Anal. Chem. 1988, 58, 2247-2251 2247

molecular weight distributions can be calculated on the basis of the the combination LALLS and UV data and also on the basis of the UV chromatogram only.

Registry No. PUSHER 500, 54182-67-1; PUSHER 700, 39341-25-8; PUSHER 1000, 54174-00-4.

LITERATURE CITED (1) Seright, R. S.; Maerker, J. M.; Hoizworth, G. Polym. Repr. Am.

Chem. SOC., Div. Powm. Chem. 1981, 22(2), 30. Paper presented at the National ACS Meeting, Division of Polymer Chemistry, New York, August 198 1.

(2) Hoagland, D. A.; Larson, K. A.; Prud'homme, R. K.; I n Modern Meth- ods of Patflck, Size Analysis; Barth, H. G., Ed.; Wiley: New York, 1984; p 277.

(3) Larson, K. A,; Prud'homme, R. K. J. Coilold Interface Sci., in press.

(4) McGowan, 0. R.; Langhorst, M. A. J. Colloid Interface Sci. 1982, 89, 94.

(5) "Determination of Molecular Weight Distributions by Combining Low Angle Light Scattering with Gel Permeation Chromatography"; Applica- tions Note LS-2, Chromatix, 1977.

(6) Hoagland, D. A. Ph.D. Thesis, Princeton University, 1985. (7) Haas, R.; Durst, F. Rheoi. Acta 1982, 21, 566. (8) Durst, F.; Haas, R.; Kaczmar, 8. U. J. Appl. Polym. Scl. 1981, 26,

3125. (9) Farinato, R. S.; Yen, W. S. J. Appi. Polym. Sci., in press.

(10) Klein, J.; Conrad, K. D. Makromol. Chem. 1978, 179, 1635.

RECEIVED for review January 27,1986. Accepted May 12,1986. A patent has been issued to The Dow Chemical Company covering this method of analysis.

Supercritical Fluid Fractionation of Petroleum- and Coal-Derived Mixtures

Robert M. Campbell and Milton L. Lee* Department of Chemistry, Brigham Young University, Provo, Utah 84602

A supercritkal fluM chromatographic system was constructed to provide separations and fraction collection on a semipre- parative scale. Columns packed with silica materials of in- termediate particle sizes (30-70 pm) were used to allow dynamic pressure programming with minimum pressure drop of the CO, mobile phase along the length of the column. A variety of complex coal- and petroieum-derlved poiycycllc aromatic compound mixtures were fractionated according to the number of aromatic rings using columns packed with an NH,modifled stationary phase bonded on silica particles. The CO, moblle phase was programed with an aiternatlng series of linear pressure ramps and isobaric intervals to effect even peak spacing and near base line resolution of compounds of differing ring number in a coal tar. A solvent refined coal heavy distillate and a crude oil were similarly fractionated. Effluents were monitored with an uitravloiet spectrophotom- eter at 254 nm and a flame ionization detector while fractions were collected In pressurized vessels for subsequent analysls by caplliary gas chromatography. Sample capacities of up to 20 mg were possible with this system.

Supercritical fluids have been used with considerable suc- c e s both as mobile phases in chromatography and as solvents in extraction processes. While reduced analysis time is the major advantage of supercritical fluid chromatography (SFC) with packed columns (I), ease of solvent removal has been one of the major benefits of extractions with supercritical C02 (2). The supercritical fluid extraction of caffeine from coffee and of nicotine from tobacco are only two of the many uses of supercritical fluids for separations which have been re- ported. An extensive review of this topic is available (2).

In 1977, Wise et al. (3) reported the liquid chromatographic (LC) separation of polycyclic aromatic hydrocarbons (PAH) according to the number of aromatic carbon atoms using a chemically bonded aminosiiane stationary phase. Each of the ring-number cuts could then be further resolved using reversed phase LC or capillary column gas chromatography (GC). While the LC ring number separation has proven to be ex- tremely useful, the method suffers from several drawbacks.

