corder pen deflection and the calibrating resistors. A good linear relationship also exists between pen deflection and solution resistances. The two solutions of highest concentration seem to deviate somewhat from the calibration curve. The possibility that this deviation is due to insensitivity of the instrument a t very high concentrations is ruled out by the fact that standard precision resistors of low resistance fall on the curve. Also, if there were such a loss of sensitivity, an increase of a x . voltage across the cell should increase its sensitivity. Consequently, Tz was replaced with a 12-volt filament transformer] and the measurements were checked. There was a slight increase in sensitivity as the dead band decreased from 0.5% to 0.25%, and the points in question (the two uppermost points in Figure 2) were moved closer to the calibration line by that amount. This still leaves a deviation from the calibration curve of l to 2.5y0and further increase of cell voltage did not improve the results. The explanation for thi? deviation probably lies in two factors. The Serfass bridge used is somewhat inaccurate a t high concentrations, and the cell used is of poor design for measurements of low resistances. I t nould appear that the former is the major cause of error, since the a x . resistance readings made with the Serfass bridge using calibrated noninductirely wound resistors with +O.lYo tolerance gave low readings with a similar magnitude of error. Resistance ranges from 0 to 1000 and
from 0 to 100,000 ohms were tried (using appropriate changes in RB), with much the same results. Theoretically, this instrument should measure the resistances of solutions of high as well as of low concentrations. I n practice, however, some sensitivity is lost a t high concentration (above 0.2.V). This may be overcome either by changing the dimensions of the cell, or, preferably] by replacing the 10turn dual Helipot (R3-R8 in Figure 1) with a 25- or 40-turn device (available from the Beckman Instrument Co.). The stability and reproducibility of the instrument were excellent. Over one period of 24 hours, no drift or deviation was observed with a 0.005X phosphate solution flowing through the cell a t various flow rates. The over-all response of the instrument could be improved considerably by reducing the over-all time constant. This would involve the reduction of the time constant (resistance times capacitance) of the rectifier and balance circuits, although the existing circuits represent nearly the optimum value possible with good filtering of the a x . It would be simpler to use a recorder having a better a x . rejection mode (a Varian recorder was used in this case), a shorter response time, and a smaller dead band. Since concentration is linear with conductance ( l / R ) , it would be convenient to have the recorder deflection linear with the reciprocal of resistance (linear with conductance). Since the resistance of the flowing solution is
being measured here, this means that one would need a hyperbolic function generator. Such a device may be constructed based upon Ohm’s law, since current is inversely proportional t o resistance, and if one keeps a constant source voltage, variation of resistance Rs will generate variable current nhich will be inversely proportional to the resistance. Unfortunately. because of the nature of the hyperbolic function, very small changes in the current output would be generated for large changes in the resistance along the horizontal asymptote of the hyperbola, thus giving very poor sensitivity in that area. For this reason no simple hyperbolic function generator was included in the instrument. LITERATURE CITED
(1) Avizonis, P. V., Triston, J. c:.) ANAL. CHEM.34, 54 (1962). (2) Cherkin, A., Matinex, F. E., Dunn, M. S., J . Am. Chem. SOC.75, 1244 (19531. (3) Creamer, R. hl., Chambers, D. H., ANAL.CHEM.26, 1098 (1954). ( 4 ) Daniels, F., Mathews, J. H., JT’illiams, J. R., staff, “Experimental Physical Chemistry,” 4th ed., pp. 454-6, McGraw-Hill, Kew York, 1949. ( 5 ) Fischer, R. B., Fisher, D. J., . ~ N A L . CHEM.24, 1459 (1952). (6) Sober, H. A., Gutter, F. J., \Tyckoff, SI. M., Peterson, E. A , . J . Am. C‘hem. SOC.78, 756 (1956). (7) Taylor, R. P., Furman, S. €I., AXAL. CHEM.24, 1931 (1952). RECEIVED for review July 6, 1961. rlccepted October 19, 1961. Work supported in part by the Elizabeth Storck Kraemer \ - - - - I
Memorial Foundation, Wilmington, Del.
Quantitative Gas-Solid Chromatog ra phic Determination of Carbonyl Sulfide as a Trace Impurity in Carbon Dioxide HARRY L. HALL The Coca-Cola Co., Atlanta 7 , Ga.
b Carbonyl sulfide at the parts per million level in carbon dioxide has been determined linearly over the investigated range of 0.5 to 2700 p.p.m. Minimum detectable corlcentration under these conditions is 0.3 p.p.m., with an estimated uncertainty at the 3-p.p.m. level of h0.07 p.p.m. The method is rapid and reliable and has been successfully applied to the determination of carbonyl sulfide in commercial carbon dioxide which produced carbonated water having offtaste and off-odor.
