Preconcentration and analysis of CH4 isotopologues by
TREX–QCLAS
Requirements for the preconcentration system
The main analytical challenge in the present work is the quantification of
the CH3D isotopologue considering its very low natural abundance. A
further constraint is given by the spectroscopic setup, as the same optical
platform is used for simultaneous measurements of the 12CH4,
13CH4 and CH3D isotopologues. This unavoidably involves
compromises regarding the spectroscopic configuration, in particular the
selected optical path length and the amount of trace gas needed to achieve
the necessary measurement precision for both isotope ratios. Simulation of
CH4 absorption spectra in the target spectral regions indicated that
optimal conditions are realized at a sample gas pressure in the range of 20
to 60 hPa and for mole fractions ranging from 600 up to 1000 ppm CH4.
Since the CH4 mole fraction in ambient air is generally in the order of
1.8 ppm, the TREX system had to be designed to selectively extract CH4
from several liter of ambient air and concentrate into a gas volume of
around 20 mL (e.g., equivalent to the amount of gas in the 0.5 L absorption
cell of the laser spectrometer at a pressure of 40 hPa). In order to fulfill
the above requirements, significant developments and innovative solutions
for both TREX and QCLAS have been accomplished.
Schematics of the preconcentration unit (TREX). The blue lines
indicate the flow of sample air and TG, i.e., ambient air CH4-mole
fractions, while red lines represent the flow of calibration gases and
desorbed air, i.e., high CH4-mole fraction. MFC 1–4 and V1–4 stand for
mass flow controllers and 2-position valves, respectively.
TREX: design
The basic technology of the TREX (Fig. 1) is based
on the “Medusa” system
(Miller et
al., 2008), which was later adopted for the preconcentration of N2O and
its subsequent isotope analysis by QCLAS
(Mohn
et al., 2010, 2012, 2013, 2014; Waechter et al., 2008; Wolf et al., 2015).
The main advantages over previously developed systems
(Brand, 1995) are the low trapping
temperatures in combination with its independence from liquid nitrogen.
Preconcentration is achieved by temperature swing adsorption on a cold trap,
filled with a specific adsorbent material. The trap is first cooled down to
a temperature at which its dynamic adsorbing capacity for the target
substance (here CH4) is sufficiently large, while the majority of the
remaining bulk gases (e.g., N2, O2, Ar) pass through. During
desorption, the trap is heated stepwise to separate the target substance
from co-adsorbed interfering compounds. To minimize kinetic fractionation
effects, it is important to adsorb and desorb the target substance
quantitatively, i.e., with nearly 100 % recovery and with a high degree of
reproducibility, as discussed below.
Given the low boiling point temperature of CH4 (112 K) as an indication
for its volatility, the original design of the preconcentration
system required major revisions in terms of cooling power to enhance its
CH4 adsorption capacity. In addition, the layout was designed to fit in
a compact and field-deployable 19′′ rack system. These two requirements led
to a novel approach for the trap assembly.
Empirical investigations on the previous preconcentration unit
(Mohn et al., 2010) with various trap
models adsorbing CH4 at different temperatures showed that for a
complete and reliable CH4 recovery, the amount of adsorbent material
(HayeSep D, Sigma Aldrich, Switzerland) had to be increased by 10-fold.
This resulted in 1.8 g of HayeSep D filled in a stainless steel tubing
(length 90 cm, OD 4 mm, wall thickness 0.5 mm, volume 6.4 cm3) and
bracketed with glass wool (BGB Analytics AG, Switzerland) and wired mesh.