Fractions collected from the NH2 column are concentrated by carefully directing a stream of nitrogen gas onto the sample until it is near dryness. Then the sample is dissolved in a solvent that is appropriate for reversed phase LC analysis. In addition to being slow and tedious, this process can result in sample losses and contamination. Quantitation can also be difficult with this procedure if the solutes of interest are volatile, photolytically unstable, or subject to air oxidation. Finally, the complex environmental samples that are fre- quently analyzed by this technique invariably, regardless of how carefully pretreated, contain material that is irreversibly adsorbed on the NH2 column. These columns are usually of the small-particle preparative scale variety and can be very expensive. After a number of fractionations, the columns tend to degrade due to these adsorbed materials, despite back- flushing with strong solvents.

An on-line multidimensional supercritical fluid chromato- graphic system was recently reported (4) which utilized both an aminosilane column in the normal phase and an octa- decylsilane column in the reversed phase mode. This system used supercritical C02 as the mobile phase for both columns and, by appropriate valving, eliminated the need for solvent removal after the NH2 column. However, the system was designed for analytical purposes and did not provide for sample collection.

It was the objective of this study to utilize the selectivity that can be obtained by programming the density of a su- percritical fluid in conjunction with columns packed with larger, less expensive particles to achieve separations similar to those that can be obtained with microparticulate HPLC columns. In several previous studies, packed columns were used with supercritical fluid mobile phases to improve the separation of closely related compounds over that which could be obtained by supercritical fluid extraction. For example, supercritical fluid fractionation (SFF) methods were used to isolate several PAH in an automobile exhaust extract (5). In another case, pressure programming of the mobile phase was used with packed columns to facilitate the separation of a series of polystyrene oligomers (6). In a study with capillary columns in SFC, where pressure drops across the column are minimal, nonlinear density programming yielded regular spacing of components of a polystyrene oligomer mixture (7).

0003-2700/86/0358-2247$0 1.50/0 0 1986 American Chemical Society

2248 ANALYTICAL CHEMISTRY, VOL. 58, NO. 11, SEPTEMBER 1986

cv1 cv2 cv3 cv4 Figure 1. Schematic diagram of the supercritical fluid extraction/ fractionation system: P1 and P2, syringe pumps 1 and 2; PH, solvent preheater/equiHbrator; EC, extraction column; SC, separation column; SV, six-port switching valve; FID, flame ionization detector; PT, pressure transducer; N,, high-pressure cylinder of nltrogen gas; CV1-4, pressurized fraction collection vessels equipped with cooling jackets.

These studies have shown that the ability to instantaneously vary the solvating power of a supercritical fluid by changing its pressure (or density) can be used to great advantage. With the selectivity of the mobile phase, compounds were resolved in SFC which would have required many more theoretical plates to resolve in gas chromatographic systems (8). This mobile phase selectivity can only be preserved by avoiding large pressure drops across the column, which can occur in columns packed with very small particles or in systems op- erated at very high flow rates (9). In the present study, columns packed with silica materials of intermediate particle sizes (30-70 pm) were used to prevent large pressure drops and to allow dynamic pressure programming for semiprepa- rative-scale separations.

EXPERIMENTAL SECTION Sample Origin. The coal tar used in this study was a me-

dium-crude coke oven tar and was obtained from the National Bureau of Standards, Washington, DC. This tar is a proposed standard reference material. The coal tar extract was obtained by eluting the crude coal tar with n-pentane after it had been adsorbed on silica gel. A solvent refined coal heavy distillate (SRC I1 HD), boiling point range 260-450 "C, was collected during the processing of a West Virginia coal from the Pittsburgh Seam, and obtained from the Fort Lewis, WA, pilot plant, which was operated by the Pittsburgh & Midway Coal Mining Co. This material is of pilot plant origin and should not necessarily be considered as representative of products that may eventually be produced on a commercial scale. The PAH fraction (IO) of the SRC I1 HD and a Wyoming Recluse Crude oil were also subjected to the SFF procedure.

Instrumentation. A supercritical fluid extraction/ fraction- ation system was constructed as diagrammed in Figure 1. The system included a 375-mL syringe pump (Isco, Lincoln, NE) and a 250-mL syringe pump (Varian 8500, Walnut Creek, CAI, each modified for pressure control at flow rates of up to 8 mL min-' (liquid), a chromatographic oven (Varian Series 2100, Walnut Creek, CA), and four 125-mL stainless steel fraction collection vessels that were fitted with cooling jackets. During fraction collection, the vessels were cooled to 3 i 2 "C via a circulating cooling bath (Grant Science/Eledronics, Dayton, OH). A six-port switching valve (Valco Instrument Co., Houston, TX) was used to effect collection of successive fractions in different collection vessels. The collection vessels were pressurized with nitrogen gas from a high-pressure tank with appropriate valving. A microm- etering valve (Autoclave Engineering, Erie, PA) was used to control the flow when the effluent was vented directly to atmospheric pressure.

