Anal. Chem. 1995,67,2461-2473
Performance and Optimization of a Combustion Interface for Isotope Ratio Monitoring Gas Chromatography/Mass Spectrometry Dawn A. MeMtt,t Katherine H. Freeman,* Margaret P. Ricci, Stephen A. Studley, and J. M. Hayes* Biogeochemical Laboratories, Departments of Chemistry and Geological Sciences, Indiana University, Bloomington, Indiana 47405
Conditions and systems for on-line combustion of effluents from capillary gas chromatographic columns and for removal of water vapor from product streams were tested. Organic carbon in gas chromatographicpeaks 15 s wide and containing up to 30 nmol of carbon was quantitatively converted to C02 by tubular combustion reactors, 200 x 0.5 mm, packed with CuO or NiO. No auxiliary source of 0 2 was required because oxygen was supplied by metal oxides. Spontaneous degradation of CuO limited the life of CuO reactors at T > 850 "C. Since NiO does not spontaneously degrade, its use might be favored, but Ni-bound carbon phases form and lead to inaccurate isotopic results at T < 1050 "C if gas-phase 0 2 is not added. For all compounds tested except CH4, equivalent isotopic results are provided by CuO at 850 "C, NiO + 0 2 (gas-phasemole fraction, at 1050 "C, and NiO at 1150 "C. The combustion interface did not contribute additional analytical uncertainty, thus observed standard deviations of 13C/12C ratios were within a factor of 2 of shot-noise limits. For combustion and isotopic analyses of CH4, in which quantitative combustion required T x 950 "C, NiO-based systems are preferred, and precision is -2 times lower than that observed for other analytes. Water must be removed from the gas stream transmitted to the mass spectrometer or else protonation of C02 will lead to inaccuracy in isotopic analyses. Although thresholds for this effect vary between mass spectrometers, differential permeation of H20 through Nafion tubing was effective in both cases tested, but the required length of the Nafion membrane was 4 times greater for the more sensitive mass spectrometer. In the late 1970s, two research groups independently reported methods using on-line combustion to facilitate precise determinations of the isotopic compositions of gas chromatographic effluents.'S2 To provide a descriptive name, Matthews and Hayes referred to the technique as isotope ratio monitoring gas chromatography/mass spectrometry and suggested the abbreviation irmGCMS.2 Improvements in the design and construction of components in the combustion unit, use of highly efficient, low-
' Present address: Environmental Chemistry Laboratory, DowElanco, Indianapolis, IN 462681053, Present address: Department of Geosciences, Pennsylvania State University, University Park, PA 16802-2711. (1) Sano. M.; Yotsui, Y.;Abe, H.; Sasaki, S. Biomed. Mass Spectrom. 1976,3, 1-3. (2) Matthews, D. E.; Hayes, J. M. Anal. Chem. 1978,50,1465-1473.
*
0003-2700195/0367-2461$9.00/0 0 1995 American Chemical Society
bleed columns, and most of all, use of highly efficient isotope ratio mass spectrometershave improved the performance of irmGCMS so that its accuracy and precision approach those of conventional isotopic In fact, irmGCMS is superior for analyses of individual compounds which are separated and combusted online. Sample requirements are often 1000 times smaller, and the risk of contamination or isotopic fractionation of samples is reduced, substantially improving accuracy. Despite the popularity of the technique and the number of successful application^,^-" the construction and operation of irmGCMS systems has been described only b r i e t l ~ . ~The - ~ design and construction of the combustion interface, optimum conditions for its operation, recognition of factors affecting analytical performance, and comparisonsbetween practical and theoretical levels of performance have not been considered. To begin a systematic examination of these issues, a recent paper focused on methods for acquisition and processing of data.12 A subsequent report focused on mass spectroscopic considerations,13namely instrumental stability and sensitivity, approaches to theoretical maximum levels of precision, and effects of background subtraction procedures. A third report considered methods for introduction of isotopic standards.14 Here, we examine the combustion interface itself, describing first the design, construction, and durability of individual interface components. Factors affecting the extent of combustion and water removal efficiency are then examined and compared, and optimal operating conditions are recommended. Finally, the analytical performance and limitations of irmGCMS are described for analyses of individual compounds over a range of sample sizes. Two different isotope ratio mass spectrometers are used, one with (3) Barrie, A; Bricout, J.; Koziet. J. Biomed. Enuiron. Mass Spectrom. 1984, 11, 583-588. (4) Freedman, P. A; Gillyon, E. C. P.; Jumeau, E. J. Am. Lab. 1988,20,114119. (5) Hayes, J. M.; Freeman, K. H.; Popp, B. N.; Hoham. C. H. Adu. 0%.Geochem. 1990, 16, 1115-1128. (6) Freeman, K. H.; Hayes, J. M.; Trendel, J.; Albrecht, P. Nature 1990,343, 254-256. (7) Jasper, J. P.; Hayes, J. M. Nature 1990,347, 462-464. (8) Rieley, G.; Collier, R. J.; Jones, D. M.: Eglinton, G.; Eakin. P. A,; Fallick, A. E. Nature 1991,352,425-427. (9) Silfer, J. A; Engel, M. H.; Macko, S. A,: Jumeau, E. J. Anal. Chem. 1991, 63, 370-374. (10) Kohnen, M. E. L.; Schouten, S.; Sinninghe Damste, J. S.; de Leeuw, J. W.; Memtt, D. A; Hayes, J. M. Science 1992, 256, 358-362. (11) Goodman, K. J.; Brenna, J. T. Anal. Chem. 1992, 64, 1088-1095. (12) Ricci, M. P.; Merritt, D. A.; Freeman, K. H.; Hayes, J. M. Org. Geochem. 1994,21,561-571. (13) Merritt, D. A; Hayes, J. M. Anal. Chem. 1994,66, 2336-2347. (14) Merritt, D. A; Brand, W. A,; Hayes, J. M. Org. Geochem. 1994,21, 573583.