HayeSep D has previously been identified as an excellent high capacity
adsorbent material for CH4 (Eyer et
al., 2014). The tubing is curled around a custom-made cylindrical aluminum
standoff (outer diameter 70 mm, height 28 mm) with an optimized wall
thickness of 0.5 mm. A thermal conductance paste (340 HSC, Dow Corning Inc.,
USA) is applied at the contact region between trap and standoff to improve
heat dissipation. To further increase the adsorption capacity of the trap,
the trap temperature had to be decreased to 100 K, which was not achievable
with the previous preconcentration unit. Therefore, we decided for a compact
Stirling cryo-cooler with a cooling capacity of > 20 W at 100 K
(CryoTel GT, Sunpower Inc., USA) gaining in terms of size, weight and
cooling performance, with respect to the standard refrigeration unit (PCC:
Polycold Compact Cooler, Brooks Automation, USA) employed in the Medusa
preconcentration device
(Miller et
al., 2008). A copper plate disk (diameter 14 cm, weight 1.4 kg) was mounted
on the cold tip of the cooler, serving as a cold plate with large heat
capacity. Furthermore, we minimized the thermal cycle time of the trap for
repeated adsorption/desorption processes through a design in which the trap
is movable by a linear actuator (ZLD225MM, VG Scienta Ltd, UK). During
cooling, the actuator pushes the aluminum standoff against the cold plate.
The contact pressure is adjusted to 100 N using a chromium-steel corrugated
spring (WF-8941-SS, Durovis AG, Switzerland) placed centrically between
actuator and standoff. The flat bottom surface of the aluminum standoff and
the copper cold plate were polished and coated with a thin layer of heat
conductance paste (340 HSC, Dow Corning Inc., USA) to improve thermal
contact. Before heating, the standoff is decoupled from the cold plate. This
approach is overall faster and yields lower trap temperatures compared to
the previous preconcentration unit because the cold plate and the Stirling
cooler are completely undisturbed during the heating process.
For thermal isolation of the system, the core parts of the unit, i.e., the
cold tip of the Stirling cooler, the cold plate and the trap are housed in
a custom-made vacuum chamber evacuated to < 10-4 mbar with a
compact turbomolecular pump station (HiCube 80 Eco, Pfeiffer Vacuum GmbH,
Switzerland). The TREX unit is controlled and monitored by a
custom-developed LabVIEW program (National Instruments Corp., USA) with a
graphical user interface. All peripherals are connected through a 16-port
serial-to-ethernet connector (Etherlite 160, Digi International Inc., USA).
TREX: preconcentration procedure
The overall CH4-preconcentration cycle can be divided into three main
phases, as illustrated by Fig. 2: CH4-adsorption
(phase I, 25 min), CH4-desorption (phase II, 15 min) and trap conditioning
(phase III, 5 min). At the onset of phase I, the trap is brought in contact
with the cold plate by the actuator. It takes about 15 min for the trap to
cool down to a temperature of 101 K, then CH4 adsorption is initiated
by switching the six-port multi-position rotary valve (Valco Instruments Inc.,
Switzerland) to the adsorption position as shown in
Fig. 2. Dehumidified (nafion drier with dew point < 230 K, PD-50T-72MSS, Perma Pure, USA), particle-filtered (2-micron
filter, SS-4FW-2, Swagelok, Switzerland) sample gas is pushed through the
cooled trap with a membrane pump (PM 25032-022, KNF, Switzerland) at a
pressure of 4000 hPa. The sample gas flow is adjusted downstream of the trap
to a flow rate of 900 mL min-1 using a mass flow controller (MFC 1,
Vögtlin Instruments, Switzerland). After 500 s, corresponding to
preconcentration of 7.5 L sample gas, the six-port rotary valve is switched to
the desorption position.
Workflow of QCLAS (top) and TREX (bottom) during a complete
measurement cycle consisting of three phases: CH4-adsorption (phase I),
CH4-desorption (phase II) and trap conditioning (phase III). During
phase I, the sample gas and CG1 are analyzed by QCLAS with intermediate
flushing, while the adsorbent trap is cooled down by coupling to the base
plate, and CH4 from ambient air is adsorbed. During phase II, CH4
desorption is initialized by decoupling the trap from the base plate and
sequential heating of the adsorbent trap. In addition, desorbed CH4 is
filled into the QCLAS gas cell. In phase III, the adsorbent trap is
conditioned (TREX), while the analysis of the sample gas is initialized
(QCLAS).
