Improved method for analysis of dic in natural water samples
1. Improved method for isotopic and quantitative analysis
of dissolved inorganic carbon in natural water samples
Nelly Assayag1,2*, Karine Rivé1
, Magali Ader1,2
, Didier Jézéquel3
and Pierre Agrinier1,2
1
Laboratoire de Physico-chimie des Fluides Géologiques, Institut de Physique du Globe de Paris & Université Paris 7 – UMR CNRS 7154, 2
Place Jussieu, 75251 Paris Cedex 05, France
2
Centre de recherches sur le stockage géologique du CO2, Institut de Physique du Globe de Paris, 4 Place Jussieu, 75252 Paris Cedex 05,
France
3
Laboratoire de Géochimie des Eaux, Institut de Physique du Globe de Paris & Université Paris 7 – UMR CNRS 7154, 2 Place Jussieu, 75251
Paris Cedex 05, France
Received 20 December 2005; Revised 19 May 2006; Accepted 21 May 2006
We present here an improved and reliable method for measuring the concentration of dissolved
inorganic carbon (DIC) and its isotope composition (d13
CDIC) in natural water samples. Our
apparatus, a gas chromatograph coupled to an isotope ratio mass spectrometer (GCIRMS), runs
in a quasi-automated mode and is able to analyze about 50 water samples per day. The whole
procedure (sample preparation, CO2(g)–CO2(aq) equilibration time and GCIRMS analysis) requires 2
days. It consists of injecting an aliquot of water into a H3PO4-loaded and He-flushed 12 mL glass tube.
The H3PO4 reacts with the water and converts the DIC into aqueous and gaseous CO2. After a CO2(g)–
CO2(aq) equilibration time of between 15 and 24 h, a portion of the headspace gas (mainly CO2RHe) is
introduced into the GCIRMS, to measure the carbon isotope ratio of the released CO2(g), from which
the d13
CDIC is determined via a calibration procedure. For standard solutions with DIC concentrations
ranging from 1 to 25 mmol LS1
and solution volume of 1 mL (high DIC concentration samples) or
5 mL (low DIC concentration samples), d13
CDIC values are determined with a precision (1s) better
than 0.1%. Compared with previously published headspace equilibration methods, the major
improvement presented here is the development of a calibration procedure which takes the carbon
isotope fractionation associated with the CO2(g)–CO2(aq) partition into account: the set of standard
solutions and samples has to be prepared and analyzed with the same ‘gas/liquid’ and ‘H3PO4/water’
volume ratios. A set of natural water samples (lake, river and hydrothermal springs) was analyzed to
demonstrate the utility of this new method. Copyright # 2006 John Wiley Sons, Ltd.
Dissolved inorganic carbon (DIC ¼ [H2CO3] þ [HCO
3 ] þ
[CO2
3 ]) represents the main inorganic carbon phase in most
natural waters. DIC can be affected by many biophysico-
chemical processes such as biogenic CO2 uptake (e.g.
photosynthesis) and release (e.g. respiration, methanic
fermentation, methane oxidation); exchange with atmos-
pheric CO2 (degassing or dissolution); carbonate precipi-
tation or dissolution; and CO2 generated by metamorphic
reactions or magmatic degassing. Its corresponding d13
CDIC
may change for two main reasons: (1) carbon isotope
fractionations taking place during these processes, and (2)
mixings between different DIC sources which have distinct
isotope signatures. Consequently, the coupled measurement
of the d13
CDIC and of the DIC concentration was recognized
as an essential tracer that provides quantitative information
on the dominant processes which affect carbon pools in
aquatic systems.1–10
Several methods for DIC concentration and d13
CDIC
measurements have previously been developed and modi-
fied. The most conventional technique consists of precipitat-
ing DIC as carbonates, collecting, drying and reacting the
precipitate with H3PO4, to release CO2(g).5,7,11–16
The gas is
then purified to remove water before analysis by mass
spectrometry.17
A ‘more direct’ technique consists of
acidifying water samples, in order to degas DIC as CO2(g).
The CO2(g) is then collected and purified on a vacuum line
before its introduction into the mass spectrometer.4,14,18–21
Recent developments have been aimed at reducing the
number of procedural steps and the considerable time
required for DIC extraction. To eliminate the step of DIC
extraction under vacuum, gas chromatography coupled to
continuous flow mass spectrometry was developed.22–28
With such methods, DIC extraction is not complete: CO2 is
partitioned between the gaseous and the aqueous phase and
a carbon isotope fractionation between CO2(g) and CO2(aq)
RAPID COMMUNICATIONS IN MASS SPECTROMETRY
Rapid Commun. Mass Spectrom. 2006; 20: 2243–2251
Published online in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/rcm.2585
*Correspondence to: N. Assayag, Laboratoire de Physico-chimie
des Fluides Géologiques, Institut de Physique du Globe de Paris
Université Paris 7 – UMR CNRS 7154, 2 Place Jussieu, 75251
Paris Cedex 05, France.
E-mail: assayag@ipgp.jussieu.fr
Contract/grant sponsor: Centre de recherches sur le stockage
géologique du CO2, Institut de Physique du Globe de Paris-
Total-Schlumberger.
Copyright # 2006 John Wiley Sons, Ltd.
2. results from this partition. As a consequence, d13
CCO2ðgÞ
measurements must be corrected from this isotope fraction-
ation in order to obtain the d13
CDIC. Among the authors
describing this method, only Miyajima et al.,22
Salata et al.23
and Capasso et al.26
reported the presence of a carbon isotope
fractionation between CO2(g) and CO2(aq). However, to
determine d13
CDIC values, they used the Henry’s Law
constant for CO2 and the carbon isotope fractionation factor
between CO2(g) and CO2(aq) for pure water. These ‘pure
water’ assumptions, which do not take salinity effects into
account, are questionable since a significant amount (a few
weight %) of H3PO4 is added to degas CO2(g) from the water
samples.