Analytical Chemistry, Vol. 67, No. 74,July 75, 7995 2461
vent
mm
Water Separator
Open Split Ion source
carrier stream
'i
mixing
Pump
Volume
variable-volume inlet Figure 1. Schematic view of the combustion interface. The positions of all valves and the volumetric flow rates of gas streams entering and leaving the interface during various operations are summarized in Table 1. a conventional ion source operating at an accelerating potential of 3 kV, and the other with a high-sensitivity source operating at an accelerating potential of 10 kV. EXPERIMENTAL SECTION (I) Instnunentation. A schematic of the flow pathway for the effluent stream in irmGCMS is shown in Figure 1. The design and construction of individual components are described below. Gas Chromatograph. The system utilized either a Series 5890A (Hewlett Packard, Avondale, PA) or a Model 3400 (Varian, Sugarland, nr)gas chromatograph. Samples were injected using either an on-column injector (Hewlett Packard) or a 1-pLinternal volume injection valve (Valco Instruments, Houston, TX). The downstream end of the chromatographiccolumn was connected to two transfer capillaries by use of a two-hole graphitized vespel ferrule on the downstream side of a microvolume union (Scienti6c Glass Engineering (SGE), Austin, TX). The first was connected to a vent controlled by an air-actuated 'T'-valve (valve A, SGE), and the second entered the combustion reactor. Combustion Reactor. The microvolume reactor consisted of a nonporous alumina tube (Bolt Technical Ceramics, Conroe, Tx; 99.7%Al203,0.5 mm i.d., 1.5 or 3.0 mm o.d., 30 cm long) packed with metal oxide. Reactors were prepared by inserting three equal lengths of copper wire plus one length of platinum wire (Aesar/Johnson Matthey, Ward Hill, MA; 20 cm long, 0.1 mm diameter, puratronic grade) into the tube so that the wires were centered end-to-end within the tube. The wires were inserted carefully in order to avoid twisting, a point discussed below in the section on the design and construction of combustion reactors. Reactors were also constructed using nickel instead of copper wire (Aesar/Johnson Matthey; puratronic grade, similar dimensions). Transfer capillaries (SGE deactivated vitreous silica capillaries, 0.32 mm id.) were inserted -1 cm into each end of the reactor and held in place by stainless steel reducing unions with graphite ferrules (SGE). The reactor was placed inside a resistively heated quartz furnace (%mm-i.d.,2km-long, 2ggauge Chrome1 A wire, 25Q cold resistance), and the furnace temperature was regulated to within k 5 "C (Model 49 temperature controller; Omega Engineering, Stamford, CT). After assembly of the reactor, the copper or 2462
Analytical Chemistry, Vol. 67, No. 14, July 15, 7995
nickel wires were oxidized in place by heating at 500 "C for 2-5 h while 02 was passed through the reactor tube. Reactors were then heated overnight at the operating temperature (850 "C for CuO or 1050-1150 "C for NiO) while 02 was passed through the reactor. Reactors were also reoxidized periodically, with the frequency of reoxidation dependent upon factors discussed below. Water Separator. Water of combustion was removed from the effluent stream by use of a tubular Nation membrane.l5,l6The Nafion tube (Permapure, Tom River, NJ; 0.6 mm i.d., 0.8 mm o.d., 12-42 cm long) was mounted coaxially inside a stainless steel or Pyrex tube (4 mm i.d., 6 mm o.d., 10-40 cm long) using reducing 'T'-unions (SGE l/4-1/l&. union). The annular volume was continuously purged by a countercurrent stream of dry helium or dry nitrogen (2-5 mL/min). The stream of dry gas was adjusted using the flow controller designated as 2 in Figure 1 (Porter Instruments, Hatfield, PA). The capillary exiting the reactor entered one end of the Nation tube, while a silica transfer capillary exited the tube. These capillaries (0.32 mm i.d, 20-50 cm long) were inserted -2 cm into each end of the Nafion tube and held in place by radial compression with 0&mm graphite ferrules, thus forming leak-tight seals. Open Split. The capillary exiting the Nation tube was inserted into an open split that was constructed using a length of stainless steel, Pyrex, or quartz tubing (5-20 cm long, 4 or 0.8 mm id., 6 mm 0.d.) connected to l/{-l/1& reducing 'T-unions (SGE). The capillary exiting the water separator was centered inside the tube. The outside diameter of the capillary leak which carried the effluent stream to the ion source (75 cm to 2.0 m long, 0.11 mm i.d.) was small enough that it could be inserted inside the capillary leading from the Nation tube. Both capillaries were held in place by the 'T-unions using 0.5" graphitized vespel ferrules. A constant stream of dry helium (2 mL/min) purged the annular volume of the open split and exited through a vent line controlled by an air-actuated valve (vent B). The rate of admission of gas to the mass spectrometer was controlled by the length of the 0.11-mmi.d. capillary. At its low-pressure end, the capillary leak entered an isolation valve through a l/d-I/dn. (15) Foulger, B. E.; Simmonds, P. G . Anal. Chem. 1979,51, 1089. (16)van Benschoten, J. J. M.S. Thesis, Department of Chemistry, Indiana University, Bloomington, IN, 1985.
Table 1. Positions of Valves and Flow Rates of Gas Streams during Different Modes of Operation
flow rates of gas streams (mL/min)
mode
entering interface exiting valve0 through flow controllers interfaceb A B 1 2 3 A B M S
normal/sampling C 0 backnush o c reoxidation C 0
0 0 2-5
2-5 2-5 2-5
2 2 2
2.8 0.2 0 0.2 2.8 0 0 5-8
0
a Valve position: C, closed; 0, open. Volumetric flow rates were measured when the flow rate of carrier gas exiting the GC column was 1.0 mL/min and the flow rate of the effluent stream entering the ion source was 0.2 mL/min (MAT 252). The flow rate of the effluent stream entering the Delta-Swas 0.6 mL/min, which modifies the entries in the table. Additionally, tabulated values will vary as a function of the rate of the carrier gas stream.
Swagelok reducing union (graphitized vespel ferrule). The length of the capillary leak was determined experimentally such that the indicated pressure in the ion source was 4-7 pTorr and the volumetric flow rate of carrier gas entering the mass spectrometer was 0.2-0.3 mL/min (1.5-2.2 m, 10-kV instrument) or 0.5-0.7 mL/min (0.75-1.0 m, 3-kV instrument). Mass Spectrometer. Carbon dioxide produced by combustion of chromatographic effluents continuously entered the ion source of the isotope ratio mass spectrometer by way of the 0.11mm capillary leak. Two different mass spectrometers were used in these experiments (Delta S, 3-kV accelerating potential, and MAT 252, l@kVaccelerating potential; Finnigan MAT, Bremen, Germany). Ions were generated by electron impact (70 ev), and the source and analyzer regions of the instruments were differentially pumped. Ion currents were measured continuously for m/z 44,45, and 46 using triple Faraday cups connected to highspeed amplifiers. Electronic offsets were also applied to the collectors by adding small, constant currents to the summing points of the electrometers. The approximate magnitudes of these offsets were 90-110,110-130, and 140-160 mV, respectively, for m/z 44,45, and 46. To provide baseline ratios near those of COz with natural isotopic abundances, the offsets were trimmed such that ratios of amplified baseline signals were 1.16-1.20 and 1.351.40 for m/z 45-44 and 46-44, re~pectively.'~ (II How ) Pathways. Pathways of gas streams were controlled using computercontrolled, air-actuated valves, designated A and B in Figure 1. The state of each valve and typical flow rates of gas streams in various modes of operation are summarized in Table 1. Normal/Sampling. In this mode, valve A was closed, vent B was open to the atmosphere, and an auxiliary stream of helium from flow controller 3, Figure 1, passed through the open split. As a result, the entire effluent stream flowed through the combustion reactor. If the flow rate was greater than that accepted by the mass spectrometer, only a portion entered the ion source (0.2 and 0.5 mL/min respectively for the 1@ and 3-kV instruments), while the remainder of the effluent and auxiliary helium streams exited through vent B. If the flow rate of the effluent stream was less than that required by the mass spectrometer, the deficit was supplied by the helium stream purging the open split. For some experiments, a trickling flow of oxygen (1%Oz in helium, 0.1 mL/min) was added at the splitting union immediately upstream from the combustion reactor.
