Exhaust Gas Analysis by Gas Chromatography Combined with Infrared Detection WILLIAM 6. HEATON and JOSEPH T. WENTWORTH Research laboratories, General Motors Corp., Detroit, Mich.
The need for analysis of individual hydrocarbons in automobile exhaust has led to the development of a more sensitive gas chromatographic technique. A readable peak can be obtained from hydrocarbon concentrations as low as 1 p.p.m., using raw exhaust gas. The improved sensitivity is obtained by passing the effluent from the chromatographic column, first through a furnace to burn the hydrocarbons to carbon dioxide and water, and then through a nondispersive infrared analyzer sensitized to carbon dioxide. The improved sensitivity obviates the need for concentrating the hydrocarbons by condensing traps. The effect of fuel composition on the hydrocarbon composition of exhaust gas hus been determined by this method. Results obtained with both commercial gasolines and pure hydrocarbons show that the exhaust hydrocarbon composition i s dependent on the composition of the fuel.
I
air pollution, particularly in cities, has caused concern abou the contribution of automobile exhaust gas. Recent attention to the organic constituents of automobile exhaust has resulted in the general acceptance that most of these are hydrocarbons. According to current theory ( g ) , hydrocarbons react in the atmosphere and form objectionable air pollutants. Haagen-Smit and Fox (3) have shown that some hydrocarbons are much more labile than others in these dilute photochemical reactions. To assess the importance of hydrocarbons from automobile exhaust in relation to other sources of hydrocarbon emission (incinerators, refineries, manufacturing plants, etc.), an analytical method which mill quantitatively determine individual hydrocarbons in exhaust gas is desirable. Automobile exhaust gas contains a complex mixture of hydrocarbons, most of which are normally present in concentrations of less than 100 p.p.m. by volume. The complexity of the mixture requires fractionation of the exhaust gas into individual components for satisfactory results. To measure the low concentrations of individual hydrocarbons, the analytical method must
i
PARTITION COLUMN NO
ADSORPTION COLUMN
TREATMENT
C O P REMOVED
X50
KCREASKG
90
6ii
30
0
30
0
TIME-MINUTES
Figure 1 . Effect of carbon monoxide and carbon dioxide removal on cruise exhaust chromatograms Column conditions in Table
have an improved method of detection or a condensed phase from a large volume of gas must be analyzed. Complications in concentrating samples, coupled with C1 and possibly Cz hydrocarbon losses, make this latter approach less desirable. A satisfactory analytical method for determining exhaust hydrocarbons should determine individual hydrocarbons quantitatively and qualitatively, have a sensitivity of better than 10 p.p.m. (volume), require small enough samples to allow analysis without concentration of the original sample, be free of interference from nonhydrocarbon components, be inexpensive, rapid, accurate, easy to calibrate and
operate, and provide results that are easy to interpret. No past methods for the analysis of exhaust gas meet all these requirements. Gas chromatography can provide the required separations but conventional thermal conductivity detectors are not sufficiently sensitive unless the sample is concentrated. Infrared analyzers, while sensitive enough, do not, in general, determine individual hydrocarbons. Mass spectrometer analyses, even when coupled with gas chromatography, do not provide low cost, ease of calibration, and ease of interpreting the results. A combination of gas chromatographic separations and sensitive infraVOL. 31, NO. 3, MARCH 1959
* 349
2.0
1.5
2 U
z Y
21.0
a d
2 J
0
O
Figure 2. Diagram tion apparatus
of gas chromatography-infrared detec-
Figure 3.
Oxygen carrier gas supply Drierite Ascarite Gas sample in 18-ml. sample tube 56-ml. sample tube Serum cap 8 . Ascarite 9. Partition column 10. Copper oxide tube 1 1 . Furnace at 410' F. 12. Ascorite 13. Furnace at 970' F. 14. Catalyst tube 15. Nitrogen back flushing gas supply 16. Drierite
17. Ascarite 1 a.
19. 20. 21. 22. 23. 24. 25. 26. 27. 2 8.
29. 30. 31.
red detection could meet all these requirements. An improved technique has been developed, although there are certain unsolved problems. Automobile exhaust gas was analyzed by this method. Hydrocarbon compositions for different fuels and engine conditions are shown and compared to those obtained by other investigators. There has been recent controversy as to whether or not the hydrocarbon composition of the exhaust gas is affected by the composition of the fuel used in the engine and new evidence is presented. DEVELOPMENT O F METHOD
Chromatographic Detectors. Because of the low concentrations of hydrocarbons normally present in exhaust gas, the sensitivity of the detector is the key to successful analysis by gas chromatography. Early experiments, by the authors, Kith gas chromatographic analysis of exhaust gas showed t h a t a thermistor-type thermal conductivity cell lacked the required sensitivity. A Liston-Becker, Model IjA, Lufttype infrared analyzer (Beckman Instruments Co.), sensitized with n-hexane, was substituted for the thermal conductivity cell in the hope that it would provide greater sensitivity. An increase was realized and there was no response to gases such as oxygen and ANALYTICAL CHEMISTRY
2C 160 200 FLCJ PATE YLJMIN.
rib
.80
3iO
Effect of flow rate on area and peak height 100
1. 2. 3. 4. 5. 6. 7.