The aim of this paper is to provide a rigorous method to
calculate correctly d13
CDIC values for solutions over a wide
salinity range, by taking CO2(g)–CO2(aq) partition and salinity
effects (essentially induced by H3PO4 addition) into account.
The improvement offered by this analytical technique was
tested and validated for several types of natural water
samples.
EXPERIMENTAL PROCEDURE
Sample preparation for DIC concentration
and d13
CDIC measurements
The first step consists of introducing 100% phosphoric acid
(H3PO4) into uncapped Labco Exetainer1
tubes29
(i.e. 12 mL
glass tubes). This task is automatically performed by a Gilson
215 liquid handler, which injects a precise and reproducible
amount of H3PO4 (Fig. 1(a)). The 100% H3PO4 was prepared
according to the recipe published by McCrea.17
H3PO4 blank
tests were performed: for H3PO4 amounts ranging between
0.1 and 1 mL, the amount of evolved CO2 (if any) was below
the detection limit of the gas chromatograph coupled to the
isotope ratio mass spectrometer (GCIRMS) (lower than
0.5 nA for the signal intensity of mass 44, corresponding to a
DIC concentration lower than 0.2 mmol L1
).
In the second step, the tube is closed with a septum and
flushed with ultra-pure He gas, at a pressure of 2 bar, via a
Gilson 22X autosampler, for approximately 2 min (Fig. 1(b)).
This He-overpressure eliminates any residual laboratory air
and avoids any contamination by air input into the tube.
The third step consists of manually sampling an aliquot of
water from the storage tube and introducing it into the He-
H3PO4 tube. (Only one aliquot of water is extracted by the
storage tube.) For this operation, we use two syringes: the
first syringe ‘He syringe’ is filled with ultra-pure He gas and
the second one ‘water sampling syringe’ is kept empty, to
withdraw the aqueous phase from the storage tube. We insert
the two syringe needles through the septum of the storage
tube (Fig. 1(c)). An aliquot of water (either 1 mL for DIC
concentrations higher than 5 mmol L1
or 5 mL for DIC
concentrations lower 5 mmol L1
) is withdrawn from the
storage tube with the ‘water sampling syringe’ while ultra-
pure He gas is injected with the ‘He-syringe’ (Fig. 1(d)). The
aim of He injection is to facilitate the withdrawal of the
aqueous phase by compensating for the vacuum that it
produces. The aliquot of water is then injected through the
septum of the He-H3PO4 tube (Fig. 1(e)). The H3PO4 reacts
with the injected aqueous phase by decreasing the solution
pH below 2, thus converting the DIC into aqueous and
gaseous CO2. At such low pH, the amounts of the other DIC
species (HCO
3 and CO2
3 ) are negligible (Fig. 1(f)).
Preparation of standard solutions
Two types of standard solutions were prepared during this
study. The first type (calcite isotope standard solutions) is
prepared from CaCO3(s) with known d13
C/PDB (Table 1) and
Figure 1. Sample preparation for DIC and d13
C measurements.
Copyright # 2006 John Wiley Sons, Ltd. Rapid Commun. Mass Spectrom. 2006; 20: 2243–2251
DOI: 10.1002/rcm
2244 N. Assayag et al.
3. is used to calibrate the d13
Csample/ref data (provided by the
GCIRMS) relative to the PDB scale (see Fig. 5 and section
‘d13
CDIC and DIC content calibrations’ below). The second
type (DIC standard solutions) is made of NaHCO3(s) from
which solutions with known d13
C/PDB and DIC concentration
are prepared and used to calibrate the DIC measurements
versus the mass 44 signal intensity (see Fig. 6 and section
‘d13
CDIC and DIC content calibrations’ below). The DIC
standard solutions were also used to determine CO2(g)–
CO2(aq) equilibration time and to evaluate the reproducibility
of the measurements (see section ‘System performance and
validation tests’ below).
Both types of standard solutions were prepared according
to the following protocol. A known weight of powder is
introduced into an empty glass tube (Fig. 2(a)): for DIC
standard solutions, the weight of NaHCO3 is adjusted
depending on the required DIC concentration of the solution;
for calcite isotope standard solutions, the weight of CaCO3
ranges between 2 and 3 mg. The tubes are then flushed with
ultra-pure He gas, at a pressure of 2 bar, via a Gilson 22X
autosampler, for approximately 2 min (Fig. 2(b)) before DIC-
free water and H3PO4 are injected (Fig. 2c and 2d)).
As stressed below, an important aspect of this method is
that, for each batch, the ‘gas/liquid’ and the ‘H3PO4/water’
volume ratios have to be kept the same for standard solutions
and samples. In this study, the ‘H3PO4/water’ volume ratio
is set at 10 (i.e. for 1 mL of injected water (sample or DIC-free
water), 0.1 mL of H3PO4 is added).
Equilibration
Following the above-mentioned preparation steps, the tubes
(both standard solutions and samples) are fixed on a stirring
plate (Bioblock Scientific Ping Pong) in a temperature-
controlled room (25 18C) and vigorously shaken for at least
15 h to ensure that CO2(aq) is in isotopic equilibrium with
CO2(g) (Figs. 1(f) and 2(e)) (see section ‘CO2(g)–CO2(aq)
equilibration time’ below).