Bacldlush. At any time during a run, the effluent stream could be diverted from the combustion interface to prevent solvent or nonessential sample components from entering the reactor. In this mode, vent A was opened to the atmosphere, while vent B was closed. Consequently, a portion of the auxiliary helium stream flowing through the open split entered the ion source, while the entire effluent stream and the remainder of the helium stream exited through vent A Oxidation. Oxygen could be introduced to the reactor by closing valve A and opening B while 02 was added to the carrier stream downstream from the column. The flow of 02 was adjusted using flow controller 1. The valve between the interface and the mass spectrometer must be closed during this operation to prevent oxygen from entering the ion source. To enhance retention of oxygen by Cu or Ni, the temperature of the reactor could be decreased to 400-500 "C during reoxidation. (111) Experimental Procedures. Samples. A mixture of a homologous series of n-alkanes (n-C17-?&5) plus pristane (2,6,10,14tetramethylpentadecane)and phytane (2,6,10,14tetramethylhexadecane) was used as a test sample. Concentrations were -20 (3-kV instrument) or 5 (10-kV instrument) nmol of C/pL per component. Methane and propane were diluted on-line either by connecting sample and helium streams to a mixing volume and adjusting the relative flow rates or by using an exponential dilution flask (EDF) .I7 Associated procedures were described previously.13 The isotopic compositions of all test materials were also determined using conventional techniques.'E Introduction of Samples and Standards. Ultra-1and Ultra-2 columns (Hewlett Packard; 50 m x 0.32 mm x 0.52 pm or 50 m x 0.2 mm x 0.11 pm) or Wr,l and XTI-5 columns (Restek, Bellefonte, PA; 60 m x 0.32 mm x 0.5 pm) were used. Injection volumes rarely exceeded 0.5 pL. Methane and propane were injected via a 1-pL internal volume injection valve (Valco Instruments, Houston, TX) onto a porous layer open-tubular (PLOT) column (Chrompack, Ratitan, NJ; Poraplot Q, 25 m x 0.32 mm x 10 pm). Methane and propane were continuously diluted in helium in the EDF so that a 1-pL aliquot contained from 40 nmol to 10.1 pmol of COZ. Several methods were used to introduce isotopic standards and have been described previously.MJ4 In the first method, a mixture containing five perdeuterated n-alkanes was co-injected with sample mixtures (25 or 5 nmol of C/pL of n-Cl6D34, CZOD~Z, C24DN, C32D66, and C36H74; MSD Isotopes, St. Louis, MO). The perdeuterated alkanes were selected as co-injectable standards because they did not coelute with naturally occuring n-alkanes but could be readily identified by their place in hydrocarbon chromatograms. In the second method, COZwas introduced directly into the ion source by introducing pure COZfrom the variable-volumeinlet or by admitting a second gas stream containing COZinto the ion source through a second capillary leak (2-m-long, 0.060-mm-i.d. deactivated vitreous silica; throughput, 0.02-0.03 mL/min) . (IV) Calculations and Presentation of Results. Notation. According to standard practice, all results are expressed in terms of 6 values: 3
13
dl3Cp~,E 10 [ ( Rsample/13Rp~B) - 11
(1)
where 13R= 13C/W and PDB refers to the Pee Dee Belemnite (17) Lovelock, J. E. Nature 1971, 230, 379-380. (18) Frazer, J. W.; Crawford, R W. Mikrochim. Acta 1963, 561.
Analyfical Chemistty, Vol. 67,No. 14, July 15, 1995
2463
primary standard. Procedures for calculation of 6 values under irmGCMS conditions have been described by Rkci et a1.12 Since only carbon isotopic data are considered in this report, 6 l 3 C pis~ ~ abbreviated here as 6 unless otherwise necessary for clarity. Results are also reported in terms of A values to express the accuracy of isotopic measurements:
A
6,
- 6,
700-
2! ?!0,
.
c 3 .
-j
0
500-
E
0 * O
300 0
100-
"
'
I
"
'
I
'
1
Length, c m
(3)
Data Collection and Processing. Events in data collection and data processing have been described in detail elsewhere12 but can be summarized briefly. The mass spectrometer was controlled by a computer with a 20-MHz 80386 microprocessor, and data were acquired under software control. Ion currents at masses 44,45, and 46 were recorded continuously by voltage-tofrequency conversion and high-precision counting. For each chromatogram, complete ion current records were stored. Files were evaluated to distinguish samplerelated signals from background. Since isotopically substituted species can be resolved chromatographically, subtle shifts in chromatographic retention times were taken into account by correcting integration times for individual ion current signals. For all peaks, total ion currents at m/z 44-46 were integrated, and background signals that were defined automatically by measuring the signal at the five points immediately preceding each peak were subtracted. Ion current ratios were converted to isotopic abundances using procedures that neutralized effects of drifts in mass spectrometric characteristics. RESULTS AND DISCUSSION
(I) Design and Performance of Interface Components. Combustion Reactors. Design and Construction. Reactors were constructed using quartz or alumina tubing with an inner diameter of 0.45-0.55 mm. The internal volume of a typical reactor was thus -50-70 yL, and the included gas volume was no more than 40-60 yL, since at least 10yL was occupied by metal oxide. The metal oxides chosen served both as oxygen donors and as catalysts mediating complete combustion. Reactors were operated in two different modes: (i) with an auxiliary flow of 02 mixed with the chromatographic effluent at the entrance to the reactor or (ii) with no auxiliary flow of 02, so that all oxygen required for combustion was supplied by thermal dissociation of the metal oxide (details are discussed below). The form of the reactor was chosen to promote laminar flow. However, even if the reactor acted as a mixing chamber, the resultant band broadening would amount to only 4%for peaks with a width of 16 s, assuming an effective flow rate of 2.6 mL/ min at an average temperature of 500 "C (for fully laminar flow under the same conditions, the expected broadening would be 0.03%). Dimensions and materials were chosen to satisfy several additional requirements: (i) the tube must be nonporous and resistant to chemical attack and breakage, (ii) the ends of the Analytical Chemistry, Vol. 67, No. 14, July 75, 1995
Furnace ( l a length m)
0
0.
