Appendix Reprints of Pertinent Literature
Reproduced from the Journal of Chromatographic Sciences
by permission of Preston Publications,
A Division of Preston Industries, Inc.
Journal of Chromatographic Science, Vol. 19, December, 1981
Determinations of Carbon Dioxide and Methane by Dual Catalyst Flame Ionization Chromatography and Nitrous Oxide by Electron Capture Chromatography
Ray F. Weiss
Scripps Institution of Oceanography, A-020,
University of California, San Diego,
La Jolla, California 92093
An automated gas chromatographic (GC) system has been developed for the measurement of carbon dioxide, methane, and nitrous oxide in air and other gases. Carbon dioxide is measured by a flame ionization detector (FID) after conversion to methane in pure hydrogen carrier gas, using palladium and nickel dual catalysts which permit direct on-line injection of oxygen-containing samples. The detector measures methane in the same sample, although the system has not been optimized for this gas. Nitrous oxide is measured in a separate aliquot of the sample using a hot electron cature detector (ECD) and argon-methane carier gas. Typical relative standard deviations of the calculated results are 0.04% for carbon dioxide, 0.4% for methane, and 0.3% for nitrous oxide
The GC measurement of trace quantities of carbon dioxide by catalytic conversion to methane, followed by FID, has become commonplace. In this method, carbon dioxide as well as carbon monoxide are reduced to methane by passing over a reduced nickel catalyst at elevated temperature in the presence of excess hydrogen (1-3). Although the method has been demonstrated to be quantitative for both CO2 and CO (3), the catalyst has been found to be susceptible to oxidation by oxygen, with a resulting drop in conversion efficiency (4). This problem has traditionally been approached by the use of in-line switching valves, which are used to shunt the oxidizing components of the sample around the catalyst instead of through it. The principal disadvantages of this approach lie in its complexity and in the effects on instrument stability of flow interruptions associated with the switching process.
the instrument described in this paper has been developed for the automated repetitive analysis of carbon dioxide, methane, and nitrous oxide in the atmosphere and in air equilibrated with ocean water. Because the magnitude of natural variations in these samples is often exceedingly small, every effort has been made to achieve the highest possible precision. In the case of carbon dioxide, a new dual catalyst technique has been developed, in which the nickel catalyst is protected by a palladium catalyst that reduces oxygen to water, thus permitting the entire sample to flow throug the system without the use of in-line switching valves. Methane is unaffected by the two catalysts, and is also measured by FID. The measurement of nitrous oxide is by electron capture chromatography (5, 6) using a separate aliquot of the sample. Because of the highly specific nature of the ECD, and because of the importance of selective interference effects in the electron capture process (7), the separating column and injection sequencing have been selected to optimize the spearation of nitrous oxide. The detailed design of the system, its performance, and maintenance are described in the following section.
The configuration of the instrument is shown schematically in Figure 1. Its major components include a custom made gas sampling module, a Perkin-Elmer (Norwalk, Connecticut) model 3920B chromatograph which has been modified extensively to provide the necessary heated areas and improve temperature stability, and a Spectra-Physics (Santa Clara, California) model SP4100 integrator which acquires, proceses, and records the data and controls the automatic operation of the system.
The automated sampling module has been designed to sample sequentially from four separate gas streams which are supplied to the instrument at ca. 0.6 atm above ambient pressure. In its usual configuration, the sequence of gases is arranged to alternate between samples and standards, each of which passes through its own drying column so that the instrument measures directly the composition of the dry gas. This approach is preferable for applications in atmospheric studies in which the composition on a dry basis is the quantity of interest, but the instrument can also tolerate the injection of moist samples without interference in the analyses. Although the compressed gas standards are already dried, they too are fitted with drying columns so that any unforeseen consequences of these columns will affect the samples and standards equally. Similarly, the pressures of the samples and standards in the drying columns should be equal and constant to minimize the effect of any partitioning between the drier and the gas phase. For the samples, this is done by constantly purging the sampling line through a back-pressure regulator which fixes the pressure at the inlet to the drying column. the samples are pressurized by a stainless steel bellows pump, and the purge rate is typically ca. 5 1/min. It is important to avoid the use of permeable materials in the plumbing of the sampling system, since the permeation rates of nitrous oxide and carbon dioxide in such materials are typically very high.
