Recovery of Submilligram Quantities of Carbon Dioxide from Gas Streams by Molecular Sieve for Subsequent Determination of Isotopic ( 13C and 14C) Natural Abundances

Anal. Chem. 1992, 6 4 , 824-827 REFERENCES Tornes, J. Aa.; Johnsen, B. A. J. Chromatogr. 1888, 467, 129-138. Bossie, P. C.; Reutter, D. J.; Sarver, E. W. J. Chromatogr. 1987, 407, Bossle, P. C.; Martin, J. I.; Sarver, E. W.; Sommer, H. Z. J. Chroma- togr. 1983, 267, 209-212. Verweij, A.; Degenhardt, C. E. A. M.; Boter, H. L. Chemosohere 1979, Verweij. A.; Boter, H. L.; Degenhardt, C. E. A. M. Science 1979, 204, Hahey, D. J.; Horning, M. G. J. Chromatogr. 1873, 79, 65. Purdon, J. G.; Pagotto, J. G.: Miller, R. K. Report DREO-936; Defence Research Establishment: Ottawa, 1985. Andersson, G.; Fredriksson, S. A,; Lindberg, G., Olofsson, B, FOA 4 Report; Forsvatels Forskningsantstalt Huvudavdeining 4, S901 82 llmen - . . . __, Sweden Tripathi, D. N.; Kaushik, M. P.; Bhattacharya. A. J.-Can. Soc. For- ensic Sci. 1987, 20 (4), 151-153. Report on identification of potential organophosphorus warfare agents; Ministry for Foreign Affairs of Finland: Helsinki, 1979 [ISBN 951-46- (1 1) Yao, C. C. D.; Zlatkis, A. Chromatographia 1987, 23 (9, 370-72. Durgesh N. Tripathi* Karuna S. Pandey Arabinda Bhattacharya Ramamoorthy Vaidyanathaswamy Analytical Services Wing Defence Research & Development Establishment T~~~~~ Road Gwalior-474 002 (MP), India RECEIVED for review August 23, 1991. Accepted January 6, TECHNICAL NOTES Recovery of Submilligram Quantities of Carbon Dioxide from Gas Streams by Molecular Sieve for Subsequent Determination of Isotopic ( 13C and 14C) Natural Abundances James E. Bauer* Department of Oceanography, Florida State University, Tallahassee, Florida 32306 Peter M. Williams Scripps Institution of Oceanography, Marine Research Division, University of California, S a n Diego, La Jolla, California 92093-0218 Ellen R. M. Druffel Woods Hole Oceanographic Institution, Department of Chemistry, Woods Hole, Massachusetts 02543 INTRODUCTION Quantitative recovery and isotopic analysis of COP generated from organic and inorganic carbon is important for a variety of natural sample types. For example, the two mast commonly used techniques for the oxidation of organic carbon to COz are the sealed-tube method' and various modifications of the flow-through combustion method., While both methods ul- timately depend upon the presence of 0, for the oxidation process, the sealed-tube method has been the method of choice for the subsequent recovery and isotopic analysis of COP This is due to the relative ease with which COz (sublimation point = -78.5 C at 760 mmHg pressure) can be cryogenically re- moved from an otherwise noncondensible, static (i.e,, non- flowing) gas mixture under vacuum. This concept has been improved upon through the use of variable-temperature cryogenic traps (VTTS)~** which allow gas components having different melting points to be sequentially frozen out and distilled off for collection of essentially pure product. In contrast, COz contained in carrier gas streams is generally difficult to collect at atmospheric pressure following organic carbon combustion since this usually entails using 0, as both the oxidant and the carrier gas. Cryogenic collection of C 0 2 at liquid nitrogen temperature (bp -195.8 C) from an 0, stream results in the condensation of liquid 0, (bp -183.OoC), a potentially hazardous situation, and possibly in incomplete C02 recovery. Pumping the liquified 0, away under vacuum, or trapping the CO, from the 0, stream at temperatures below * Corresponding author. the boiling point of 0, using VWs, can also result in incom- plete recovery and isotopic fractionation of CO,. In fact, even when N, is used as the carrier gas at atmospheric pressure and liquid nitrogen temperature, intermittent condensation of the N2 stream may O C C U ~ though this is much less severe than for 0,. Finally, systems which are designed to collect CO, from gas streams (especially 0,) cryogenically or using partial vacuums can attain considerable complexity. A method is therefore needed which allows CO, to be quantitatively recovered from 0, and N2 gas streams at at- mospheric pressure. For isotope studies, a further requirement is that no fractionation of the carbon isotopes in the CO, occurs. This led us to develop a simple, convenient method which is based on the highly selective absorptive properties of aluminosilicate molecular sieves for different gases. While molecular sieves have been employed to a limited extent in some commercial organic carbon analyzers, in some cases for 613C determinations (e.g., see ref 61, a thorough assessment of the analytical capabilities of sieves has, to our knowledge, not been made. The method presented here has a low asso- ciated blank (<0.02 pmol of C), is robust over a wide range of 0, and N2 carrier gas flows (at least 100-2000 cm3.min-l) and quantities of CO, (at least 0.3-50 wmol of C), and is nonfractionating with respect to carbon isotopic (13C and 14C) natural abundances. Molecular sieves should have a variety of applications for C 0 2 quantification and isotopic analysis. EXPERIMENTAL SECTION The specific molecular sieve used in these studies was a syn- thetic sodium aluminosilicate zeolite having the molecular formula 0 1992 American Chemical Society