where the subscripts m and e refer to measured and expected (or "true") values of 6. To summarize the quality (combining both accuracy and precision) of a set of n measurements of 6 values, we report "rms A", the root-mean-square A:
2464
1l
9
I-
rms A = ( x A 2 / n ) 0 , 5
I
900 1
0
10
20
30
time, hr
Figure 2. (a) Temperature profile for a typical reactor when the temperature at the center of the furnace was controlled at 800 "C. Squares indicate temperatures inside the reactor when the GC oven was set to 60 "C,whereas filled diamonds indicate the case in which the temperature of the GC oven was 320 "C. (b) Ion current at m/z 32 as a function of time in the effluent streams of heated CuO reactors. The 0 2 represented by the ion current derives from the thermal dissociation of CuO. For these observations, the 3-kV instrument was used, and 60% of a 1 mUmin gas stream was transmitted to the mass spectrometer.
tube must extend at least 3 cm outside of the heated zone (>800 "C), (iii) the tube must be long enough so that the first 4-5 cm of the reactor can be positioned inside the chromatographic oven, and (iv) if the furnace was to be operated without auxiliary 02, the volume of included oxide must be great enough to supply 02 levels and catalytically active surfaces adequate for complete combustion of effluents, including column bleed, for at least 12 h prior to reoxidation. Requirement ii prevented exposure of the compression fittings to temperatures greater than 400-450 "C. Requirement iii ensured that effluents were not exposed to cold spots from the time they exited the column until they were converted to CO2. As demonstrated in Figure 2a, both requirements are met-the temperature at the ends of the reactor did not exceed 450 "C, and the temperature at the inlet of the reactor was always at least 25-30 "C greater than the temperature inside the chromatographic oven. Because experience suggested 10 cm as the m i n i u m length for the combustion zone, these constraints dictate a minimum length of 16-18 cm. For convenience, reactor lengths approaching 30 cm (metal oxide beds of 20 cm) were used. The first reactors were constructed using quartz tubes packed with oxidized copper wire. Quartz tubing, although fragile, served
well when reactor temperatures did not exceed 800 "C. However, since quartz and cupric oxide form a eutectic phase at temperatures greater than 850 "C, reactors failed quickly when exposed to higher temperatures. Reactors constructed using nonporous alumina (AlzO3) were more durable, easier to install, had less variable inner and outer diameters, and could be used at higher temperatures (TI1300 "C). Attempts to increase furnace performance by use of greater quantities of metal oxide failed. Reactors containing more than four wires (one Pt plus three Cu or Ni) inevitably developed restrictions to flow after only a few oxidation cycles. This problem probably arose because CuO and NiO occupy a greater volume than the unoxidized metal wires. Moreover, the wires break during repeated oxidation cycles, and migration of particles can lead to flow restrictions. Flow restrictions can also develop if wires are twisted inside the reactor. Such problems are not often encountered when fewer than four Cu or Ni wires were used and when wires were loaded carefully into reactor tubes. Durability. Oxygen is released within the reactors by a chemical equilibrium,
2MO -- 2M + 0, where M represents either Cu or Ni. For the system involving Cu, traces of CUZO,CuO, and OZ,I9 log PO, (TOIT) = 12.8026 - 13686.3/T
(4)
where T is the absolute temperature, in kelvin. At 850 "C, therefore, the equilibrium partial pressure of 02 in a combustion reactor containing CuO will be -5 Torr; at 950 "C, the equilibrium p 0 ~will be -50 Torr. Given the connection to the mass spectrometer, the conversion of cupric oxide to copper metal and the release of oxygen can be monitored by observing ion currents at m/z 32. This is demonstrated in Figure 2b for a CuO reactor operatingwithout any auxiliary flow of 02. Not shown is an initial burst of oxygen released during the first 20-30 min following the return to higher temperatures after oxidation; ion currents at m/z 32 were >30 nA during this period. This burst likely represented sorbed 02. Subsequently, 02 produced from thermal dissociation of CuO predominated. Assuming a sensitivity equal to that for C0213and correcting for the split ratio, the initial rates of oxygen release after the desorption period corresponded to 0.58, 2.9, and 4.0 nmol of OZ/Sfor reactors operated at 800, 825, and 850 "C, respectively. At these rates, a reactor containing 56 mg of CuO (the amount corresponding to three fully oxidized Cu wires, 0.1 mm x 20 cm) would lose its entire oxygen supply within 160,32, and 23 h, respectively, assuming a constant flux of oxygen. Results summarized in Figure 2b roughly confirm these estimates-baseline levels of 02 (0.25 nA) were obtained within >30 and -26 h for reactors operated at 825 and 850 "C, respectively. From these same considerations and observations, it follows that, at >825 "C, thermal decomposition is more important than combustion as a sink for 02. This comparison was made by analyzing hydrocarbon mixtures sequentially during a l&h period (19) Tatievskaja, E. P.; Cufarov, G. I. Izu. Akad. Nauk SSSR Otdel. Tech. 1946, 1005-1013; cited in Gmelins Handbuch de Anorgankchen Chemie; Verlag Chemie: Weinheim, 1958; Vol. 60,p 31.
(seven analyses). No auxiliary flow of 02 was employed. Column bleed during temperature-programmedanalyses (60-320 "C at 3 "C/min, hold at 320 "Cfor 30 min) produced a maximum signal of 0.25 V at m/z 44 using the 3-kV instrument. Multiplying by 1.7 to correct for the split ratio prevailing at those conditions and assuming an efficiency of 3 x ion~/molecule,~~ this corresponds to a COn production rate of 0.5 nmol/s. Assuming that the COZ derives from the combustion of SiO(CH3)2,the corresponding 02 demand was 1 nmol/s during the 30-min, hightemperature holds, and the total amount of 02 consumed during seven such intervals was 13 pmol. Additionally, for n-alkanes eluting during these analyses, a total of 6 (nmol of C/component) x 19 (components/ijection) x 7 (injections) = 0.8 pmol of COZ was produced, corresponding to an 02 demand of 1.6 pmol. Together, these sample and column bleed requirements (the latter, notably, much the larger) do not approach the observed, integrated 02 loss of 170 pmol (2.9 nmol/s x 57 600 s) due to thermal decomposition of CuO. It is, therefore, not surprising that reoxidation requirements were generally independent of sample loads but strongly dependent on temperature. For NiO, no 02 was produced from thermal decomposition even when reactors were heated to > 1300 "C. Specifically, m/z 32 ion currents did not increase as reactor temperature was increased. Additionally, partially used reactors, even when fully effective in the combustion of organic material and operated at >1150 "C, were