C
ARBONYL sulfide has been
found as a trace impurity in some commercial carbon dioxides in this country and abroad. In contact with water this compound slowly hydrolyzes to form hydrogen sulfide. A 1-p.p.m. carbonyl sufide content in carbon dioxide produces a discernible off-taste in some carbonated beverages. A faint odor of hydrogen sulfide will be present in most carbonated beverages if the carbon dioxide contains as much as 5 p.p.m. of carbonyl sulfide. A rapid, reliable, direct, and specific method of analysis
was required and provided the motivation for this work. Carbonyl sulfide has been determined in various gases by a number of methods, including spectrophotometric (9), potentiometric (3)’ colorimetric ( I I ) , selective solvent extractive (@, and gravimetric ( 1 ) . Xone of these methods satisfy all of the abovementioned requirements. A recent gasliquid chromatographic method ( l a ) was sensitive a t the 25-p.p.m. concentration level but with a +l5-p.p.m. uncertainty. Gas-solid chromatographic separation of carbon dioxide VOL. 34,
NO. 1, JANUARY
1962
61
from other gases was done as early as 1955 by Patton and coworkers (IO),and more recently carbon dioxide was analyzed for certain trace gases by Timms and coworkers (13). However, the gassolid chromatographic separation of carbonyl sulfide was not found in the literature. The use of large sample volumes in gas chromatographic trace analysis often calls for a column attachment designed to remove portions of the major component to reduce the separation problem and aid in the detection of the minor component by increasing its relative concentration. Brenner and Ettre (2) used a condensing column and samples as large as 40 liters, Lawrey (7) used a splitter column, and Farrington and coworkers (6) treated a 16.4liter sample by passing it through a drying agent. Such sample treatment is preferably avoidable, but is sometimes necessary for the detection of trace components. The procedure described here involved no special treatment of the sample to increase the concentration of the trace component. Rather the resolution and detectability were adequate for direct analysis of the amount of sample involved.
A Pye argon chromatograph with a 0to 10-mv. range Bristol Dynamaster recorder and a Lovelock beta ionization detector (8) was used. Detector voltage was 2000 and signal amplification was minimum (X10) unless otherwise indicated. Chart speed was 6 inches per hour. The 4foot glass column was 4 mm. in internal diameter, filled with 40- to 60-mesh activated silica gel and operated a t 25' C. Argon pressure was 10 p.s.i. with a corresponding flow rate of 200 cc. per minute. A 115-ml. glass-sampling valve was constructed especially for this determination. This valve, which fits into a ground-glass fitting a t the top of the regular column, was made so that a small rubber serum cap would tightly fit into one end, so that it could be evacuated via a hypodermic needlerubber tubing connection to a vacuum pump. A Hamilton 0.1-ml. gas microsyringe was used for measuring small volumes of gases. REAGENTS
Carbonvl sulfide. minimum 97% The MatKeson Co. ' Carbon dioxide. minimum 99.99%. with no carbonyl'sulfide or other de: tectable impurities. , - I
EXPERIMENTAL
The detector was calibrated with standard solutions of carbonyl sulfide in carbon dioside, using 0.05- to 0.40-ml. samples. The results are shown in Figure 1. For the range below 50 p.p.m. the signal was amplified by a ANALYTICAL CHEMISTRY
/
/
PElK
HEIGHT,
Calibration curve
factor of 3 (X3), giving increased noise but with continueigood reproducibility. Although a few inches of silica gel would have been adequate for this calibration separation, 4 feet were used to provide uniformity with the length used in the work with larger samples. The calibration data indicated that if satisfactory peak height were to be obtained in analysis a t the parts per million concentration level, large volumes of sample would be required. Additional standard solutions containing 0 to 10 p.p.m. of carbonyl sulfide in carbon dioxide were prepared using an 18.2-liter glass carboy fitted with a rubber serum stopper. Twenty milliliters of small glass beads in the carboy proved sufficient to mix the contents in a minute of vigorous agitation. Aliquots were withdrawn from the carboy by evacuating the sampling valve to below 5 mm. of mercury and allowing it to equilibrate via the rubber serum cap with the contents of the carboy. This procedure results in the removal of 114.3 ml. of the original sample, assuming complete evacuation of the sampling valve, I n this work, after the removal of one such aliquot, a new standard solution was prepared. When it is desired to sample the carboy repeatedly, the number of milliliters, V , of the original sample withdrawn a t n number of samplings is expressed by the following equation:
I I
i
200
100
Figure 1.
