356
Anal. Chem. 1988, 60,356-362
of latex were kept low to obtain greater sensitivity, to prevent light scattering in the spectrophotometric measurements, and to inhibit possible agglutination of the latex. LITERATURE C I T E D (1) Engvall, E. I n Methods in Enzymology; Van Vunakis, H.-L., Langone, J. J., Eds.; Academic: New York, 1980; Vol. 70, pp 419-439. (2) Rubensteln, K. E.: SchneMer. R. S.;Ullmann, E. F. Blochem. Biophys , Res. Commun. 1972, 47, 846-851. (3) Green, N. M. Biochem. J. 1963, 8 9 , 599-809. (4) Wllchek, M.; Bayer, E. A. Immunol. Today 1984, 5 , 39-43. ( 5 ) Freda-Pletrobon, P. J. Dissertation, Lehlgh, University, 1983. (8) O’Sulllvan, M. J.; Gnernml, E.; Chieregatti, G.; Slrnmonds, R . D.; Simmons, M.; Bridges, J. W.; Marks, V. Anal. Biochem. 1979, 100, 100-108. (7) Bergrneyer, H. U.; Grabl, M.; Walter, H.-E. I n Methods of Enzymatic Analysls, 3rd ed.; Bergmeyer. H. U., Ed.); Verlag Chernie: Welnheim, 1978; Vol. 2, p 179
( 8 ) Schray, K. J.; Gergits, F.; Nledbala, R. S. Anal. Biochem. 1985, 149,
225-228. (9) Ngo, T. T.; Lenhoff, H. M.; Ivy, J. Appl. Biochem. Biotechnol. 1982, 7 . 443-454. (IO) Niedbala, R. S.;Schray, K. J. Clin. Chem. (Winston-Salem, N . C . ) 1985, 31,903. (11) Engvall, E.; Perlrnann, P. J . Immunol. 1971, 109, 129-135. (12) Ngo, T. T. I n Enzyme-medlatedlmmunoessay:Ngo. T. T., Lenhof, H. M., Eds.; Plenum: New York, 1985; pp 3-32. (13) Engvall, E.; Jonsson, K.; Perlmann. P. Biochim. Siophys. Acta 1971, 251, 427-434. (141 Kato. K.; Harnaauchi, Y.; Fukui, H.: Ishikawa. E. f u r . J. Biochem. 1976,6 2 , 285-j92. (15) Hibi, N.; Shima, K.; Tashiro, K.; Tsuzuki, K.; Yutaka, T.; Hirai, B. J. Neurol. Sci. 1984, 6 5 , 333-340.
for review
29, 1987. Accepted October 22,
1987.
Supercritical Fluid Chromatographic Determination of Hydrocarbon Groups in Gasolines and Middle Distillate Fuels Robert M. Campbell, Nebojsa M. Djordjevic, Karin E. Markides, and Milton L. Lee* Department of Chemistry, Brigham Young University, Provo, Utah 84602
A supercrktcal fluW chromatographk method to determlne the percentage of saturates, olefins, and aromatics In gasollnes and middle dktlRate fuels was developed. A microbore (1 mm Ld.) sHka gel column was used to Isolate and determlne the aromatics, while a sllver-Ion-loaded strong catlon exchange slllca gel mlcrobore column was used to separate the saturates from the detkrs and a m t l c s . oletlns were detennlned by difference or by uslng a column swltchhkg technlque whlch allows complete separatlon of saturates, oleflns, and aromatics In a slngle run. A mlxed moblle phase of 10% CO, In SF, (mde percent) was used for both columns, and a flame lonlzatlon detector provided unlform, hear response of the hydrocarbons, allowlng quantltatlon wlth mlnbnal callbratlon. The average absolute error was I S % ,while the standard deviatbn of the area percent of the aromatlcs of one gasaline sample was 0.5%. Analysls tlmes of under 60 mln were achieved.
