H
=
height of a theoretical plate
(HETP)
p i n
k
= = =
minimumjl~~~ reduced HETP (Equation 5) column capacity coefficient = amount of solute per unit length in liquid phase divided by amount per unit length in gas phase column length eddy diffusion parameter, A =
L
x
= =
hT
= number of theoretical plates
T
= radius of capillary
in column
% t,
= standard deviation = retention time = linear gas velocity = value of u at column outlet = linear gas velocity a t minimum
v
=
W
= =
U
U0
w
LITERATURE CITED
(1) Carman, P. C., “I$Jw of
2hd,
tR U
designing and providing the integration and peak height measuring unit, and to the British Petroleum Co. Ltd., for a gift in aid of the research. Without this assistance the work could not have been carried out.
HETP reduced linear gas velocity (Equation 5 ) peak midth (Equation 10) parameter in gas phase mass transfer coefficient (Equation 4) ACKNOWLEDGMENT
The authors are very much indebted t o Bruce Peebles Ltd., Edinburgh, for
Gases through Porous Media, Butterworths, London, 1956. (2) Bohemen, J., Purnell, J. H., J . Chem. SOC.1961, 360. (3) Bohemen, J., Purnelf,, J. H. “Gas Chromatography 1958, Ed. Desty, p. 6, Butterworths, London, 1958. (4) Dal Nogare, S., Chiu, J., ANAL.CHEhf. 34, 890 (1962). (5) Deemter, J. J. van, Zuiderweg, F. J., Klinkenberg, A , , Chem. Eng. Sci. 5 , 271 (1956). (6) Deemter, J. J. van, Gas Chromatography Symposium, U. S . Public Health Service, Cincinnati, Ohio, 1957. (7) Desty, D. H., Geach, C. J., Goldup, A., “Gas Chromatography 1960,” Ed. Scott, p. 46, Butterworths, London, 1960. (8) Desty, D. H., Goldtp, A . , “Gas Chromatography 1960, Ed. Scott, p. 162, Butterworths, London, 1960. (9) Giddings, J. C., Ax.4~. CHEM. 34, 1186 (1962).
(10) Giddings, J. C., J . Chromatog. 2, 44 (1959). (11) Giddings, J. C., S u t u r e 184, 357 (1959). J. Chromatog. 5 , 61 (1961). (12) Giddings, J. C., Robison, R. rl., ANAL.CHEY.34, 885 (1962). (13) Giddings, J. C., Seager, S. L., Stucki, L. R., Stewart. G. H.. Ibzd.. 32, 867 (1960). (14) Gilliland, E. R., Ind. Eng. Chem. 26, 681 (1934). (15) Golay, &I.J. E., “Gas Chromatography 1958,” Ed. Desty, p. 36, Butterworths, London. 1958. (16) Jones, 1s’. L., AUAL. CHEV. 33, 829 (19611. (17) Kieselbach, R., I b i d . , 33, 23 (1961). (18) Kieselbach, R., I b i d . , 33, 806 (1961). (19) Littlewood, A. B., “Gas Chromatography 1958,” Ed. Desty, p. 23, Butterworths, London, 1958. (20) Norem, S. D., ANAL. CHEW 34, 40 (1962). ( S:; alsq D.al Nogare, S., Juvet, R. S., Gas-liquid Chromatography,’] p. 140, Interscience, Sew York, 1962.) (21) Scott, R. P. W., Private communication. (2’2) Scott, R. P. W., Hazeldean, G. S. F., Gas Chromatography 1960,” Ed. Scott, p. 144, Butterworths, London, 1960. RECEIVEDfor review December 11, 1962. Accepted February 5, 1963. Presented a t the International Symposium on Advances in Gas Chromatography, University of Houston, Houston, Texas, January 21-24, 1963.
Use of Catalytic Combustion Filaments for Qualitative Gas Chromatography 0.F. FOLMER,
Jr., KANG YANG, and GERALD PERKINS, Jr.
Research and Development Department, Continental Oil Co., Ponca City, Okla.
b Commercially available catalytic combustion detectors have been examined to determine whether they can b e made to respond selectively i o different molecular species. A selective response was found, but it was masked in many cases b y a residual thermal conductivity signal. Though present detectors may b e used inconveniently for qualitative discrimination, they are not a t all satisfactory for selective quantitative analysis. Quantitative accuracy is too adversely affected b y the residual thermal conductivity signal. A new mode of detection is described which measures the charge released during reaction. This latter detector has no residual signal, i s more sensitive than present detectors, and gives a more selective response.