The selection of an absorbent for the drying columns is most critical for carbon dioxide because this gas is measured most precisely, and is involved in acid-base reacions. For example, the neutral absorbent anhydrous calciumn sulfate performs adequately as a dying agent for methane and nitrous oxide measurements, but initially removes small quantities of carbon dioxide from the sample stream. As the absorbent is gradually moistened, this CO2 is apparently reliberated, and the sample stream becomes slightly enriched in CO2. Such difficulties are avoided by the use of an acidic absorbent such as P2O5. After some experimentation, the best solution for this application has been found to b Aquasorb (Mallinckrodt, St. Louis, Missouri) consisting of a solid support impregnated with P2O5 and containing a colored indicator, which gives a dew point of -96°C. With Aquasorb drying columns packed in Pyrex tubes so that the colored indicator can be observed, the results for carbon dioxide, nitrous oxide, and methane have been found to be reproducible and unaltered within the analytical limits of the instrumentation.
Inside the thermostated sampling module, each of the four sample inlets is connected through a metering valve to a Carle Instruments (Anaheim, California) model 2025 4-port stream selector valve, fitted with an 8-position rotary electric actuator which allows the valve to be indexed to the four ports and four intermediate off-port positions. After the stream selector valve, the sample gases flow through a Brooks (Emerson Electric, Hatfield, Pennsylvania) flowmeter, which has been rebuilt to remove dead volume in the end fittings, and is used to set the four inlet metering valves to give the same flow. The sample gases then flow through two Carle model 2021 6-port rotary gas sampling valves, each of which is fitted with a 1 ml sample loop and an electric actuator. The effluent gas is then vented to the atmosphere through an isolation coil made of 2 in X 2.1 mm i.d. stainless steel tubing.
Prior to each sample injection, the stream selectro valve is indexed to the next sample port, and the system is puged with sample at 75 ml/min for 90 sec. Considerably less sample could be used, but even at these excessive flushing rates, a single tank of standard will last for over one year of continuous operation. At the end of the flushing period, the stream selector valve is indexed to the next intermediate off-port position, and the gas in the two sampling valves is allowed to equilibrate with atmospheric pressure through the isolation coil before injection.
The effectiveness of the isolation coil was tested by injecting pure N2 samples and adjusting the equilibration time following the flushing to see if back-mixing of the trace constituents of room air could be detected in the analyses. No such contamination was found at the longest tested equilibration time of 150 sec. The time required for pressure equilibration was tested by measuring the effect or peak area of reducing the equilibration time using normal air samples. No effect was observed with equilibration times as short as 2 sec. From the results of these tests, a 10 sec equilibration time was selected for routine use.
Because the amount of gas in the sampling loops depends directly on the absolute temperature of the sampling loops and on barometric pressure, the accuracy of the results is limited by the changes in these quantities which occure between the injection of a sample and the two adjacent standards. Variations in barometric pressure can enerally be neglected, but precise temperature control is essential. In addition, conventionalpressure regulators and needle valves used to control the flow of carrier gas are significantly affected by temperature variations. These components are therefore installed within the thermostated sampling module.
The temperature of the sampling module is maintained at ca. 45°C by a high velocity air bath. The recirculation time for air withing the closed sampling module is about 1 sec, and the temperature within the module is sensed by a micro-thermistor bead mounted near the two sampling loops. The thermistor is connected through a bridge and a high gain amplifier to a zero-crossing solid state switch which controls the power to the heating element. Because the thermistor has a low thermal mass and the circuit is sensitive to changes of a few thousandths of a degree, the cycling time for the entire heating system is also on the order of 1 sec. thus, components within the sampling module which are larger than the thermistor and have longer thermal exchange times will experience correspondingly smaller temperature variations. The esimated temperature stability of the gas sampling loops is better than 0.02°C.
Standards and Carrier Gases
For applications involving atmospheric analyses, the instrument is calibrated using compressed air working standards doped to contain a range of concentrations of the measured constituents. Although standards prepared in nitrogen, helium, or other diluent gases may also be used, it is preferred practice, especially with detectors such as the ECD which are prone to interference effects (7), to use standards which reflect the composition of the unknown samples as closely as possible. The standards used with this instrument are prepared by compressing clean marine air at La Jolla and are stored in aluminum cylinders. by using selective absorbents, or by doping the cylinders prior to the addition of the compressed air, it is possible to vary the trace gas composition of this air over a wide range. Typically, the low range working standard is prepared with near-ambient concentrations of CO2 (≈335 ppm), CH4 (≈1.5 ppm), and N2O (≈300 ppb), and the high range standard is prepared with these constituents enriched by 50 to 75%. In addition, a suite of comparison standards covering a wide range of concentrations is used to check linearity. All the compressed air standards are dried to a dew point of less than -80°C at 1 atm.