The fluorescenceindicator adsorption (FIA) method, ASTM D 1319 ( I ) , has been used for many years by the petroleum industry to determine saturates, olefins, and aromatic hydrocarbons in petroleum products. This method suffers from many problems (2,3), including the following: (a) poor precision due to the operator’s inability to distinguish the borders of the colored bands, (b) long analysis time, (c) poor resolution between the hydrocarbon groups, (d) limited applicability to colored samples (i.e., higher-molecular-weight samples), (e) limited applicability to samples containing pentane or lighter hydrocarbons, and (f) lot-to-lot variations in dye composition. Due to these and other drawbacks of the FIA method, better methods for the determination of hydrocarbon groups have been sought. Among the most promising alternative methods have been nuclear magnetic resonance (NMR), mass spectrometry (MS), and high-performance liquid chromatography (LC). NMR gives results in terms of the number of aromatic
or olefinic carbon (or hydrogen) atoms, instead of weight percentage of molecules containing aromatic or olefinic functionality and, therefore, has not been as useful to the petroleum industry (4). While requiring expensive instrumentation, MS methods have proven to be very successful in determining hydrocarbon group types. However, heteroatomic species (S, N, or 0) present in the sample have been reported to cause interferences in the calculations of the hydrocarbon groups from the spectral information (5). Several LC methods for the determination of hydrocarbon types have been employed and have demonstrated short analysis times and excellent separation between the hydrocarbon classes. Suatoni and Swab (6) used a silica column with refractive index (RI) detection and an n-hexane mobile phase to determine saturates, aromatics, and polars. Aromatics were back flushed and polars were retained on the column and determined by difference. A cyanopropyl column has been coupled to a silica column and used to perform hydrocarbon group Beparations (7). The dtraviolet absorbance NJV) and RI detectors were used again with an n-hexane mobile phase. A similar system utilized a bonded aminosilane stationary phase to separate saturates, mono-, di-, and polyaromatics, and polars in kerosene and diesel fuels (8). One of the best of the LC methods utilized an activated silica column with a low-polarity, fluorinated hydrocarbon mobile phase and RI detection (9-11). Calibration was required, but the separation between saturates, olefins, and aromatics was excellent. This method was less successful with heavier samples; the resolution between high-molecular-weight saturates and olefins was only marginal. The high cost of the mobile phase was an additional drawback to this method. In an effort to improve the resolution between saturates and olefins, silver-impregnated silica columns were used (22-25) in LC systems. Such columns were used with carbon tetrachloride and infrared detection (IR) to fractionate gasdines into the hydrocarbon classes (26). Nonuniform response
0003-2700/88/0360-0356$01.50/0@ 1988 American Chemical Society
ANALYTICAL CHEMISTRY. VOL. 60. NO. 4. FEBRUARY 15. 1988 * 357
1 . 1 . 1
I
Figure 1. Schematic diagram of tlm single cdumn supercrltlcalfluid
fractionation system. factors hampered quantitation in this case as well. The main disadvantage of the LC methods is detection. Unwanted high specificity was observed with UV, RI, and IR detectors, while poor sensitivity was a further hindrance. Two other detectors have emerged with potential for application to hydrocarbon group analysis, the b e ionization detector (FID), and the dielectric constant detector (DC). The FID, although not compatible with most solvents, has been used in LC. Ita use with samples of high volatility is extremely limited, hut it demonstrates uniform response for the different hydrocarbon groups and wide-molecular-weight ranges. Recently, the Dc detedor has been demonstrated in hydrocarbon analysis to have uniformity of response when used with mobile phases that have high dielectric constants (17, 18). The separation of hydrocarbon types in this case involved the use of a strong cation exchange column which was loaded with silver ions. A Freon mobile phase (FC 123) was used in conjunction with hack flushing to effect the separation of saturates, aromatics, and olefins in less than 8 min. While demonstrating uniformity of response for both the different hydrocarhon groups and for hydroearhons of the same group type hut different molecular weight, the DC detector was not extremely sensitive. Detection limits for ole& were reported to be only 0.5%. The Freon mobile phase was also reported to he very expensive. I n an effort to utilize the advantages of the FID, supercritical carbon dioxide was investigated as a potential mobile phase for hydrocarbon group analysis (2,19,20). I n 1984, Norris and Rawdon reported such a system where saturates, olefins, and aromatics were separated on a silica column and a silver-impregnated silica column in series (2). We were unable to reproduce this work in our lab. Another approach to hydrocarbon group analysis utilizing the FID was recently reported by Schwartz and Brownlee (21). This method involved the use of supercritical sulfur hexafluoride (SFs) as mobile phase, which has a very low solvent strength, similar to that of pertluorinated hydrocarbons which were reported earlier to he successful for hydrocarbon separations in LC (9). Unmodified silica columns were used in the adsorption chromatography mode to effect the separation of the hydrocarbon types, and the flame ionization detector provided g d quantitative results. Analysis times were about 20 min. However, resolution between the saturates and the olefm was only marginal; 1-hexenecoeluted with decane. The method was not applicable to higher boiling distillates such as jet fuels. It was the objective of the present study to develop an accurate, reproducible, and rapid SFC method to separate hydroearbon mixtures into the chemical classes of saturates, olefms, and aromatics with quantitation provided by the FID. The only useful solvents which were compatible with the FID were CO1 and SF,. It was also desired that the method be applicable to samples in the carbon number range of C, to, C.,
Flgurb 2. Schematic diagram of the column-switching system for hydrocarbon group analysis by muitldimensionat supercritical fluM
chromatography.