The fractions containing the more volatile compounds were collected either by bubbling the effluent through methylene chloride or by venting the effluent through a 60-cm length of in. i.d. Teflon tubing which was immersed in a dry ice/isopropyl alcohol bath (-78 "C). To recover the collected materials, the Teflon tubing was removed from the bath and rinsed with a small amount of an appropriate solvent. An extraction column (10 cm X 4.6 mm i.d.) and a separation column (25 cm X 4.6 mm i.d.) were placed in the oven, and effluents were monitored with a UV-absorbance detector (Hitachi Model 100-10, Toykyo, Japan) equipped with a high-pressure flow cell (Hewlett-Packard,

Pmrruro (atm) 7 2 9 6 9598 106 1,OS 1,20 240

30 80 do Time (mid

Figure 2. Supercrttical fluid fractl0naMn of a coal tar with UV detection at 254 nm. Numbers refer to fractions that were collected.

Avondale, PA). A small fraction of the column effluent was split into a flame ionization detector (FID) through a 15 cm length of tapered 25 pm i.d. fused silica tubing. The FID block was shortened by half and heated to 400 "C with heat tape. All parts of the extraction/fractionatjon apparatus were constructed of stainless steel. Fractions were analyzed by capillary column gas chromatography using an HP 5880 gas chromatograph equipped with a 20 m X 0.2 mm i.d. fused silica capillary column coated with SE-54 (0.25 pm film thickness).

Procedure. A coal tar was fractionated by adding 1 mg of tar in 50 p L of methylene chloride to the top of the extraction column, which had been dry-packed with 40-63 pm silica (Sigma No. S-0507, St. Louis, MO). The column end fittings (equipped with 2-pm frits) were tightened and the column was installed in the oven by tightening the appropriate fittings. The separation column was dry-packed with NHz-Adsorbosil (3e70 pm, Applied Science, State College, PA). The oven temperature was raised to 40 O C and held there for the duration of the fractionation. The pressure of the COz mobile phase (SFC Grade, Scott Specialty Gases, Plumsteadville, PA) was brought to 72 atm and then raised to 95 atm at 8 atm min-'. The pressure was held at 95 atm until phenanthrene began to elute (fraction 3), whereupon it was raised at 1.5 atm min-* to 98 atm. As soon as the fluoranthene/pyrene peak (fraction 4) started eluting, the pressure was again pro- grammed at 1.5 atm min-' to 130 atm. At this point, as the chrysene peak (fraction 5 ) was finishing, the pressure was pro- grammed at 5 atm min-' to 198 atm and held for about 10 min. Fractions were collected as marked on the chromatogram in Figure 2. The other materials were similarly fractionated. After the C02 was removed from the collection vessels by venting to at- mospheric pressure, the solutes were removed from the vessels by rinsing with a small amount of solvent.

RESULTS AND DISCUSSION Detailed analyses of complex mixtures of polycyclic aro-

matic compounds (PAC) are important for several reasons. Many PAC have been shown to be mutagenic or carcinogenic and while some members of a closely related group of isomers may be biologically active, other members of the series may be inactive. Resolution of these isomers is therefore necessary, especially since selective detection methods for closely related isomers are extremely rare. Detailed structural information is also important for evaluation of products and process conditions of liquefaction, thermal cracking, and other pro- cesses. Many mixtures of PAC exhibit a wide molecular weight range and almost all are extremely complex. Some type

ANALYTICAL CHEMISTRY, VOL. 58, NO. 11, SEPTEMBER 1986 2240

Trmp ('C)

3'0 6'0 Time (mln)

F W e 3. CapHiary gas chromatogam of the fractions obtained during the supercrltical fluid fractionation of a coal tar. Numbers refer to fractions shown in Figure 2.

of fractionation is important for simplification of the mixture prior to analysis by high-resolution separation techniques such as capillary GC and capillary GC/mass spectrometry (GC/ MS). For example, ring-number fractions of complex mixtures of PAH have recently been obtained for analysis by GC with a liquid crystalline stationary phase which exhibits excellent selectivity for the separation of geometric isomers (11-13).