350
80
40
,D.D.m. '
Needle valve Ascarite Hopcalite Adsorption column Ascarite Soap bubble flowmeter Carbon dioxide detector Sample cell Reference cell Chopper Sources Infrared analyzer Amplifier Recorder
nitrogen which do not absorb energy in the infrared region. Adequate response to individual hydrocarbons was obtained from samples of exhaust gas collected during deceleration, when the hydrocarbon concentration is highest. Although encouraging, very little could be distinguished in samples collected during idle, and practically no response was obtained from samples collected during steady speed operation (cruise). It was apparent that for a detector to be practical, it must have enough sensitivity to respond to hydrocarbons from exhaust gas samples collected a t all engine conditions. Another problem with the n-hexane detector was that its response was different for different hydrocarbonsthe response to methane &-asmuch less than the response to an equal amount of n-hexane on either a volume or a weight basis. The response to acetylene was almost nil. Thus separate calibrations for each component would be required. Combustion-Carbon Dioxide Analyzer Technique. Martin and Smart ('7) described a gas chromatographic technique using an infrared carbon dioxide detector. Hydrocarbons were burned to carbon dioxide, over hot copper oxide as they emerged from the chromatographic column. Increased sensitivity was obtained by detecting this carbon dioxide with a nondispersive, infrared analyzer sensitized with carbon dioxide.
(ethylene) C2Hcadsorbent column
Besides increased .m*ni; ity, Martin and Smart found 1111 prcblem due to d s e r e n t responses I r a n different hydrocarbons. The c,iiIImtion problem n as solved because bl! hydrocarbons nere burned to c a r l m dioxide before they were measured. Thus only the peak area and carton number were needed to determine the hydrocarbon concentration. The method of Martin and Smart was adopted with minor changes. Oxygen carrier gas and a catalyst tube (PermaTherm, Burrell Corp at 970" F. were used in place of the nitrogen carrier gas and hot copper oxide tcb ensure complete burning of all hydrocarbons Analyses were made of a gas sample containing 400 p.p,m. of n-hexane in nitrogen, to compare the sensitivity of the combustion-carbon dioxide technique to that obtained from the thermal conductivity cell and n-hexane detector. The sensitivity using the combustioncarbon dioxide technique was more than 20 times that of the n-hexane detector. The thermistor thermal conductivity cell, using helium carrier gas, showed no response to the 400-p p.m. hexane sample. Further analyses of lower concentration sample gases and exhaust gas samples collected at all engine conditions showed that the combustioncarbon dioxide technique had the required 10 p.p.m. sensitivity. Exhaust Hydrocarbon Separations. Satisfactory separations of exhaust gas hydrocarbons through C, by gas chromatography required ta o separate columns. An adsorption column containing silica gel was used t o separate C1 through C3hydrocarbons plus the C1 paiaffins. This column, when operated under conditions necessary for adequate separations of the C1 through Ca components, would not pass C b olefins in less than 2 hours and hydrocarbons with higher carbon numbers (CSthrough C,) had even longer elution times. A partition column was used to separate
the C4 olefins and Cg through C, hydrocarbons. While this column would not separate any of the components separated on the adsorption column, adequate separations of Cb compounds through C7 olefins and including some CSparaffins could be made in less than 2 hours. The operating conditions are shon n in Table I. Chromatograms obtained using the columns and conditions listed in Table I and the combustion-carbon dioxide technique are shown in Figure 1. The peaks were identified on the basis of retention time data obtained with pure hydrocarbons. I n the Cg through C, hydrocarbon range, the identifications are tentative because of the many possible isomers. Where more than one compound is listed for one peak, it is understood that both (or all) compounds are possible components. Peaks emerging after benzene are labeled to give an idea of the carbon number and type of component with the retention time shown. The identifications do not imply that these are the only possible components of these peaks. Large peaks, such as the carbon monoxide and dioxide peaks in Figure 1, A and B , were attenuated to keep the trace on scale and the factors are shown on the chromatograms. For example, in Figure 1, A and B, the carbon monoxide and dioxide peaks were attenuated by a factor of 50. Problems with Combustion-Carbon Dioxide Technique. It is apparent from Figure 1, A and B , t h a t the carbon monoxide and dioxide in the exhaust gas sample interfere by masking out some hydrocarbon peaks. Two methods were tried to overcome this. The first method was to monitor the water, rather than the carbon dioxide, formed by the combustion of the hydrocarbons. An infrared detector sensitized with water was tried. The sensitivity was lower than the carbon dioxide detector and interference from the hydrogen present in the exhaust sample was as much of a problem as was the carbon monoxide-carbon dioxide interference with the carbon dioxide detector. Also, the water system gave erratic results, probably because of the tendency of &-ater to adsorb on surfaces, or even condense, before it passed through the analyzer. Thus, it was concluded that the combustion-water technique was unsatisfactory. The second possible solution to the carbon monoxide-carbon dioxide interference problem was to remove the carbon monoxide and dioxide and was more successful. A ii-inch('/d-inch outside diameter) column packed with Ascarite successfully removed the carbon dioxide (and water) but no appreciable amount of hydrocarbons. Figure 1, D and C, shows chromatograms obtained with the carbon dioxide removed. To ensure that the water present in the exhaust gas sample did not interfere with the activity of the column packings, the Ascarite was placed ahead of the chromatographic columns.
Table 1.