APPARATUS
62
3000 I
MY.
for carbonyl sulfide
Vn = 115 X
where V1, VI, and V 3 represent milliliters of original sample withdrawn at the first, second, and third samplings, respectively. After equilibration with the carboy contents, the valve was closed off and placed on top of the column. After the base line was established, the contents were swept into the column. The flow rate of 200 cc. per minute gave a satisfactory peak shape, as shown in Figure 2. Carbonyl sulfide retention time was 16 minutes a t a flow rate of 200 cc. per minute. For these dilute standards, the peak height us. carbonyl sulfide concentration relationship was linear. Using the 114.3-ml. samples, the peak height for the 3-p.p.m. carbonyl sulfide peak was 9 mm. From the data in Figure 1 a peak height of 19.1 mnl. was obtained using small concentrated samples. Thus in this case the use of a 114.3-ml. sample resulted in 9 peak height loss of 5370. RESULTS A N D DISCUSSION
Figure 2 shows that carbon dioxide and air are detected negatively by the Lovelock detector. Any gas having an ionization potential above the excitation potential of argon will generally be
1
CARBONYL SULFIDE
Table I.
Compound (Air) 02, P\T1 Cot
so2 cos
CzHz HIS Unknown CS2
Figure 2. Chromatogram of 3 p.p.m. of carbonyl sulfide in carbon dioxide Column. &foot silica g e l Flow rate. 200 ml./min. Temperature. 25' C. Sample volume. 114.3 ml. Carbonyl sulfide p e a k height.
9 mm.
tictected in this manner ( 4 ) . The shape of the carbon dioxide peak is symmetrical up to volumes of 0.5 ml., after 11 hich it becomes distorted. Large amounts have a strong sharp negative peak, followed by a weaker negative peak which tails. The air peak which s h o m in Figure 2 was admitted into the -tream from air pockets in connections leading to the sampling valve. Since this small amount of air did not interfere with the analysis, no effort was made to exclude it completely. Water, which is also negatively detected, did not manifest itself during the course of this work, and is apparently retained on the column. At one point, a 115-mI. sample of moist air was passed through the column with no effect on the base line. The column was used satisfactorily for as long as a week before the t ontents were sn-ept overnight a t 200" C. to remove moisture and maintain optimum condition. If impurities are present, the silica gel may be washed with concentrated hydrochloric acid without damage. Silica gel from different suppliers varies markedly in adsorption characteristics. One supplier's product would not separate carbonyl sulfide from acetylene, while the superior adqorption characteristics of another supplier's product provided a 100% increasc in peak height over that reported in this work. In the course of this work, acetylene and carbon disulfide trace impurities
Determination of Carbonyl Sulfide in Carbon Dioxide
Retention Time, Minutes
Peak and Detection Characteristics
1 . 5 (25' C.) 2.7 6.0 16.0 23.6 32 0 24.9 (50' C.) 46.0 (50" C.)
Neg.-sym.-linear Neg.-asym.-overload Neg. Pos.-sym.-linear Pose-sym.-linear Pos. - linear Pos.-sym.-linear Pos.-sym.-linear
were encountered and determined by this procedure. Other impurities were encountered but were not identified because of time limitations. The carbon dioxides analyzed in this work were manufactured by natural gas combustion processes. Table I shows data on the identified contaminants as well as on several others which could possibly be encountered. The hydrogen sulfide peak has at times been symmetrical and a t other times distorted with a sharp negative fall-off from the base line. The cause of such variation is not at this time known. 9 need for further study is indicated. The determinability of sulfur dioxide by this method has not been established because of time limitations. K'either sulfur dioxide nor hydrogen sulfide has been encountered as impurity in carbon dioxide in this work. The separation of air, carbon dioxide, sulfur dioxide, carbonyl sulfide, and acetylene is satisfactory a t 25" C. At 50" C. the separation of these compounds is not satisfactory. Because they have rather high retention times a t 25" C., the higher boiling compounds such as carbon disulfide are best chromatogrammed a t 50" C. The long retention time range a t 25' C. indicates a need for temperature programming. It is possible to increase the operating temperature from 25" to 50" C. in 23 minutes with very little base line shift. Increasing the operating temperature from 25" to 80" C. in 45 minutes results in a positive, gradual base line shift of only 28% of scale, thereby permitting preliminary runs to ascertain the elution range and number of compounds present. Isothermal operation between 25" and 50" C. was not possible because of lack of temperature selection and control between those two temperatures. All identifications were made by retention time comparisons and peak enrichment procedures using the pure compounds. It is believed that a longer column of greater diameter would permit the insertion of even larger volumes of sample,
Mininiuni Detect able Concentration, P.P. 11.
0 3 T'./v. 0 . 1 v./v.