duction of the sarnple in 15 to 20 seconds leads to considerably higher sensitivities (in terms of the amount of sample required) than is obtained in conventional batch inlet mass spectrometry. Suitable mass spectral signals can be obtained with gram of a compound passing through the capillary column, and in special ca.ses, identification can be made with to IO-" gram. However, the smaller limits require that a very favorable niass hpectral pattern is formed, and in general, interpretation of a mass hpectral pattern may require as much as 30 t o 100 times the signal required for detection. The column bleed patterns shown in Figure 5 are typical and will vary a modest amount from colunin ta column and with uhe of the column. However, holm that less rigorous conditioning permits escessive bleed a t lT5" to 200" C. for most columns and frequently obscures the mass spectral pattern of the separated material.
ACKNOWLEDGMENT
The authors thank . J o ~ I I'cruzzi ~ for excellent cooperation in the) (Jonstruction of the instruments and I). C. Patterson for his help. LITERATURE CITED
(1) Brunnee, C., Jenckel, I,, Kronentierger, K.,10th Annual Jleeting of .4ST&I Committee E-14, S e i \ Orleans, T,a .Tune - ~~~- 1962 -( 2 )Buttery, R . G., lIcF:tdden, LV. H., Teranishi, I t , , Kealv, 31. P., l l u n , T. I Coree, J. LV., AI(.: Fadden, LV. H., Black, 1). It,, Llorgan, h. I., J . Food Sci. 28, 478 (19G3). (16) Teranishi, R., S i n i m o , C. C., Corse, J., ;\S.%I..CHEM. 32, 1384 (1960). RECEIVEDfor review Jlarch 26, 1964. 29, 1964. 2nd Interi i i r r i on Advances in Gas , University of Houston, Houston, Texas, ;\larch 2 - 2 6 , 1964. Reference to a company or product, name does not imply approval or recommendation of the product hy the C . S. Ikpartrnent of Agriculture to the esclusion of others that may he suitable.
Simulated Distillation by Gas Chromatography L. E. GREEN, L. J. SCHMAUCH, and J. C. WORMAN Research and Development Department, American Oil Co ., Whiting, Ind.
b Applications of gas chromatography to obtain distillation-type data have been reported, but efforts thus far have been aimed at reproducing low efficiency distillations. This paper describes a gas chromatographic method fo robtaining a boiling pointdistribution curve, equivalent to that obtained from a 1 00-theoreticalplate analytical distillation. An automated, temperature programmed instrument is coupled with an electronic integrator to print the per cent off on a paper tape at regular time intervals. The instrument is calibrated with a known blend of pure hydrocarbons. Only 1 pl. of sample is required. An average analysis takes one hour, which is only 1% of the time required for a distillation of comparable precision; furthermore the initial and final boiling ranges are better defined, and the method accounts for the total sample. One operator can use as many as three instruments simultaneously. The method has been applied to petroleum distillates with boiling points up to 1000" F., and to samples containing as much as 85% nonvolatiles. It has been used widely in bench-scale and pilot plant studies of hydrocarbon reactions, in process studies of crude oil distillation, in fuel blending, and in the analysis of natural gas condensates. 151 2
ANALYTICAL CHEMISTRY
S
R A L RAPIII but relatively crude distillation methods, such as ASTM 11-86 (I), are used commonly to determine t,he boiling range of ,petroleum fractions. 1Iany correlations have been made in an attempt to establish useful relationships between product properties or Ilrocess operating conditions and .\SIX boiling range. but their usefulness is restricted by the inherent error of drawing conclusions from limited data. Such one-plate distillations are also inaccurate in the initial and final boiling ranges and approach a true picture only near the midpoint. I3oiling range is determined much more accurately by a precise analytical distillation, commonly called true boiling point (1'13P) distillation. This term has many connotations, for there is no standard T13P distillation. For the purl)ose of this paper it is defined as a distillation on a column operating at an efficiency of 100 theoretical plates, xith data observations a t 1YG fractions. .I typical distillation requires at least 100 hours. The data have been used estensively for 1)rocess design and pilot 1)lant studies, and they nould be extremely useful for control of manufacturing processes. However, the time required makes the method of no value for control, and the expense greatly limits its use for researcah. 1,ike XSTlI D-86, the lengthy TI3T'