ANALYTICAL SYSTEM
Gas chromatography/isotope ratio mass
spectrometry
The GCIRMS, an Analytical Precision 2003 (today entitled
GV 2003 and provided by GV Instruments,30
Manchester,
UK), runs under He-continuous flow. The room temperature,
where the GCIRMS is installed, is held constant at 25 18C
(air conditioning).The outlet pressures on the He (carrier gas)
and CO2 (reference gas bottle) are maintained at 4 bar.
The He-CO2(g) gas mixture is removed from the headspace
gas (of the equilibrated samples) automatically via a Gilson
22X autosampler. This gas passes through a Nafion
membrane to remove any remaining H2O, and is then
directed through a Valco 6-port valve into a sample loop
(300 mL). After the sample loop is charged with gas, the valve
changes position and transfers the gas aliquot to a He stream
that flows into a gas chromatograph (GC). The GC column
(stainless column 60
1/800
2 mm, packed with HayesepQ
60/80 mesh, Chrompack) separates the CO2 gas from the
other residual gas components (e.g. N2, O2, Ar). The purified
CO2(g) passes through a non-adjustable open split (split ratio
1%), and is transferred via a capillary to the mass
spectrometer (Fig. 3). The GC He flow rates are set to
predefined values for this type of GCIRMS, by the instrument
supplier, and cannot be modified: 15 mL/min for the GC
column and about 60 mL/min for the Nafion membrane.
The mass 45/44 and 46/44 signal intensity ratios of CO2(g)
are measured by the IRMS and compared with those from the
CO2 reference gas, yielding the raw data. Craig’s corrections
are applied to these raw data and d13
Csample/ref is calcu-
lated.31
For each extraction tube, four successive GCIRMS
gas analyses (i.e. replicates) are carried out. The ‘final’
d13
Csample/ref results from the mean of the four GCIRMS gas
analyses.
Table 1. Carbon isotope composition of solid standards
(determined using the method described by McCrea17
)
Standard names d13
CPDB (%)
CaCO3 STANDARD 1 9.77 0.05
CaCO3 STANDARD 2 0.38 0.05
NaHCO3 STANDARD 3 4.40 0.05
CaCO3 STANDARD 4 8.78 0.05
Figure 2. Protocol for the preparation of standard solutions for the calibration of the DIC isotope
compositions and DIC concentrations.
Copyright # 2006 John Wiley Sons, Ltd. Rapid Commun. Mass Spectrom. 2006; 20: 2243–2251
DOI: 10.1002/rcm
Isotopic and quantitative analysis of DIC in water samples 2245
4. Prior to running the analyses, the stability and the linearity
of the GCIRMS ion source are checked. The ion source is
considered as stable if, for a constant CO2 reference gas
pressure (i.e. a constant reference gas pulse height), the
standard deviation s on isotopic ratios (e.g. 45/44, 46/44) is
less than 0.01% (n ¼ 10). It is considered as linear if, for
different CO2 reference gas pressures (i.e. the reference gas
pulse heights range between 1 to 8 nA), the difference
between isotopic ratios (e.g. 45/44, 46/44) for the two
extreme pulse heights (1–8 nA) is less than 0.06%/nA. The
ion source parameters (e.g. high tension, X-Z steer plates,
electron energy . . .) may be readjusted if necessary, in order
to improve the stability and the linearity of the ion source.
Correction for the ‘non-linearity’
of the d13
Csample/ref analysis
For water samples with very low DIC concentrations
(1 mmol L1
), the volume of CO2(g) (in the 300 mL sample
loop) injected into the GCIRMS is very small. In such a case,
we noted that the measured isotopic ratios were over-
estimated, as was also reported by Miyajima et al.22
An
isotope fractionation of the injected CO2 is thought to occur
mostly in the GC column (different elution time of 13
CO2 and
12
CO2 molecules) and in the ion source (ionization efficiency
differences between the masses 44, 45 and 46 of CO2(g)), and
to be responsible for the non-linearity of the GCIRMS when
the concentration of CO2 in the gas stream is too low.
To model this ‘non-linearity effect’, we ran a set of calcite
standards of various weights ranging from 0.5 to 3 mg (in this
case, we followed the protocol shown in Fig. 2 but we did not
add DIC-free water). In Fig. 4, the linearity correction
(D ¼ d13
Csample/ref d13
C
) is plotted versus the mass 44
signal intensity, ranging between 9 to 1 nA (d13
C
is the mean
value of the measured d13
Csample/ref between 6 and 9 nA). The
linearity correction ‘D’ is almost null for mass 44 signal
intensity, ranging between 6 and 9 nA, and gradually
increases when the mass 44 signal intensity is lower than
6 nA. This figure provides the non-linearity curve for the
whole system, including sampling steps, gas chromatog-
raphy and mass spectrometry. For a given mass 44 signal
intensity, the measured d13
Csample/ref is corrected from this
non-linearity effect according to the following relationship:
‘true d13
Csample/ref’ ¼ d13
Csample/ref D.
Figure 4 shows that the non-linearity curves overlap over a
period of 3 months; the maximum linearity correction ‘Dmax’
(1.2%) remains constant through time. The invariance of
the non-linearity curve shape and of ‘Dmax’, over a certain
period, demonstrates the stability of the GCIRMS system.
Nevertheless, before each sample batch analysis, it is
necessary to run a set of calcite standards of various weights
in order to check the shape of the non-linearity curve. (The
non-linearity curve shape may vary from one GCIRMS to
another.)
Figure 3. Gas chromatograph coupled to the mass spectrometry module.
Figure 4. ‘Non-linearity’ curve of the GCIRMS. The linearity
correction (D ¼ d13
Csample/ref–d13
C
) is plotted versus the ma-
ss 44 signal intensity, ranging between 9 to 1 nA (d13
C
is the
mean value of the measured d13
Csample/ref between 6 and
9 nA).
