CO2] is the partial pressure ratio of N2O to CO2. The symbols
represent the average of repeated measurements, their
standard error is given as error bars (�1s/Hn). Least-squares
regression leads to: dm45CO2¼ (7.11� 0.01) � (268�
10)[N2O]/[CO2]; dm46CO2¼ (�4.22� 0.01) � (378� 19)[N2O]/
[CO2]. The present [N2O]/[CO2] mixing ratio is indicated in the
figure as a vertical line; it is computed from the most recent
(2001) concentrations and trends given for both species by
Houghton et al.24
Figure 3. Isotope measurements on series BN N2O/CO2
mixtures. Least-squares regression leads to: dm45CO2¼
(6.79� 0.01) � (267.7� 6.0)[N2O]/[CO2]; dm46CO2¼ (�0.36�
0.01) � (320� 16)[N2O]/[CO2]. For further description, see
Fig. 2.
Figure 4. Isotope measurements on series BE N2O/CO2
mixtures. The apparent change in isotopic composition
(d45CO2, left axis, and d46CO2, right axis) is shown as the
deviation from the individual CO2 gas composition present in
each cylinder before the addition of N2O. The error bars give
the standard errors of repeated measurements after N2O
spiking. Least-squares regression leads to: (dc45CO2�
dm45CO2)¼ (583�33) [N2O]/[CO2]; (dc
46CO2�dm46CO2)¼ (910�
90) [N2O]/[CO2].
1844 C. Sirignano, R. E. M. Neubert and H. A. J. Meijer
Copyright # 2004 John Wiley & Sons, Ltd. Rapid Commun. Mass Spectrom. 2004; 18: 1839–1846
on natural mixtures, as is expected from mass balance
considerations. This testifies to the absence of any provable
secondary effect due to the N2O admitted to the source of a
mass spectrometer together with the sample.
CONCLUSIONS
In order to achieve very accurate atmospheric CO2 isotopic
determinations, a very accurate N2O correction has to be
applied. We did not find any secondary effects produced
by N2O different from the barometric additions on m/z¼ 44,
m/z¼ 45, m/z¼ 46, as shown by the behaviour of enriched
CO2 samples spiked with N2O. However, evidence of non-
linear response to the inlet pressure of the pure gases and a
different behaviour of these compared with mixtures of
CO2 with traces of N2O suggest a careful determination of
the relative ionisation efficiency used as parameter in the
N2O correction algorithm. In particular it is recommended
to estimate the N2O correction to be applied for each machine
from mixtures of well-known mixing ratios in the range of the
measured samples, and using the same settings of the IRMS
source as for the actual measurements. This procedure is pre-
ferable over the application of the theoretical algorithms with
the E parameter being derived from pure gas measurements.
For our VG SIRA series II, we find as best value for the ionisa-
tion efficiency ratio E¼0.75. This results in a correction of
0.23% for d45CO2 and of 0.30% for d46CO2, for atmospheric
samples with 0.85% N2O/CO2 mixing ratio. Regarding our
results, we conclude that it is unlikely that N2O effects cause
the difference between pure gas and whole air CO2 results
between CIO and SIO. The N2O corrections made for
d45CO2 were almost the same (0.217 and 0.210% for CIO
and SIO, respectively, for 1992–1994 air with mixing ratios
around 0.87%15). To explain the difference of over 0.05%in d45CO2, we would have to assume drastic errors in these
values, and thus in the E determination of the IRMS systems
involved. Furthermore, as our results show that secondary
effects of N2O are absent, the differential drifts between
pure CO2 and atmospheric CO2 that many groups observe
cannot be blamed on N2O either. However, the CIO-SIO
comparison shows that, besides the cryogenic extraction,
there must be other sources of error and deviation, yet to be
identified.
Figure 5. Differences between the true d45CO2 values and
those computed applying Eqn. (6), using different E values.
Reported are the results for E ranging from 0.60 to 0.90, in
steps of 0.02: (a) simulation on series AN; (b) simulation on
series BN; and (c) simulation on series BE.