distillation fails to establish the initial or final boiling l~oints accurately. Seither method completely accounts for the total sample charged, and as the amounts of sample boiling below room temperature and above 500" F. incwase, theze deficiencies become more >eriou>. I%ecauseof dihtillation column holdup. false bottom>-high boiling compounds-must be added to the sample if all of the eaml)le is to br distilled. Even so. it is difficult to determine exactly where the sample ends and the false bottoms hegin. To prevent cracking of sample components, the distillation pot temperature niust not be allowed to rise above 500" I?. Khen that temlierature is reached, the distillation must be completed under reduced pressure. Thus, pressure variations, loss of noncondensable light ends, loss during pressure reduction and vacuum operation, and column holduIi very severely limit the confidence to be plaoed in Tf3P distillations of broad-boiling sanil~les. Gas chromatography offers a completcly neJV means of obtaining boiling ~)oint-distribution data. and several authors have re1)orted such apl)lications. Rggprtsen, Groennings, and Holst ( 3 ) dercribe a gas chromatographic method whicah yields distillation (lata rquivalrnt to that obtained from I O t o 20
EXPERIMENTAL
Apparatus. Figure 1 shows a photograph of the simulated diatillation instrument. Essential components are: chromatographic column, hydrogen flame ionization detector, electronic integrator, temperature programmer, digital voltmeter, printer, and recorder. CHROMATOGRAPHIC COLUMN. The column consists of stainless steel tubing, packed with 3% S E 3 0 silicone gum rubber on 30- to 60-mesh Chromosorb W, and is heated electrically by means of its own resistance. Columns as long as 20 feet have been used but the standard length is 6 feet. Dual column operation is most satisfactory because it cancels out base-line drift due to column bleed. Figure 2 shows details of the columndetector assembly. For simplicity only Figure I , Simulated distillation instruone column and one-half the detector ment are shown. Two identical U-shaped columns are attached to the bottom of theoretical-plate distillation columns. the detector base by tubing connectors However, the curves shown apparently threaded directly into the detector base. Stainless steel rods, silver soldered to are in error, because temperatures by the middle of the U-band a t the bottom both gas chromatography and column of each column, serve as electrodes in dist,illation are lower than corresponding the column heating circuit. Control ASTM values throughout the boiling and measuring thermocouples are silver range, Ilarras and lloyle (2) descrihr soldered directly to the columns. a similar gas ehromatograghic method, An insulating and coo1in.q jacket, but their data also shows the same formed by two concentric aluminum discrepancy between ASTM values tubes, surrounds the column assembly. and those oht,ained by gas chromatogThe aluminum tubes are closed at each raphy. Gaylor, et al. (4) used gas end by grooved aluminum plates. The chromatography in conjunction with whole assembly is held together by four esternal rods, threaded at both ends and a time-of-flight mass spectrometer to spring loaded to provide for expansion obtain composition data on crude oils. when heated, The detector, the upper Serne ( R ) used the same chromatoends of the columns, and the two lower graphic system to study optimization electrodes are insulated electrically of an atmospheric crude distillation from the aluminum jacket hy Teflon unit. He reported values closely apinsulators. The outer aluminum tube proximating those ohtained from ASTM is wrapped with asbestos paper. Chilled distillations. Petrocelli, Puzniak, and dry air is circulated thmugh the cooling Clark (7) describe a n automatic process jacket and column chamber to cool gas chromatographic analyzer for boiling t,he column rapidly to the starting point characterization of hydrocarbon temperature. These columns can be used with streams also equivalent to that obtained samples having a boiling range up to by ASTM D 86. Maier, Bossart, and 850" F. For samples having a final Heller (5)describe a commercial version boiling point above 850" F. 15-inch of this process analyzer for use with columns, packed with 0.2'%. of SC30 samples having an end point up to on ion- to 120-mesh g l w micro b e d s , 750" F. are used. The gas chromatographic method HYDROGEN FLAME IONIZATION DEhas been improved and instrumentation TECTOR. Figure 2 also shows a cutdewlolied to provide automatic simulaaway diagram of one half of the hytion of boiling-point distrihution curves drogen flame ionization detector. The equivalent t,o those ohtained from a fitainless steel base, housing the detector, contains two sample-charging 100-theoretical-plate analytical distillaand vaporizing passages, and is heated tion. Only 1 of sample is required, by a small electrical cartridge heater. and the average time of anlysis is one The vaporizing passages are filled with hour, which is only 1% of the time reuncoated 50-mesh glaB micro beads. quired for a distillation of comparable A tubing connector, containing a precision. The initial and final boiling silicone rubber disk, is screwed into the ranges are better defined, and the tal, opening of each passage, and the method accounts for the total sample. connector attached to the chromatoThe instruments are aut,omated, thus graphic column inlet is screwed directly enabling simultaneous use of three into the bottom. This provides a sysinstruments by one operator. The term tem of minimum volume and large heat, capacity for flash vaporization of simulated distillation (S.D.) has been the samplr. Nitro-n carrier gas, assigned to this method to distinguish it, reKulated by parallel flow controllers, from other less efficient methods.