Copyright # 2006 John Wiley Sons, Ltd. Rapid Commun. Mass Spectrom. 2006; 20: 2243–2251
DOI: 10.1002/rcm
2246 N. Assayag et al.
5. d13
CDIC and DIC content calibrations
The change of scale to obtain the d13
Csample/ref values relative
to the PDB scale is achieved using a calibration based on a set
of calcite isotope standard solutions with known d13
C vs.
PDB values (Table 1). The d13
CDIC calibrations are shown in
Fig. 5; each calibration (i.e. equation) is specific to the ‘gas/
liquid’ and ‘H3PO4/DIC-free water’ volume ratios and can
be used to calculate the d13
CDIC values of samples prepared
with the same ‘gas/liquid’ and ‘H3PO4/water’ volume
ratios.
Mass 44 signal intensity is used to determine DIC
concentration, as it is proportional to the CO2(g) amount in
the headspace gas. Only the mass 44 signal intensity of the
first GCIRMS gas analysis is used to evaluate the DIC
content, because the CO2 pressure (in the headspace gas)
decreases with the number of gas samplings. The mass 44
signal intensity-DIC concentration relationship is calibrated
using two sets of DIC standard solutions. The first set
corresponds to the low DIC concentration range: 0.9–
4.9 mmol L1
and large solution volume (V ¼ 5 mL). The
second set corresponds to the high DIC concentration range:
5.0–19.0 mmol L1
and small solution volume (V ¼ 1 mL). In
both cases, good linear relationships (r2
0.995) are obtained
(Fig. 6). The precision of DIC determination is better than 5%
for DIC concentrations larger than 5 mmol L1
(for 1 mL
solution volume), or larger than 2.5 mmol L1
(for 5 mL
solution volume).
As mentioned above, the GC parameters do not require to
be readjusted daily. However, slight fluctuations of
unknown causes (e.g. room temperature fluctuations)
occurring at different time scales (daily, monthly) may affect
slightly the GCIRMS system (analytical drift). In order to
cope with such fluctuations, it is necessary to include a set of
DIC and calcite isotope standard solutions, at the beginning
and at the end of each daily sample batch run.
CO2 PARTITION AND CARBON ISOTOPE
FRACTIONATION BETWEEN CO2(G)
AND CO2(AQ)
As indicated above, the H3PO4 acidification of the aqueous
phase converts the DIC into CO2(g) and CO2(aq). Because
carbon isotope fractionation occurs between CO2(g) and
CO2(aq), the measured d13
C of the CO2(g) d13
CCO2ðgÞ
must be
corrected to obtain the d13
C of the DIC (d13
CDIC).32–37
In order
to describe this problem, we first present a mathematical
expression of the partition effect. This will lead us to suggest
a new procedure adapted for any type of solution.
This partition effect can be described with the following
equations:
(i) the carbon isotopes conservation law:
d13
CDIC ¼ XCO2ðgÞd13
CCO2ðgÞ þ ð1 XCO2ðgÞÞd13
CCO2ðaqÞ (1Þ
Figure 6. DIC calibrations. Linear relationship between
mass 44 signal intensity and DIC concentration. For 1 mL:
y ¼ 2.64 103
(0.03 103
)x 4.13 104
(1.5 104
),
r2
¼ 0.9998; For 5 mL: y ¼ 8.26 103
(0.04 103
)x
2.15 104
(1.6 104
), r2
¼ 0.9953.
Figure 5. d13
C calibrations. Linear relationship between the
d13
Csample/ref measured by the GCIRMS and the d13
Csample/
PDB: For a volume of 1 mL: y ¼ 1.0154 (0.023)x 34.12
(0.66), r2
¼ 0.9985; For a volume of 2.5 mL: y ¼ 0.995
(0.009)x 33.74 (0.26), r2
¼ 0.9998; For a volume of
5 mL: y ¼ 0.980 (0.016)x 33.46 (0.45), r2
¼ 0.9995.
These equations structurally include the conversion to the
PDB scale and the correction from the isotope fractionation
associated with the CO2(g)–CO2(aq) partition (‘A’ correction
term of Eqn. (4)).
Copyright # 2006 John Wiley Sons, Ltd. Rapid Commun. Mass Spectrom. 2006; 20: 2243–2251
DOI: 10.1002/rcm
Isotopic and quantitative analysis of DIC in water samples 2247
6. where XCO2ðgÞ and (1 XCO2ðgÞ) are the molar fractions of
CO2(g) and CO2(aq), respectively;
(ii) the carbon isotope fractionation ðD13
CCO2ðgÞ CO2ðaqÞÞ
between CO2(g) and CO2(aq):
D13
CCO2ðgÞCO2ðaqÞ ¼ d13
CCO2ðgÞ d13
CCO2ðaqÞ (2Þ
(iii) the NCO2ðgÞ=NCO2ðaqÞ molar ratio (N is the mole num-
ber):38
NCO2ðgÞ
NCO2ðaqÞ
¼
Vg
V1RTa
(3)
where a is the Henry’s Law constant for CO2 in water, Vg and
Vl are gas and liquid volumes, respectively, R is the ideal gas
constant, and T is the temperature.
