Development of automated preparation system for isotopocule analysis of N 2 O in various air samples

Nitrous oxide (N2O), an increasingly abundant greenhouse gas in the atmosphere, is the most important stratospheric ozone-depleting gas of this century. Natural abundance ratios of isotopocules of N2O, NNO molecules 10 substituted with stable isotopes of nitrogen and oxygen, are a promising index of various sources or production pathways of N2O and of its sink or decomposition pathways. Several automated methods have been reported to improve the analytical precision for the isotopocule ratio of atmospheric N2O and to reduce the labor necessary for complicated sample-preparation procedures related to mass spectrometric analysis. However, no method accommodates flask samples with limited volume or pressure. Here we present an automated pre-concentration system for gas samples of various amounts and various N2O 15 concentrations. The shortest processing time for a single analysis of typical atmospheric sample is 40 min. Precision values of isotopocule ratio analysis are < 0.1‰ for N (average abundances of NNO and NNO relative to NNO), < 0.2‰ for O (relative abundance of NNO), and < 0.5‰ for SP (difference between relative abundance of NNO and NNO). That precision is comparable to those of other automated systems, but better than that of our previously reported manual measurement system. 20


Introduction
Long-term monitoring of trace gases that are increasingly abundant in the atmosphere is fundamental for the analysis of the imbalance of their sources and sinks and for the prediction of future environmental change on Earth.Nitrous oxide (N 2 O) is one such trace gas, with global warming potential that is 220 times as great as that of carbon dioxide (CO 2 ), and is the most important stratospheric ozone-depleting gas of this century (Myhre et al., 2013;Ravishankara et al., 2009).Its globally averaged concentration, given as a mole fraction, was about 324 nmol mol −1 (10 −9 moles per mole of dry air) in 2011 (Hartmann et al., 2013) and increases by 0.73 nmol mol −1 a −1 (Ciais et al., 2013).Sources of N 2 O include natural and agricultural soils, aqueous environments such as oceans, rivers, and lakes, industrial processes such as fossil fuel combustion, biomass burning, and animal and human wastes (Ciais et al., 2013); its major sink is photochemical decomposition in the stratosphere.
Although concentration analyses yield quantitative information related to trace gases straightforwardly, it is often difficult to differentiate the sources contributing to the increase of such gases in the atmosphere, especially for N 2 O. Natural abundance ratios of stable isotopes of the elements that compose trace gas molecules have qualitative information related to the origin and production-decomposition processes of the gases because isotope ratios are generally different among different substrates.Moreover, they can change during physical, chemical, and biological processes.Regarding N 2 O, measurements of the nitrogen isotope ratio ( 15 N / 14 N) for the atmosphere and various sources since the 1980s have revealed that the imbalance of isotopically light N 2 O from surface sources and isotopically heavy N 2 O refluxed from the stratosphere after its partial decomposition causes a progressive decrease in the 15 N / 14 N isotope ratio of tropospheric N 2 O (Ishijima et al., 2007;Röckmann et al., 2003a;Sowers et al., 2002).Furthermore, a technique developed for measuring isotopomers of N 2 O ( 14 N 15 N 16 O and 15 N 14 N 16 O) expanded conventional isotopic analysis to isotopocule analysis by which ratios of NNO molecules substituted with stable isotopes of nitrogen or oxygen at any site relative to 14 N 14 N 16 O are obtained and by which production and decomposition pathways can be differentiated in greater detail (Toyoda et al., 2015 and references therein).
Compared to concentration analysis, stable isotope and isotopocule analyses require (1) larger sample amounts, (2) more time and labor to extract and purify the target compound from the sample, and (3) larger and more expensive apparatus.Although recently developed tunable diode laser absorption spectroscopy (TDLAS) relaxes some of the requirements above, and although it has some potential for onsite monitoring of stable isotope / isotopocule ratios of trace gases (Harris et al., 2014;Mohn et al., 2012;Tuzson et al., 2008), mass spectrometry combined with flask sampling still holds advantages for high-precision isotopic monitoring at polar regions or remote areas and flight observation using a balloon or an airplane because of smaller sample volume requirements.
In most currently used mass spectrometric analytical methods for N 2 O isotopocules, air samples are first passed through chemical adsorbents to remove CO 2 and water vapor.Then, N 2 O is concentrated on chemically inert adsorbents or inner walls of narrow tubes at liquid nitrogen temperatures.It is further purified on a capillary column of a gas chromatograph (GC) and is introduced directly into an isotope ratio monitoring mass spectrometer (IRMS).The analysis of a single sample takes 30-60 min.The precision reported in earlier studies is typically 0.1-0.5 ‰ for 1 nmol of N 2 O (e.g., Toyoda et al., 2001) (See Sect. 2.4 for notation of isotopocule ratios), which is worse than the ultimate precision expected from the shot-noise limit of the IRMS (Potter et al., 2013) and which is insufficient to resolve the secular trend of atmospheric N 2 O isotopocule ratios.This low precision is partly caused by incomplete separation of interfering components such as CO 2 and fluorinated hydrocarbons, or by imprecise manual handling during sample preparation.
To improve the precision of the isotopocule ratio analysis of atmospheric N 2 O and to reduce the labor for complicated sample preparation procedures for mass spectrometric analysis, several automated methods have been reported.Röckmann et al. (2003b) improved the precision of fragment ion (NO + ) analysis by modifying the gas chromatographic purification of N 2 O from interfering species such as halocarbons and less volatile compounds (Röckmann et al., 2003b).Röckmann and Levin (2005) and Potter et al. (2013) reported further improvement in the precision by partially or fully au-tomating sample preparation steps and by slightly increasing the sample size.
In addition to the mass spectrometric method, an automated sample preparation system has been reported, which can be coupled to a quantum cascade laser absorption spectrometer (QCLAS) for the monitoring of atmospheric isotopocules of N 2 O (Harris et al., 2014;Mohn et al., 2010).However, previously reported automated methods entail several shortcomings.For example, they are designed to measure pressurized air samples such as ambient air drawn by pumps or air collected into glass bottles or metal cylinders using pumps.For that reason, they are not applicable to samples at ambient or subambient pressure.
Here we present an automated preconcentration system that offers enhanced flexibility in terms of sample gas pressure and N 2 O concentration.The novel system encloses a vacuum line and a computer program that controls valves to inject samples of a designated amount.