Figure 2 shows a UV chromatogram of the coal tar sample. Figure 3 shows capillary gas chromatograms of the fractions of the coal tar extract obtained on the SFF system. The polar amino bonded phase in conjunction with supercritical fluid pressure programming provided good selectivity for separation according to ring number. Alkylated species eluted at or near the same times as their parent compounds, while compounds of different ring numbers were widely separated.

A UV chromatogram of a supercritical fluid fractionation of the aromatic fraction of an SRC I1 HD is shown in Figure 4. GC chromatograms of these fractions are displayed in Figure 5. Again, while a higher degree of alkylation was present in this sample as compared to the coal tar, excellent separation according to number of aromatic rings was ob- tained. Fraction 1 contained the two- and three-ring com- pounds such as naphthalene and phenanthrene. Fraction 2 was compaed of the peri-condensed four-ring compounds such as fluoranthene and pyrene. Fraction 3 included the cata- condensed four-ring compounds such as chrysene, and fraction 4 contained the five-ring peri-condensed compounds (e.g., benzopyrenes).

To demonstrate the applicability of the SFF system to samples containing materials other than PAC, a crude pe-

4,O 1 eo 265

Prorruro (rtm)

30 6-0

Figure 4. Supercritical fluid fractionation of the aromatic fraction of an SRC I1 HD. Numbers refer to fractions that were collected.

Tim. (mln)

72 9s ep~os 190 200

I ! 3

40 265 2'0 40 do

me (mid Figure 5. Capiiiary gas chromatograms of the fractions obtained during the supercritical fluid fractionation of an SRC I1 HD sample. Numbers refer to fractions shown in Figure 4.

troleum oil was fractionated. In this case, the separation was mainly by boiling point, since the aliphatic materials had little interaction with the polar NH2 stationary phase. Boiling point fractionations are useful because the narrow molecular weight fractions obtained can then be easily further fractionated into chemical classes. A flame ionization detector was used in addition to the UV detector in this fractionation to monitor the effluents as shown in Figure 6. The gas chromatograms of the resolved fractions (Figure 7) demonstrate the exactness of the separation; little overlap of isolated fractions was ob- tained.

Standard compounds were used to study the effects of mobile phase density on resolution of closely related com-

2250 ANALYTICAL CHEMISTRY, VOL. 58, NO. 11, SEPTEMBER 1986

uv

60 4p Tmmp CC)

2'0 4'0 Tim. (mln)

9 2 160 { lo 1 i O 2bo Proaowo (rtm) I

Figwe 6. Supercritical fluid fractionation of a Wyoming Recluse crude oil. Numbers refer to fractions that were collected.

Table I. Resolution of Acenaphthalene and Biphenyl at Various Pressures of COz at 40 "C and Using a 30-70 pm NH2-Silica Column

pressure, atm resolution pressure, atm resolution

85 1.1 140 0.9 90 1.1 160 0.9

100 0.8 180 0.7 120 0.9 198 0.5

pounds. The resolution of biphenyl and acenaphthylene was measured at a number of different mobile phase densities, and these data are listed in Table I. It was found that, in general, the resolution of a pair of closely related isomers could be improved by as much as 100% by decreasing the mobile phase pressure at constant temperature on a given column. However, it should be noted that the particle size of the packing in the column has a large effect on resolution as well. Column ef- ficiency, which is dependent on the particle size of the column packing material, may have a greater effect on resolution than the selectivity of the mobile phase in supercritical fluid sys- tems as well as in liquid chromatography. Furthermore, analysis times can be shortened by going to smaller particle diameter column packings. Although large presssure drops across columns packed with very small particles can cause significant selectivity losses, the gain in efficiency in many cases may offset the loss in selectivity.