EXPERIMENTAL SECTION Instrumentation. Schematic diagrams of the instrumentation used in this study are shown in Figures 1and 2. A Varian 8500 syringe pump, modified for pressure control as previously described (22) was used to deliver the mobile phase at flow rates up to 8 mL/min. Samples were introduced without dilution using a 0.2-pL internal sample loop valve (Valco C14W, Houston, TX). For rapid screening of different stationaryphases, large particle size packings (3W70 pm) were dry-packed in 25 an X 2.1 mm i.d. stainless steel columns. With this particle size adsorbent, sufficient efficiency was obtained by using columns which were dry-packed. The column effluent was split by use of a l/ls-in. Swagelok "tee" with a Swagelok-to-capillary conversion union (SGE SSMF/165, Austin, TX) on the FID end of the "tee". A piece of fused silica tubing (0.5 m X 50 pm i.d.) joined the "tee" to the restrictor mounted in the FID. The restrictor was one of three types: (a) fritted 50-Mm-i.d. fused silica (LeeScientific, Salt Lake City, UT), (h) tapered 50-pm-i.d. fused silica with a 300-pm-id. fused silica sheath, or (c) crimped platinum/iridium tubing (120-pm i.d., Hewlett-Packard, Avondale, PA). In all cases, the restriction was adjusted to give 15-100 mL/min (pas)into the FID in the pressure ranges used. A zero dead volume capillary union (SGE VSU/005) connected the restrictor to the fused silica transfer line. The tip of the restrictor was positioned in the FID such that it was 1to 2 em below the tip of the flame jet, which was a packed column jet, or one which had a larger orifice to accommodate larger flows. The FID hydrogen flow rate was 60 mL/min, the air flow rate was 425 mL/min, and the FID temperature was set at 325 OC. The gas chromatographic oven was equipped with an FID, but with no injector (HP 5890, HewlettPackard, Avondale, PA). The FID collector was gold-plated, and the combustion exhaust gases were removed hy suction through an aspirator when SF, was used as the mobile phase. The remainder of the column effluent was vented though a t a p e d fusedsilica restrictor which was trimmed to give a column linear velocity of approximately 0.5 cm/s. H e a t i i the restrictor to =300 "C prevented plugging and irregular flows associated with excessive cooling of the expanding COSor SF, mobile phase. When studies were conducted with microparticulate (5-pm particle diameter) stationary phases, glsss-lined microbore columns (25 em X 1mm id.) were used. The instrumentation for this was identical with that described above except that the 50-wm fused silica transfer line to the FID was connected directly to the outlet of the column, and the entire effluent from the column was directed into the FID. The direct connection of the fused silica transfer line to the 25 cm X '/I6 in. 0.d. microbore column was accomplished by use of a graphitized Vespel ferrule (SGE GVFOO5) with a 'filled in" Valco 'Ils in, nut, i.e. a 1/16-in.nut that had it's center plugged with silver solder, after which a hole was drilled in it just large enough to accommodate the 50-pm4.d. fused silica tubing (400-500 pm in diameter). A Teflon spacer in front of the ferrule eliminated the extra dead volume in the '/ls-in. female fitting at the outlet end of the microbore column. When reversal of flow in the column was necessary, a hack flushing valve (C6W, Valco) was used (see Figure 1). In this case,
358
ANALYTICAL CHEMISTRY, VOL. 60, NO. 4, FEBRUARY 15, 1988
the 50-pm fused silica transfer line to the detector was installed in the outlet port of the back flushing valve in the same manner as it was installed in the column outlet, as described above. A thin stainless steel fritted 2-pm filter (Mectron, City of Industry, CA) was added to the base of the fitting in the valve. This prevented plugging of the restrictor in the FID with small particles that had collected on the frit a t the head of the column and which were swept out of the frit during back flushing. A schematic diagram of the column switching system is shown in Figure 2. An Ag+ loaded column was placed outside the oven and was wrapped with heating tape, so that its temperature was maintained a t 40 "C. A silica column was placed in the gas chromatographic oven a t 50 "C. The columns were connected by three six-port valves, allowing us to switch columns out of the flow, or to reverse the flow through the silver-loaded column. All other parts of the system were as previously described for the single column system. The analog signal from the FID was fed to an electronic integrator (HP 5880, Hewlett-Packard) where it was converted to area slice data which was processed with a BASIC program that allowed the user to define cut points on the chromatogram and integrate the total area of each fraction without regard to peak shape. The base line, as established at the beginning of each run, was subtracted from each fraction, and the corrected area of each fraction was printed as a percentage of the total corrected area. Procedure. A 25 cm X 1 mm i.d. glass-lined Stainless steel tube in. 0.d.) was packed with 5-pm, 60 A average pore diameter silica (ES Industries, Marlton, NJ), installed in the supercritical fluid chromatographic system as shown in Figure 1 and described above, and evaluated by using supercritical SF6 mobile phase (Air Products, Tamaqua, PA). Saturates and olefins were eluted at 230 atm and 50 "C, while aromatics were eluted by programming the temperature to 150 "C as described previously (21). Three gasoline samples of known compositions, as determined by mass spectrometry, were analyzed. The compositions of these samples were reported to be accurate to within 2% absolute error. A series of six more petroleum-derived samples, including several middle distillate materials, provided by the ASTM committee on SFC Hydrocarbon Type Analyses were also used to further test the method. These six samples were (a) a high olefinic gasoline, (b) a low olefinic gasoline, (c) an aviation gasoline containing little or no olefins, (d) a jet fuel (kerosene), and (e and f) diesel fuel middle range distillates, which differ mainly in the types of aromatics present. In order to study metal complexation as a means to separate olefins from aliphatics, a number of transition-metal salts were selected, and each was deposited on silica (40-63 