In phase II (CH4 desorption), the linear actuator decouples the trap
from the copper cold plate with the six-port rotary valve set to the
desorption position (Fig. 1). Step-wise desorption
enables quantitative separation of the target substance CH4 from more
volatile gases (e.g., traces of N2, O2) and less volatile trace
gases, e.g., CO2 and N2O. To avoid that the latter gases, which are
mainly adsorbed on the front part of the trap, are released when the ends of
the trap heat up, the flow direction in the desorption step is forward. The
trap temperature during phase II is stepwise increased. Immediately after
decoupling, its temperature increases from around 106 to 113 K without
heating. Then, the trap temperature is raised first to 145 K and then to 175 K by heating with a round flexible polyimide heat foil (diameter 62.2 mm,
100 W, HK5549, Minco Products Inc., USA) placed centrically at the bottom of
the aluminum standoff and controlled by a PID temperature controller (cTron,
Jumo Mess- und Regeltechnik AG, Switzerland). During this period, mainly
volatile bulk gases (e.g., N2, O2, Ar) with low boiling points (77 to 90 K) are desorbed from the trap and vented through the QCLAS multipass
cell. The CH4 desorption is initiated by increasing the trap
temperature to 258 K and purging with 2 mL min-1 high-purity synthetic
air (20.5 % O2 in N2, purity 99.999 %, Messer Schweiz AG).
In parallel, a two-way solenoid valve (series 9, Parker Hannifin Corp., USA)
at the outlet of the evacuated QCLAS gas cell is closed; the desorbed
methane is thus accumulated in the cell. When the pressure in the QCLAS
absorption cell reaches 39.64 ± 0.04 hPa (Baratron 700B, MKS
Instruments, USA), the solenoid valve at the inlet of the cell is closed,
isolating the desorbed CH4 in the cell for subsequent analysis.
Measured absorption spectra for the determination of δ13C- (left) and δD-CH4 (right) along with the spectral
fit using Voigt profiles and the corresponding line-strengths from the
HITRAN database. Potential interferences are expected mainly from N2O
and H2O. The spectral line of N2O is divided by a factor of 1000
to fit in the graph, evidencing that even N2O-mole fraction of around
300 ppb can cause severe interference.
After CH4 desorption, phase III (conditioning) is initiated, in which
the residual, less volatile trace gases are removed from the HayeSep D trap
to assure reproducible starting conditions for each preconcentration cycle.
This is achieved by heating the trap to 323 K and purging it for 5 min at
reduced pressure (via V3, N920APE, KNF, Switzerland) with 25 mL min-1
high-purity synthetic air in backward flow direction. Thereby, residual gas
compounds such as H2O, N2O, CO2 and VOCs are removed. The
preconcentration cycle is completed by turning the six-port rotary valve to
isolate the HayeSep D trap.
QCLAS
The laser spectrometer is a modified version of a previous dual-QCL
instrument (QCL-76-D, Aerodyne Research Inc., USA) with a multi-pass cell of
76 m optical path length and a volume of 0.5 L, originally developed for
CH4, N2O and NO2 eddy flux measurements
(Tuzson et al., 2010). To comply with the demanding
requirements of high-precision isotope ratio measurements, critical elements
of the hardware electronics were upgraded as described in the following.
Because laser operation stability is of outmost importance, ultra-low noise
laser drivers (QCL1000, Wavelength Electronics Inc., USA) were installed to
minimize laser intensity variations and frequency jitter. The long-term
performance was improved by tighter and more precise control of the laser
heat-sink temperature by deploying high-precision controllers (PTC5K-CH,
Wavelength Electronics Inc., USA). A new pair of continuous wave DFB-QCL
laser (Alpes Lasers SA, Switzerland) was installed. Figure 3 shows the
covered spectral range at wavenumbers of 1295.7 and 1307.0 cm-1,
selected for δ13C- and δD-CH4, respectively. The
spectral regions were chosen to offer maximum sensitivity for the less
abundant CH3D isotopologues
(∼ 10-22 cm-1/(molecule × cm-2)), comparable line-strength for 13CH4 and
12CH4 to avoid saturation and are relatively free from spectral
interferences by other molecular species. The susceptibility to spectral
interferences could be further reduced by decreasing the pressure in the
laser spectrometer gas cell. These conditions could not be fulfilled within
the tuning capabilities of a single DFB-QCL, therefore, the simultaneous
measurement of δ13C- and δD-CH4 required a dual-laser
configuration
(McManus et al., 2011). The measured
absorption spectra were analyzed using commercially available software
(TDLWintel, Aerodyne Research Inc., USA). In terms of precision and
long-term stability, the instrument performance was characterized using the
Allan variance technique (Werle, 2010).