One additional factor that should be considered when choosing an optimum particle size for column packings in SFF is that large pressure drops across columns packed with small particles can reduce the molecular weight range of the solutes that can be eluted from the column. The maximum column pressure obtainable in a given system is equal to the maximum pressure rating of the pump minus the pressure drop across the column. Therefore, the use of columns with low pressure drops increases the maximum solvent strength of the mobile phase on the column and the maximum molecular weight of solutes that can be eluted from the column without adding a modifier. Particles in the 30-70 pm range were chosen for the separations that were studied here. The trade-off between column efficiency and pressure drop was a t or near optimum with this particle size. Smaller particles resulted in large pressure drops, while larger particles resulted in poor efficiency ( I , 6, 9). While LC with microparticulate columns provides

1

2

4

Tomp PC) 40 150 290

'do 40 60 Tim. (mid

Figure 7. Capillary gas chromatograms of fractions obtained during the supercritical fluid fractionation of a Wyoming Recluse crude oil. Numbers refer to fractions shown in Figure 6.

higher resolution for PAC mixtures than SFF with 30-70 pm diameter particles, the advantages of lower cost and higher sample capacity make the 30-70 pm packing an attractive alternative for many separations.

Samples were best introduced into the SFF system by ap- plying them in a small amount of solvent to the head of the extraction column with a syringe. This method resulted in narrower bands than were obtained when the sample was distributed over the entire extraction column. An external sample loop injection valve was not used in the system because at the lower pressures used at the beginning of a fractionation run, all sample components were not dissolved, resulting in plugging of the frit a t the head of the separation column.

UV monitoring of the column effluents was essential to enable precise cuts during fractionation on a routine basis as well as during development. Solutes in the fractions were best collected by cooling the collection vessels to 3 f 2 "C. This created a two-phase (gas/liquid) region in the collection vessels, causing the solutes to collect in the C02 liquid phase. The collection of fractions by use of pressurized vessels was not sufficient to contain very volatile solutes such as naph- thalene. Therefore, fractions containing very volatile com- ponents were collected by bubbling the effluent through methylene chloride or n-pentane. As mentioned in the Ex- perimental Section, very volatile hydrocarbons were also trapped by venting the effluent through a piece of Teflon tubing which was immersed in a dry ice/isopropyl alcohol bath. Typical analysis times for separations of components ranging from two to five rings were between 1 and 1.5 h.

With this pumping system, the sample capacity appears to be about 10 to 20 mg. per run on a 6.2 mm i.d. column. Sample capacity was double or triple that which could be obtained on a microparticulate column of equal dimensions. This can be explained by the fact that with larger particle size packings, efficiency is lower and the peaks are broader. This means that the solutes are in contact with more stationary phase, resulting in greater sample capacity. This phenomenon is discussed by Snyder (14).

A number of advantages were realized with this SFF system. The use of COz allowed an FID to be used to monitor the fractionations and to provide quantitative information, while

Anal. Chem. 1986, 58, 2251-2255 2251

C02 was easy to remove from the fractions that were collected. The selectivity produced by varying the density of the su- percritical fluid enabled the use of larger particle packings. The lower cost of these packings eliminated the need for extensive sample cleanup prior to analysis by SFF. In ad- dition, this procedure may be used to fractionate thermally labile materials, since C02 has a very low critical temperature. Overall, it was found during this study that semipreparative SFF shows great potential for high-quality separations with easy solvent removal.

LITERATURE CITED (1) a r e . D. R.: Board. R.: McManlaill. D. Anal. Chem. 1982. 54.

Graham, J. A.; Rogers, L. B. J . Chromatogr. Sci. 1980, 18, 75-84. FJeklsted, J. C.; Jackson, W. P.; Peaden, P. A.; Lee, M. L. J . Chroma- togr. Sci. 1989, 21, 222-225. Sie, S. T.; Rijnders, W. A. Sep. Sci. 1967, 2 , 755-777. Peaden, P. A.; Lee, M. L. J . Li9. Chromatogr. 1982, 5 (Suppl. 2),

Later, D. W.; Lee, M. L.; Bartle. K. D.; Kong, R. C.; Vassilaros. D. L. Anal. Chem. 1981, 53, 1612-1620. Nishioka. M.; Chang, H.C.; Lee, M. L. submitted for publication in En- vlron . Sei. Technol. Markldes, K. E.; Nlshbka, M.; Tarbet, B. J.; Bradshaw, J. S.; Lee. M. L. Anal. Chem. 1985, 57, 1296-1299. Nlshioka, M.; Whiting, D. G.; Campbell, R. M.; Lee, M. L. Anal. Chem.