Combining Eqns. (1), (2), and (3), a relationship between
d13
CDIC and d13
CCO2ðgÞ can be written:
d13
CDIC ¼ d13
CCO2ðgÞ
D13
CCO2ðgÞCO2ðaqÞ
1 þ
Vg
V1RTa
(4)
Numerical application of this relationship (4) requires
knowing both a and D13
CCO2ðgÞCO2ðaqÞ. Using the
literature values of a and D13
CCO2ðgÞCO2ðaqÞ (3.345
104
mol m3
Pa1 39
and 1.07%,34,35,37
respectively) for
pure water at 298.15 K, 1 mL of injected water þ 0.1 mL of
H3PO4, in a 12 mL glass tube (Vg 10.9 mL; Vl 1.1 mL),
provides carbon molar fractions of 92% for CO2(g) and 8% for
CO2(aq). In this case, the corrective (second) term of the Eqn.
(4) is very small (0.08%). Accordingly, it is reasonable to
use the approximation: d13
CDIC d13
CCO2ðgÞ. In the case of a
greater volume, 5 mL of injected water þ 0.5 mL H3PO4
(Vg ¼ 6.5 mL; Vl 5.5 mL), the carbon molar fraction of
CO2(g) decreases to 59% and that of CO2(aq) increases to
41%. In this case, the corrective term of Eqn. (4) is not
negligible (0.44%) and Eqn. (4) must be used to calculate
the d13
CDIC.
The limitation of this formal approach is that published
values of a and D13
CCO2ðgÞCO2ðaqÞ) apply for pure water but
not for H3PO4-acidified solutions. Indeed, numerical values
of a and D13
CCO2ðgÞCO2ðaqÞ depend on the ionic strength of
the solutions.34,36,39
For example, a is known to decrease with
the ionic strength of the solutions: a salinity change of 3.5%
from that of the pure water (0%) to that of seawater (3.5%)
decreases a by 18%.39
Consequently, the correction term of
Eqn. (4) is not well constrained for any type of solution. This
lack of knowledge of a and D13
CCO2ðgÞCO2ðaqÞ may be solved
by taking this correction term as a constant ‘A’ (Eqn. (5)) and
appraising ‘A’ globally, rather than evaluating individually
D13
CCO2ðgÞCO2ðaqÞ and a.
A ¼
D13
CðCO2ðgÞCO2ðaqÞÞ
1 þ
Vg
V1ðRTaÞ
(5)
The numerical value of ‘A’ remains unchanged if the ‘gas/
liquid’ volume ratio (Vg/Vl), a and D13
CCO2ðgÞCO2ðaqÞ are
constant and the temperature is kept constant. In order to
maintain a and D13
CCO2ðgÞCO2ðaqÞ constant, it is essential that
the salinity differences between standard solutions and
samples be negligible. This can be achieved by reproducing
the same ‘H3PO4/water’ volume ratio for standard solutions
and samples. Indeed, the high ionic strength of the H3PO4
buffers the ionic strength of standard solutions and samples
to a common high value, and therefore erases salinity
differences between standard solutions and samples.
This procedure allows us to be free of poorly known
parameters (a and D13
CCO2ðgÞCO2ðaqÞ, for H3PO4-acidified
solutions) and to keep constant the numerical value A. In
addition, there is no need to specifically evaluate the constant
‘A’; the correction term ‘A’ for the DIC partition into CO2(g)
and CO2(aq) is included as a part of the additive constant in
the equation for isotopic scale conversion (see legend to
Fig. 5).
SYSTEM PERFORMANCE
AND VALIDATION TESTS
CO2(g)–CO2(aq) equilibration time
In order to evaluate the time required for DIC to equilibrate
as CO2(g) and CO2(aq) after H3PO4 acidification, about 20 DIC
standard solutions with DIC concentrations of 4, 10 and
20 mmol L1
were analyzed for equilibration times ranging
from 6 to 36 h. As shown by Fig. 7, equilibration times
ranging between 15 and 24 h yielded good reproducibility of
the d13
CDIC values (1s ¼ 0.06%, n ¼ 12).
For shorter equilibration times (15 h), d13
CDIC values are
less reproducible and more dispersed probably because
CO2(g)–CO2(aq) equilibration is not fully achieved. For longer
equilibration times (24 h), the d13
CDIC values are also less
reproducible, probably as the consequence of CO2 diffusion
across the septa, out of the overpressured glass tube.
The standard solutions and the samples are prepared the
day before analysis on the GCIRMS and the CO2(g)–CO2(aq)
equilibration time takes place during the night. On the
following day the samples and standard solutions are both
analyzed on the GCIRMS. For each sample, the four
Figure 7. Effect of CO2(g)–CO2(aq) equilibration times on the
measured d13
CDIC. Equilibration between CO2(g) and CO2(aq)
tested for different time lengths: 6, 10, 15, 18, 21, 24, 30
and 36 h, and for three different DIC concentrations: 2, 4
and 10 mmol L1
. The shaded area represents the dis-
persion of d13
C values for equilibration time between 15
and 24 h.
Copyright # 2006 John Wiley Sons, Ltd. Rapid Commun. Mass Spectrom. 2006; 20: 2243–2251
DOI: 10.1002/rcm
2248 N. Assayag et al.
7. replicates are analyzed in 7 min; therefore, up to 70 analyses
(50 samples þ 16 standard solutions (2 4 calcite isotope
standard solutions þ 2 4 DIC standard solutions)) can be
run per 8 h day.
Reproducibility of isotopic measurements
In order both to evaluate the reproducibility of the method
over a wide range of DIC concentrations and to further check
that d13
CDIC values remain reproducible independently of
the injected sample volume, a set of DIC standard solutions
with DIC concentrations ranging from 2 to 20 mmol L1
, for
solution volumes of 1 and 5 mL, was run after a CO2(g)–
CO2(aq) equilibration time of 18 h (Table 2). Analysis of 50
DIC standard solutions yielded an average d13
CDIC value of
4.41% with a standard deviation (1s) of 0.09%, indicating
good analytical precision. This average d13
CDIC value
(4.41%) is consistent with the d13
C value of the NaHCO3(s)
(4.40%).