Figure 2.
Column and detector detail
enters at either side UL LIE uetector near the top of each vaporizer passage. Parallel supplies of hydrogen are provided through passages drilled in the metal base. The dual jets are formed from l/le N P T pipe plugs to which Stupakoff insulators and platinum burner tips are silver soldered. The metal base also is tapped for Tefloninsulated shielded connectors for polarizing voltage leads and for two stainless steel rods, which serve as the upper electrodes for electrical heating of the columns. Opposing voltages of +300 and -300 VDC are applied through shielded cables to each jet. A single collecting electrode is mounted in the stainless steel chimney 1 cm. above the burners and centered equally between them. Air for combustion is supplied through an opening at the bottom of the burner cavity. ELECTRONIC INTEGRATOR. The eketrunic integrator consists of an operational amplifier with a capacitor feedback path to the input of the amplifier. Current through the feedback path balances ion-current from the detector while charging a capacitor. As the charge accumulates in the capacitor, the amplifier output voltage attains values which are integrals of the timeion-current products. That fraction of the output voltage across a voltage divider drives the digital voltmeter. Peak chromatograms also may be obtained by differentiating the output voltage to drive a strip chart recorder. When a simulateddistillation run is completed, the integrator is reset to 0 voltage by discharging the capacitor through a switch. T E M P E a A m R E PROGRAMMER. The temperature programmer is a proportional-contmller system with a motor driven reference voltage. Column temperature is sensed with a thermorouplr, VOL 36. NO. 8. JULY 1964
.
1513
850 -
800 -
750 700 650 600 550 500 450 400 350 300 -
// /*
I CI I
2
I50
I
I
cl4
.L
2
PRINT NUMBER
Figure 3.
whose voltage then is compared to that of the reference. As the reference voltage is increased to program the column temperature upward, a small error results from the comparison. This error is amplified with voltage amplifiers and a power amplifier, which in turn supplies more current through a transformer to raise the column temperature. The reference voltage is provided by a resistance bridge and reference thermocouple circuit similar to those used in temperature recorders. A 1000-ohm helipot in one arm of the bridge is driven by a synchronous motor to increase the temperature of the column at a rate of 9' F. per minute. Two low-level magnetic amplifiers provide enough amplification of error signals to drive a silicon controlled rectifier type power amplifier. A transformer matches the low resistance column to the output resistance of the power amplifier. A safety device is provided by a pyrometer and relays to stop the program automatically a t a preselected 15 14
ANALYTICAL CHEMISTRY
Calibration curve
masimum column temperature. Also included in the programmer chassis is a cycle timer for periodically giving print commands. I t consists of a clock motor, a cam, and a switch that is closed every 10 seconds. This function is automatically started every time the program is initiated. DATAREADOUTSYSTEM. The data readout system consists of a digital voltmeter, a printer, and a recorder. The digital voltmeter monitors the integrator output and, on remote control from the cycle timer in the temperature programmer, area counts are printed out on a paper tape. The printer also is modified to number each print in sequence on the tape. The recorder monitors the chromatographic peaks. Procedure. Dry air, a t the rate of 4 cu. ft./minute, is passed through the aluminum jacket and out the bottom of the column chamber until the column temperature is reduced to 120' F. Then a Dewar flask, containing liquid nitrogen, is raised around a coil of copper tubing, located upstream from
the inlet to the cooling jacket. The air flow is reduced to 1 cu. ft./minute, and cooling is continued until a column temperature slightly below -4' F. is achieved. The temperature programmer is adjusted to control isothermally a t -4' F. and the columns are allowed to equilibrate for 5 minutes. The sample (0.5 to 1.0 pl.) is injected into the analyzer column with a micro syringe. The programmer is switched to run position, and the analysis cycle proceeds automatically. As the temperature of the column is raised, the area under the chromatogram and the print number are printed out at 10-second intervals. The print number a t which the first significant rise in area count (greater than 5 per print number) occurs is chosen as the initial boiling point. The first print number a t which the area count becomes steady a t the end of the run is chosen as the end point. Samples containing nonvolatile materials are not completely characterized, and the run is stopped a t the print number corresponding to a boiling