In combination with the TREX technique the laser spectrometer is operated in
a batch mode; i.e., the multipass cell is either filled with preconcentrated
sample or with calibration gas. Before each preconcentrated sample (ambient
or pressurized air), the cell is purged for 2 min with high-purity
synthetic air at 25 mL min-1 flow rate and reduced pressure (8 hPa) and
then evacuated to a pressure of 0.5 hPa. Similarly for the calibration gas
measurements, the cell is first purged and then flushed with calibration gas
dynamically diluted with high-purity synthetic air to the desired CH4
concentration at a total gas flow of 25 mL min-1. The cell pressure is
set to around 40 hPa (±0.04 hPa).
Interlaboratory comparison campaign
The intercomparison campaign took place from 6 to 22 June 2014 at the Empa
campus, located in the densely populated area of Dübendorf, Switzerland
(47∘24′11′′ N/8∘36′48′′ E, elevation
432 m a.s.l.). A main road passes 100 m south and a highway around
750 m north of the sampling site. Air was continuously sampled from the
rooftop of a five-story building at a flow rate of 25 L min-1 through a
25 m long unheated polyethylene-coated aluminum tubing (ID 9 mm,
Synflex-1300) using a piston pump (Gardner Denver Thomas GmbH). At the inlet
of the sampling pump the air was branched off to the different analyzers, as
indicated in Fig. 4. The purpose of the campaign is
to validate the TREX–QCLAS system under unattended operation conditions
comparable to a “field campaign”. Flask and bag sampling as well as
calibration of the commercial available laser spectrometers, however, were
operated manually.
Calibration gases and target gas
The calibration gases CG 1 and CG 2 were prepared at Empa from pure fossil
(99.9995 %, Messer, Switzerland) and biogenic CH4 (> 96 %, biogas plant Volketswil, Switzerland), diluted with high-purity
synthetic air. The exact amounts of added CH4 were determined using a
high precision flow measurement device (molbox1, DH Instruments Inc., USA),
and the dilution with synthetic air was controlled gravimetrically (ID 3,
Mettler Toledo GmbH, Switzerland). Before use, the biogenic CH4 was
purified from major contaminants, mainly CO2 and H2O, by flushing
it through an Ascarite/Mg(ClO4)2 trap. The δ13C and
δD-CH4 values of the reference gases CG 1 and CG 2, as well as
of a cylinder with pressurized air used as the target gas were calibrated
against the calibration scales of the Stable Isotope Laboratory of the
Max Planck Institute (MPI) for Biogeochemistry in Jena, Germany
(Sperlich et al., 2012, 2013; P. Sperlich, personal communication, 2016). It should be noted that the isotopic
composition of the measuring gas is outside the range covered by the
calibration gases CG1 and CG2 for δ13C and δD-CH4,
which may create problems for analytical techniques with a non-linear
response to isotope ratios. This, however, is assumed to be compensated by a
correction of results of all analytical techniques/laboratories for the
offset in the target gas between assigned value determined by MPI and
respective measured values. Results of all analytical techniques/laboratories were corrected for the offset in the target gas between
assigned value determined by MPI and respective measured values.
Schematics of the sampling setup used in the interlaboratory
comparison campaign. Ambient air was continuously sampled from the rooftop
of the building, and split from the main line to the batch sampling unit
(bags and flasks), to the TREX–QCLAS system and to the continuous flow CRDS
and OA–ICOS spectrometers. The laser spectrometers were additionally
supplied with the calibration gases CG 1, CG 2 and the target gas to
determine calibration factors and repeatability.
List of CH4 mole fractions and isotopic composition
(δ13C and δD-CH4) of laboratory standards used in the
intercomparison campaign. The indicated uncertainty is the 1 σ
standard deviation for repeated analysis of the respective measurement
system.