observed to operate as 02 sinks when 02 was mixed with the gas stream entering the reactor. Therefore, lifetimes of NiO reactors were limited mainly by the amount of organic material to which they were exposed. Under most conditions, their effective lifetimes significantly exceeded those of CuO reactors. Regeneration of capacity by flushing spent reactors with 02 for 4-6 h at 500 "C was highly effective. When reactors containing CuO were operated at ~ 8 0 0"C, reoxidation was required at 3day intervals, even when reactors were not exposed to organic material. During most irmGCMS analyses, where reactors were typically operated at 850 "C, CuO reactors were reoxidized daily. For NiO reactors not exposed to high-temperature column bleed (see Characteristicsof NiO Reactors, below), reoxidation was required at most once or twice per week when the interface was operated for 8-16 h/day, regardless of reactor temperature. When CuO reactors were not reoxidized regularly or if too much organic material passed through the reactor, regions of bare copper (visible in quartz reactors) were formed. In some cases, sample components were partially pyrolyzed at the head of the reactor, and combustion was incomplete. This was typically indicated in ion-current or ratio traces by excessive peak tailing, by secondary or ghost peaks, and by unusual variations in ion current ratios in which the commonly observed positive and negative deflections were followed by a second positive deflection. Calculated contents of 13Cwere, in such cases, systematically in error. Values of S were too high by as much as 5Oh, presumably due to contributions from multicarbon species to the m/z 44-46 multiplet. The m/z 45/44 ratio in propane, for example, is 3 times larger than that for COz, and admixture of traces of C3Hs would quickly shift computed values of &I4 In certain cases, nonvolatile pyrolysates accumulated and were observed as blackened or discolored surfaces in the first few centimeters of the reactor. Often, this material was removed during reoxidation, but the Analytical Chemistry, Vol. 67, No. 14, July 15, 1995
2465
process appeared to interfere with regeneration of CuO. Copper surfaces exposed by removal of pyrolysates occasionally were converted to CUZOrather than CuO. When sufficient quantities of CUZOformed (evident as brilliant-red or reddish-purpleregions), reactor efficiency was significantly affected, regeneration was not effective, and replacement was necessary. For nickel reactors, formation of multiple oxides was not a problem, and when auxiliary flows of 02 adjusted to deliver 02 at the equilibrium partial pressure characteristic of CuO at the operating temperature were employed, particularly stable operation was obtained with no requirement for reoxidation of the reactor bed. The efficiency and lifetime of the reactors were degraded by exposure of the reactors to elements that formed salts with copper or nickel. This parallels observations made in studies of the water-gas shift reaction?O in which ppm levels of chlorine and sulfur were found to poison copper- or nickel-based reactor systems. When organic materials contain these elements, accumulation of involatile products can be observed visually (quartz reactors), and the effective lifetime of the reactor is shortened. For example, replacement was required after only 15 analyses of mixtures containing long-chain alkanethiols. Water Separator. Trace amounts of water are generated from the combustion of chromatographic effluents. If HzO remains in the carrier stream, COZin the ion source can be protonated to form H12C160~t and H13C1602+, species that are indistinguishable from isotopically substituted forms of COZin isotope ratio mass spectrometers. Due to this enhancement of signals at m/z 45 and 46, computed values of 6 will be systematically in error. The seriousness of this effect will depend on the COZ-HZOcollision frequency, and that, in turn, will depend on lifetimes of charged and neutral particles in the ion source as well as on concentrations of COZand HzO in the gas stream. The required completeness of water removal is thus dependent on ion source characteristics. Design and Construction. Use of a cold trap for removal of water from the effluent was avoided after it was observed2that a cold trap at -120 "C and atmospheric pressure broadened and delayed COz peaks. Instead, water was removed from the effluent by diffusion through a Nafion membrane. The membrane itself was tubular and formed a portion of the gas flow pathway downstream from the combustion reactor. A small diameter (0.5 mm) was used to minimize broadening of chromatographic zones. The Nafion tube was mounted coaxially inside a Pyrex or stainless steel tube where a countercurrent flow of dry gas maintained a near-zero partial pressure of HzO in the annular volume. Ideally, HzO should pass through the wall of the Nafion tube easily, while COZ instead stays quantitatively within the tube. Additionally, any interaction of the COZ with the surface of the Nafion should not be great enough to degrade chromatographic resolution or to lead to memory effects. In preliminary tests in which an exceptional length was chosen deliberately in order to exaggerate any nonideal effects, a Nafion tube 2.9 m in length removed HzO as efficiently as equilibration with Mg(C10&, transmitted 96 f 2%of the COZ, did not measurably increase the width of 24s chromatographic peaks, and did not alter the carbon isotopic composition of COZ that passed along its length.15J6 Therefore, for the lengths of Nafion tubing employed in irmGCMS systems (10-40 cm), 56 nmol of C/compound, reactor capacity was significantly higher, and reactor capacity did not deteriorate as quickly as a function of time. Different minimumtemperatures were required for combustion of different types of organic material. For example, while n-alkanes (10 nmol of C) were combusted quantitatively at 750 "C, similar quantities of aromatic compounds (represented by phenylhexane and 1,2:5,6dibenzanthracene) required higher temperatures. Data provided in Table 2 indicate that both compounds were combusted quantitatively at T > 775 "C since mean values of A were near zero, but reproducibility improved until temperatures reached 875 "C. Reproducibility again decreased at 900 "C, probably due to loss of oxidative capacity due to thermal decomposition of CuO. Recommended Operating Temperatures. Experiments indicate an optimal reactor temperature of 825-850 "C for most compounds combusted in CuO reactors, although this represents a compromise between reactor preservation and combustion efficiency. For methane analyses, minimum combustion temperatures were 900-950 "C (CuO), and, in the absence of auxiliary flows of 0 2 , consequent reactor lifetimes were 2-3 h. Oxygen Source for Combustion. In experiments aimed at development of improved reactors, n-alkane test mixtures were (21) Memtt, D.A: Hayes, J. M. I. Am. SOC.Mass Spectrom. 1994,5, 387-397.