Preparation system
The preparation system developed in this study consists of a sample injection unit, cryogenic concentration unit, purification unit, and cryofocusing unit (Fig. 1).It is placed in a steel rack (60 cm width, 80 cm depth, 150 cm height) with wheels attached, and is connected to a gas chromatograph-isotope ratio monitoring mass spectrometer (GC-IRMS).Details of each unit are presented below.
A sample flask (made of either glass or ss) or a gas cylinder is manually connected to one of the switching valve ports with an ss connector (Cajon Ultra-Torr or Swagelok; Swagelok Company, Solon, OH).The tubing between the diaphragm valve (V1, V2, or V3) and the flask/cylinder valve is evacuated by manually operating the valves and vacuum pump via control software (see below).Then, all the diaphragm valves are closed, the flask valve is opened by hand, and a computer program for sample preparation (see Sect. 2.5) is started.First, sample gas pressure in the flask is measured using the pressure gauge by expanding the sample gas into the vacuum line until V4.Based on the pressure and the volume of the flask and the sample size to be injected, the "sample expanding option" and final pressure of the sample injected into the vacuum line is calculated.Seven options exist for sample expansion into the calibrated volume in the vacuum line from 100 cm 3 (option no. 1) to 510 cm 3 (option no. 7).This expansion is realized by a combination of the three glass bottles (C, D, and E in Fig. 1) with different volumes.
Next, an aliquot of the sample in the flask is expanded by sequential open-close operation of diaphragm valves.The pressure is monitored using the manometer.When the pressure agrees with the precalculated value within ±5 %, either V1 or V3 (when sample is analyzed, Fig. 1) or V2 (when standard gas is analyzed) and V4 are closed, the pressure is recorded, and the injected sample amount is calculated.
First, the concentration trap is purged with ultra-pure He (> 99.9999 %, Japan Air Gases Ltd., Tokyo, Japan) at 100 cm 3 min −1 for > 10 s by switching SV2 to the "inject" position.The He is purified in advance through a column packed with molecular sieves 5A, active charcoal, and molecular sieves 13X in series (PC in Fig. 1).Next, SV2 is switched to the "load" position and the trap is cooled with liquid nitrogen in an ss dewar which is driven up and down by a custom-made air-actuated stage and which is filled with liquid nitrogen from an automatic liquid nitrogen supply system (Koshin Ltd., Tokyo, Japan).Then V5, V15, and valves relevant to the sample injection option (V9-V14) are opened.The sample gas in the calibrated volume is transferred to the concentration trap through the chemical trap by He carrier gas at 30 cm 3 min −1 .The He is purified in a similar manner to that described above.When more than two glass bottles are filled with the sample gas, the transfer is conducted sequentially.The transfer time is set so that the total volume of He which flows through the bottle is twice the bottle volume.
S. Toyoda and N. Yoshida: Development of automated preparation system for isotopocule analysis of N 2 O
Initially, SV3 and SV4 are set to the load position.After the sample concentration step is completed, SV2 is switched to the inject position.The concentration trap is heated to 70 • C by lowering the liquid nitrogen dewar and turning on a sheathed electric heater attached to the trap.The concentrated trace gases are transferred to the GC column with purified He at 20 cm 3 min −1 .When 2 min have passed after the GC injection, the cryofocusing trap is cooled with liquid nitrogen by moving up another ss dewar.Three minutes later, SV4 is switched to the inject position.Purified N 2 O from the GC is focused on the trap for 2 min.