Chromatographic Column Data
Carrier gaB, oxygen. Column temperature, 75" F. Adsorption Column Partition Column Separates CI-C3, plus C, paraffins C6-C7,plus Cc olefins Packing Silica gel, 60-200 mesh 31 wt. % alkylated naphthalene (Shell Oil Co., T-533) on 30-80 (Davison, 950) mesh Celite 545 (Johns-Manville CO.) Column, inch 3 X 1(,(O.D,) (copper 7 X 8 , ' ~ (O.D.) (stainless steel tubing) tubing) 48.5 Flow rate, ml./min. (NTP) 146 Sample volume, at 760 mm. 18 56 Hg, ml.
0
-
I
I
I 0
A
A
I
METHANE ETHYLENE ACETYLENE PROPANE PQOPYLENE n-BUTANE
\
i-
ETHYLENE PROPYLENE PROPANE "-BUTANE ISOPENTANE "-HEXANE
0
A
A
-
0 0
L
-
1OW P.PM (VOLUMfl
2
4
6
100 PPM. (VOLUME)
CARBON NUMBER
Figure 5. I
2
3
Partition column calibration
4
Conditions in Table I
Figure 4. tion
Adsorption column calibraConditions in Table I
Several methods of removing the carbon monoxide were tried, but only the conversion to the dioxide and removal of this carbon dioxide prior to burning the hydrocarbons, was successfd. A 4inch ('/(-inch outside diameter) tube filled with active Hopcalite (Mine Safety Appliances Co.) would oxidize the carbon monoxide to carbon dioxide a t room temperature and yet would pass low molecular weight hydrocarbons without difficulty. The Hopcalite was followed by a 4-inch ('Ir-inch outside diameter) tube of Ascarite to remove the carbon dioxide formed from the carbon monoxide in the Hopcalite. Because water reduces the efficiency of Hopcalite, it had to be placed somewhere after the initial Ascarite used to remove the carbon dioxide and water present in the original sample. -4 chromatogram, made after these changes were incorporated in the adsorption column section of the apparatus, is shown in Figure 1, E. Interferences from carbon monoxide and dioxide were completely removed and all the light hydrocarbons through butane were readily identified and measured.
Hopcalite was not satisfactory for the partition column runs because it retained heavier hydrocarbons. A substitute was found in hot copper oxide (about 410' F.) which burned carbon monoxide satisfactorily. The copper oxide tube was placed after the partition column with a valve so that it could be bypassed. A partition column analysis using the copper oxide treatment followed by Ascarite is shown in Figure 1, F. Methane through n-butane are lumped into the first peak. The shoulder on the first peak is due to Cd olefins. After the Cq olefins, the gain could be increased, and hydrocarbon peaks, starting with isopentane, were easily measured. The fact that none of the hydrocarbon peaks in Figure 1, F , were visibly affected by the copper oxide treatment was provisionally accepted as evidence that no hydrocarbons were being lost. Most of the data dealing with exhaustfuel relationships were obtained with copper oxide in the circuit. However, it was occasionally suspected that olefins were being lost in the copper oxide. Subsequent investigation revealed that varying proportions of 1-pentene, 4methyl-2-pentene, diisobutylene, and VOL. 31, NO. 3, MARCH 1959
351
n-heptane were lost by combustion in the copper oxide. It was concluded that the copper oxide treatment has serious limitations. As shown in Figure 1, D, however, results can be obtained with the partition column without removing the carbon monoxide, although the carbon monoxide tail interferes with the C d olefin and C5paraffin peaks. Combined Chromatographic-Infrared Apparatus. The apparatus is shown in Figure 2. The oxygen carrier gas Tvas passed through Drierite and Ascarite, and then through a bypass sample system made with copper tubing and Hoke toggle valves. The two sample volumes (18 and 56 ml.) were filled in series, without interrupting carrier gas flow, by mercury displacement from a 250-ml. sampling bottle. The sample pressure was adjusted to 1 a t m . The tn-o columns \\-ere arranged with reversing valves so that while a separation was being made in one column, the other was back-flushed with dry, carbon dioxide-free nitrogen. The copper oxide furnace was installed after the partition column with a valve so that it could be bypassed during an analysis. The effluent from either column then passed through the catalytic furnace, the infrared analyzer, and a soap bubble flowmeter. The Liston-Becker, Model 15A, infrared analyzer was equipped with a 5-inch sample cell with sapphire windows. The output was applied to a Speedomax (Leeds 8: Northrup Go.) variable span recorder across a 2.5-ohm shunt. Standard gain was obtained with the infrared analyzer gain set a t 100 and the recorder span at 4 mv. For large peaks either the analyzer or recorder gain was reduced to keep the trace on scale. Calibration. The carrier gas flow rate was a n important variable controlling peak area and, incidentally, peak height. A 100-p.p.m. sample of ethylene in nitrogen TT as run a t various flow rates and the results are shown in Figure 3. The area under t h e peak was of primary interest because it was used t o determine t h e quantity of hydrocarbon in t h e exhaust analyses. Because area depends on flow rate, t h e flow rate was frequently checked, during the runs, Kith the soap bubble flowmeter. Calibration runs were made with a number of gases containing known amounts of various hydrocarbons. The results of the calibration runs 1%-iththe adsorption column are plotted in Figure 4. The area under t,he peak at standard gain is plotted against carbon number. Both Hopcalite and Ascarite were in the circuit as shown in Figure 2. Area is proportional to carbon number for a constant concentration. Or, area divided by carbon number is proportional to concentration. Thus, calibration is very simple. If the carbon number of a n unknown compound can be deduced from its retention time, its