Using the present method and this GCIRMS system,
d13
CDIC values can be measured over a range of DIC
concentrations from 1 to 25 mmol L1
. Raising the trap
current (i.e. increasing the number of electrons crossing the
ion source and CO2 ionization yield) from 250 to 350 mA
allows us to lower the detection limit and to measure DIC
concentrations lower than 1 mmol L1
. However, the
precision of the isotopic measurements is degraded (0.2–
0.3%).
Salinity
As the concentration of the total phosphate in H3PO4-
acidified solutions is high (19.4 mol L1
), this protocol is
expected to be valid for water samples in a wide salinity
range. A set of tests was carried out to check if the injected
H3PO4 amount (‘H3PO4/water’ volume ratio set at 10) is
sufficient to buffer the salinity of any type of solution. DIC
and calcite isotope standard solutions, with NaCl concen-
trations ranging from 0 to 4 mol L1
, for solution volumes of
1 and 5 mL, were run after a CO2(g)–CO2(aq) equilibration
time of 18 h.
Table 3 shows that, despite the variations in NaCl
concentrations in the DIC standard solutions, the standard
deviation of d13
CDIC values (1s 0.06%) did not vary
beyond the analytical precision (0.1%). Figure 8 shows that
the d13
CDIC calibration lines remain essentially unchanged
for NaCl concentrations ranging from 0 to 4 mol L1
.
Therefore, any type of solution, in this specific range of
NaCl concentrations, can be compared with standard
solutions prepared using DIC-free water (i.e. deionized
water). As explained above, the ionic strength of H3PO4 is
very high compared with that of standard solutions and
samples. As a consequence, salinity differences between
standard solutions and samples become negligible when
H3PO4 is added. This validates the calibration procedure
described above.
Table 2. DIC concentrations and d13
CDIC values of DIC stan-
dard solutions (d13
CNaHCO3 ¼ 4.40%)
DIC concentration
(mmol L1
) n
Volume
(mL)
d13
CDIC
(%)
Sample
reproducibility
1.2 2 1 4.46 0.06
2.9 2 1 4.47 0.08
5.1 3 1 4.43 0.09
6.2 2 1 4.50 0.04
7.2 3 1 4.49 0.03
8.2 3 1 4.47 0.04
10.5 3 1 4.31 0.02
20 3 1 4.32 0.08
4.1 6 5 4.37 0.07
5.2 4 5 4.29 0.04
6.3 14 5 4.32 0.04
7.3 3 5 4.36 0.11
8.2 3 5 4.56 0.07
Mean d13
CDIC S4.41
Standard deviation 1s W0.087
Sample reproducibility: standard deviation of the four replicate
analyses of each sample.
Table 3. DIC concentrations and d13
CDIC values of DIC standard solutions with NaCl concentrations ranging from 0 to 4 mol L1
(d13
CNaHCO3 ¼ 4.40%)
DIC concentration (mmol L1
) n NaCl concentration (mol L1
) Volume (mL) d13
CDIC (%) Sample reproducibility
10.1 3 0 1 4.39 0.02
3.1 3 0 5 4.42 0.02
10.1 3 0.2 1 4.45 0.02
3.2 3 0.2 5 4.28 0.03
10.2 3 0.4 1 4.37 0.08
3.2 3 0.4 5 4.28 0.03
10.0 3 0.8 1 4.38 0.01
3.0 3 0.8 5 4.29 0.04
10.3 3 1.4 1 4.37 0.08
3.5 3 1.4 5 4.26 0.04
10.2 3 2 1 4.33 0.05
3.1 3 2 5 4.42 0.03
10.3 3 4 1 4.33 0.05
3.1 3 4 5 4.32 0.06
Mean d13
CDIC S4.35
Standard deviation 1s W0.063
Sample reproducibility: standard deviation of the four replicate analyses of each sample.
Copyright # 2006 John Wiley Sons, Ltd. Rapid Commun. Mass Spectrom. 2006; 20: 2243–2251
DOI: 10.1002/rcm
Isotopic and quantitative analysis of DIC in water samples 2249
8. APPLICATIONS
On the sampling site, water samples are filtered with 0.20 mm
Sartorius sterile filters, to remove organic matter and mineral
particles, and loaded up to the brim (to avoid any air
bubbles) in 12 mL glass tubes closed with rubber septa
(Labco Exetainer1
tubes). Duplicate water samples are
collected; they correspond to the same ‘sample’ stored in two
different Labco Exetainer1
tubes. In the case of long storage
times (3 months), 2 mL of saturated HgCl2 solution is added
per mL of sample, to stop bacterial activity which may affect
DIC concentrations and d13
CDIC values. The tubes are kept in
the dark, at 48C.28
A set of natural water samples from a lake, rivers and
hydrothermal waters (Fig. 9 and Table 4) was analyzed to
validate this method of measurement. The lake samples are
from a depth profile achieved in Lake Pavin (French Massif-
Central). The river samples are either flowing in a carbonate
setting in France (River Seine, Paris, laboratory tap water
extracted from the River Vanne and rivers in the Jura
Mountains), or in an andesitic volcanic setting in Guade-
loupe, in the French West Indies (River Bambou). The
hydrothermal waters (deep well and hydrothermal springs)
are from the Soufriere volcano of Guadeloupe.