INTRODUCTION
Quantitative recovery and isotopic analysis of COP generated from organic and inorganic carbon is important for a variety of natural sample types. For example, the two mast commonly used techniques for the oxidation of organic carbon to COz are the sealed-tube method' and various modifications of the flow-through combustion method., While both methods ultimately depend upon the presence of 0, for the oxidation process, the sealed-tube method has been the method of choice for the subsequent recovery and isotopic analysis of COP This is due to the relative ease with which COz (sublimation point = -78.5 "C at 760 mmHg pressure) can be cryogenically removed from an otherwise noncondensible, static (i.e,, nonflowing) gas mixture under vacuum. This concept has been improved upon through the use of variable-temperature cryogenic traps (VTTS)~** which allow gas components having different melting points to be sequentially frozen out and distilled off for collection of essentially pure product. In contrast, COz contained in carrier gas streams is generally difficult to collect at atmospheric pressure following organic carbon combustion since this usually entails using 0, as both the oxidant and the carrier gas. Cryogenic collection of C02 at liquid nitrogen temperature (bp -195.8 "C) from an 0, stream results in the condensation of liquid 0, (bp -183.OoC), a potentially hazardous situation, and possibly in incomplete C02 recovery. Pumping the liquified 0, away under vacuum, or trapping the CO, from the 0, stream at temperatures below * Corresponding author. the boiling point of 0, using VWs, can also result in incomplete recovery and isotopic fractionation of CO,. In fact, even when N, is used as the carrier gas at atmospheric pressure and liquid nitrogen temperature, intermittent condensation of the N2 stream may O C C U~,~ though this is much less severe than for 0,. Finally, systems which are designed to collect CO, from gas streams (especially 0,) cryogenically or using partial vacuums can attain considerable complexity.
A method is therefore needed which allows CO, to be quantitatively recovered from 0, and N2 gas streams at atmospheric pressure. For isotope studies, a further requirement is that no fractionation of the carbon isotopes in the CO, occurs. This led us to develop a simple, convenient method which is based on the highly selective absorptive properties of aluminosilicate molecular sieves for different gases. While molecular sieves have been employed to a limited extent in some commercial organic carbon analyzers, in some cases for 613C determinations (e.g., see ref 61, a thorough assessment of the analytical capabilities of sieves has, to our knowledge, not been made. The method presented here has a low associated blank (<0.02 pmol of C), is robust over a wide range of 0, and N2 carrier gas flows (at least 100-2000 cm3.min-l) and quantities of CO, (at least 0.3-50 wmol of C), and is nonfractionating with respect to carbon isotopic (13C and 14C) natural abundances. Molecular sieves should have a variety of applications for C02 quantification and isotopic analysis.

EXPERIMENTAL SECTION
The specific molecular sieve used in these studies was a synthetic sodium aluminosilicate zeolite having the molecular formula ANALYTICAL CHEMISTRY, VOL. 64, NO. 7, APRIL 1, 1992 825 THERMOCOUPLE CLEAN 0 2 I N2 Flguro 1. Schematlc drawing of the gas flow/vacuum line system used In the present study for evaluation of Type 13X molecular sieve. The flow of gases In the system Is from left to right. Na20A1203Si02 (Type 13X, 1/16-in. pellets, Linde Division, Union Carbide Corporation, Danbury, CT). The sieve aperture is 10 A in diameter, while the molecular diameters of C02, 02, and N2 are 3.34,2.98, and 3.15 A, respectively. The number of molecules of a specific compound which are retained in the sieve cavities via the sieve apertures is strongly dependent on both the molecular diameter and the temperature of the sieve.' Molecular sieves are selective for specific gas molecules not only on the basis of pore size but also on the basis of molecular polarity and boiling point. In the case of Type 13X, the Na+-containing sieve is strongly polar, resulting in a preferential retention of COP over other, less polar gases (e.g., O2 and N2). It is likewise imperative that molecules which could potentially displace C02 (e.g., H20) be prevented from entering the sieve. Type 13X is commonly used in industrial applications for the removal of C02 gas and H20 vapor from hydrocarbon gas mixtures.