352
0
ANALYTICAL CHEMISTRY
was 11% higher than the average, it was assumed that its actual concentration was 111 p.p.m. and the daily calibrations were based on this corrected figure. Because different flow rates and sample sizes were used, the area obtained for a given concentration and carbon number is approximately five times as great with the partition as with the adsorption column. This must be remembered when comparing relative peak sizes from chromatograms obtained with the two columns. Ultimate Sensitivity. Samples containing 10 and 1 p.p.m. of ethylene in nitrogen were also run t o determine t h e ultimate sensitivity of t h e apparatus. The 10-p.p.m. sample gave a good, readable peak when introduced into t h e adsorption column at t h e conditions listed in Table I. However, the peak due to t h e 1-p.p.m. sample could not be distinguished from the noise present in t h e output of the analyzer. With a 56- instead of t h e
concentration can be determined from the area under the peak and the calibration obtained with other, known hydrocarbons. The entire quantitative calibration for a column could be determined from a single run with one sample whose concentration and carbon number are known. The calibration data for the partition column are plotted in Figure 5. These data were obtained with the copper oxide furnace bypassed. Again, the various calibration gases checked one another fairly well. Isopentane was about 11% higher than the average calibration value and n-heptane was about 10% lower. One standard gas in each column was run as a daily calibration check The 100-p.p.m. ethylene sample was used for the adsorption column and the isopentane sample was used for the partition column. These gases were chosen because their short retention times allowed rapid daily calibration. Because the isopentane concentration
PARTITION COLUMN
ADSORPTION COLUMN
j
ACCELERATION
I
DECELERATO N
I 90
I
60
10 TIME-MINUTES
Figure 6. Effect of engine operating conditions on exhaust hydrocarbon composition Column conditions shown in Table I
18-ml. sample normally used with the adsorption column, a readable peak was obtained with the 1-p.p.m. ethylene gas sample. The lowest concentration found with the standard, 18-ml. sample size was a 3-p.p.m. ethane impurity in the 100-p.p.m. ethylene sample. The ultimate sensitivity was even better with the partition column. This is partly due to the larger sample used (56 ml.), and partly to the increase in peak height obtained a t the lower flow rate. I n addition, higher sensitivity to heavier hydrocarbons is obtained due to the increased volume of carbon dioxide obtained per molecule with the combustion-carbon dioxide technique. Liquid Analysis. For some tests described below, it was desired to analyze samples of gasoline using the same detector t h a t was used for the exhaust hydrocarbon analyses. Because of the high sensitivity of the detector, it was difficult to reproduce a liquid injection t h a t was small
enough using a hypodermic needle. To overcome this, the sample was diluted in a pure hydrocarbon, tertbutylbenzene, to a concentration of 0.4% (volume). Liquid injections of 0.02 ml. of the mixture allowed reproducible sample injections. The tertbutylbenzene was selected as the diluent because the retention times of tertbutylbenzene, and its impurities, were greater than the range of retention times of sample components of interest. Diluting the liquid samples with a high boiling liquid had an undesirable effect on peak shape and long tails and broad peaks were obtained. It is believed that these were caused by retention of the hydrocarbons in the tert-butylbenzene. The use of a capillary dipper liquid sample introduction system (10) might obviate the need for liquid sample dilution. EXHAUST-FUEL RELATIONSHIP
By using the apparatus and procedure described, the relationship between the
PARTITION COLUMN GASOLiNE A
-
ADSORPTION COLUMN
FUEL