Lake Pavin is a meromictic lake: the mixolimnion, i.e. the
upper water layers (from 0 to 60 m depth), is affected by
seasonal overturns,40,41
whereas the monimolimnion, i.e. the
deeper waters layers (from 60 to 90 m depth), is not affected
by seasonal overturns and remains isolated from the upper
part. The d13
CDIC depth profile of Lake Pavin, for November
2002 (Fig. 9), highlights the main biogeochemical processes
that affect the DIC. The mixolimnion is itself divided into two
parts: from 0 to 20 m depth, the superficial zone, where the
light penetration is the most important, is essentially affected
by the photosynthetic activity. The preferential assimilation
of 12
C of CO2, by photosynthetic organisms, leads to an
increase in d13
CDIC values. Between 20 and 60 m depth, the
photosynthetic activity decreases and organic matter decay
becomes dominant: d13
CDIC values decrease due to the
release of 12
C-enriched CO2 by the falling decayed organic
matter. In the deeper water layers, the d13
CDIC values
increase mostly due to a mixing between different DIC
sources: a methanogenic contribution (d13
CDIC 10%) and a
CO2 magmatic input (d13
CDIC 5%).42
The d13
CDIC values of river samples flowing in a carbonate
setting present homogenous values ranging between 11.5
and 14.7%. Kendall and Doctor43
proposed a scenario to
explain these d13
CDIC values. The CO2 mostly comes from the
organic matter decay in soils (d13
CCO2 26%).5
CO2
diffusion through soil comes with a carbon isotope
Figure 8. Plot of the 1s error ellipses for the d13
CDIC cali-
bration lines (i.e. d13
Csample/ref vs. d13
Csample/PDB) obtained for
a wide salinity range (NaCl concentrations ranging between 0
and 4 mol L1
). All the ellipses overlap, confirming the
absence of salinity effect.
Figure 9. d13
CDIC depth profile in Lake Pavin (Massif Central,
France) – November 2002.
Table 4. d13
CDIC values of natural water samples measured by the analytical protocol presented here
Type of natural water samples d13
CDIC (%) DIC concentration (mmol L1
)
River Seine (carbonate setting) 11.45 3.72
Tap water – River Vanne (carbonate setting) 12.68 4.37
Tap water -River Vanne (carbonate setting) 12.99 4.30
Jura River 1 (carbonate setting)47
14.69 0.70
Jura River 2 (carbonate setting)47
14.37 0.65
Jura River 3 (carbonate setting)47
14.59 0.63
Jura River 4 (carbonate setting)47
13.65 0.66
River Bambou (andesitic volcanic setting) 5.23 1.81
River Carbet (andesitic volcanic setting) 6.00 0.61
Hydrothermal spring (Soufriere volcano) 27.07.02 2.76 2.13
Hydrothermal spring (Soufriere volcano) 31.10.03 2.85 2.05
Deep well – 70 m depth (Soufriere volcano) 6.82 1.98
Copyright # 2006 John Wiley Sons, Ltd. Rapid Commun. Mass Spectrom. 2006; 20: 2243–2251
DOI: 10.1002/rcm
2250 N. Assayag et al.
9. fractionation of þ4.4%.44
Since H2CO3 is the predominant
carbonate species at typical soil pH values 5, CO2(g)
dissolution in water as H2CO3 occurs with a second carbon
isotope fractionation D13
CCO2ðgÞH2CO3
of þ1%.45
Both
carbon isotope fractionations lead the initial d13
CCO2 value
to a d13
H2CO3 value of about 22.6%. The total DIC resulting
from the dissolution of calcite (d13
CDIC 0%)5,8
by H2CO3
(d13
CDIC 22.6%) has a d13
CDIC value 11%. This d13
CDIC
value is consistent with that of the River Seine. The other
values (River Vanne and Jura rivers) may be explained by a
higher contribution of the organic matter to the DIC.
The d13
CDIC values of hydrothermal waters of the Soufriere
volcano: springs (2.8%) and deep well (6.8%), are
influenced by the local magmatic CO2 (3.5%).46
In the
same way, the d13
CDIC values of the River Bambou (5.2%)
and the River Carbet (6.0%) indicate a contribution of
magmatic CO2 to the riverine DIC in a volcanic setting.
CONCLUSIONS
In addition to this protocol offering several advantages
(automatic system, rapidity, precision, low sample volume),
the distinctive feature of this study is the rigorous treatment
of the CO2(g)–CO2(aq) partition. Attention should be dedi-
cated to this problem since it may lead to a systematic error
towards higher d13
CDIC values, particularly for the large
‘gas/liquid’ volume ratio required for low DIC concentration
samples.
Previous studies have treated the CO2(g)–CO2(aq) partition
by a theoretical approach based on questionable assump-
tions; numerical values of a (the Henry’s Law constant for
CO2) and D13
CCO2ðgÞCO2ðaqÞ (the carbon isotope fractionation
between CO2(g) and CO2(aq)) used are those determined for
pure water. The protocol presented here proposes a
calibration procedure which includes the isotope fraction-
ation associated with the CO2(g)–CO2(aq) partition and takes
salinity effects into account. Its key point is that standard
solutions and samples have to be prepared using the same
‘H3PO4/water’ and ‘gas/liquid’ volume ratios.
This protocol was validated for natural and synthetic
water samples in a wide salinity range (NaCl concentration:
0–4 mol L1
). For standard solutions with DIC concen-
trations ranging from 1 to 25 mmol L1
, d13
CDIC values are
determined with a precision better than 0.1% (1s).
Acknowledgements
We are grateful to Michel Girard, for the maintenance of the
mass spectrometer, and Nicole Vassard, for help during the
sample preparation. We thank Pierre Cartigny for helpful
discussions. We thank the two anonymous reviewers for
their constructive comments. This study was supported
financially by the Centre de recherches sur le stockage
géologique du CO2, Institut de Physique du Globe de
Paris-Total-Schlumberger ADEME partnership (Number
contribution 2137).