To evaluate the molecular sieve for quantitative collection and release of C02, as well as the isotopic fidelity of the trapped and released COP, a sieve trap was interfaced to a gas flow/vacuum extraction line. The components of the system in which Type 13X was used are shown in Figure 1. The quartz U-tube (dimensions: height, -150 mm; i.d., 9 mm; o.d., 12 mm) contains approximately 12 g of sieve which is immobilized by plugging the tube with prebaked quartz wool. The U-tube has sealing O-ring stopcocks at each end which allow it to be removed from the main line as needed without exposing the sieve to air. The sieve should be exposed to air for a minimum of time when transferring it to the U-tube in order that absorption and adsorption of atmospheric gases and contaminanta are minimized. Water vapor is the most deleterious of these components, and it may reduce the efficiency of the sieve if exposure is chronic (G. Rand, personal communication). When the unit is attached to other instrumentation a trap of magnesium perchlorate upstream from the sieve trap has been found to be adequate for eliminating the H20 before it reaches the molecular sieve. Care should also be taken to keep the sieve free of organic contaminants during handling because of the potential for their slow oxidation at elevated temperatures (see below).
When a new batch of sieve is used it should be conditioned for at least 24 h by opening the U-tube to full vacuum (l@ atm) while heating at 550 OC with a heating jacket (Glas-Col, Terre Haute, IN). This temperature is well below that which purportedly causes degradation of the zeolite (-700 "C; G. Rand, personal communication), but it is high enough to eliminate absorbed C02 and H20 from the sieve and to volatilize and combust any initially adsorbed organic Contaminants. In operation, the sieve is allowed to equilibrate at room temperature (22-25 OC) before C02 collection. At least temperatures, O2 and N2 molecules pass freely through the sieve apertures and occupy very little of the sieve's total available cavity space.7 Since the sieve will normally be at full vacuum following the analysis of a prior sample, it is then brought to atmospheric pressure by closing the system to vacuum and then introducing ultra-high-purity O2 or N2 (400 cm3-min-') into the U-tube by slowly opening the upstream stopcock A. The gas is then allowed to flow through the remainder of the system and to vent by opening stopcocks B-E. At this point, if the sieve is to be used for collecting C02 from a remote instrument, it may be isolated by closing stopcocks A and B, and flowmeter 1 and the U-tube are removed as a unit from the flow/vacuum line. For purposes of this study, however, it was maintained intact on the flow/vacuum line.
The recovery of C02 from O2 and N2 streams was evaluated as follows. The carrier gas was adjusted to the desired flow rate with the molecular sieve trap maintained at room temperature. A quantity of pure C02 of known isotopic composition was injected through heavy-walled latex rubber tubing using a precision microliter syringe (Micrometrics, Shorewood, IL). The total time for all injections was 2.0 min. A period of 5 min was allowed to elapse after injection, and flow was then stopped by closing stopcocks A and E. The entire system was then brought to full vacuum. Horibe trap 1 was immersed in a bath consisting of dry ice in 2-propanol (-78.5 "C) to serve as a water trap. Horibe trap 2 was immersed in liquid nitrogen to serve as the C02 trap. The Horibe traps were employed here because their design, with increased dead volume for greater residence time and glass frits for increased surface area, allow for quantitative collection of C02 even under full dynamic vacuum? The stopcocks to vacuum were then closed so that the system was under static vacuum. Preliminary h t s of the retention efficiency of the sieve demonstrated that no C02 leaked past the U-tube following injection and none desorbed from the sieve while at room temperature under full dynamic vacuum for at least 1 h. The heating jacket was then placed around the U-tube, and the C02 was transferred to Horibe trap 2. Following transfer, Horibe trap 2 was evacuated again, and its contents were cryogenically transferred to the measured volume under static vacuum. The volume of gas was measured using a Baratron absolute pressure transducer (MKS Instruments, Andover, MA) which had previously been calibrated with known COP gas standards.