exhaust hydrocarbons and the fuel a t various engine conditions was studied. Exhaust gas samples were collected during actual road tests using both commercial gasolines and pure hydrocarbons as fuels. Two 1957 automobiles in good condition were used and no adjustments were made between runs. Exhaust gas samples were collected through unheated copper tubing tapped into the exhaust pipe ahead of the muffler. After the automobile had warmed up, the fuel line was switched from the gasoline tank to a can of test fuel and the carburetor was allowed to run out of fuel twice to ensure that all of the warm-up fuel had been replaced by the test fuel. The automobile was then run approximately 15 minutes, on the test fuel, before sampling, to ensure equilibrium. Exhaust gas samples were collected in evacuated 250-ml. glass bottles. Samples were analyzed as soon as possible after sampling, usually within 1 or 2 hours. Effect of Engine Operating Condition. The relation between exhaust gas hydrocarbon composition and engine operating condition for a commercial gasoline is shon-n in Figure 6. All four runs were made in the same car. The idle and 30 miles per hour cruise runs were made under steadv conditions. The acceleration sample was collected as the car passed 30 m.p.h. on a full throttle acceleration from a standing start. The deceleration sample was collected as the car passed 30 m.p.11. on a closed throttle deceleration from 45 m.p.h. All runs were made on a leyel road. The same general hydrocarbon pattern mas obtained, with the partition column, for all four conditions. The reproducibility of the chromatograms obtained with the partition column suggests that these components (in the gasoline molecular n eight range) result from unreacted fuel. Except for butane, the components shown in the chromatograms obtained with the absorption column represent cracked products not present in the original fuel. Concentrations of the individual hydrocarbons shon-n in Figure 6 are listed in Table 11. The values for per cent unburned hydrocarbons were calculated as weight per cent of the supplied fuel from the formula: ( R t . so of supplied fuel) = Z[hydrocarbon (vol. %) X mol. w t . ] X av. of carbon number of fuel x 100 total carbon (vol. yo) x av. mol. wt. of
% unburned hydrocarbons
-
fuel
1IME MINU iE5
Figure 7. Comparison of fuel and 30-mile-per-hour cruise exhaust composition for commercial gasolines Column conditions shown in Table
The average molecular weight of the gasolines was assumed to be 98, the average carbon number, 7 (C,HI1). VOL. 31, NO. 3, MARCH 1959
353
The unburned hydrocarbons have been divided into two groups, unreacted fuel and cracked products. The percentages of each, as a weight per cent of the total exhaust hydrocarbons, are listed at the bottom of Table 11. Cracked products were defined as components not present in the original fuel; all other exhaust hydrocarbons were assumed to be unreacted fuel. The justification for this arbitrary division will be evident in a later section. The unreacted fuel was calculated by dividing the total C4 through C7 hydrocarbons (less the Cq olefins) by the total hydrocarbons. Similarly, the cracked products were defined as all CI through Cs hydrocarbons plus the C d olefins. The Cd olefins were assigned to the cracked products because only very small amounts were found in the fuel. The assumption was made that the propylene-n-butane peak was due to equal volumes of each component and was therefore divided between unreacted fuel and cracked products. Total exhaust hydrocarbon content
Table II.
was found to be the lowest a t acceleration (670 p.p.m.) and highest a t deceleration (6534 p.p.m.). The ratios of hydrocarbon concentration at the four engine operating conditions are cruise 1, acceleration 0.73, idle 2.0, and deceleration 7.1. The total hydrocarbon concentration obtained a t deceleration is highly dependent upon intake manifold vacuum (9), which in turn depends on the car speed. Therefore, the 30 m.p.h. deceleration value shown is not necessarily typical. At acceleration, cruise, and deceleration the percentages of unreacted fuel are about equal, while a t idle it is somewhat lower. The percentage a t deceleration is surprising because i t had been assumed that the exhaust hydrocarbons a t deceleration were mostly unreacted fuel. The analyses shown here, however, are limited to hydrocarbons below Ce and if higher molecular weight hydrocarbons were to be included, the per cent of unreacted fuel would be greater. While the ratios of concentrations of individual hydrocarbons are fairly con-
Exhaust Hydrocarbon Concentrations at Various Engine Conditions
[P.p.m. (volume)]
30 M.P.H. Idle Components Methane Ethane Ethylene Acety 1ene Propane
729 52 287 483
...
90 ... 62
Isobutane
Caolefins
Isopentane n-Pentane 2-Pentene 2-Methyl-1-butene 1-Pentene 2-Methyl-2-butene) 2-Methylpentane 2,3-Dimethylbutane) 3-Methylpentane +Hexane 1-Hexene Methylcy clopentane 2-Methylhexane 2,3-Dimethylpentane) Benzene 3-Heptene 3-Methyl-2-hexene) Total unburned hydrocarbon Total unburned as carbon, vol. Orsat analysis, vol. 70
70
CO2 0 2
co
Total carbon, vol. 70 Air-fuel ratio (weight) Unburned hydrocarbons (wt. Yo of supplied fuel) Unreacted fuel (wt. Yo of exhaust hydrocarbons) Cracked products (wt. Yo of exhaust carbons) 0 Less than 1 p.p.m.
354
0
ANALYTICAL CHEMISTRY
198 27
165 140 11
41 6.6
Cruise 309 36 248 121 Trace" 57
...
Deceleration 1890 334 1405 1500
...
493
...
15 6.7
38 19
143 87
8.0
5.9
12
95
Trace
4.0
...
59
9.8 5.0 7.1 ... 4.0 4.7 54 7.4
8.1 5.2
12 6.2 8.0
87 38 59
2.8 2.2 25 2.6
5.2 3.8 38 5.9
232 4i
12
i
Acceleration
3.6
...
...
6.0 31 28
919 6534 670 1815 0.2105 1.518 0.1503 0.3596 3.30 16.40 13.7
13.00 0.65 2.80 16.01 13.8
8.70 2.80 7.50 17.72
0.92
1.33
8.58
9.35 0.45 8.05 17.76 11.7
12.95
2.04
0.20
...