Injection to GC-IRMS
After the cryofocusing step is completed, SV4 is switched to the load position, the liquid nitrogen dewar is moved down, and the cryofocus trap is heated to 70 • C similarly, as in the case of the concentration trap.The N 2 O is injected into another GC (GC6890; Agilent Technologies Inc., Santa Clara, CA, GC2 in Fig. 1) with He (2 cm 3 min −1 ).It is further purified with the GS Carbon PLOT column (0.32 mm inner diameter (i.d.), 3 µm film thickness, 30 m; Agilent Technologies Inc.) maintained at 35 • C. The purified N 2 O is finally into an IRMS (MAT252; Thermo Fisher Scientific K.K., Yokohama, Japan) via an interface that includes a gas dryer with a permeation tube and two open split interfaces for the sample and reference gas (GC-combustion interface; Thermo Fisher Scientific K.K., slightly modified).
Mass spectrometric analysis of N 2 O isotopocules is conducted as described elsewhere (Toyoda and Yoshida, 1999;Toyoda et al., 2015).Briefly, molecular (N 2 O + ) and fragment (NO + ) ions of N 2 O are analyzed in independent runs.Solving the following equations and applying correction for the rearrangement or scrambling reactions during fragmentation, the isotopocule ratios are obtained as delta values. (4) In Eqs. ( 1)-( 6), 45 R and 46 R respectively denote the measured ion-beam intensity ratios of m/z 45/44 and 46/44 in molecular ion analysis; 31 R shows a 31/30 ratio by fragment ion analysis; 15 R α , 15 R β , 17 R, and 18 R respectively denote the abundance of ions 14 N 15 N 16 O + , 15 N 14 N 16 O + , 14 N 14 N 17 O + , and 14 N 14 N 18 O + relative to 14 N 14 N 16 O + .In Eq. ( 4), A = 0.00937035 and γ = 0.516 (Kaiser et al., 2003).In Eq. ( 5), δ 15 N bulk denotes the average isotope ratios for 15 N / 14 N.The subscripts sample and standard respectively denote the isotope ratios for the sample and the standard.SP denotes site preference.International standards for N and O isotope ratios are atmospheric N 2 and Vienna Standard Mean Ocean Water (VSMOW), respectively.National Instruments Corp., Austin, TX) were used to activate the solenoid valves that regulate compressed air for airactuated shut-off valves and the air-actuated up-down stages, the multi/two-position switching valves, the vacuum pump system, the automatic liquid nitrogen supply system, and the temperature controller for the heaters.The PC also received analogue data from the pressure and vacuum gauges and from the manometer, received the TCD signal, and synchronized the GC-IRMS data acquisition with the end of the sample preparation procedure.The timing of each regulation function is presented in Fig. 2. The program developed in this study includes a special algorithm to adapt the sampling procedure to the prevailing sample pressure and N 2 O concentration.As described briefly in Sect.2.1, it has a user interface to obtain information related to the sample: the flask volume and the sample size to be injected.When the actual sample gas pressure is measured and obtained, it automatically determines the optimal procedure for sample injection.Figure 3 shows a flow chart for the algorithm.