T
o ACHIEVE a qualitative analysis, an instrument must respond to some unique combination of properties of each molecular species. I n chro454
ANALYTICAL CHEMISTRY
matography, the column usually performs this function. Unfortunately, those properties which lead to differences in retention time are seldom sufficiently unique to give an adequate qualitative analysis. With the samples encountered in the petroleum industry, massive interference is generally the result of such analyses. I n these situations the detector may be enlisted as a second dimension. If it is capable of responding uniquely to various molecular properties, it may be used in conjunction with the column t o extend the qualitative information available. Detectors which provide such information are constantly sought. So far, few have been found. Those which have been developed rely upon selectivity to achieve their qualitative function. They respond not to individual molecular differences but to differences between molecular species and are able to selectively discriminate among various homologous series in complex samples. Recently Schay, SzBkely, and Traply
( 5 ) ,in a study of the quantitative characteristics of the catalytic combustion detector, noted a differential response of this detector to certain classes of compounds. They found that alcohols and aromatic compounds reacted at a filament temperature of 150’ C; whereas, methane and chlorinated hydrocarbons did not react until 200’ C. This observation made no claims for a discriminatory response of the detector but its inclusion in their report would indicate that they attached significance to the phenomenon. Such a response could be exploited and extended to give discrimination among some, at least, of the species present in petroleum samples, and the compact, commercially available, combustion detectors might be readily utilized for this purpose. Accordingly, the present work was undertaken to study the qualitative, or discriminatory, characteristics of presently available catalytic combustion detectors. The results, as will be shown, were not entirely satisfactory when the detectors were employed conventionally.
However, a new mode of operation has been developed which does allow successful discrimination. The catalytic combustion detector should prove valuable in extending the range of the chromatographic technique, for, though its response depends upon the rate of the catalyzed reaction of each detected species, it depends, more fundamentally, upon the chemical and physical properties of the reacting species. On the other hand, the action of the chromatographic column depends mainly upon the physical properties of the species being separated. These two, then, reacting to both physical and chemical properties, complement each other, and the information derived from the combination should be more useful than that obtained with each alone. The catalytic combustion detector has been used for many years to detect combustible gases in mines but, as noted above, there has been little reported formal study directed toward its application. I n 1944 ( 2 ) it was applied to the detection of combustible gases in drilling muds. Then, in 1957, a chromatographic column was added and the detector was first used in gas chromatography. At this time Turkel’taub and Zhukhovitskii ( 8 ) reported the quantitative application of this combination t o mud logging and petroleum exploration. I n 1958 Greenbrier Instruments. Inc., built and sold the first of many process gas chromatographs using catalytic Combustion filaments as detectors. ?\lore recently, Slhdecek (6) described the application of this type of detector to quantitative laboratory analysis and, during 1962, Schay, SzBkely, and Traply (5) studied the relationship of certain operating parameters to the quantitative response of the detector, n ith particular reference to trace analJ sis. That no one has reported previously an evaluation of the qualitative properties of this type of detector is surprising considering the large volume of work in reaction kinetics, where the catalytic detector could become an important ekperiniental device. It is predictable from a knowledge of kinetics that the energy necessary to initiate combustion should vary from one molecular species to another. Mechanisms and nonequilibrium distribution of products should also vary. Results with the flame detector (4, ?) have indicated differences in the reactions of hydrocarbons, alcohols, and carbonyl and carboxyl compounds with oxygen. These differences might be amplified and exploited with a catalytic combustion detector. Lack of evidence of previous interest in this detector is also surprising in view of the fact that the response of this detector arises from chemical reaction and,