23.0
30.6
35.0
35.6
77.0
69.4
65.0
64.4
stant for the unreacted fuel components a t different engine operating conditions, there are wide differences in the ratios of the cracked products. For example, a t cruise there was twice as much ethylene as acetylene, while at idle there was almost twice as much acetylene as ethylene. Exhaust-Fuel Relationship Using Commercial Gasolines. A comparison of t h e exhaust hydrocarbons obtained at 30 m.p.h. cruise for two commercial gasolines is shown in Figure 7. T h e chromatograms for t h e fuels are shown above t h e chromatograms for the exhaust hydrocarbons. The similarity of t h e fuel and exhaust compositions a s obtained from t h e partition column runs for each gasoline supports t h e belief t h a t t h e exhaust hydrocarbon composition, in this range, is controlled by the fuel composition. The correlation is most evident in the comparison of the Ce paraffin peaks. The correlation is not so clear for the higher hydrocarbon peaks because of tailing and the broad peaks obtained with liquid injections of the gasoline. The concentrations of individual hydrocarbons in the exhaust obtained from the two commercial premium gasolines shown in Figure 7 are tabulated in Table 111. Fluorescent indicator adsorption (FIA) type analyses (ASTPVI D 1319) of the gasolines were included for reference. Analyses of cruise exhaust obtained from a variety of commercial gasolines shoiTed no correlation between fluorescent indicator adsorption-type analysis results, the total concentration of unburned hydrocarbons, or the per cent of unreacted fuel in the exhaust. Earlier similar conclusions have led some investigators to assume that the exhaust composition is independent of fuel composition. Figure 7, however, shows that the exhaust hydrocarbon pattern is related to the fuel composition. K i t h about 40% of the exhaust hydrocarbons composed of fuel components, i t appears that the exhaust hydrocarbon composition can be influenced by proper selection of the components comprising the gasoline. A wide range of unburned hydrocarbon concentrations a t cruise have been reported by other investigators and Table IV compares some of the results. Values reported in this paper are of the same order of magnitude as those obtained with both infrared and mass spectrometer analyses. Exhaust-Fuel Relationship Using Pure Hydrocarbon Fuels. Further evidence t h a t fuel controls t h e exhaust composition is shoivn in Figure 8. The exhaust hydrocarbons from three pure hydrocarbon fuels conclusively show t h a t a large portion of the exhaust hydrocarbons is unreacted fuel. A comparison of t h e exhaust
hydrocarbons produced from t h e three fuels shown in Figure 8, a Cg paraffin, a Cg olefin, and a Cg aromatic, also shows t h a t fuel type affects t h e amount and distribution of cracked products. The absence of tailing on the 4methyl-2-pentene peak of the fuel run is notable. The 4-methyl-2-pentene liquid was injected bv dipping the needle of a hypodermic syringe in a sample of the fuel without drawing any into the syringe. The needle was then inserted through the serum cap briefly and enough fuel evaporated from the needle to produce the chromatogram
shown in Figure 8. This procedure is not recommended, however, because it is impossible to inject a reproducible amount and fractionation of gasoline samples can occur. Table V lists the concentrations of individual hydrocarbons in the exhaust, at 30 m.p.h. cruise, for the three pure hydrocarbon fuels. A different division of the exhaust hydrocarbons into cracked products and unreacted fuel was made, assuming all hydrocarbons, other than the input fuel, to be cracked products. The wide difference between the amount of cracked products obtained
PARTITION COLUMN Z,3-DIMETHYL BUTANE
ADSORPTION COLUMN
- FUEL
I
Table 111. Exhaust Hydrocarbon Concentrations from Two Commercial Gasolines
[P.p.m. (volume a t 30 m.p.h. cruise)] Gasoline Brand Components Methane Ethane Ethylene Acetylene Propane Propylene) n-Butane Isobutane C4 olefins Isopentane %Pentane 2-Pentene 2-Methyl-lbutene
309 185 35.8 39.4 248 208 121 109 Trace& 12.0 58 3
I
i
38 b 19 1
103 Trace 43 0 33 3
12.2
2.4
Butene 2-Methylpe~tane 11.8 2,3-Dimethylbutane I 3-Methylpentane 6 2 8.0 n-Hexane Alethj-lc\-clo5 2 Dentane 2-hethy1-5hesene 2,3-Dimethy1-3pentene j 2-Methylhexane'j 3.7 2,3-Dimethyl pentane Benzene 37.5 2,2,+Trimeth!-lpentane ... 3-Heptene 3-Methy1-2hesene 2,4-Dirnethj-lhexane , . .
\
\
' ' '
,I
CMETHYL 2-PENTENE
-
FUEL
Total unburned hydrocarbons Total unburned as carbon, vol. yo Orsat analvsis. vol.
cop 0 2 co
B
h
920
13 0 6 0
3 2 7 9
Trace
5.2 41.0 2.1 1.0
5.1 82 1
0.2105
0.2155
13 00 0.65 2 80
13 00
16.01
15.32
13.8
14.0
% l "
Total carbon, vol. % Air-fuel ratio (wt.)
Unburned hydrocarbons, wt. yo of supplied fuel Unreacted fuel, wt. 70 of exhaust hydrocarbons Cracked products, wt. yo of exhaust hydrocarbons
1 33
0 80
2.10
1 44
35 0
41
65 0
58 1
45.4
59.B
li.3 37.3
17.3
(1
F I A type analysis, 9n
eo
bo
JO
0
30
0
I,YE--YINUT€S
Figure 8. Comparison of fuel and 30-mile-per-hour cruise exhaust for pure hydrocarbons
Pakiffins Olefins Aromatics a
23.1
Less than 1 p.p.m.
Column conditions shown in Table I
VOL. 31, NO. 3, MARCH 1959
355
from benzene, 106 p.p.m. (328 - 222 p.p.m.) and the amount obtained from 2,3-dimethylbutane, 714 p.p.m. (860 146 p.p.m.) shows the effect of fuel type on the amount of cracked products in the exhaust. The amount of unburned fuel, however, is greater in the case of benzene, 222 p.p.m. compared to 146 p.p.m. from 2,3-dimethylbutane. This unreacted fuel represents 83.5% of the total exhaust hydrocarbons in the case of benzene and 33.7% for 2,3-dimethylbutane. The analyses showed that the lowest
Table IV.