Sample injection to the system
The time required for sample injection depends on the sample expanding option (see Sect. 2.1).It takes about 5 min when a 300 cm 3 aliquot of air, which contains ca. 4 nmol of N 2 O in the case of ambient air (ca.320 nmol mol −1 ), is injected from a 1 L flask pressurized to about 2.5 atm (option no. 7).When the flask volume or inner pressure is lower, more time is needed because the number of repetitive sample  diffusion steps increases and one valve was operated 5 s after actuating another valve to equilibrate the pressure in the inlet line and to avoid potential fractionation of isotopocules.When a smaller sample volume with high N 2 O concentration is measured, sample injection is completed in a minute or less.However, the performance of quantitative sample injection becomes poor for samples with more than 10 µmol mol −1 N 2 O because the system cannot fully adjust the introduction of a small amount of sample (< 10 cm 3 ).Moreover, the relative error of the pressure measurement increases for low pressure.Such highly concentrated samples are better introduced after dilution with N 2 or He.Modification of one glass bottle (e.g., bottle C) to enable manual injection with a microsyringe is also possible.

Concentration, purification, and cryofocusing of N 2 O
During cryogenic concentration of N 2 O, the flow rate and flowing time of the He carrier gas should be optimized carefully to ensure quantitative recovery of N 2 O and thus also minimize contamination of subsequent analyses (blank values).Our preliminary tests showed that each glass bottle (C, D, or E) is purged completely when the total volume of He is more than twice the bottle volume.This result indicates that laminar flow is predominant in the bottle.Turbulent flow, which is expected to cause exponential dilution and to result in the consumption of a larger amount of He to sweep out the initial sample gas, is negligible.
The main purpose of the purification step is separation of N 2 O from CO 2 and compounds that are less volatile than N 2 O. CO 2 is 1000 times more abundant than N 2 O in ambient air samples.Its isotopocules have the same mass as those of N 2 O. Therefore, it often interferes with mass spectrometric analysis of N 2 O molecular ions.We tested two column packing materials for this purpose: Porapak Q and silica gel (dimension of the column was identical to that of Porapak Q, 60-80 mesh; GL Sciences Inc., Tokyo, Japan).Although silica gel has the unique property of eluting N 2 O before CO 2 , their separation took longer than in Porapak Q.The separation was not complete, even at 50 • C with the flow rate of 15-50 cm 3 min −1 .We also strove to separate CO 2 and N 2 O without using chemical adsorbents, which revealed a condition under which CO 2 and N 2 O are separated almost completely in preliminary experiments using a thermal conductivity detector and a mixture of CO 2 and N 2 O in N 2 bath gas (mixing ratios of CO 2 and N 2 O were ca.250 µmol mol −1 ).However, a small CO 2 peak was observed on the m/z 44 chromatogram after separation by the second GC.Mass ratios 45 R and 46 R showed dependence on the area of CO 2 peak, which indicates that separation on the first column was insufficient for precise isotopocule ratio analysis of N 2 O. Therefore, we inserted the chemical CO 2 trap before cryogenic concentration.for i = 1 to 7

Volatile compounds such as halocarbons and hydrocarbons
Open V1, Read PG1(1) . Flow chart of the algorithm for sample injection.n denotes the sample size in cubic centimeters at 25 • C and 1 atm.n meas or n meas (j ) is the actual sample size calculated from measured pressure.v b is the partial volume (cm 3 ) of the vacuum line indicated by B in Fig. 1. v(i) is the calibrated volume that corresponds to the sample expanding option i. v S is sum of the volume of sample flask and partial volume A. PG1(j ) and PG2(j ) are outputs of pressure gauges 1 and 2 (in kPa) (Fig. 1).P calc (i) and PG2 calc (i) are calculated pressures that correspond to the sample expanding option i. have longer retention time than that of N 2 O on columns typically used for N 2 O analysis.Some of them are known to hamper the chromatography of successive runs caused by their very slow elution (Röckmann et al., 2003b).Similar to previous studies, such compounds were prevented from being transferred to the next step and were backflushed to vent by switching the flow path in the present system.The cryofocusing step was necessary to inject the N 2 O purified in the high-flow system to the low-flow capillary GC-IRMS system.To achieve a quantitative N 2 O recovery, the timing of the cryofocusing step was optimized to trap the eluent from the first column only while N 2 O was released.