The compressed air standards are callibrated independently, and their concentrations are reported as mole fractions of the dry gas. For CO2, the calibrations are performed in the laboratory of C.D. Keeling at Scripps, using nondispersive infrared analysis and manometric techniques (8). Alternatively, the chromatograph itself can be used to caligrate the compressed air standards against the Keeling CO2 standards. In either case, the CO2 calibrations have an accuracy of up to ±0.1 ppm or about 0.03 relative percent at near ambient concentrations, depending upon how directly the comparisons are tied to the Keeling primary manometric standards. The N2O concentration in the standards is given by the product of th CO2 concentration and the N2O-CO2 ratio as measured by ultrasonic phase shift GC (9). This method offers the benefit of a nonspecific detector which is relatively immune to interference effects, and also allows N2O in the 300 ppb concentration range to be calibrated using mixtures of N2O in CO2, which need not be diluted more than about 1 ppt. Using this "bootstrap" technique, the calibrations of ambient N2O concentrations have an accuracy of about ±0.2 relative percent. The calibrations of CH4 in the standards are still in progress using classical manometric and static dilution techniques.
the approximately 50 ml/min flow of hydrogen carrier gas for the flame ionization measuremnt of CO2 and CH4 is prepared electrolytically from distilled water by a General Electric (Wilmington, Massechusetts) hydrogen generator. The gas is further purified by an Aadco (Rockville, Maryland) hydrogen purifier employing a palladium diffusion thimble operated at 420°C. Without this final step of purification the precision of the methane analysis is significantly degraded.
The carrier gas for the electron capture measurement of N2O is a 95% argon/5% methane mixture, supplied at a rate of ca. 20 ml/min through a molecular sieve 5A (Coast Engineering, Gardena, California) trap, which is activiated at 300°C and is operated at room temperature. Initally, a chemical oxygen scrubber ws also used to purify the Ar-CH4 carrier gas. However, the small amounts of oxidant apparently present in most cylinders of high purity Ar-CH4 was found to enhance the sensitivity and the stability of the N2O measurement; thus, best results are obtained with the molecular-sieve trap alone.
Separation, Catalysis, and Detection
the separation of the CO2 and CH4 peaks is carried out at ca. 55°C on a 3 m x 2.2 mm id. stainless steel column packed with 80/100 mesh Porapak Q (Waters Associates, Framingham, Massachusetts). The CO2 separation can be carried out equally well, with a lower pressure drop but without adequate resolution for CH4, on a 1.8 m column of Porapak T at 50°C.
Following the separation column, the gases enter the palladium catalyst for the conversion of oxygen to water by reaction with the hydrogen carrier (10). the catalyst is prepared using -60 mesh palladium metal (Alfa Inorganics, Danvers, Massachusetts) mixed in a 1.4 volume ratio with 80/100 mesh acid-washed Chromosorb W (Johns-Manville, denver, Colorado) as an inert support to prevent the fine palladium powder from blocking the gas flow. A 37 cm length of this mixture is packed into a 2.2 mm i.d. stainless steel tube. By using pure palladium metal (11), the catalytic surfaces are kept strongly reducing, not only by gas phase H2 but also by hydrogen dissolved in the metal. The reaction is further enhanced by operating the catlyst at elevated temperature, 240°C having been selected as a convenient choice because it is the temperature of the molecular sieve separating column used in the N2O measurement. Because of the high temperature of this catalyst and of the nickel catalyst which follows, the water produced by the reaction i srapidly eluted through the two catalysts and the detector before the elution of the methane and carbon dioxide peaks from the separating columns.
The reduced nickel catalyst for conversion of CO2 to CH4 is prepared from nickelous nitrate deposited on Chromosorb W (3). A 13 cm length of this catalyst is packed into a 2.2 mm i.d. stainless steel tube. After first oxidizing the catalyst by purging with air at ca. 500°C for approximately 30 min, the nickel surface is reduced by purging with H2 at 500 to 525°C for several hours. The catalyst is then operated at 400°C. Because of the high temperatures involved, quartz wool, rather than the usual Pyrex wool, is used to hold the packing inside the stainless steel tube. If the machine is shut down, or occasionally after prolonged use, it is necessary to reactivate the catalyst using the oxidation and reduction procedures described above.