Composition
CH4[ppm]
δ13C-CH4 c [‰]
δD-CH4 c [‰]
CG 1
Fossil/biogenic CH4 in synthetic air
938.8 ± 3.5a
-46.60 ± 0.10
-250.46 ± 1.05
CG 2
Fossil CH4 in synthetic air
1103.8 ± 3.5a
-36.13 ± 0.10
-180.58 ± 1.05
TG
Pressurized ambient air
2.3523 ± 0.0002b
-48.07 ± 0.10
-120.00 ± 1.05
CH4 mole fractions were measured by CRDS a after dilution by a
factor of 1:500 or b by direct measurement.
c Isotopic values were analyzed by IRMS at MPI.
The CH4 mole fractions of CG 1 and CG 2 were analyzed with QCLAS
against commercial standards for CH4 mole fractions (1000 ± 20 ppm CH4 in synthetic air, Messer, Switzerland), while the target gas
was analyzed by WCC-Empa against the NOAA/GMD scale by CRDS (G1301, Picarro
Inc., USA). Table 1 summarizes the CH4 mole fractions and δ values of TG, CG 1 and CG 2.
A complete measurement cycle consist of three main sequences:
(a) three consecutive measurements of preconcentrated discrete ambient air
samples, (b) one measurement of preconcentrated pressurized air (target
gas), followed by the calibration phase (c). The latter is used for the
determination of calibration factors for δ13C-CH4 and
δD-CH4 and the dependence of isotope ratios on elevated
CH4 mole fractions. The calibration gases are dynamically diluted to
the indicated CH4 mole fractions as described in Sect. 2.2.2. All measurements are bracketed by the
analysis of CG 1 (anchor) at 635 ppm CH4 to drift-correct the
measurements.
TREX–QCLAS
During the intercomparison campaign a measurement cycle of 220 min duration
was applied (Fig. 5), including the measurement of
three different types of calibration gases (CG 1 at 635 and 745 ppm, CG 2 at
635 ppm) as well as repeatability measurements with preconcentrated target
gas (TG). This sequence allowed the measurement of up to 20 ambient air
samples per day.
Raw isotope ratio measurements were at first corrected for their dependence
on the laser frequency position followed by a drift correction based on
regular measurements of CG 1 at 635 ppm. Calibration factors for referencing
isotope ratios to the international standard scales as well as correction
factors to account for the dependence of isotope ratios on CH4 mole
fractions were determined by taking the mean of the calibration gas
measurements in intervals of 16 to 48 h and applying a linear regression
analysis. Note that the calibration gases were not preconcentrated, thus
violating the identical treatment principle. This was compensated, however,
by referencing the results to pressurized ambient air (TG) measurements.
The δ13C values of preconcentrated samples were corrected for a
2.3 ‰ offset, which was caused by an increase in O2
mole fractions to 40 ± 2 % during preconcentration as discussed in
Sect. 3.1.2. The δ13C-offset value
was shown to be constant for a large range of CH4 mole fractions and
the full range of δ-values covered by this study. For δD-CH4 no significant effect could be observed; most likely, its
magnitude was within the uncertainty of the system.
CH4 mole fractions in both ambient air and target gas were determined
based on the analysis of preconcentrated CH4 mole fractions
(12CH4), divided by the preconcentration factor. This factor was
computed for each cycle from the gas volume in the multipass cell and the
volume of preconcentrated air. The latter is derived from the sample gas
flow and the adsorption time. As the trap additionally adsorbs small amounts
of N2 and O2 (up to 4 % of the preconcentrated sample volume,
depending on the trap temperature), variations in the trap temperature also
need to be considered. Finally, the CH4 mole fraction measurements were
linked to the WMO-X2004 calibration scale
(Dlugokencky et al.,
2005) through calibration of the target gas against NOAA reference standards
at Empa.
Commercial laser spectrometers
During the campaign, an off-axis integrated cavity output spectrometer
(OA–ICOS, δ13C-CH4 and CH4 mole fraction,
MCIA-24e-EP, Los Gatos Research, USA) provided by Utrecht University (UU),
and a cavity ring-down spectrometer (CRDS, δ13C-CH4,
δ13C-CO2, CH4 and CO2 mole fraction, G2201-I,
Picarro Inc., USA) provided by Eawag, were deployed. The OA–ICOS analyzer
operated in the MIR spectral region, while the CRDS instrument comprises a
NIR laser source. OA–ICOS and the CRDS isotope analyzers were calibrated
twice per day using the calibration gases CG 1 and CG 2 (Table 1) for 30 min
each. These standards were diluted with high-purity synthetic air by a
factor of 1:500, to 1955.3 ± 6.8 ppb CH4, which is close to the
ambient mole fraction (Fig. 4). The dependencies of
δ- values on CH4 mole fraction were linear up to a
concentration of around 2500 ppb and determined to be -6.35 and 1.18 ‰ ppm-1
for OA–ICOS and CRDS, respectively. Variations over the duration of the
campaign were not significant and therefore a constant factor was applied.