REFERENCES
1. Aravena R, Schiff SL, Trumbore SE, Dillon PJ, Elgood R.
Radiocarbon 1992; 34: 636.
2. Yang C, Telmer K, Veizer J. Geochim. Cosmochim. Acta 1996;
60: 851.
3. Wachniew P, Rozanski K. Geochim. Cosmochim. Acta 1997; 61:
2453.
4. Atekwana EA, Krishnamurthy RV. J. Hydrol. 1998; 205: 265.
5. Aucour AM, Sheppard SMF, Guyomar O, Wattelet J. Chem.
Geol. 1999; 159: 87.
6. Barth JAC, Veizer J. Chem. Geol. 1999; 159: 107.
7. Caliro S, Panichi C, Stanzione D. J. Volcanol. Geot. Res. 1999;
90: 219.
8. Telmer KH, Veizer J. Chem. Geol. 1999; 159: 6.
9. Helie JF, Hillaire-Marcel C, Rondeau B. Chem. Geol. 2002; 186:
117.
10. Marfia AM, Krishnamurthy RV, Atekwana EA, Panton WF.
Appl. Geochem. 2004; 19: 937.
11. Gleason JD, Friedman I, Hanshaw BB. USGS Prof. Pap. 1969
650-D, D 248 250D.
12. Deines PD, Langmuir D, Harmon RS. Geochim. Cosmochim.
Acta 1974; 38: 1147.
13. Barnes I, Downes CT, Hulston RJ. Am. J. Sci. 1978; 278: 1412.
14. Hassan AA. USGS Water-Resour. Invest. 1982; 82: 6.
15. Bishop PK. Chem. Geol. 1990; 80: 251.
16. Kusakabe M. Geochem. J. 2001; 41: 217.
17. McCrea JM. J. Chem. Phys. 1950; 18: 849.
18. Mook WG. PhD thesis, University of Groningen, The
Netherlands, 1968.
19. Tan TC, Pearson GJ, Walker RW. In Bedford Institute of
Oceanography, Canada, Rep. Series BI-R-73-16. 1973.
20. Games LM, Hayes JM. Anal. Chem. 1976; 48: 130.
21. Favara R, Grassa F, Inguaggiato S, Pecoraino G, Capasso G.
Geofisica Internacional 2002; 41: 313.
22. Miyajima T, Yamada Y, Hanba YT. Limnol. Oceanogr. 1995;
40: 994.
23. Salata GG, Roelke LA, Cifuentes LA. Mar. Chem. 2000; 69:
153.
24. Saint Jean G. Rapid. Commun. Mass Spectrom. 2003; 17: 419.
25. Torn MS, Davis S, Bird JA, Shaw MR, Conrad ME. Rapid.
Commun. Mass Spectrom. 2003; 17: 2675.
26. Capasso G, Favara R, Grassa F, Inguaggiato S, Longo M.
Annals Geophys. 2005; 48: 159.
27. Spotl C. Isotopes Environ. Health Stud. 2005; 41: 217.
28. Torres ME, Mix AC, Rugh WD. Limnol. Oceanogr. Methods
2005; 3: 349.
29. Available: http://www.exetainer.com/.
30. Available: http://www.gvinstruments.co.uk/GV2003.htm.
31. Craig H. Geochim. Cosmochim. Acta 1957; 12: 133.
32. Deuser WG, Degens ET. Nature 1967; 215: 1033.
33. Vogel JC, Grootes PM, Mook WG. Z. Phys. 1970; 230:
225.
34. Mook WG, Bommerson JC, Staverman WH. Earth Planet.
Lett. 1974; 22: 169.
35. O’Leary MH. J. Phys. Chem. 1984; 88: 823.
36. Zhang J, Quay PD, Wilbur DO. Geochim. Cosmochim. Acta
1995; 59: 107.
37. Szaran J. Chem. Geol. 1998; 150: 331.
38. Michard G. Chimie des eaux naturelles. Published: Paris,
2003;
39. Weiss RF. Mar. Chem. 1974; 2: 203.
40. Camus G, Michard G, Olive P, Boivin P, Desgranges P,
Jézéquel D, Meybeck M, Peyrus JC, Vinson JM, Viollier E,
Kornprobst J. Bull. Soc. Géol. France 1993; 164: 767.
41. Aeschbach-Hertig W, Hofer M, Kipfer R, Imboden DM,
Wieler R. Geochim. Cosmochim. Acta 1999; 63: 3357.
42. Assayag N, Matter J, Jézéquel D, Ader M, Agrinier P. Carbon
stable isotopes to monitor the reactivity of injected CO2, in
geological reservoirs, Sixth International Symposium on
Applied Isotope Geochemistry, Prague, Czech Republic, 2005,
Abstracts, p 10.
43. Kendall C, Doctor DH. In Surface and Groundwater, Weath-
ering and Soils, Treatise on Geochemistry, Drever JI (eds).
Elsevier: Amsterdam, 2005; 319–364.
44. Cerling TE, Solomon DK, Quade J, Bowman JR. Geochim.
Cosmochim. Acta 1991; 55: 3403.
45. Deines P, Langmuir D, Harmon RS. Geochim. Cosmochim.
Acta 1974; 38: 1147.
46. Allard P, Hammouya G, Parello F. C. R. Acad. Sc. Paris 1998;
327: 315.
47. Chanchah Do Vale S. Masters Report, University Paris VII,
2003.
Copyright # 2006 John Wiley Sons, Ltd. Rapid Commun. Mass Spectrom. 2006; 20: 2243–2251
DOI: 10.1002/rcm
Isotopic and quantitative analysis of DIC in water samples 2251