By using this new dual catalyst technique, it is possible to carry out repetitive on-line measurements of CO2 in oxygen-containing gases by methane conversion without need of column switching or backflushing during the analysis. The water formed by the palladium catalyst passes through the nickel catlyst without reaction, carbon dioxide is unaffected by the palladium catalyst, and methane passes through both catalysts without reaction.
The CH4 and CO2 as CH4 peaks are detected by the Perkin-Elmer 3920B FID using the approximately 50 ml/min flow of H2 carrier gas as the source of combustion fuel. The usual hydrogen inlet port of the detector is blocked off in this application. Hydrocarbon-free oxidizing gas is provided at ca. 550 ml/min by an Aadco model 737-1-B "clean air generator," a preparative chromatograph which purifies room air (12). This instrument also enriches the O2 concentration in the effluent to ca. 40%. The flows of carrier gas and combustion air have been selected to optimize combustion efficiency. The use of normal air for the detector, as opposed to oxygen enriched air, reduces the detection sensitivity by ca. 6%. Although the method described here is greatly simplified, as opposed to the usual flame ionization technique which requires separate supplies of carrier gas and fuel hydrogen, the choice of carrier gas flow rate is fixed by the flame burning characteristics, and is not necessarily the ideal choice for the best column performance. Fortunately, the separations performed here are not terribly demanding, and a relatively high flow rate for this column diameter is tolerable. If necessary, the flow rate in the separating column can be reduced by introducing additional fuel hydrogen before the palladium catalyst.
The separation of the N2O peak is carried out at approximately 240°C on a 1.8 m x 2.2 mm i.d stainless steel column packed with 60/80 mesh molecular sieve 5A. This choice of column packing offers several advantages over Porapak or Porasil columns often used for this analysis (6). The molecular sieve 5A column is highly selective for N2O, thus reducing the probability that other trace constituents with high electron capture cross sections or significatn interference effects will be eluted together with N2O. Unlike most other column packings, molecular sieves do not produce progressively longer retention times with increasing molecular size. Molecules which are too large to interact with the structure of the molecular sieve pass through the column relatively rapidly. Thus, the interference of such compounds during subsequent analyses is minimized in applications which require repetitive analyses without the benefit of backflushing. The Perkin-Elmer 3920B 63Ni ECD is operated at elevated temperature (350°C) for enhanced sensitivity to N2O (5, 6).
Data Processing and System Automation
A Spectra-Physics SP4100 computing integrator with BASIC program control is used to integrate the peaks with a sensitivity of 0.5 µV-sec per count. Valve operations required for the sampling procedure are controlled through zero-crossing solid state switches activated by the integrator. In addition to sampling and injection, this includes returning the sampling valves to the load position 30 sec after the start of the analysis so that the next sample can be prepared for injection while the analysis is being completed. Because the analyses are highly repetitive, the integeration parameters are specifically tailored for each of the three peaks. The instrument begins each analysis by recording the signal from the FID. Following the elution of the CO2 peak, the integrator activates reed switches to transfer the input signal to the ECD before the elution of the N2O peak.
At the end of each run, the sample identification, injection time and date, retention times, peak areas, and baseline correction codes are printed and recorded automatically by a solenoid-controlled Hitachi (Tokyo, Japan) model D-980 magnetic cassette recorder interfaced to the integrator using frequency shift keying at a rate of 1200 baud. In the laboratory, the tapes are replayed through a similar tape deck which is fitted with a decoding circuit so that the data may be played directly into a computer through an RS-232C interface for final processing and interpretation. For real time observation of the data, the integerator calculates the concentrations of the unknown samples using the two bracketing standard calibration runs. The results of multiple standard calibration have shown that the performance of the detectors is best represented by a linear fit of the sensitivity against concentration, which is equivalent to fitting peak area against concentration with a quadratic fit forced through the origin. The integrator also calculates the concentration intercept of a linear fit to the two standards as a way of monitoring detector linearity. The linearity of the responses and tuning for optimum linearity are discussed below.
Results and Discussion
A typical chromatogram for a 1 ml air sample, measured under shipboard conditions, is shown in Figure 2. The peaks have been recorded at various attenuation factors as labeled along the baseline and opposite each peak. The labeled peak areas are in units of 0.5 µV-sec per count for the unattenuated signal connected to the 1.0 V full-scale output of the electrometers. The retention times for CH4, CO2, and N2O are about 110, 195, and 360 sec, respectively. Under automatic operation an analysis is repeated every 450 sec, or exactly eight times per hour. Thus, the cycle of the four inlet gases requires 30 min. For samples which do not require the measurement of N2O, the measurement of CH4 and CO2 can be repeated every 220 sec.