Thereafter, for both analyzers a drift and a two-point calibration
correction for δ13C-CH4 was performed based on the
measurements of CG 1 and CG 2. Finally, 30 min averages of sample data were
calculated, resulting in 550 measurement points for the CRDS over the
2-week period of the intercomparison campaign. The repeatability of
OA–ICOS and CRDS for δ13C-CH4 was assessed based on
repeated analysis of the target gas (pressurized air) every 6 h for 30 min.
Bag and flask sampling
In addition to the in situ optical analyzers, manual sampling in glass
flasks and Tedlar bags for subsequent IRMS laboratory analysis was
performed. Glass flasks were purged for 10 min with dehumidified
(Mg(ClO4)2, Sigma-Aldrich, Switzerland) sample gas at 2 L min-1 using a
membrane pump (KNF, Netherlands) and then filled to 2000 hPa. Air samples collected in glass flasks were analyzed for δ13C-CH4, δD-CH4 and CH4 mole fraction at the
Institute for Marine and Atmospheric research Utrecht (IMAU) of Utrecht
University (UU) and a selection of flasks were also analyzed by the Stable
Isotope Laboratory of Max Planck Institute (MPI) for Biogeochemistry.
Parallel to the glass flask sampling and through the same sample line, 3 L
Tedlar bags (SKC Ltd., USA) were filled and subsequently analyzed for
δ13C-CH4 by IRMS and CH4 mole fraction by CRDS
(G1301, Picarro Inc., USA) at the Greenhouse Gas Laboratory, Department of
Earth Sciences (GGLES) of the Royal Holloway University of London (RHUL). In
total, 81 flask and 48 bag samples were taken at different intervals,
usually at least twice per day. Additionally, intensive sampling was
performed on 13 June and from 20 June 12:00 to 22 June 12:00 LT (local time),
when both flask and bag samples were filled every one to 3 h.
IRMS analysis of δ13C-CH4 and δD-CH4 in flask
samples at UU
Both δD and δ13C of CH4 were measured by
continuous flow IRMS (Thermo Finnigan Delta plus XL)
(Brass and Röckmann, 2010). First a 40 mL
stainless steel (SS) sample loop is filled with sample or reference air at
atmospheric pressure. The air is flushed by a flow of helium carrier gas
(purity 99.9999 %) to the preconcentration unit (1/8′′ SS tube filled with
6 cm HayeSep D 80–100 mesh) cooled to 137 K, where the CH4 is retained
and separated from the bulk air. The CH4 is released by heating the
adsorbent trap to 238 K and focused on the cryo-focus unit (25 cm PoraPLOT Q, 0.32 mm ID, 117 K). For δD analysis, the CH4 is injected (by
heating the cryo-focus trap to 198 K) into a pyrolysis tube furnace (1620 K), where CH4 is converted to H2 and carbon. The H2 enters
the IRMS, after passing a 2 m CarboPLOT column at room temperature (RT) and
a nafion dryer, via the GasBench interface. No krypton interference could be
determined in this setup. The repeatability for δD-CH4 is
better than ±2 ‰, based on 10 consecutive
analyses of standard air. A detailed inter-laboratory comparison between UU
and MPI is presently ongoing, and a preliminary scale offset of
4 ‰ has been used for the present evaluation.
For δ13C, the CH4 is injected from the cryo-focus unit
into a combustion oven with nickel wire catalyst at 1373 K, where the
CH4 is converted to CO2 and H2O. The resulting gas mixture
passes a nafion dryer and a 5 m PoraPLOT Q column (RT) to eliminate an
interference from co-trapped krypton (Schmitt et
al., 2013) before entering the IRMS via the GasBench interface. The
repeatability of δ13C is better than 0.07 ‰.