pm, 60-A pore size) and tested for retention of olefins with SF6 and/or cop mobile phases. Silica (40-63 pm, 60-A pore size, Sigma) was dry-packed in a 25 cm X 2.1 mm i.d. stainless steel column. After being packed, the column was installed in the supercritical fluid chromatographic system as described above and shown in Figure 1. The retentions of l-hexene and toluene were determined with supercritical SF6mobile phase a t 350 atm, with temperatures of 50 and 150 OC. Unmodified silica was tested first as a reference on which to compare the other stationary phases. These other phases were prepared, packed, and tested as described for silica above. As the final choice for separation of aliphatics and olefins, silver ions were loaded on a 25 cm X 1 mm i.d. column packed with Nucleosil5-rm SA (sulfonic acid) silica (catalog no. 71224, Alltech Associates, Deerfield, IL) by flushing with aqueous AgNO, until silver ions were observed to elute by testing with chloride. Formation of a white precipitate (AgC1) indicated the presence of silver ions. The column was then washed with 10 mL of water, rinsed with methanol, installed in the system (Figure l),and allowed to equilibrate with the CO, mobile phase. The system was operated as described above. A mixture of 10% (mol % ) COBin SF, was investigated as a mobile phase that could be used for both columns. The mixture was prepared in the following way. The first of two Varian 8500 syringe pumps was filled with SF6, isolated by closing both valves leading to the cylinder, and pressurized to 72 atm. Both pumps were thermostated at 2 "C with a circulating cooling bath. The volume of SF, in the first pump was noted. By use of this volume, the densities of CO, and SF, at 7 2 atm and 2 "C, and the molecular
weights of the two gases, the volume of CO, necessary to make a 10% (mole percent) mixture of C02in SF, was calculated. The second Varian syringe pump was then filled with an excess of CO,, pressurized to 72 atm, and vented until the correct volume as calculated above was attained. This pump was then isolated by closing the two valves associated with it. The calculations were made such that the total final volume of the mixture did not exceed the volume of one pump (250 mL). A valve on each pump was opened, which led to an interconnecting line, and the contents of pump 1 (SF,) were pumped over to pump 2. The entire contents of pump 2 were then similarly transferred to pump 1. This process was repeated several times to ensure that the fluids were well mixed. The final empty pump was isolated and the mixed fluid was delivered to the chromatographic column by opening the outlet valve on the full pump. The generalized equations used to calculate the C02 volume in the mixture are as follows: V-
v, = v,
- VI
(2)
where VT is the total volume of the mixture, VI is the volume of fluid 1 (SF,) in the mixture, % is the desired mole percent of the mixture (or moles of fluid 2 divided by moles of fluid l),dl is the density of fluid 1, MW1 is the molecular weight of fluid 1, d, is the density of fluid 2, MW, is the molecular weight of fluid 2, and V , is the volume of fluid 2 (CO,) in the mixture. The multidimensional system utilizing column switching techniques was operated as follows. The sample was first injected onto the silica column, which was maintained at 50 "C, and the aromatics were retained. The saturates and olefins eluted from this column into the second column, which was an Ag+-loaded strong cation exchange column maintained at 40 "C. In this column, the olefins were retained and the saturates were eluted and detected with the FID. The silica column, on which the aromatics were still retained, was switched out of the system, and the olefins were back flushed off the silver column. Finally, the silver column was switched out of the system, the silica column was switched back, and the aromatics were eluted by programming the temperature to 150 "C. All this was accomplished with one mobile phase, 10% C02 in SF,.
RESULTS AND DISCUSSION The adsorption chromatographic methods for hydrocarbon group separations were studied by evaluating a number of stationary phases for retention of olefins. Sulfur hexafluoride (SF,) was chosen as the mobile phase based on previous results (21). It is a very weak solvent, and early results showed C 0 2 to be too strong for separations of gasoline range samples with unmodified silica. The best results obtained previously in the literature were on silica (9,21);therefore, it was chosen as a reference. Other stationary phases tested included magnesia, an aminopropyl bonded silica, a sulfonic acid bonded phase, and a potassium hydroxide modified silica. However, none of the stationary phases tested were found to be superior to silica for separating olefins from aliphatics. The best results were obtained when a high surface area silica was used after drying by heating. Adding other materials to silica to modify the surface resulted in decreased surface area and decreased retention and selectivity for olefins. Since selectivity increased with surface area, it is postulated that the mechanism for retention of olefins on silica involves acidic sites on the surface such as active surface silanol groups or silicon atoms in the siloxane matrix having partially positive charges. This mechanism is supported by the fact that olefins are Lewis bases. After evaluation of the evidence, high surface area silica was chosen as the best adsorbent for hydrocarbon group separations. Table I lists retention data for selected standard compounds on the microbore silica column with SF6mobile phase. As can be seen, l-hexene eluted after n-dodecane. The best separation of the three group types that was obtained by using
ANALYTICAL CHEMISTRY, VOL. BO, NO. 4, FEBRUARY 15, 1988
Table I. Retention of Selected Standard Compounds on a Silica Microbore Column compound
carbon no.
tR"
ktb
' a
ad
6 8 10 12 15 22 5 6 6
1.55 1.80 2.07 2.48 3.45 6.83 4.27 3.08 19.0
0.85 1.1 1.5 1.95 3.1 7.1 4.1 2.7 21.6
0.58 0.75 1.00 1.33 2.1 4.8 2.8 1.84 14.8
0.60 0.77 1.00 1.33
n-hexane
n-octane n-decane n-dodecane n-pentadecane
n-docosane 2-pentene 1-hexene benzene
1.16 1.09 3.34
"Retention time in min, SF6mobile phase at 230 atm and 50 "C. * k ' = ( t R - to)/to,to = 0.84 min. e a = &'of solute/k'of n-decane. d a from ref 21.