Figure 6. Correlation coefficients (on the right axis) and
slopes (on the left axis) associated with the regression lines
fitting the computations shown in Figs. 5(a)–5(c). The data
are plotted against E, in steps of 0.01, with a higher resolution
than in the previous plots, but for the same simulated range
of E.
N2O influence on isotopic measurements of atmospheric CO2 1845
Copyright # 2004 John Wiley & Sons, Ltd. Rapid Commun. Mass Spectrom. 2004; 18: 1839–1846
REFERENCES
1. Meijer HAJ, Neubert REM, Visser GH. Int. J. Mass Spectrom.2000; 198: 45.
2. Keeling CD, Whorf TP, Wahlen M, van der Plicht J. Nature1995; 375: 666.
3. Ciais P, Denning AS, Tans PP, Berry JA, Randall DA,Collatz GJ, Sellers PJ, White JWC, Trolier M, Meijer HAJ,Francey RJ, Monfray P, Heimann M. J. Geophys. Res. 1997;D102: 5857.
4. Levin I, Graul R, Trivett NBA. Tellus 1995; 47B: 23.5. Trolier M, White JWC, Tans PP, Masarie KA, Gemery PA.
J. Geophys. Res. 1996; D101: 25897.6. Francey RJ, Tans PP, Allison CE, Enting IG, White JWC,
Trolier M. Nature 1995; 373: 326.7. Nakazawa R, Aoki S, Tanaka M. J. Geophys. Res. 1997; D102:
1271.8. Recommendations of the participants, Annex A4 of report
of the 11th WMO/IAEA meeting of experts on carbondioxide concentration and related tracer measurement tech-niques, Tokyo, Japan, 25-28/9/2001, Toru S, Kazuto S (eds).WMO-TD No. 1138, WMO-GAW, Geneve, 87–92.
9. Allison CE, Francey RJ, Meijer HAJ. IAEA-TECDOC 1995;825: 155.
10. Neubert REM, Spijkervet LL, Schut JK, Been HA, MeijerHAJ. J. Atmos. Ocean. Technol. 2004; 21: 651.
11. Francey RJ, Steele LP, Langenfelds RL, Lucarelli MP,Allison CE, Beardsmore DJ, Coram SA, Derek N, de SilvaFR, Etheridge DM, Fraser PJ, Henry RJ, Turner B, Welch ED,Spencer DA, Cooper LN. Baseline Atmospheric Program(Australia) 1993–1996; Bureau of Meteorology and CSIRODivision of Atmospheric Research, Melbourne, Australia;8–29.
12. Keeling CD. Geochim. Cosmochim. Acta 1961; 24: 277.13. Gemery PA, Trolier M, White JWC. J. Geophys. Res. 1996;
101: 415.14. Sturm P, Leuenberger M, Sirignano C, Neubert REM, Meijer
HAJ, Langenfelds R, Brand WA, Tohjiama Y. J. Geophys. Res.2004; 109: D04309.
15. White JWC, Ferretti DF, Vaughn BH, Francey RJ, AllisonCE. IAEA-TECDOC 2002; 1268: 3.
16. Keeling CD, Mook WG, Tans PP. Nature 1979; 277: 121.17. Allison CE, Francey RJ. IAEA-TECDOC 1995; 825: 131.18. Friedli H, Siegenthaler U. Tellus 1988; 40B: 129.19. Mook WG, Jongsma J. Tellus 1987; 39B: 96.20. Mook WG, van der Hoek S. Isotope Geosci. 1983; 1: 237.21. Ferretti DF, Lowe DC. J. Geophys. Res. 2000; 105: 6709.22. Craig H, Keeling CD. Geochim. Cosmochim. Acta 1963; 27:
549.23. Tanaka N, Rye DM, Rye R, Avak H, Tadashi Y. Int. J. Mass
Spectrom. Ion Processes 1995; 142: 163.24. Houghton JT, Ding Y, Griggs DJ, Noguer M, van der Linden
PJ, Dai X, Maskell K, Johnson CA. IPCC 2001; 881.
1846 C. Sirignano, R. E. M. Neubert and H. A. J. Meijer
Copyright # 2004 John Wiley & Sons, Ltd. Rapid Commun. Mass Spectrom. 2004; 18: 1839–1846