!mint of 1000’ F. In that case the ))el cent off is determined by comparing the total area under the chromatogram to that of an internal standard. If this type sample is analyzed frequently, backflushing of the columns ma); be required occasionally to prevent buildup of heavy components in the column. Cumulative weight per cent off a t each successive print number is obtained by dividing the area count a t that print by the total area count a t the end point. From a calibration curve, a temperature is assigned to each print number. Thus the distillation curve-cumulative weight per cent us. boiling point-may be plotted. If a computer is used for the calculations, the tape is submitted directly to the computing center. CALIBRATION. A known blend of hydrocarbons, covering the boiling range expected in the sample, is used as a reference standard to relate print numbers to boiling points. The mixture is run in the manner described above. A reference curve is drawn by plotting the print number a t each plateau point us. the boiling point of that component. The plateau point is taken as the print number a t which the maximum area counts per print interval is obtained. If the chromatographic peaks are monitored simultaneously on the recorder, this point will correspond roughly to the peak maximum for each component. The reference curve and reference chromatogram are compared in Figure 3. Three aromatic hydrocarbons-benzene, toluene, and o-xylene --are included in the reference mixture to show that hydrocarbon type does not affect the curve significantly. RESULTS AND DISCUSSION
Precision. The precision of simulated dihtillation can be no better than the ability to reproduce the calibra-
Table
a
Lowest
F.
- 44
I.
Print number Highest
7
23 1 258
85 96
421 456 488 548 602 65 1 736 809
179 198 217 252 286 316 372 419
Precision of Calibration -~
Average 8 31 53 63 75 87 97 109 118 140 160
10
34 55 65 77 88 99 110 120 141 161 181 20 1 220 255 290 321 378 425
180
200 219 254 288 318 375 422
tion curve. T o test the precision, data from daily calibration runs were compared. These runs were made routinely by five separate operators over a period of sis weeks. The data are summarized in Table I. These data indicate that repeated runs on the same sample should agree within 1 2 ’ F. a t any point across the boiling range normally encountered in most petroleum fractions. The effect of column stability is also shown in these data. S o significant variation in retention times occurred during this six-week period. Column life routinely has been from 6 months to one year before replacement was required. Accuracy. The accuracy of simulated distillation was determined by analyzing a synthetic naphtha consisting of 37 hydrocarbons, ranging from Cs through Cj8) and including
Standard deviation Print so. “F.
2
0,9 0.9
3
1. o
2
1. o
2
paraffins, olefins, naphthenes, and aromatics. Table I1 shows the composition and boiling range of this mixture. Figure 4 shows a comparison of the simulated distillation curve with a theoretical curve, constructed from the composition and the boiling point of each component. The excellent agreement of the two curves indicates that the boiling point-distribution data ob-
Table
II.
Composition of Synthetic Naphtha
Component Wt 54 Cyclopentane 1 08 2,2-D1methylbutane 1 86 2,3-Dimethylbutane 2 79 4-Methyl-2-pentene 1 70 3-Methylpentane 1 88 1 95 1-Hexene Hexane 3 76 Methylcyclopentane 3 31 4 97 Benzene Cyclohexane 3 32 2,i-Dimethylpentane 3 86 8-Heptene 1 06 Heptane 3 88 2,2,4-Trimethylpentane 4 89 Methylcyclohexane 3 30 Toluene 6 13 1-Octene 3 08 2,2,5-Tr1methylhexane 2 03 Octane 5 07 Ethylbenzene 2 49 3 65 p-Xylene o-Xylene 3 76 Nonane 5 30 1 26 Isopropylbenzene 1 27 n-Propylbenzene 1 3,5-Trimethylbenzene 1 25 1,2,4-Trimethylbenzene 1 30 1-Decene 2 12 Decane 3 18 n-Butyl benzene 1 28 Irndecane 3 19 1-Dodecene 1 13 2 17 Dodecane Tridecane 2 18 1 15 1-Tetradecene 2 20 Tetradecane 1 20 Hexadecane
BdP., F. 121 121 136 137 146 146 156 161 176 177 177 208 209 211 214 23 1 250 255 258 277 281 292 303 306 319 328 337 339 345 362 385 416 421 456 484 488 548
7
t
6oo
~
4
IO
20
&
40
50
60
io
AMOUNT DISTILLED, WT % Figure 4.