S/aturates
Aromatics
Olefins I
r 5c
L
I
I
I,
Temperature ("C) 150
10
20
30
40
50
Time (min)
Flgure 3. Hydrocarbon group separation of standard gasoline 1: column, 25 cm X 1 mm i.d., Nucleosii 5-pm silica; mobile phase, SF, at 350 atm.
only one column during this study is shown in Figure 3. The resolution between the saturates and olefins was slightly better than that which was reported by Schwartz and Brownlee (21). The silica material used in the earlier work had an average pore diameter of 80 A (23) and therefore had a somewhat lower surface area than the silica used here. Despite the improved selectivity in this work as compared to the earlier work (21),it was observed that the resolution between the saturates and the olefins was still not sufficient for many applications and was especially inadequate for analyses of middle distillates such as jet fuels. Therefore, a study of the ligand exchange methods for hydrocarbon group analysis was initiated. Silver was known to cause retention of olefins as discussed in the introduction, but extreme irreproducibility was experienced for previous analyses which used silver-based separations. The present study was an attempt to develop an accurate and reproducible ligand exchange method for hydrocarbon group separations. The following metal salts were tested for their affinity for unsaturated hydrocarbons: copper(I1) chloride, copper(1) chloride, nickel(I1) nitrate, palladium(I1) chloride, silver(1) nitrate, silver(I), acetate, mercury(I1) nitrate. The metal salts were chosen
359
Table 11. Retention of Selected Standard Compounds on a Silver-Loaded Strong Cation Exchange Bonded Silica Microbore Column compound
tR"
n-octane
1.34 18.23 18.36
1-octene toluene
k 0 12.60 12.70
ORetention time in min, C02 mobile phase at 350 atm and 40 bk' = ( t , - to)/to,t o = tR of n-octane.
OC.
based on literature values for formation constants with olefins and aromatics (24))positions in the periodic table relative to silver, and reported or suspected soft acid/base properties (25). Most of the metals showed only weak interactions with olefins and aromatics; only mercury(I1) and silver(1) showed significant retention of the unsaturated hydrocarbons. Based on previously published results (18)and on results of the survey described above, it was decided to use silver(1) cations for hydrocarbon group separations in the ligand exchange mode. Silver showed the strongest interaction with olefins and aromatics and it appeared to have lower potential for catalytic activity than the other metals. Although olefins and aromatics could be readily eluted from silver-loaded columns with supercritical SF6 at 150 "C, it was determined that it was best to avoid heating the column in order to prevent degradation of the silver species and to prolong column life and reproducibility of retention times. On the basis of previous results (18)and results of the above studies, silica bonded with sulfonic acid groups was selected as the support material for the silver ions. The strong cation exchange silica was chosen because it had high surface area, was rigid (could withstand high pressures), and provided the best potential for reproducible loading of the metal on the supporting particles. In addition, the use of the cation exchange silica ensured that no silver ions would be washed out with supercritical fluid mobile phases. Silver could only be removed by flushing the column with a strong aqueous acid. Retention data of 1-octene and toluene on the silver-loaded cation exchange column with C02mobile phase are shown in Table 11. Alkanes were observed to elute almost unretained as established previously (2), while unsaturated hydrocarbons were strongly retained. Since unsaturates were strongly retained, back flushing was employed to reduce analysis time and improve peak shape. Recoveries of olefins and aromatics, and quantitation, were excellent with this column system. Although unsaturated compounds are cleanly separated from saturates on the ligand exchange system, olefins could not be separated from aromatics. However, on the silica column with SFs (adsorption mode), aromatics were well separated from both the saturates and olefins. Therefore, the two methods were used to determine the amounts of saturates, olefins, and aromatics in the hydrocarbon samples. Each sample was run on both of the columns, and percent saturates were determined on the silver-loaded column, while aromatics were determined separately on the silica column. Percent olefins were determined by adding the percent saturates to the percent aromatics and subtracting the sum from 100. A mixture of 10% (mol 70)COz in SF, was investigated as a mobile phase which could be used with both columns. Retention data for standard compounds on each of the two columns with the mixed mobile phase are shown in Table 111. The three known gasoline samples were analyzed to determine the accuracy of the method. Results are reported in Table IV. Resolution between the fractions was excellent, making cut points easy to define. The percent saturates of gasoline 2 was low. Some of the lighter saturates, such as the pentanes, probably evaporated during shipping or storage. With the
380
ANALYTICAL CHEMISTRY, VOL. 60, NO. 4, FEBRUARY 15, 1988
Olefins Aromatics
Table 111. Retention Data for Standard Compounds with 10% COz in SF, (mol %) on Silica and Silver-Loaded Strong Cation Exchange Silica Columns
silican compound
carbon no.