analyzed under three sets of conditions: (i) no metal oxide present in reactors (02 was supplied by an auxiliary gas stream, and Pt wire was present as a catalyst); (ii) reactors contained metal oxide as the only oxygen donor, NiO at 1150 "C being examined for comparison with CuO; and (iii) reactors contained metal oxide and supplementary 02. Results are summarized in Table 3. For tests in which 02 was the exclusive oxygen source, a stream of 1%02 in He was mixed with the carrier stream (1 mL/ min each) at the inlet of the reactor. The resulting mole fraction of 0 2 in the reactor gas, -0.005, approximated that in equilibrium with CuO at 850 "C, and the flux, 6.7 nmol of O Z / S ,would theoretically serve for combustion of about 3.5 nmol organic of C/s. These trials used the 3-kV instrument. Due to the lower sensitivity of the 3-kV instrument, the sample size chosen was 5 nmol of C/component. The average chromatographic peak width was 15 s, so maximal fluxes were well below the level expected to draw down completely the supply of 02. As shown in Table 3, systematic errors were absent, but the precision was poor, rms A values at least 2-fold better being commonly observed with the 3-kV instrument. In spite of the narrowness of the chromatographic zones, the peaks of C02 produced by the reactor were >20 s in width and tailed badly. Moreover, peak areas were variable and 50-75% smaller than expected. The absence of systematic errors in such circumstances is unique to this data set. Commonly, shifts of more than 1% are observed in cases of incomplete combu~tion.~~ When NiO served as the oxygen source, inaccurate results were obtained when NiO reactors were operated without supplemental 0 2 at temperatures below 1050 "C. Accuracy and precision both improved as the reactor temperature was increased from 1050 to 1150 "C. Best results were obtained when metal oxides and 0 2 were both present (NiO/Oz or CuO, 850 "C). In order to provide a supplemental flux of 0.67 nmol /s to NiO reactors, 0.1 mL/min of 1%02 in He was added to the effluent stream (1 mL/ min) at the inlet of the reactor. The resulting mole fraction, 9 x approximates that in a CuO reactor operated at 780 "C. However, as shown in Table 2, the rms A value for a CuO system at 800 "C (0.20??)is considerably poorer than that characteristic of the NiO/Oz system at 1050 "C (Table 3,0.09??),indicating that temperature of combustion can play an important role in controlling the precision of irmGCMS analyses. Additionally, comparison between NiO reactors operated with and without supplemental 0 2 at 1050 "C indicates that the presence of traces of gas-phase 02 can be important. These issues are considered in more detail below. Characteristics of NiO Reactors. Pyrolysis, rather than combustion, was often observed when NiO reactors were operated Analytical Chemistty, Vol. 67, No. 14, July 15, 1995
2469
1.5
1.4
oc m *
1.3
"
I
2000
I
3000
4000
5000
6000
7000
Time, sec Flgure 6. A chromatogram demonstrating the failure of an NiO reactor operated without an auxiliary source of 02 at 850 "C.The dip in the m/z45 signal and the corresponding distortion of the 45/44 ratio trace indicate the onset of pyrolytic-as opposed to oxidative-processes when 0 2 demands increased as a result of inputs of column bleed. The reactor temperature was increased to 1000 "C at 6000 s. As the temperature stabilized at 1000 "C (6200-6250 s),the oxidizing capacity of the reactor was restored, and yields of CO2 improved.
without supplemental 02 at temperatures below 900 "C. The severity of this problem was dependent on rates of input of organic carbon. For example, when reactors were operated at 850 "C in the absence of 02,pyrolysis began to occur when levels of column bleed entering the reactor became significant (7.7 nA for m/z 44, corresponding to 0.36 nmol of U s ) . Effects of pyrolysis appear dramatically in Figure 6, which shows a fall-off of combustion efficiency together with marked distortion of the ratio trace (due to contributions by pyrolysis products to the m/z 45 ion current) at a retention time of 5200 s. When the reactor temperature was reset to lo00 "C (at t = 6000 s, Figure 6), ion current and ratio traces gradually returned to normal levels. Pyrolysis also occurred when NiO reactors were operated between 900 and lo00 "C, although its effects were not apparent during inspection of ion current or ratio traces. The problem was instead indicated by systematic disagreement between 6 values based on comparison to external COZstandards and those based on co-injected organic standard^.'^ When external COZstandards were used, A values for all compounds were shifted to more positive values, with the magnitude of the shift increasing almost linearly as a function of background COz from 1.5% (0.3 nA) to 25% (5.3 nA). Attribution of the positive shift to contributions by pyrolysis products to the m/z 45 beam is based on several points: (i) the offset was not observed when co-injected organic compounds were used as isotopic standards because the ion current ratios of those standards were also shifted by contributions from their pyrolysis products, and the effects roughly canceled, (i) comparison of the COz standardswith the cc-injected standards confirmed that ion current ratios for the latter were shifted, and (ii) the effect was eradicated by increasing reactor temperatures to >lo50 "C. 2470 Analyfical Chemistry, Vol. 67, No. 14, July 15, 1995
When NiO reactors were operated between lo00 and 1050 "C, the first few isotopic analyses obtained using a new reactor were often wildly in error, being offset by as much as 500h. Peak areas associated with these results were typically 5-20% smaller than expected, and 6 values were shifted to more negative values. The magnitude of this shift was extremely variable and was exaggerated when a smaller amount of material passed through the reactor. Systematic errors decreased rapidly as more C passed through the reactor, a nearly constant offset of -0.5% being observed by the 5th to loth compound that entered the reactor (1 nmol of C/compound). Successive analyses demonstrated that the shift in A continued to decrease gradually so that it was less than -0.1% by the 7th injection. These observations imply that stable Ni-C compounds (possibly nickel carbonyls) formed in reactors when organic materials were combusted at temperatures below 1050 "C and that Ni-I3C bonds formed, or once formed, survived preferentially. To test this hypothesis, the temperature of a reactor that had been operated at lower temperatures was increased to 1050 'C to determine whether nickel-bound carbon could be released. Large amounts of COz and CO (>35.7 nA at m/z 44-46 and 28-30) were released during the first 5 min following the temperature increase, and ion currents settled to the baseline only after the reactor was heated for 45-60 min. No additional COz and CO was released from the reactor after this period even when the temperature was increased from 1050 to 1300 "C, and, when hydrocarbon mixtures were analyzed at these temperatures, values of 6 were not shifted to more negative values, even for the first few compounds passing through a new reactor. On balance, particular care is required when using NiO reactors. The formation of nickel-bound carbon and the possibility
of incomplete combustion can be avoided by adding a trickling stream of 02 to the reactor during analyses. For example, when 0.67 nmol of OZ/S was added, the reactor temperature could be lowered to 1O00-1050 "C without causing systematic errors when isotopic calibration was based upon external COZ. At 1050 "C, the admixture of 02 improved rms A values from 0.22 to O.W? (Table 3). Repeated exposure to high-temperaturecolumn bleed sometimes caused irreversible loss of capacity for absorption and subsequent donation of 02. When such exposure could not be avoided, CuO + 02 yielded more reliable long-term performance. For analyses of 15N and CH4,21,22 the advantages of NiO are of great practical value. (111) Overall Performance. At the shot-noise limit, precision and sample size are related as follows:13
+ R)'/EmNAR
ad2= 2 x 106(1
0.5
,
l
I
a ........
-0.5 0.5
I I
~
-
Q-0............
00 -0.5
-
-1.0
-
I
I
._u-.....
b -
e-
........... 9.....................
0'0
...0..... .................. .........
0...
----I_________
0
0 0 '
-1.5
9."
'
.........
0.5
I
I
I
-
c .
0
....-..
-..
-a%..-._ ..........
........
0
. ....,.
Computed values of US represent the performance levels that would be observed in systems completely free of noise associated with signal processing or sample handling. Real instruments can only approach these theoretical minima. For the l@kV system examined here, values of US observed in trial analyses of COZcame within a factor of 2 of the theoretical minimum; for the 3-kV system, the approach was closer (1.5x), probably because noise associated with sample handling was minimized through use of larger ~amp1es.I~The question now arises, if performance is retested using procedures including combustion instead of direct introduction of COZ,to what extent will observed values of ud increase? In other words, how much additional noise is associated with combustion? Propane. Propane provides an interesting case since it is representative of the easily combusted hydrocarbons and can be handled as a gas, simplifyingproduction of a wide range of sample sizes by use of exponential dilution. For the 3-kV instrument, results obtained over a range of sample sizes are summarized in Figure 7a. All 6 values were calibrated versus COZintroduced from the variable-volume inlet. The broken lines in Figure 7a indicate the envelope expected to include 99%of all observations (22) Menitt, D. A.; Hayes, J. M.; Des Marais, D.J. J Geophys. Res., in press.