do
do
)
Agreement with known composition ~~~~~
VOL. 36, NO. 8, JULY 1964
1515
tained by simulated distillation are accurate. Agreement with Distillation. Simulated distillation was compared with 'I'BP distillation on' four types of petroleum distillates. Distillations were made on 100-theoretical-plate spinning band columns (6, 9) with data points observed a t 1% fractions. All fractions were weighed. Where feasible saniples were chosen with initial boiling points above 120" F. and end points below 800' F. in an effort to minimize problems encountered in T B P distillation. Even so, some sample was lost in distillation. However, recoveries of sample were a t least 96%, and losses were prorated across the entire boiling range. =i comparison of distillation curves obt,ained with a heavy naphthakcrosene fraction from a Middle East crude is shown in Figure 5 . The solid line is the best smoothed curve plot'ted through the TBP data. Each data point obtained in a single simulated distillation run is plotted along the TI3P curve by the circles. The agreement between the two curves is excellent across the entire boiling range. S o false bottoms were added, and 1.4y0of the sample could not be taken overhead in the TBP distillation because of column holdup. Figure 6 similarly compares distillation curves of the products obtained from cracking of hesadecane. rlgreement of the two curves is very good above the midpoint, and loss of light components probably accounts for the shift of the T B P curve a t the front end. Yo false bottoms were added, and lOy0 of the sample was not taken overhead in the T B P distillation. Although the sample contained approsimately 30y0 unreacted hesadecane, the S.D. curve defines the long 548" F. plateau
700T-----600
i-
1
'
1
OOL
Figure
3b
20
IO
6.
40 50 60 70 AMOUNT DISTILLED, w T %
just as accurately as the T13P curve. I n fact over 50y0 of the sample boils within a range of less than 50" F., yet the S.D. curve characterizes this portion with the same precision as the T B P curve. Similar comparisons were made with samples of natural gas condensate an'd a heavy cracked naphtha. Agreement between S.D. and T13P curves was as good as shown in Figures 5 and 6. Applications. During the past four years over five thousand samples, including all types of petroleum dist,illates boiling below 850" F., have been analyzed routinely by simulated distillation. The method has been used for control of small-scale bench reactors as well as process units and has provided data for process design calculations. By the use of internal
~
-!
A
I
20
Figure 5.
1516
1
I
4'0 50 60 70 00 90 AMOUNT DISTILLED, WT % Distillation of heavy naphtha-kerosene fraction 30
ANALYTICAL CHEMISTRY
I
I
80
90
140
Distillation of products from cracking hexadecane
6001
IO
I
b
700
0
1
100
standards, the method has been extended to determine the amount of material boiling below 1000O F. in such samples as crude oils, reduced crude, coker gas oils, and heavy polymers that contain as much as 85y0 nonvolatile material. A domestic patent is pending on the simulated distillation instrument, which will be manufactured by Wilkens Instrument and Research, Inc., Walnut Creek, Calif., as licensee. ACKNOWLEDGMENT
The authors acknowledge the valuable assistance of H. M.Grubb in designing the electronic integrator and temperature programmer. LITERATURE CITED
(1) "ASTM Standards on Petroleum Products and Lubricants," Vol. I, p. 8, 1962. ( 2 ) Barras, R. C., Boyle, J. F., Or1 Gas J., 60, S o . 31, 167, July 30, 1962. (3) Eggertsen, F. T., Groennings, S., Holst, J. J , ANAL. CHEM. 32, 904 (1960). (4) Gaylor, V. F., Jones, C. S . , Landerl, J. H., Hughes, E. C., Sixth World Petroleum Coneress. Frankfort. Germany, June, 1 9 6 . ' ( 5 ) Maier, H. J., Bossart, C. J., Heller, H.,18th Annual Instrument Society of America Conference, Chicago, Sept. 1963. ( 6 ) Serheim, A. G., Dinerstein, R . A., ANAL.CHEM.28, 1029 (1956). ( 7 ) Petrocelli, J. A., Puzniak, T. J., Clark, R. O., I b i d . , 3 6 , 1008 (1964). (8) Serne, R. W,, 18th Annual Instrument Sorietv of America Conference. Chicago, Sept., 1963. ( 9 ) Winters, J. C., Dinerstein, R. A., ANAL.CHEM.27, 546 (1955). RECEIVED for review February 12, 1064. Accepted May 18, 1964. Presented at 2nd International Symposium on Advances in Gas Chromatography, I-niversity of Houston, Houston, Texas. March 23-26, 1964.