tRc
n-hexane n-octane n-decane n-dodecane n-pentadecane eicosane docosane 1-octene 1-octadecene benzene toluene
6
1.46 1.46 1.65 1.75 1.91 2.21 2.43 2.00 3.12 4.00 5.25
8
10 12 15 20 22 8
18 6 7
Saturates t Olefins
silver-loaded SA silicab
k'd
tRc
k 'd
1.53 0 1.53 0 0.12 2.10 0.37 0.19 2.53 0.67 0.31 3.57 1.33 0.52 7.15 3.67 0.67 9.15 4.98 0.37 30 24.9 1.13 >30 >25 1.93 17.4 10.4 2.84 12 9.3 0 0
*
25 cm X 1 mm i.d. column, 5-pm silica, 350 atm, 50 O C . 25 cm 1 mm i.d. column, 5-pm silver-loaded strong cation exchange column, 350 atm, 40 "C. c t =~retention time of solute. d k ' = (tR - to)/to,to = t~ of n-hexane. X
Table IV. Results of Hydrocarbon Group Analysis of Three Gasolines of Known Compositions
comDosition ligand exchange columnd diffr actual8
\-.-
f Backflush
wt %
sample
typea
adsorption columnb
1
S
56.9'
31.1
43.1 64.4c
68.ge 28.1
S
35.6 66.6'
71.9' 33.2
0 A
33.4
66.8e
0 A
2
S 0
A
3
31.1 25.8 43.1 28.1 36.3 35.6 33.2 33.4 33.4
31.0 24.3 44.7 32.2 32.6 35.2 31.9 34.8 33.3
i
4
errorh 0.1 1.5 -1.6 -4.1 3.7 0.4 1.3 -1.4
a
6
Time (min)
-
B
Temperature ("C)
50
exception of sample 2, agreement with the known values fell within 1.5% absolute error. Chromatograms of gasoline 1on the two different columns are shown in Figure 4. Triolefins were found to elute with or near benzene in the adsorption chromatography system with the 10% COz in SF6 mobile phase as well as with pure SFG,but they are not expected to be present in significant quantities in petroleum-derived samples and are not expected to be a problem. Accurate results were obtained for standard gasoline 3, even though it contained unusually large (10% ) amounts of triolefins. Although the FID response factors of some compounds can deviate from unity by as much as 10 or 15%, the response factors average out to unity for complex mixtures (21). Since olefins were determined by difference, their detection limit was dependent on the precision obtained in the determination of the saturates and aromatics. The standard deviation of the area percent of three determinations of a gasoline sample on the adsorption system was 0.56% for the aromatics fraction, which is close to the precision reported for a similar system (21). A series of five consecutive injections of an octane/ toluene standard mixture on the silver-loaded cation exchange column with back flushing showed a standard deviation of 0.12% for the area percent of toluene. Therefore, the detection
5
10
Time (min)
Figure 4. Hydrocarbon group separations of standard gasoline 1: columns, 25 cm X 1 mm Ld., (A) 5-hm silver-loaded strong cation exchange silica and (B) 5-hm silica; mobile phase, 10% CO, in SF, (mole %), 350 atm.
Table V. Hydrocarbon Group Analysis of Petroleum Distillate Samples
0.1
nKey: S = saturates, 0 = olefins, A = aromatics. b25 cm X 1 mm i.d. column, 5-pm silica, 10% COBin SF6 (mol %), 350 atm, 50 "C, then temperature program to 150 "C. % saturates + % olefins. d25 cm X 1 mm i.d. column, 5-pm silver-loaded strong cation exchange silica, COB,350 atm, 40 "C. e % olefins + % aromatics. f % olefins = 100 - [ % saturates (from ligand exchange) + 70 aromatics (from adsorption method)]. #Determined by MS, *2%. hAbsolute error.
150
sample description high olefinic gasoline low olefinic gasoline aviation gasoline jet fuel diesel fuel diesel fuel
adsorption group typea columnb S 0 A S 0 A S 0 A S 0 A S 0 A S
wt 7 0 ligand exchange columnd
63.7'
46.9
36.3 59.OC
53.1' 52.6
41.0 63.9c
47.4e 64.1
46.9 16.8 36.3 52.6 6.4 41.0 64.1
36.1 78.OC
35.9e 78.2
36.1 78.2
22.0 68.7c
21.8e 68.5
31.3 69.3'
31.5O 65.9
30.7
34.1e
22.0 68.5 0.2 31.3 65.9 3.4 30.7
0.0 0.0
0
A
differencef
"Key: S = saturates, 0 = olefins, A = aromatics. *25 cm X 1 mm i.d. column, 5-rm silica, 10% co2in SF6,350 atm, 50 "c then temperature program to 150 "C. % saturates + % olefins. d25 cm x 1 mm i.d. column, 5-pm silver-loaded strong cation exchange silica, COz, 350 atm, 40 OC. e % olefins + % aromatics. f % olefins = 100 - [% saturates (from ligand exchange method) + % aromatics (from adsorution method)l. limit for olefins is estimated to be about equal to the sum of the standard deviations obtained on the two columns separately (about l .5 % ). The difference method was further evaluated by analyzing the series of six ASTM samples described above. A summary
ANALYTICAL CHEMISTRY, VOL. 60, NO. 4, FEBRUARY 15, 1988
361
Saturates
Saturates
+
Olefins
Aromatics
Aromatics
I
JL f Backflush I
I'
10 20 Time (min)
Flgufe 5. Hydrocarbon group separations of a diesel fuel middle range distillate. Condltions are the same as in Figure 4, except neat COz mobile phase In (A).