Two instruments of this design have been operated a total of more than 8000 hr of continuous analyses under conditions ranging from an air-conditioned shorebased laboratory, to shipboard operation in the tropics with substantial vibration, ship motion, and laboratory temperature variations. The precision of these measurements, expressed as the relative standard deviation of the calculated results as the relative standard deviation of the calculated results for the sample analyses, ranges from 0.02 to 0.08% for CO2, 0.2 to 0.7% for CH4, and 0.2 to 0.5% for N2O. The individual analyses, as opposed to the calculated values which include the results of both sample and standard analyses, are more precise by a factor of 1.4. Although extreme variations in environmental conditions can affect instrumental precision adversely, the instrument generally performs as well under shipboard conditions as it does in a well controlled shorebased laboratory. The typical relative standard derivations of the calculated results are 0.04% for CO2, 0.4% for CH4, and 0.3% for N2O.
The very high precision which has been achieved for CO2 is due to the precision of the gas sampling operations, the quality of the integrating electronics, and the integrating nature of the FID. Unlike concentration detectors such as the thermal conductivity detector, the integrated response of the FID does not show a first-order dependence on carrier flow rate. Rather, the flow of carrier gas is only important insofar as it affects ionization and collection efficiencies. The precision for CH4 is limited by the small sample size and by less than ideal column separation, both of which are selected soley to optimize the precision for CO2. The precision achieved for N2O is equivalent to the best precisions for this analysis reported in the literature. Unlike the FID, the ECD most closely approximates a concentration detector, and its precision is therefore dependent on flow control. The precision for N2O is probably also limited by the signal-to-noise ratio of the detector.
The linearity of the instrument has been investigated in detail for the measurement of CO2 in the concentration range of 320 to 450 ppm through analysis of a number of standad tanks provided by the Keeling laboratory. If the detector is properly cleaned and tuned, the results are completely linear. There is no significant difference at the 0.1 ppm confidence level between linear and quadratic fits of the peak area data against concentration for these standards. However, experience has shown that after several months of continuous operation the FID can become significantly nonlinear, either by contamination of the collector surface or by misadjustment of the air inlet orifice. In this case, as discussed above, the results are still closely represented by a quadratic fit of peak area against concentration. The linearity of the detector is easily restored by cleaning the collector in strong base and then in strong acid or by adjusting the air inlet orifice.
Occasionally, after months of operation, the nickel catalyst has been found to deteriorate in a way which will not respond to the usual activation procedures. This deterioration results in a broadening of the CO2 peak and a reduction in CO2 precision. This problem is best cured by replacement of the nickel catalyst. A search for alternate forms of nickel catalyst is now in progress.
The long-term performance of the ECD has been excellent despite tis high operating temperature. Although the linearity of this detector has not been studied as carefully as the FID, no significant nonlinearities have been detected within the limits of instrumental precision over the 300 to 600 ppb concentration range.
An example of instrumental performance for continuous atmospheric analyses is shown in Figure 3. The data were measured aboard ship in the equatorial Pacific Ocean during a 24 day period in the fall of 1979. Dry air mole fractions are plotted against time (Figure 3) with 48 separate measurements plotted for each gas each day. Approximate latitudes are labeled along the upper axis, showing that the ship's track is divided into northbound and southbound segments, crossing the Intertropical Convergence Zone at about five degrees north three separate times. Thus, the data demonstrates not only the precision of the measurements over an extendede period at sea, but also show reproducibly the interhemispheric differences in CO2 and CH4 concentrations. The existence of a possible small interhemispheric gradient in N2O is not well demonstrated by the limited geographical coverage of these data, but is now being investigated using similar data covering a broader geographical range.
The dual catalyst GC technique for the measurement of CO2 has been demonstrated to be comparable in precision to the best infrared analysis techniques now in use. In addition, the chromatograph has several distinct advantages over the infrared method. The 1 ml sample size, as compared to as much as 1 liter for the infrared analysis, makes possible the analysis of small samples, such as those equilibrated with, or extracted from, small samples of seawater or other natural waters. Efforts to exploit this capability are now in progress. Another consequence of this smaller sample size is that the consumption of standards in long-term monitoring is no longer a practical limitation. The chromatographic analysis is rapid and easily automated. The linearity of the chromatographic response exceeds that of the infrared analyzer, making calibration less tedious, especially when samples cover a wide range of CO2 concentrations. Finally, the response for CO2 is essentially independent of the gross composition of the gas, so that there are no corrections analogous to the pressure broadening correction which is required by the presence of oxygen in the infrared analyzer.