Reference
, .
(11)
percentage of unreacted fuel was obtained with the olefin fuel. During the analysis of the exhaust from the olefin fuel, the copper oxide was bypassed before the unreacted fuel was eluted from the partition column. It is possible, however, that C4, and some CS, olefinic cracked products were lost because of the copper oxide and such losses would make the per cent of unreacted fuel even smaller. I n the case of the 4-methyl-2-pentene exhaust, the amount of unreacted fuel was reduced with time. The 4-
Hydrocarbon Content Comparison to Other Investigations
Cruise Condition
en&e meed 30-m.p.h.
Present work 30 m.p.h. MS. Mass spectrometer. IR. Infrared analysis.
No. of Cars
1 1
Method of Analysis MS IR MS MS MS
Unburned Hydrocarbons ’ Wt. % supplied fuel P.p.m. (vol.) 2.2 ... 107 (as‘hesane) 5.3 3170 . .. 2240 ... 1.5
IR IR
0.94 0.80 1.39
2 fuels
190 (as gasoline) 158 (as gasoline) 870
Table V. Exhaust Hydrocarbon Concentrations Obtained from Pure Hydrocarbon Fuels
[P.p.m. (volume) a t 30 m.p.h. cruise] Fuel 2,34Dimethyl- Methyl-2butane pentene Component Methane Ethane Ethylene Acetylene Propane Propylene 1 n-Butane 1 Isobutane C, olefins Isopentane Bhlethyl-%butene\ 1-Pentene J Phlethyl-1-pentene 2,3-Dimethylbutane 2-Methylpentane PMethyl-2-pentene 3-Methyl-Bpentene 1-Hexene Methylcyclopentane 3-Methylhexane Benzene 3-Heptene 3-Meth yl-2-hexene Total unburned hydrocarbon Total unburned as carbon, vol. % ’ Oreat analysis, vol. yo
}
}
co2 co
156 40 155 88 15.6 177
356
ANALYTICAL CHEMISTRY
118 ...
51.2
28.2 ... 30.8 32.0 ... ...
9.7 .. 47.7 12.6 Trace5 . . . (Bypass CuO) . . . 23.0 6.1 ...
...
11.4
...
12.4 ...
... ... ... ... 1.5 ... ...
85.1
...
146
2.7
...
...
Tiace 1.2
2.5
222 Trace
... ...
...
860 743 328 0.2612 0.1765 0.1574
12.45 1.40 1.30 Total carbon, vol. yo 14.01 Air-fuel ratio (wt.) 14.6 Unburned hydrocarbons, wt. % ’ supplied fuel 1.86 Unreacted fuel, wt. % ’ total unburned hydrocarbons 33.7 Cracked products, wt. yototal unburned hydrocarbons 66.3 0 Trace amount, less than 1 p.p.m. 0 2
245 65 143
Benzene
12.00 0.70 2.55 14.73 13.6 1.22 28.4 71.6
14.55 1.30 2.55 17.28 12.4 0.93
83.5
16 5
methyl-2-pentene exhaust shown in Table V was anaIyzed about 1 hour after sampling. A subsequent analysis of a duplicate sample 4 days after sampling showed only 23 p.p.m. of unreacted fuel compared to 85 p.p.m. found 1 hour after sampling. Reactions betneen the olefin and nitrogen dioxide in the sample bottle may account for this reduction. The proportion of unreacted fuel obtained with the pure hydrocarbon fuels varied from 28.4 to 83.5%; corresponding values obtained from the commercial gasolines were 35.0 and 41.9%. The values for the gasolines are well within the range found for the pure hydrocarbons. This supports the assumptions made in dividing the hydrocarbons from the gasoline runs into cracked products and unreacted fuel. On the other hand, the presence of small amounts of cracked products in the Cj through C, range obtained from the pure hydrocarbon fuels shows that not all of the cracked products are in the C1 through C, group. Therefore, the division used in the case of the gasoline runs was not exact and must be regarded as an approximation. Other investigators have also studied thp effect of fuel nn the exhaust gas composition. Richards (8) and Magill, Hutchison, and Stormes (6) concluded that exhaust hydrocarbon composition was independent of the composition of the fuel. Both studies n-ere made using gasolines only, and it is possible that the range of gasoline compositions studied n-as too narrow to show a n effect. Data reported by Rounds, Bennett, and Xebel (9) for exhaust hydrocarbons from pure hydrocarbon fuels showed that appreciable amounts of unreacted fuel were present. The exhaust from aromatic fuels contained the highest per cent of unreacted fuel, and the exhaust from olefin fuels contained the lowest. Hurn ( 4 ) ,using a gas chromatographic analysis of condensed exhaust gas, reported that R hen using iso-octane as the fuel, 45% of total hydrocarbons in the exhaust was unreacted iso-octane. Hurn also reported that aromatics pass through the engine with less cracking than nonaromatics. He further reported a correlation between the olefin content of the exhaust and the olefin content of the fuel. These results agree, qualitatively, with the present work. I n addition, Mader and Chambers (6), by changing the olefin concentration of test gasolines from 0.7 to 32.5%, showed significant differences in the olefin concentration of the exhaust. Their experimental evidence showed that the olefin content of the exhaust increased as the olefin content of the fuel was increased.