of the results of the analyses is given in Table V, and chromatograms of one of the diesel fuels (e) on both the adsorption and ligand exchange column systems are shown in Figure 5. The determination of the hydrocarbon groups in these samples demonstrates the applicability of the method to middle distillates, in addition to the lighter materials. The excellent results obtained by using the difference method reported here led to the construction of a two-dimensional column switching system as shown in Figure 2. When used as described in the Experimental Section, results as shown in Figure 6 are obtained. A complete separation of saturates, olefins, and aromatics can be achieved in a single run.The analysis of standard gasoline 1gave 32.0% saturates, 23.6% olefins, and 44.4% aromatics, which is very close to the results reported in Table IV. In this example, the aromatics were allowed to elute through the silica column, instead of back flushing them off in a shorter time period. With elution, some aromatic structural information could possibly be obtained (mono-, di-, triaromatics, etc.). It is easy to see that such an analysis could be automated for routine determinations. Various organic solvents were observed to have effects on solute retention and on the stability of the silver columns. Solvents having a bonds, such as toluene or acetonitrile, either were used, by some investigators, to apply silver nitrate to the silica support or were used as solvents in slurry packing procedures for silver-loaded HPLC columns. These solvents formed complexes with the silver ions, deactivating the column toward retention of unsaturated analytes. Columns resulting from these procedures produced widely varying solute retention times while the mobile phase gradually removed the solvent which had complexed with the silver ions. When silver nitrate was loaded on an amino-bonded phase silica, no retention of olefins or aromatics resulted when the column was
I
I
0
30
Tirne[rnin)
Figure 6. Hydrocarbon group separation of standard gasoline 1 using the two-dimensional column switchkg system. For conditiins, see text.
tested. All organic compounds containing nitrogen are expected to complex strongly with silver cations. While alcohols are commonly used as solvents for silver nitrate, these compounds and other oxygenated species including aldehydes, ketones, and glycols, can react with silver speciea at higher temperatures (26). When methanol was used as a solvent, silver nitrate/silica mixtures were gray after heating overnight at 110 "C, while silver nitrate/silica mixtures in water were white after similar heat treatment. The gray color was due to silver metal adsorbed on the silica. Oxidationlreduction reactions between silver species and oxygenated compounds are well-known, although they are not particularly synthetically useful (27). Halogenated organic compounds can also react with silver nitrate to form the silver halide and nitrate esters (25). Reactivity is highly dependent on structure, with tertiary halides being much more reactive than primary halides. The secondary halides, R2CHC1,are reported to be of intermediate reactivity. Reactions of halides of marginal reactivity can be facilitated by heating. Clearly, certain types of compounds should be avoided when using silver-treated columns. Among the classes of compounds to avoid are amines, aldehydes, glycols, alcohols, sulfides, and halogens. Above all, silver-loaded columns should never be subjected to high temperatures. Fortunately, these types of compounds are not generally found in petroleum-derived samples to be analyzed. If such compounds are present in the sample, the silver-loaded column could be protected through the use of a silica guard column, although this was not found to be necessary in the present study. When the columns were prepared and used as described in this paper, results were accurate and reproducible. Hundreds of injections were made on both the silica column and the silver-loaded column without noticeable losses of efficiency or resolution.
ACKNOWLEDGMENT The authors wish to thank Vince Giarrocco and Larry Altman (Mobile), Wolf Schultz (Exxon), Paul Hayes (Jet Propulsion Lab), and Lane Sander (National Bureau of
362
Anal. Chem. 1968, 60,362-365
Standards) for helpful discussions, suggestions, and samples for analysis.
LITERATURE CITED Manual on Hydrocarbon Analysis, 3rd ed.; American Society for Testing and Materials: Philadelphia, PA, 1977. Norris, T. A.; Rawdon, M. G. Anal. Chem. 1984, 5 6 , 1767-1769. Norris, T. A.; Shively, J. H.; Constantin, C. S. Anal. Chem. 1961, 33,
1556-1558. Petrakis, L.; Alien, D. T.; Gavalas, G. R.; Gates, B. C. Anal. Chem. 1983, 5 5 , 1557-1564. Ozubko, R. S.;Clugston, D. M.; Furimsky, E. Anal. Chem. 1981, 5 3 , 183-187. Suatoni, J. C.; Swab, R. E. J . Chromtogr. Sci. 1975, 13, 361-366. Aifredson, T. V. J. Chromtogr. 1981, 218, 715-728. Cookson, D. J.: Rix, C. J.; Shaw. I.M.; Smith, B. E. J. Chromatogr. 1984, 312, 237-246. Suatoni, J. C.; Garber, H. R.; Davis, B. E. J. Chromatogr. Sci. 1975, 13, 367-371. Miller, R. L.;Ettre, L. S.; Johansen, N. G. J. Chromafogr. 1983, 259, 393-412. Miller, R. L.; Ettre, L. S.; Johansen, N. G. J. Chromatogr. 1983, 264. 19-32. DiSanzo, F. P.; Uden, P. C.; Siggia, S. Anal. Chem. 1980, 5 2 , 906-909. Heath, R. R.; Tumlinson, J. H.; Doollttie, R. E.; Proveaux, A. T. J. Chromafogr. Scl. 1975, 13, 380-388. McKay, J. F.; Latham, D. R. AnalChem. 1980, 5 2 , 1618-1621. Eganhouse, R. P.; Ruth, E. C.; Kaplan, I.R. AnalChem. 1983, 5 5 , 2 120-2 126.