For 613C determinations the above procedure was followed, and C02 standard gas samples were cryogenically collected in 6-mm glass tubing which was flame-sealed. The 613C of the samples was measured with a Nuclide 6-60RMS mass spectrometer. For AI4C determinations, oceanic humic matter of known A1% was dissolved in ultra-high-purity artificial seawater and combusted to C02 using a high-temperature total organic carbon analyzer. The unit was interfaced to the sieve U-tube which trapped the C02 from the O2 gas stream. The C02 was purified and collected as above for subsequent A14C analysis by tandem accelerator mass spectrometry.

RESULTS AND DISCUSSION
A wide range of carrier gas flow rates and quantities of C02 was used to assess the conditions under which Type 13X sieve could be used. In addition, various operating parameters (desorption temperature, desorption time) were also evaluated in order to develop an optimized protocol for routine use of the sieve. Initial teats using O2 and N2 as carrier gases revealed that recoveries of C02 were identical using these two gases.  Subsequently, O2 was used as the carrier gas for further experimental evaluation both because it is normally used for flow-through organic carbon combustion and because it is more problematic in cryogenic applications. The system blank using Type 13X was insignificant (always C0.02 pmol of C), and hence sample values required no correction. The CO, was trapped with 100% efficiency by Type 13X molecular sieve for quantities of C 0 2 ranging over 2 orders of magnitude (0.3-50 pmol, ? = 1.O00, n = 12). The trapping efficiency of CO, gas was independent of the flow rate of 0, through the sieve and was also 100% in all cases (Table I). This demonstrates the extremely high affinity of Type 13X for C02 since the residence time of gas in the U-tube was less than 1 s at the greatest flow rate used here (2000 cm3.min-'). The equivalent range of C02 concentrations over all tests was 2-130 pmol per liter of carrier gas. The sieve retained 100% of the CO, (10 pmol!, even when followed by prolonged periods (3 h) of continuous 0, flow (Table I). The CO, was also retained with 100% efficiency when it was introduced to the sieve after the sieve was subjected to prolonged periods of continuous 0, flow (Table I). Therefore, applications which require extended periods of time to collect adequate amounts of CO, for a given analysis are assured of quantitiative recoveries.
The time dependence of C02 release from Type 13X sieve under vacuum was evaluated a t three temperatures ( Figure  2). At room temperature (25 "C), prolonged (1 h) application of full vacuum to the sieve trap resulted in no loss of absorbed COP At 200 OC, 100% release took up to 45 min. However, a t 425 OC, complete release was achieved within 5 min of heating. For most applications, rapid release is preferable in order to avoid potential isotopic fractionations resulting from the kinetic effects of slow release. We routinely heat the sieve rapidly to 425 "C by maintaining the heating jacket at this temperature, even when it is not around the U-tube. Although the sieve can be heated to -700 "C, its structure may begin  The lowest blanks are attained when the sieve is kept at 425 "C, especially when it is not being used for extended periods of time.
The stable isotope ratio of COz trapped and released by the sieve was identical to that of the stock COz (Table 11). Triplicate injections of 10 pmol of C 0 2 at an O2 flow rate of 150 cm3.min-' yielded 613C = -35.8 f 0.1% while the known value of stock C 0 2 was 613C = -35.9 f 0.2% indicating that fractionation of COS upon passage through the sieve did not occur. Although stable isotope ratios were measured for 10pmol samples of CO, only, accurate ratios are expected for lesser and greater quantities of C02 and at different flow rates since recoveries were 100% in all cases. However, for very small or very large amounts of COz the accuracy of isotopic values should be checked against similarly sized known gas standards which have been processed through the sieve trap. When the sieve was interfaced to a high-temperature organic carbon analyzer with O2 as the carrier gas, the Al4C signatures of the C 0 2 (8-10 pmol) derived from combustion of oceanic humic substances were indistinguishable from the known signatures determined by sealed-tube combustion (Table 11). Thus, both 613C and A14C may be accurately determined by collection of C02 from the gas stream using the sieve trap. It should also be noted that for these short-term collections the associated blank was insignificant.
The simplicity of the trap and the range of conditions over which Type 13X molecular sieve can be used provide several advantages over systems which use cryogenic collection of COP The sieve can be used at atmospheric pressure and at the rmm temperature of most laboratories to collect CO, from either 0, or N2 (and possibly other gases) carrier gas streams. The upper limits examined in this study were 2000 cm3.min-' flow rate and 50 pmol of CO,. These ranges may well be greater in the system used here but in any case could be expanded by using collection tubes which hold a greater volume and mass of sieve or by using multiple traps arrayed in series. Under very rigorous flow conditions or extremes in amounts of C 0 2 collected, retention efficiencies, 89 well as stable isotope ratios, should be checked periodically to ensure that the capacity of the sieve is not exceeded. We have used a single batch of Type 13X in a U-type continuously for over 12 months with no signs of performance deterioration, resulting in significant savings in both cost and maintenence over cryogenic collection systems.