Although the electron capture results for N2O are of high precision, the specific nature of the detector and is susceptibility to interference effects argue against its use of long-term monitoring unles the composition of standards, samples, and carrier gases with respect to all major and minor components can be rigidly maintained. These long-term measurements are best carried out by more direct, nonspecific methods such as the ultrasonic phase shift technique (9). However, the precision of the electron capture technique makes this method ideal for the measurement of spatial or short-term variations such as those made from a moving platform
After the measurement of tens of thousands of samples in the laboratory and on ships around the world, the reliability and serviceability of the instrumentation have been well demonstrated. The data recovery rate for both the flame ionization and electron capture systems has been well above 99%, the only losses having been due to the failure of two power supply capacitors. The instrumentation has also proven to be relatively portable and well suited for extended operation in remote areas, not only because of its low rate of consumption of standard gases, but also because it generates its own hydrogen carrier gas and hydrocarbon-free combustion air. The dual catalyst flame ionization technique is now being implemented for the measurement of atmospheric carbon dioxide at remote, clean-air global monitoring sites.
The author is indebted to a number of persons who have contributed to the development of this instrumentation. The shipboard operations and much of the laboratory testing were carried out by F. A. Van Woy, whose patience and perserverance have led to a better understanding of the system and its idiosyncracies. The calibration of the instrument for CO2, and the verification of its performance for this gas, have depended directly on the generous assistance of C. D. Keeling, P. Guenther, and D. Moss of the Scripps CO2 Laoratory. Finally, the author thanks N. Andersen and L. Machta for their interest and support. This work was supported by the Marine Chemistry and Atmospheric Chemistry Programs of the U.S. National Science Foundation. Additional testing and development work during the NORPAX Equatorial Experiment was supported by the U.S Department of Energy.
- F.W. Hightower and A.H. White. Synthesis of methane from water gas. Ind. Eng. Chem. 20: 10-15 (1928).
- K. Porter and D.H. Volman. Flame ionization detection of carbon monoxide for gas chromatographic analysis. Anal. Chem. 34: 748-49 (1962).
- F.W Williams, F.J. Woods, and M.E Umstead. Determination of carbon dioxide in the parts-per-million range with gas chromatography. J. Chromatagr. Sci. 10: 570-72 (1972).
- Carle Instruments, unpublished operating instructions for the "Methanizer" catalytic converter.
- W.E. Wentworth and R. Freeman. Measurement of atmospheric nitrous oxide using an electron capture detector in conjunction with gas chromatography. J. Chromatogr. 79: 322-23 (1973).
- R.A Rasmussen, J. Krasnec, and D. Pierotti. N2O analysis in the atmosphere via electron captue-gas chromatography. Geophys. Res. Lett. 3: 615-18 (1976).
- R.E. Sievers, M.P. Phillips, R.M. Barkley, M.A. Wizner, M.J. Bollinger, R.S. Hutte, and F.C. Fehsenfield. Selective electron-capture sensitization. J. Chromatogr. 186: 3-14 (1979).
- C.D. Keeling, R.B. Bacastow, A.E. Bainbridge, C.A. Ekdahl Jr., P.R. Guenther, L.S. Waterman, and J.F. Chin. Atmospheric carbon dioxide variations at Mauna Loa Observatory, Hawaii. Tellus. 28: 538-51 (1976)
- R.F Weiss, C.D Keeling, and H. Craig. The determination of troposheric nitrous oxide. J. Geophys. Res. 86: 7192-202 (1981).
- M. Krejci, K. Tesarik, and J. Janak. Rapid chrmoatogrphic separation and determination of helium-neon-argon and nitrogen-krypton-methane mixtures and determination of argon in the presence of oxygen at room temperature. In Gas Chromatography. H.J. Noebels, R.F. Wall, and N. Brenner, eds. Academic press, New York, New York, 1961, pp. 255-61.
- R.F. Weiss and H. Craig. Precise shipboard determination of dissolved nitrogen, oxygen, argon, and total inorganic carbon by gas chromatography. Deep-Sea Res. 20: 291-303 (1973).
- W.H. King Jr. Generation of clean air to permit gas chromatography without cylinders. Anal. Chem. 43: 984 (1971).
Manuscript received February 2, 1981;
revision received July 1, 1981.