(16) Matsushita, S.;Tada. Y.; Ikushige, T. J. Chromatogr. 1981, 208, 429-432. (17) Hayes, P. C., Jr.; Anderson, S. D. Anal. Chem. 1985, 5 7 , 2094-2098. (18) Hayes, P. C., Jr.; Anderson, S. D. Anal. Chem. 1986, 5 8 , 2384-2388. (19) Lundanes, E.; Greibrokk, T. J. Chromatogr. 1985, 349, 439-445. (20) Lundanes, E.; Iverson, B.; Greibrokk, T. J. Chromatogr. 1988. 366, 391-395. (21) Schwartz, H. E.; Browniee, R. G. J. Chromatogr. 1986, 353, 77-93. (22) Fjeldsted, J. Ph.D. Dissertation, Brigham Young University, Provo, UT, 1985. (23) Schwartz, H. E., personal communication, 1986. (24) Hartiey, F. R. Cbem. Rev. 1973, 7 3 , 163-190. (25) March, J. Advanced Organic Chemlstiy: Reactions, Mechanisms and Structure, 2nd ed.; McGraw-HIII: New York, 1977; pp 236-238. (26) Fetizon; Goifier C.R. Acad. Sci., Ser. C 1968, 267, 900. (27) Shriner, R. L.; Fuson. R. C.; Curtin, D. Y.; Monill, T. C. The Sysfemafic Idenfiflcation of Organic Compounds : A Laboratory Mannual”, 6th ed.; Wiley: New York, 1963; pp 170, 178, 202-204.
RECEIVED for review August 12,1987. Accepted October 15, 1987. This work was supported by the U.S. Department of Energy, Office of Health and Environmental Research, through Grant No. DEFG02-86ER60445, Exxon Research and Engineering, Exxon Education Foundation, and the Dow Chemical Co.
Gravity-Augmented High-speed Flow/Steric Field-Flow Fractionation: Simultaneous Use of Two Fields Xiurong Chen,l Karl-Gustav Wahlund? and J. Calvin Giddings* Department of Chemistry, University of Utah, Salt Lake City, Utah 84112
The large stable of subtechnlquw of fleld-tlow fractlonatlon (FFF) avalbble from the COmMnath of known drlvlng forces and operating modes can be further expanded by d n g two or more drlvlng forces at the same tkne. However, the new subtechnlques created are superHuous unless they demonstrate capabllltles not avallable from a slngle drlvlng force alone. I n thls context, the flrst productive association of external forces used In FFF Is reported: the comblnatlon of crossflow-based forces with gravlty. I t Is shown both theoretically and experlmentalty that these forces, applled to particles In the 5-50 pm dlameter range, yleld better peak spaand resolution than elther force acting singly. When applled to polystyrene latex standards, thls comblnation generates excetlent resolution of seven different bead diameters in about 100 s run tlme.
An essential feature of all forms of field-flow fractionation (FFF) is a driving force acting in a direction perpendicular to the walls of a thin flow channel ( 1 , 2 ) . The driving force may have its origin in a number of externally applied fields or gradients. In some cases internally generated forces, such as those produced by shear, may contribute to the driving force. Among the primary (externally applied) driving forces ‘Permanent address: I n s t i t u t e of Chemistry, Academia Sinica, Bei‘ing, China. zlpermanent address: D e p a r t m e n t of A n a l y t i c a l Pharmaceutical Chemistry, U n i v e r s i t y of Uppsala Biomedical Center, S-751 23 Uppsala, Sweden.
that have been used or could potentially be used in FFF are those associated with sedimentation (including both centrifugation and gravity), thermal diffusion, crossflow, electrical fields, dielectrophoretic phenomena, magnetic fields, concentration gradients, photophoresis, and other physicochemical gradients. The combination of these driving forces with the different operating modes of FFF (normal FFF, steric FFF, hyperlayer FFF, etc.) provides a very large and diverse collection of subtechniques for application to particulate and macromolecular separation problems (3, 4). The choice of an appropriate subtechnique depends upon many factors. Along with the availability or potential availability of equipment, we must consider the different selectivities of the different subtechniques as well as their useful mass (molecular weight) range and their applicability to aqueous versus nonaqueous solutions/suspensions and related considerations. With such a large stable of subtechniques potentially available, one or more can usually be identified that will provide a suitable combination of properties for the solution of a particular problem. Quite obviously one could greatly multiply the number of potential subtechniques by simultaneously applying two or more independent driving forces. However, with such a large assortment of subtechniques made available by single driving forces, there is no a priori virtue in creating new subtechniques simply for the purpose of having a larger subtechnique base. In view of the fact that most combinations of driving forces are much more difficult to implement than single driving forces, and the results are more difficult to interpret, we believe that any given pairwise combinations of driving forces cannot be considered as a viable subtechnique unless some
0003-2700/88/0360-0362$01.50/00 1988 American Chemical Society