Snyder, L. R. Rinclples of Adsorption Chromatography; Marcel Dek- ker: New York, 1968.

179-221.

1986, 58, 2251-2255.

. I . . - . . , 736-740. .

(2) Randall, L. G. Sep. Scl. Technol. 1982, 17, 1-118. (3) Wise. S. A.; Chesler. S. N.: Hertz. H. S.: HilDert. L. R.: Mav. W. E.

RECEIVED for review February 3, 1986. Accepted April 17, 19%. work Was supported by the Department of Energy, . .

Anal. Chem. 1977, 49. 2306-2310.

togr. Commun. 1985, 8 , 824-828. (5) Jentoft, R. E.; Gouw, T. H. Anal. Chem. 1978, 48, 2195-2200.

Office of Health and Environmental Research, Contract No.

Energy Program, Contract No. DE-FG22-83PC60807.

(4) Christensen. G. HRc cc. J . Hk7h Resolot. Chromatogr. Chroma- DE-AC02-79EV10237, and the Department of E ~ ~ ~ ~ , ~ ~ ~ ~ i l

Supercritical Fluid Fractionation and Detailed Characterization of the Sulfur Heterocycles in a Catalytically Cracked Petroleum Vacuum Residue

Masaharu Nishioka, David G. Whiting, Robert M. Campbell, and Milton L. Lee*

Department of Chemistry, Brigham Young University, Provo, Utah 84602

The sulfur heterocycles In a catalytically cracked vacuum resldue were Isolated by use of a newly developed ligand exchange chromatographic procedure, and they were then fractionated according to the number of aromatic rings by uslng a supercrttlcal fluid fractlonatlon system with supercri- tical COP as the fractionation solvent. Compounds In each fraction were Identified by capillary column gas chromatog- raphy and gas chromatography/mass spectrometry. Char- acteristic mass spectral fragmentation patterns were used to confirm Identlflcatlons, particularly for the alkyl-substituted compounds. The sulfur heterocycles In this sample were annellated on both sides of the thbphene ring and were highly alkylated; the most abundant compounds contained alkyl chalns with three carbon atoms. Several four-rlng cOmpOunds containing two sulfur heteroatoms were also detected. This example Illustrates the effectiveness of supercritical fluid fractlonaflon according to the number of aromatic rings for highly alkylated samples.

The petroleum industry has continually been troubled with various problems related to sulfur in petroleum, such as product odor and storage stability, corrosion of processing equipment, and pollution during usage. With regard to pollution, sulfur oxides produced during fossil fuel combustion are major contributors to air pollution (I) and can lead to acid rain damage to forests (2). For these reasons, hydro- desulfurization during refining is important. It is essential to identify the structures of sulfur compounds in crude oils and petroleum-derived products in order to more effectively

0003-2700/66/0358-2251$01.50/0

optimize these desulfurization processes. Desulfurization of heavy oils is more difficult than that of

light oils. This is because highly aromatized polycyclic aro- matic sulfur heterocycles (PASH) in heavy oils decrease the reactivity of desulfurization (3-5). These compounds are also known to poison catalysts used in catalytic desulfurization processes (6). High-molecular-weight sulfur compounds in high-boiling-point fractions and heavy residues are difficult to determine and have not yet been adequately characterized. This is primarily because they are highly alkylated and their identification in such complex mixtures is difficult, even when using high-resolution mass spectrometry.

The structures of sulfur compounds found in petroleum and the analytical methods for different types of sulfur compounds were reviewed by Drushel (7) and Dean et al. (8). Isolation or concentration of sulfur-rich fractions is essential for ac- complishing the required detailed analysis. Although a totally nondiscriminating separation method for sulfur heterocycles has not been found, a new isolation method for compounds with two to six annellated rings, based on ligand exchange chromatography, was recently developed by us ( 9 , I O ) . Further subfractionation based on molecular weight or number of aromatic rings is necessary in order to further simplify complex PASH fractions for final characterization (11).

In this study, PASH in a catalytically cracked petroleum vacuum residue were isolated by ligand exchange chroma- tography, followed by ring-number separation using a new supercritical C02 fractionation system. The resultant PASH fractions were then analyzed using capillary column gas chromatography (GC) and gas chromatography/mass spec- trometry (GC/MS) in order to provide detailed structural information.

0 1986 American Chemical Society