IRMS analysis of δ13C-CH4 and δD-CH4 in flask
samples at MPI
At the Stable Isotope Facility of MPI Jena (“BGC-IsoLab”) methane isotopes
from air samples have been analyzed using a new custom made twin-mass
spectrometer analysis system (Delta V+, Thermo-Fisher, Bremen, Germany)
with cryogenic preconcentration and GC separation (W. A. Brand, personal communication, 2016). The system allows analyzing δ13C and δD
simultaneously in an automated and fully calibrated fashion. For every air
sample, a reference air sample is analyzed concurrently. Only the difference
between the reference and sample air is used for calibration. While the ion
currents are analyzed on the same mass spectrometers, reference and sample
air pass through dedicated cryogenic acquisition lines. The isotopic
relation between these lines is established daily using four complete
analyses with reference air sent through the sample preconcentration duct.
Using small-volume flow controllers, cryogenic acquisition is made at 5 mL min-1 over 20 min, thereby consuming 100 mL air for each isotope
measurement. Prior to methane concentration, CO2 is removed
cryogenically using a permanent liquid nitrogen bath. The cryo traps for
methane retention consist of 1/8′′ stainless steel tubes filled with
HayeSep-D polymer for specific absorption of CH4 at 143 K. The latter
temperature is generated by compression coolers (Cryotiger, Brooks
Automation, Jena, Germany), which can operate down to 123 K at a heat
digestion capacity of around 30 Watt.
After acquisition, the acquired methane is transferred to a cryogenic focus
trap of similar design, from where gas chromatographic separation is
initiated by rapid heating. The methane peaks are heart cut
(Deans, 1968) for combustion (δ13C) and pyrolysis
(δD), respectively. CH4-derived CO2 is separated from
non-reacted CH4 and from the co-trapped krypton with a post-reaction
gas chromatographic separation before being introduced to the respective
mass spectrometer via open split coupling. An entire sample carousel with 18 analyses (13 sample analyses net) takes about 27 h.
The system is in continuous operation since July 2012. The overall precision
including all instrument failure times is ±0.15 ‰
(δ13C) and ±1.14 ‰ (δD), as
determined through daily measurement of a QA (quality assurance) sample air.
Removing the times of instrumental malfunction improves the precision to
±0.10 ‰ (δ13C) and ±1.05 ‰ (δD) over the entire period of operation
(3 years). The precision for 10 repeated measurements of standard air is
typically 0.07 ‰ (δ13C) and 0.7 ‰ (δD).
Allan variance plots for δ13C-CH4 (left)
and for δD-CH3 (right) using 750 ppm CH4. The upper plot
shows the corresponding time series of δ-values recorded at 1 second temporal resolution. At 600 s spectral averaging, the square root of
the Allan variance indicates a precision of 0.1 ‰ for
δ13C-CH4 and 0.5 ‰ for δD-CH4.
IRMS analysis of δ13C-CH4 in bag samples at
RHUL
δ13C-CH4 was measured using a modified GC-IRMS system (Trace
Gas and Isoprime, Isoprime Ltd.). This system uses a modified trace gas
preparation system in dynamic mode, whereby the original catalyst is replaced
by palladized quartz wool in a wider 4 mm ID ceramic furnace tube.
Conversion of methane to CO2 and H2O is completed at 1063 K using
oxygen in the air sample as the oxidant. A highly modified and automated
inlet system
(Fisher et
al., 2006) was applied consisting of an auto-sampler including a six-port
rotary valve (Valco Instruments Inc.) with a 75 cm3 Swagelok stainless
steel sample volume and four samples, one standard gas and a vacuum line
attached. The 75 cm3 sample volume is evacuated up to the solenoid
valve directly before the bag valve, then the air moves from the bag into
the sample volume maintaining ambient atmospheric pressure. This air is then
pushed through the preparation system with a flow of helium gas set to a
pressure of 758 hPa. Individual sample analysis lasts approximately 19 min
and all sample measurements were made in triplicate. Repeatability based on
10 consecutive analyses of standard air is ±0.05 ‰ or better. δ13C-CH4 values of
RHUL are offset corrected by -0.3 ‰ based on
intercomparison measurements with NIWA
(Lowe et al., 2004).