ACKNOWLEDGMENT
We thank D. Des Marais for comments on the manuscript, A. Witter for help in preliminary tests, D. Des Marais and A.

INTRODUCTION
Advances in analytical instrumentation have led to new designs and application of on-line, real-time process monitoring,l+ including methods which rely on chromatographic2 or spectroscopic With the advent of less costly, compact, and less complex mass spectrometer^,^ increasing efforts have been directed at developing on-line mass spectrometric technique^,'"'^ because they offer both qualitative and quantitative information about the chemical process with high sensitivity. Mass spectrometry is particularly suitable for sampling gaseous streams. Many applications of on-line gas chromatography-mass spectrometry interfaces'O have been reported. Additionally, direct gas sampling may be achieved by use of a pulsed electromagnetic valve connected to a mass spectrometer."J2 The use of a pulsed valve with a Fourier transform ion cyclotron resonance mass spectrometer" or a quadrupole ion trap mass spectrometer'* has proven effective for intermittent introduction of volatile components. More recently, several examples of mass spectrometric liquid sampling have been demonstrated. Most rely on atmospheric pressure ionization (API) sources13 or membrane separa-tor~'"'~ to accommodate the potentially high gas load when a liquid is sampled into a vacuum. Unfortunately, API sources are mechanically complex and involve additional stages of differential pumping. Although the membrane-based devices are versatile and less complex, they exhibit long or variable response times due to slow diffusion of the analyte through the membrane.
In some cases, a more direct liquid-sampling system is preferred to a membrane interface, especially if quantitative sample transfer or faster response times are needed. Microliter quantities of liquid solutions can be introduced into a mass spectrometer via a conventional direct insertion probe,m but this technique involves use of a vacuum interlock and is not suitable for on-line continuous-monitoring applications. In this report, we demonstrate for the first time direct on-line monitoring of gas or liquid samples by using an inverse sampling valve coupled to a mass spectrometer.
Standard versus Inverse Liquid-Sampling Valves. Standard Liquid Sampling. Liquid-sampling valves such as those made by A13B21 or MATz2 operate on the principle of metering microliter quantities of a liquid into a carrier gas stream and then vaporizing the liquid rapidly and completely into a gas chromatograph (GC). To our knowledge, these liquid-sampling valves have never been used to transfer samples directly to mass spectrometers. Figure 1 is a cross-sectional view of a portion of a MAT valve attached to a gas chromatograph. Liquid sample from the process stream is directed to the valve through port 1. The sample fills the grooved slot, 2, of the piston, 3, and is transferred to the heatad vaporization zone, 4, upon actuation of the piston. The seals, 5, ensure that only sample in the slot (-1 pL) is transferred. Hot carrier gas, 6, flashes the sample and directs it into the GC column, 7.
Inverse Sampling. Inverse sampling23 is a recent modification of the liquid-sampling valve, wherein the heated vaporization section is removed and the remainder of the valve is welded to the vessel holding the material to be transferred. For our purposes, this vessel is a short welded section of 2-in. pipe, which servea as the sample chamber. The valve is further modified to expose one of its seals into the sample chamber.
The liquid-transfer section of the standard valve becomes an evacuation chamber or vaporization section in the modified valve. Finally, the mode of operation is opposite to the standard liquid sample; i.e., the inject position becomes the load position and vice versa. The sampling piston captures sample from the chamber or process stream into its l14-pL groove while in the extended (load) position and transfers it in the retracted (inject) position. Transference takes place by evacuation of the sample-filled groove into the analytical instrument, with or without a supplementary carrier gas stream. When the sample is a liquid, the "vaporization zone" must be heated and a carrier gas is recommended. These added steps are unnecessary for gas samples.
The advantage of inverse sampling is that the valve can be attached directly to the process, obviating the need for divert streams or pumps. This sampler can be used for transference of any vaporizable substance and is particularly useful for molten streams which are ordinarily difficult to sample.

EXPERIMENTAL SECTION
Two spectrometers were used, a Finnigan quadrupole ion trap mass spectrometer (ITMS)" located at the University of Texas for gas sampling and a Finnigan 4600 quadrupole mass spectrometer located at Dow U.S.A. for liquid sampling. Pentachloropyridine (PCP) was obtained from Dow Chemical. All other