Optimization of GC-IRMS analysis and measurement precision
We tested two fused-silica capillary columns for the separation of N 2 O from other constituents in the second GC, a porous polymer PLOT column (HP PLOT Q, 0.32 mm i.d., 20 µm film thickness, 30 m; Agilent Technologies Inc.) and a PLOT column with a monolithic carbon layer (GS Carbon PLOT).The latter column was found to have benefits for the separation of CO 2 , N 2 O, and other interfering compounds such as fluorinated hydrocarbons (Fig. 4).A shortcoming was that the retention time of N 2 O at the optimized condition became longer than that obtained with the porous polymer column, which was used in previous studies (Potter et al., 2013;Röckmann et al., 2003b).The degree of precision of the measurements was evaluated with the standard deviation of repeated analyses (n = 3) of synthetic air (349 nmol mol −1 N 2 O) pressurized in an aluminum cylinder that had been calibrated against the international isotopic standard and which was used as a working standard (Toyoda et al., 2013).As presented in Fig. 5, precision of δ 15 N bulk , δ 18 O, and SP values measured on a single day are typically better than 0.1, 0.2, and 0.5 ‰, respectively, when more than 4 nmol (which corresponds to about 300 cm 3 of the synthetic air) of N 2 O is injected.The peak area of major ions m/z 44 and 30 showed good linearity with respect to the sample size (data not shown).The N 2 O concentration was obtained by comparison of the peak area normalized to the specific sample size between the sample and the laboratory standard.The resulting precision of the concentration measurement is better than 0.5 % (coefficient of variation, n = 3).In addition, measured isotopocule ratios are independent of the sample size of 4-8 nmol.Results show that routine analyses of atmospheric air samples can be conducted with samples of 320 cm 3 so that measurements of each sample are sandwiched by those of the working standard (Table 2).
The performance of the developed system is presented along with that of previous works in Table 1.The precision and required sample size of this work is comparable to that of similar automated GC-IRMS systems.It takes 40 min for a single run, which means that a total of 80 min is necessary to obtain a single set of δ 15 N bulk , δ 18 O, and SP on some mass spectrometers that are incapable of simultaneous monitoring of five ions (m/z 44, 45, 46, 30, and 31).This might be a shortcoming of the present system, but it presents advantages in terms of flexibility of the sample pressure and sample size.a Obtained with 420 cm 3 air; b estimated from the reported precision for δ 15 N α , δ 15 N β , or δ 15 N bulk ; c not available or not described.

Conclusions
A fully automated sample preparation system was developed for the measurement of concentrations and isotopocule ratios of N 2 O in both pressurized and subatmospheric pressure samples.An ambient atmospheric sample of 320 cm 3 can be analyzed in 40 min with a precision of < 0.5 % (coefficient of variation) for concentration, < 0.1 ‰ (1 standard deviation) for δ 15 N bulk , < 0.2 ‰ for δ 18 O, and < 0.5 ‰ for 15 N site preference (SP).The system, not being limited to use for mass spectrometric analysis, can also be applied to concentration or isotopic analyses of other trace gases such as CO 2 and CH 4 by replacing the chemical trap, GC columns, and cryogenic concentration/focusing traps and by re-optimizing the temperature, flow rate, and flow switch conditions.Unlike previously reported systems, this system enables analysis of grab-sampled air samples that are collected into a pre-evacuated container at atmospheric pressure.This capability is particularly valuable when compressors or pumps cannot be used for sampling because of logistic reasons such as electric power or weight.

Figure 1 .
Figure1.Schematic portraying the sample preparation system developed in this study: CT, chemical trap; FM, flow monitor; GC, gas chromatograph; IRMS, isotope ratio mass spectrometer; LN, liquid nitrogen; MFC, mass flow controller; NV, needle valve; OS, open split interface; P, pressure regulator; PC, purification column; PG, pressure gauge; SV, electrically actuated switching valve; T, trap; V, air-actuated diaphragm valve; VG, vacuum gauge; VP, vacuum pump.A-E denote parts of the vacuum line or glass bottles that are used to expand the sample, the volume (cm 3 ) of which is also shown.

Figure 2 .
Figure 2. Time sequence of the sample preparation procedure.Horizontal arrows on the top indicate the periods for evacuation of the inlet line (a), sample injection (b), cryogenic concentration of N 2 O on T1 (c), purification of N 2 O by GC1 (d), cryofocusing of N 2 O on T2 (e), and injection of N 2 O into GC2 (f).See also Fig. 1.

Figure 4 .
Figure 4. Typical chromatogram of background air sample obtained in a fragment-ion (NO + ) monitoring run.The x axis shows elapsed time after the start of heating of the cryofocusing trap.After the peaks of reference N 2 O injected from GC-IRMS interface (no.1no.4), sample N 2 O peak appears (no.5).The peak elution about 100 s later is only detected on m/z 31 trace and is CF + derived from a fluorinated carbon species.

Figure 5 .
Figure 5. Precision of isotopocule ratio measurements as a function of sample size.

Table 1 .
Comparison of the analytical precision obtained in this study and those values reported in the literature.

Table 2 .
Example of measurement results conducted on a single day.
a Difference/2 is shown.