-1.0
............. ........... b"-....... aP- .....+.--Q ......... ...................... . 9-0_.._
,...Oo.O.P
0.0 -0.5 -
where A is the integrated signal (V s, m/z 44) derived from introduction of m moles of sample, Rf is the feedback resistance of the ion current amplifier, and qe is the electronic charge. Combining these equations, inserting values for fundamental constants and instrumental parameters (Rf = 3 x 108 Q), and considering details of sample versus standard comparisons, the following relationship is obtained:13
1
o0n ,O
(5)
where US is the expected standard deviation in 6, R is the ion current ratio, E is the efficiency (or sensitivity) of the mass spectrometer system, expressed in ions collected per molecule introduced, m is the sample size (imoles), and NAis Avogadro's number. The expression can be cast in terms of observed signal strengths by considering an additional relationship,
1
.............. 0. - a0@_.-...? ....0D.........
_..
'
.......
...... .--.
.a-
I
I
10
100
I 1000
C,pmol Figure 7. Results of combustion tests using exponential dilution series and CuO reactors. In each case, the difference between observed and expected values of 6 is plotted as a function of sample size. The envelopes (broken lines) that would enclose 99% of all observations if the system were operating within a factor of 2 of the shot-noise limit are also shown. (a) Propane, 3-kV instrument, 850 "C. (b) Methane, 3-kV instrument, 900 "C. (c) Methane, 10-kV instrument, 900 OC.
if the system is performing within a factor of 2 of the shot-noise limit (that is, the lines are at US, where UShas been calculated, with due attention to conditions of standardization, using eq 7). Relative to the theoretical maximum, performance is, in fact, poorer for large samples. The cause is almost certainly difticulty in obtaining complete combustion when oxygen demands are high. Particular care should be taken, therefore, to avoid peaks containiig much more than 10 nmol of C at the high-temperature ends of chromatograms, when oxygen demands associated with column bleed are likely to be high. For samples smaller than 10 nmol of C, however, the evidence indicates that the precision of irmGCMS measurements is degraded only marginally, if at all, by the combustion process itself. Methane. In contrast to propane, methane provides a worst case, being the least easily combusted substance found thus far. Effects of combustion on precision are likely to be maximal. Values of A from a series of measurements using the 3-kV instrument with a CuO furnace operating at 900 "C are plotted as a function of sample size in Figure 7b. The significance of the envelope is the same as that in Figure 7a, so the results do, indeed, indicate that performance in analyses of methane is not as good as that obtained with more easily combustible substances. For Analytical Chemistry, Vol. 67, No. 14, July 15, 1995
2471
samples smaller than 10 nmol of C, there is no evidence for systematic errors. For large samples, however, systematic errors appear clearly. In fact, such samples could not be combusted quantitatively in CuO reactors even when the temperature was increased from 1000 to 1100 "C and a freshly oxidized reactor was employed. Since reactor capacity decreased quickly as CuO was lost through thermal decomposition, combustion of CH4 was often incomplete after reactors had been operated for more than 3 h at temperatures greater than 900 "C. Larger samples might be combusted quantitatively if the oxidation capacity of the reactor was increased either by packing reactors with cupric oxide granules or by significantly increasing the length and number of copper wires inside reactors. Both options are unsatisfactory. Microvolume reactors cannot accommodate more than four 0.1mm-diameter wires, and additional metal oxide, whether added as granules or oxidized wires, could only be accommodated using larger reactors. Although such reactors could offer higher capacity, chromatographic resolution would likely degrade unless makeup gas was added. The corresponding dilution of COZin the reactor effluent would be self-defeating, since the resulting higher split ratio would decrease the fraction of sample actually reaching the ion source. Similar trial analyses of CH4 were made using the l@kV instrument, which, due to its sensitivity, has a great advantage in this application, allowing the use of particularly small samples while still preserving a useful dynamic range. As shown in Figure 7c, results that were nearly within a factor of 2 of the shot-noise limit were obtained for samples ranging in size from 8 to 800 pmol of c . The key to broadly satisfactory isotopic analyses of methane seems to be provided by the use of NiO reactors. Although no results for exponential dilution series are available, two specific comparisons can be made. For six replicate analyses of 400pmol samples, the observed standard deviation was 0.05%, only 2 . 5 ~ greater than the shot-noise limited minimum and as good as or better than the performance summarized for the CuO reactor in Figure 7c. For nine replicate analyses of 95pmol samples, the observed standard deviation (0.12Oh) exceeded the related shotnoise minimum by a factor of 3.2. Far more importantly, the NiO reactors not only work as well as the CuO reactors (Le., provide equal precision), but they operate fur longer, being stable indefinitely at the very high temperatures that promote efficient combustion of CH4.21s22 n-Alkanes. Precise assessment of performance relative to theoretical limits is more difticult for compounds introduced by injection of solutions. Sample size is relatively dficult to control, and m, a key variable in eqs 5-7, is often not well known. Nevertheless, the n-alkanes represent a particularly interesting case since all related analytical procedures (sample handling, chromatographic separation, definition, and integration of peaks) mirror those employed in analyses of the larger compounds to which irmGCMS has been most prominently applied thus far. Here, homologous series of n-alkanes were analyzed and results compiled in order to allow comparison of observed levels of precision to theoretical limits. Three sets of data were obtained under optimal conditions: (i) CuO at 850 "C, 3-kV instrument; (ii) CuO at 850 "C, 1@kV instrument, and (iii) NiO with supplemental 02 at 1050 "C, 1@kVinstrument. Together, these results allow comparison of performance furnished by the low2472 Analytical Chemistry, Vol. 67, No. 14, July 15, 1995
Table 4. Accuracy and Precision of Analyses of mAlkanesa
n-alkane (no. ofC)
3-kV, CuO, 850
6
(%Ib
17 -25.46 18 -34.62 19 -29.02 20 -28.19 21 -27.48 22 -33.98 23 -27.15 24 -31.65 25 -28.08 26 -34.65 27 -25.44 28 -27.45 29 -30.89 -27.50 30 31 -28.85 32 -26.61 -30.42 33 34 -27.50 -26.41 35 mean A (%o) rms A (Oh)
meanA
Ud
0.00 0.09 -0.04 0.02 0.04 0.00 0.10 0.11 0.02 0.07 0.05 0.20 0.07 0.14 0.13 -0.01 0.01 0.01 0.03 0.05 0.11
0.14 0.02 0.01 0.10 0.16 0.02 0.14 0.10 0.08 0.02 0.16 0.09 0.03 0.09 0.05 0.13 0.03 0.11 0.13
l@kV , CuO, lGkV, NiO + 850 02, 1050 meanA Ud mean A Ud 0.00 0.02 0,oo 0.10 0.10 -0.10 -0.08 -0.15 0.02 -0.15 -0.16 -0.03 0.02 -0.02 -0.05 -0.03 -0.02 0.04 0.05 -0.02 0.10
0.07 0.06 0.06 0.05 0.04 0.09 0.04 0.14 0.08 0.07 0.03 0.06 0.03 0.14 0.09 0.04 0.03 0.10 0.09
0.06 0.06 0.01 0.07 0.00 -0.03 -0.10 -0.09 -0.05 -0.04 0.00 0.02 0.06 -0.04 0.01 -0.08 -0.03 -0.09 0.08 -0.01 0.09
0.08 0.04 0.07 0.05 0.14 0.08 0.12 0.07 0.07 0.10 0.06 0.09 0.02 0.13 0.10 0.06 0.19 0.17 0.09
a Mean deviation and standard deviation (all in O h ) of three replicate analyses. Column headings designate mass spectrometer,oxidant, and temperature of combustion (in OC). Accepted 6 value for standard alkane.