consequently, permits a choice of that reactant other than the sample constituents. This might allow, within a simple, convenient detector, a degree of selectivity now available only by means of external chemical tests or auxiliary reactors. The selective removal of interfering response would allow not only more convenient and accurate quantitative analyses, but also qualitative identification from retention data to be applied to more complex systems. The catalytic combustion detector, as presently employed, has a complex response which depends immediately upon the heat transferred to the filament during the combustion reaction and the heat removed from the filament by thermal conduction. Depending upon the carrier gas, these two factors may be additive or subtractive a t the moment of sample passage. Ordinarily, the heat liberated by the reaction is far in excess of any effect of thermal conductivity and, as a good approximation, may be considered the source of signal. The actual recorded signal, in a manner identical to thermal conductivity detectors, arises from the imbalance of a bridge occasioned by the changed resistance of the sensing filament. To secure a signal which is both measurable in magnitude and proportional to the sample size requires that the rate of reaction be rapid. This is realized by the catalytic initiation and subsequent chain reaction. At moderate temperatures and in the absence of catalysts, the gas-phase oxidation of many hydrocarbons and other organic molecules is immeasurably slow. The reactions are chemically quite complex. When catalysts are employed a t the lower temperatures, the products are usually few in number, with a chain length only slightly different from that of the original molecule. Depending upon the catalyst, one product usually predominates. I n the case of hydrocarbons, as the temperature is increased, the reactions produce a greater diversity of products which begin to include considerable amounts of carbon monoxide and mater. The behavior of hydrocarbons in particular is discussed a t length by Benson ( 1 ) and may be summarized as falling into three categories: (a) slow combustion which is immeasurable in our case; (b) orderly combustion which is measurable; and ( c ) explosion, which, in the catalytic combustion detector, results in a spike of immeasurable area. The limits of each type of behavior are rather well defined for any particular molecular species and a given set of conditions. Not all combustions present these three phases. I n the case of hydrogen there is a region of normal reaction, which may or may not have a measur-
able rate, and a region of explosion, resulting from a branching chain reaction. Most cases will present a region of useful reaction and a region of explosion. The general picture is one in which, under specific conditions, ( a ) no response is encountered, (b) measurable response is encountered, or (c) an, as yet useless, explosion is encountered. With sufficient control of conditions it should be possible to operate either with situation (a)or ( b ) . If conditions can be chosen to give condition (a) for a species A and condition (b) for a species E , the detector will discriminate between species A and B. Variables which are most readily controlled with the type of catalytic detector now commercially available are pressure, cell temperature, filament temperature-Le., filament current-gas flow rate and nature, and concentration of the reactant other than the sample. Of these, the filament current is most easily controlled and is most likely to provide selective action for any particular choice of reactants; furthermore, it was the variable noted by Schay, SzBkely, and Traply as producing some discriminatory response. Accordingly, for this investigation, it was chosen to demonstrate whether filacontrol of filament current-Le., ment temperature-will provide a selective response in catalytic combustion detectors and whether this selectivity can be put to use in analyses. Filament temperature itself \vas not measured because the introduction of sensing elements into the combustion chamber might have altered actual conditions.
EXPERIMENTAL
Apparatus. A conventional chromatograph with a four-filament thermal conductivity cell was modified by the addition of a catalytic combustion cell and a n inlet for a second gas stream. Figure 1 shows the arrangement of the modified instrument and the gas paths. Most of the work was done with argon as carrier gas. The argon was introduced through the carrier-gas inlet of the chromatograph. Only argon passed through the column and thermal conductivity detector. Oxygen was added to the argon stream between the two detector cells and this mixture flowed through the catalytic combustion cell to vent in the atmosphere. The composition of the gas mixture was cal~ of argon alone culated from the f l o rate and the flow rate of the mixture of argon and oxygen. These flow rates were measured with a soap bubble flowmeter at the vent. The catalytic combustion filaments were made by the Davis Emergency Equipment Co., Inc., and were obtained from Greenbrier Instruments, VOL. 35, NO. 4, APRIL 1963
455
X500~
NEEDLE VALVE PRESSURE ~
VARIABLE 2 rRESTRlCTlON
50 K 30K
CATALYTIC CELL
Figure 1.