The preponderance of evidence clearly indicates t h a t the exhaust hydrocarbon composition is dependent on the composition of the fuel. ACKNOWLEDGMENT
The authors thank the members of the Fuels and Lubricants Drpartment, Research Laboratories for their valuable criticism and advice. LITERATURE CITED
(1) Chandler, J. M., Cannon, W. A., Keer-
man, J. C., Rudolph, Arthur, J . A i r Pollution Control Assoc. 5 , 65 (1955).
IS.A., Rogers, L. H., Air Pollution Foundation (Los Angeles), Rept. 21 (October 1957). (3) Haagen-Smit, A. J., Fox, M. M., Ind. Ens. Chem. 48, 1484 (1956). (4) Hurn, R. W., Fifth Annual Anachem Conference, Association of Analytical Chemists, Dearborn, Mich., October 1957. (5) Mader, P. P., Chambers, L. A., First Technical Meeting, West Coast Section, Air Pollution Control Association, Los Angeles, Calif., March 1957. (6) Magill, P. L., Hutchison, D. H., Stormes, J. H., Proc. Second National Air Pollution Symposium, Stanford Research Institute: Menlo Park, Calif. (1952). ( 2 ) Faith, W. L., Renzetti,
(7) Martin, A. E., Smart, J., Nature 175, 422 (1955). (8) Richards, L. M., First Technical Meeting, West Coast Section, Air Pollution Control Association, Los Angeles, Calif.. March 1957. (9) Rounds, F. G., Bennett, P. A., Nebel, G. J., S . A . E. Trans. 6 3 , 591 (1955). (10) Tenny, H. & Harris, I., R. J., ANAL. CHEM.29, 317 (1957). (11) Twiss, S. B., Teague, D. M., Bozek, J. W.. Sink, M. V.. J . A i r Pollution Control Assoc. 5 , 71 (1955). RECEIVED for review January 24, 1958. Accepted October 13, 1958. Division of Petroleum Chemistry, 133rd Meeting, ACS, San Francisco, Calif., April 1958.
Gas-Liquid Chromatography Determination of Column Efficiency H. W. JOHNSON, Jr., and F. H. STROSS Shell Development
Co., Emeryville, Calif.
b Gas-liquid chromatography (GLC) can b e used to determine physical constants as well as for quantitative analyses. For the latter, only peak areas need be accurately known, but accurate determination of physical constants, such as partition and activity coefficients, requires that certain corrections b e applied to peak position and width for detector volume. Even in calculating column efficiency such corrections are important. The equations relating detector concentration to volume of carrier gas are derived for a model apparatus, which is shown to represent an actual gas-liquid chromatography apparatus with sufficient accuracy for this purpose. The necessary corrections for the detector volume within wide limits of retention volume, column efficiency, and detector volume can b e calculated by the use of graphs.
G
chromatography (GLC) is widely used to separate the components of chemical mixtures. The apparatus normally includes a detector and automatic recorder, so that the separated components are indicated on a chart as a series of peaks. 811 quantitative information must be obtained from the interpretation of these peaks. Each peak has three quantitatively useful parameters: peak height or area, peak position, and peak width. Peak height is a function of effluent concentration, sample identity, detector design, and other experimental variables (1,10). Peak height is not relatJed to separation. AS-LIQUID
Peak position and peak width both affect separation and their relationship is demonstrated in Figure 1. Gas-liquid chromatography is versatile because the peak position over wide ranges can be controlled by changing the liquid substrate (4-6, 8). However, with a n y given liquid substrate there will be materials with peak positions very close together and in such cases the peak width becomes very important. The actual peak width obtained on the chart determines the efficiency of the apparatus with the given column. However, such efficiency data are not sufficient. Chromatographers will be more interested in the performance of the partition column than in duplicating the entire apparatus. Even if there is interest in comparing apparatus, the actual peak width is of little value. It would be necessary to duplicate the column with the new apparatus under test, and it would still be difficult to decide if differences in peak widths were due to slight differences in the columns, or in the remainder of the apparatus. Gas-liquid chromatography is also used to determine physical constants (8, 9). Present uses are based on the linear relationship between peak position and the partition coefficient. However, peak width may also become important in the field of physical constant determination. The theoretical paper by van Deemter, Zuiderweg, and Klinkenberg (11) relates peak width to diffusivity and other kinetic quantities. Experimental evaluation and application of such a theory require accurate values for
A . U r r e s n l v e d !'ears
Figure 1. Relationship between peak position and peak width
the peak width due to the column alone. It would be desirable, therefore, to report the peak width due to the column independently from the peak width due to the remainder of the apparatus. This paper discusses the separation of peak ri-idth into column and noncolumn effects and the magnitude of errors that may arise when the entire peak width is attributed to the column alone. A comparison of the repeatability of peak n-idths with that of partition coefficients is made and a method for estimating column peak widths is described. PEAK WIDTH
Definition. Under ideal conditions. gas-liquid chromatography peaks a r e very close approximations t o normal distribution curves. These latter approach t h e base line asymptotically. VOL. 31, NO. 3, MARCH 1959
357