and high-sensitivity instruments and by the major alternative combustion reactors. For analyses performed using the 3-kV instrument, combustion of -10 nmol of C/component yielded, taking account of split ratios, -4 nmol of COz/component in the ion source. Similar levels of perdeuterated n-alkanes were co-injected kith the sample for isotopic calibration. Results are summarized in Table 4. Mean values of A for each of three injections do not differ significantly from zero, and rms values of A only slightly exceed the overall average standard deviation. Both of these observations indicate an absence of systematic error. For the integrated signal levels observed, the theoretical minimum value for ud calculated using eq 7 was 0.05%. The observation that the observed average ud exceeds this value by a factor of 1.6 indicates that the performance obtained is very close to that obtained for analyses of COZalone. Therefore, chromatographic separation and combustion are not signiticant sources of additional analytical uncertainty under these conditions. Smaller amounts were injected for analysis on the l@kV instrument, specitically, 2.5 nmol of C (0.4 nmol of COz transmitted to the ion source)/component. Similar levels of perdeuteriocarbons were co-injected with the sample for isotopic calibration. For the integrated signals observed (not markedly larger than those on the 3-kV instrument since the amount of COa reaching the ion source was l o x smaller), the theoretical minimum value of ad was 0.04Oh. Results are summarized in Table 4. Accuracy is superior to that observed using the 3-kV instrument, and the precision is equal even though the samples are smaller. Since the observed average values for ub (0.10, 0.09%) exceed the theoretical minimum by a factor of 2.3-2.5, the theoretical versus observed comparison is less favorable than that with the 3-kV instrument, but this difference duplicates that observed in previous analyses of pure C o n and thus again fails to provide any evidence
that combustion-based analyses of multicarbon compounds are beset by additional, significant sources of analytical uncertainty. The performance of the NiO/Oz system did not differ signiscantly from that of the CuO system. Performance Observed for Complex Samples. Conditions for the analyses just summarized were optimal not only in terms of conditions of combustion but also in terms of chromatographic resolution. All sample peaks were well resolved, and background ion currents were either small (‘2 &I m/z 44)or varied smoothly and predictably. As a result, assignments of peak boundaries and background levels were unambiguous. For real samples, these same advantages, and thus, these levels of performance, may be difficult to duplicate. Consequently, isotopic data are often substantially less accurate and p r e c i ~ e , ~although , ~ J ~ recent advances in irmGCMS software, specifically a new algorithm for automatic background correction,’2 may improve the situation dramatically. CONCLUSION Quantitative combustion is the most important key to accuracy in the analysis of I3C in organic material. The present results demonstrate that it can be obtained in a variety of ways. Not surprisingly, results indicate that the most robust performance is obtained when temperature is maximized and 02 is present as a gas-phase reactant. By itself, CuO can provide these conditions, but not for long because the spontaneous degradation of CuO occurs so rapidly. Partial pressures of 02 generated by the decomposition of NiO appear negligibly low, but the analyst can elect to provide 02 from some external source. Here, good performance has been achieved through use of a system in which a trickling flow of 02 is added to the gas stream in the reactor. The resulting mole fraction is low enough that it poses no threat to the integrity of the heated filament in the ion source of the mass spectrometer,but it appears to be high enough that it serves in some critical way as a gas-phase oxidant. Moreover, observation of Orrelated ion currents shows that, when metal-bound supplies of 02 have been drawn down by use of the reactor, the trickling gas-phase flow serves to recharge the metal oxide reservoir continuously. In this way, any need for periodic reoxidation of the reactor bed is avoided, and operation of the system is simplied. Results obtained show that when analytes are handled optimally (i.e., in ways precluding contamination or isotopic fraction-
ation and thus guaranteeing isotopic integrity; for example, direct admission of materials from an exponential dilution system), the scatter of analytical results does not worsen when combustion is added to the analytical sequence. Therefore, the combustion reactor itself is not a “noise source” that contributes signiscantly to analytical uncertainty. An exception seems to be provided by CH4, for which attainment of complete combustion is particularly difficult, but this point bears further study. The efficiency of the Nafion-based water removal system varied as expected: lengthening the separator and taking care that the gas purging the “dry” side of the membrane contained as little water as possible yielded improvements that have made the unit fully adequate for use with an instrument intolerant of HzO. On the basis of these developments, it can now be claimed with some justification that irmGCMS combines the speed, convenience, and sensitivity of GC/MS with the precision of conventional isotopic analyses. ACKNOWLEWMENT We are indebted to former co-workers in the Biogeochemical Laboratories, particularly James Collister, Glenn Hieshima, and Louis Brzuzy, for valuable discussions and observations conceming the practical operation of these instruments. We are most grateful to Dr. W. Brand, Finnigan MAT, for his suggestion that NiO would serve as a suitable oxidant and to Dr. R Carlson, Chevron Petroleum Technology Co., for supplying the n-alkane and perdeuterated alkane standards used in this work. We express our appreciation to Dr. K Habfast at MAT and to Dr. M. Schoell at Chevron for their continuing interest and scientific collaboration. We thank Finnigan MAT for making available the MAT 252 and Chevron and MAT for supporting the purchase of the Delta S and development of irmGCMS. Partial support of personnel costs associated with this work has come from NASA (NAGW-1940 and a fellowship awarded to KH.F.) and from a GE fellowship awarded to DAM. Received for review December 13, 1994. Accepted April 20, 1995.@ AC941205R @Abstractpublished in Advance ACS Abstracts, June 1, 1995.
Analytical Chemistry, Vol. 67,No. 14, July 15, 1995
2473