CELL
l o 2
& A
a"x 2 1
Gas flow paths in the apparatus
IK
RECORDER .a
Inc. These filaments are very similar to coiled-wire thermal-conductivity filaments in appearance; the chief difference is that the wire is of larger diameter and is made of platinum. The resulting resistance is of the order of 0.5 ohm when the filament is cold. X-ray inspection indicates that the filament surface is metallic platinum and microscopic examination reveals that the surface is smooth. Electrically, the two filaments serve as arms of a Wheatstone bridge. This bridge was supplied with power from a 12-volt storage battery through a variable resistance (Figure 2 ) . The signal from the bridge was fed through an attenuator circuit to one input (5-mv. full scale) of a Bristol two-pen recorder. The signal from the thermal conductivity bridge was fed into the other input (1-mv. full scale). The thermal conductivity detector was operated a t a bridge current of 100 ma. and only indicated the presence of components for comparison purposes. The apparatus and operating conditions were such that less than 10 seconds elapsed between the recording of a peak from the thermal conductivity detector and the recording of the same peak from the catalytic combustion detector. The two-pen recorder then gave essentially simultaneous indications from each detector. The column used was a six-foot length of l/d-inch 0.d. aluminum tubing packed with 13.2 grams of 60- to 80-mesh Celite coated with 28.5% of Apiezon K. This column had previously been used at 320" C. and m-as well conditioned. Most sample injections were made with a 10-pl. Hamilton syringe. Because, at the operating conditions, 1.5 pl. of sample was extracted from the needle of the syringe, samples of 1 pl. or less required the use of a 1-pl, Hamilton syringe. Samples of larger size than 10 p!. were injected with a 50-p1. Hamilton syringe. To detect changes in electrical conductivity within the combustion chamber-Le., production of ions during combustion-a modified detector was constructed in the laboratory. The catalytic filament was enclosed by a cylindrical platinum electrode, and the auxiliary circuit shown in Figure 3 was employed to measure current flow between the filament and the cylinder. The electrometer was a Keithley Model 415, Keithley Instruments, Inc., Cleveland, Ohio.
456
ANALYTICAL CHEMISTRY
Figure 2. Schematic of the power supply, bridge, and attenuator with the catalytic combustion filaments
Reagents. All hydrocarbons were Phillips pure grade or better (Phillips Petroleum Go., Bartlesville, Okla.). The oxygenated compounds were all reagent grade. The gases, argon, oxygen, and helium, were obtained from the Linde Air Products Division of Union Carbide Gorp., New York, and were not further purified. When air was used as carrier gas, i t was drawn from the laboratory compressed air line and was passed through a filter of Linde 5rl Molecular Sieves. Procedure. T o determine the influence of the filament current on selectivity, a specific set of conditions was chosen for the operation of the detector. These were designed as about what would normally be encountered in a typical chromatographic analysis. The chosen conditions were as follows: temperature of cell, 70" C.; gas flow rate (measured a t cell exit under quiescent conditions), 40 ml. per minute; reactant gas composition, 20% O2 to 80% Ar; and sample size, about 1 pl. per component. With these conditions constant, a sample mixture of methanol, methyl ethyl ketone, diethyl ether, and n-hexane was repetitively analyzed a t different values of detector filament current. Sub-
-
300
b ELECTROMETER
TO
Figure 3. detector
RECORDER
Simplified circuit of ion
sequently, each condition was varied through a range generally met in practical situations, and variations in response were measured. Following this, results were obtained for certain interactions of variables n-hich might be expected in normal applications. Finally, some practical mixtures were analyzed with and without detector discrimination. As a consequence of the results obtained m-ith the conventional detector and to eliminate response due to thermal conductivity changes in the gas mixture, an entirely new mode of operation was employed. I n this case, the detector was modified as previously described, and the electrical conductivity of the gaseous mixture was recorded. RESULTS AND DISCUSSION
For characterization of the applicability of present catalytic coinbustion detectors to a qualitative function, combustion thresholds-Le., the lowest filament current at which the catalytic combustion filaments respond to a substance-were chosen as a criterion of selectivity. These w r e determined by repetitively injecting samples of the test mixture (methanol, methyl ethyl ketone, diethyl ether, n-hexane) while successively lowering the bridge current between each analysis. The response obtained in these experiments generally followd the pattern shown by curves a and b, Figure 4. For each compound there exists a small region of filament current values through which the response falls rapidly. If this region occurs for low values of filament current (about 700 ma.) , the response falls vrry nearly to zero. If the region occurs for higher values of filament current, a lingering and definite response occurs donm to low filament currents. This residual response may sometimes be negative-Le., of opposite polarity. Such a case is shown in Figure 8a. This residual signal is in
Table I.
Combustion Threshold Ranges for Components of Test Mixture
Combustion threshold," ma. Detector temp.. ' C. 100
70
40
Flow through detector ml. per min. 80
40
20
yo vol.
Oxygen cone., 30 20
n-Hexane
950150 9501.50
950150 950150 % O f 5 0
bIethano1
7501.10
790+10
790110 790110 790110 850150 950k50
720120
720120 Tf30+20
Sample size, 11
10
pl.
5 5
950150 550110 950k50 8 8 5 1 1 5 950150 950150
2.7 >1200
i90&10
7 6 8 3 ~ 5 8201.20
790110 800120
810110
720120
7301.30
720120 7 6 0 f 2 0
650110
630130
io0
630130
720f20
7603~20 720120