2696
Anal. Chem. 1986, 58, 2696-2699
ence Publishers: Ann Arbor, MI, 1981, Vol. 1, p 167. (23) Pellizzari, E. D.; Tomer, K. B.; Moseley, M. A. In Advances in the Identification and Analyss of Organlc Polluants in Water; Keith, L. H., Ed.; Ann Arbor Science Publlshers: Ann Arbor, MI, 1981; p 197. (24) Riedo, F.; Fritz, D.; Targan. G.; Kovats, E. sz. J . Chromtogr. 1976, 126, 63. (25) Haken. J. K.; Ho, D.K. M. J . Chromatogr. 1977, 142, 203.
(26) Zielinski, W. L., Jr. Ph.D. Thesis, Georgetown University, Washington, DC, 1972.
RECEIVEDfor review January 15, 1986. Resubmitted June 9, 1986. Accepted June 9, 1986.
Identification of Polycyclic Aromatic Hydrocarbons in Extracts of Diesel Particulate Matter by Supercritical Fluid Chromatography Coupled with an Ultraviolet Multichannel Detector Kiyokatsu Jinno*
School of Materials Science, Toyohashi University of Technology, Toyohashi 440, Japan Tadao Hoshino
School of Medicine, Keio University, Tokyo 160, Japan Toshinobu Hondo, Muneo Saito, and Masaaki Senda
JASCO Japan Spectroscopic, Co., Ltd., Hachioji 192, Japan
The identnlcatkn d POrycycUc aromatic hydrocarbons (PAHs) present in the fraction of an extract from diesel particulate matter has been performed in order to demonstrate the performance of supercritkai fluid chromatography with carbon dioxide as the mobile phase coupled with an ultraviolet multichannel detection system. As a result, 11 of 16 EPA priority pollutants, PAHs and benzo[elpyrene, were identitied by performing two actual analyses assisted by a microcomputerized retention predictlon system without any triai-anderror experhrents. Thls work clearly strows that the approach discussed herein has a high potential for analysis of various klnds of complex mixtures of poiyaromatlcs.
The resolution and identification of the components in complex mixtures continue to represent difficult analytical tasks in spite of improvements in instrumentation and methodology. It appears that the increasing complexity of analytical problems will exceed the current state-of-the-art in separation and analysis. Since chromatograms of complex samples are always very complicated, the problem of overlapping peaks becomes a nearly universal concern in the practice of analytical separations. T o improve this situation, it is valuable to recognize that the use of multichannel detectors in chromatography can significantly increase the number of independent informational degrees of freedom in the measurements (1-6). In order to expand the possibility of ultraviolet (UV) multichannel detectors, a coupling with supercritical fluid chromatography (SFC) (7) is a promising direction in practical analysis, since SFC has several advantages compared to gas chromatography (GC) and liquid chromatography (LC). As is well-known, the properties of a supercritical fluid are intermediate between gases and liquids. Solute diffusivities are about 100 times higher than those in the liquid phase and
viscosities are similar to those in the gas phase. Furthermore, the greater density of supercritical fluids compared with gases imbues the mobile phase with solvating powers, which can readily be controlled by application of pressure and temperature. In addition, common supercritical fluids such as carbon dioxide have high transparency in UV wavenumber regions. As a result, these properties should enable greatly enhanced chromatographic efficiency compared to LC, shorter analysis time than in LC, and the possibility of separating high-molecular-weight and thermally labile compounds that cannot be separated by GC. In order to demonstrate the capability of SFC coupled with UV multichannel detector, identification of priority pollutants polycyclic aromatic hydrocarbons (PAHs) in extracts of diesel engine particulate matter has been described in this communication. The supercritical fluid chromatography-ultraviolet (SFC-UV) multichannel detector system was assessed for approximate identification of PAHs contained in a sample extract by the microcomputer-assisted retention prediction system developed in our previous study (8).
EXPERIMENTAL SECTION The SFC system used here was a JASCO (Tokyo, Japan) Model Super-100 directly coupled SFE/SFC system (supercriticalfluid extraction/supercritical fluid chromatography). The system consists of a microscale extraction apparatus and supercritical fluid chromatograph, that allows direct introduction of the SFE extract of a sample into the SFC section of the system. The detection was performed by JASCO MULTI-320 UV photodiode array detector, which is controlled by an if-800 microcomputer (Oki Electrie, Tokyo, Japan). Data processing was performed by using an if-800 and NEC PC-9801 VM2 (16 bits, Nippon Electric, Tokyo, Japan). The column was stainless steel packed with Develosil ODS-5(Nomura Chemicals, Seto, Japan, 4.6 mm i.d. x 15 cm long). The mobile phase was carbon dioxide, pressurized between 100 and 250 kg/cm2, and the temperature was controlled at 40 "C. Standard UV spectra of 16 EPA priority pollutants PAHs were previously stored on a floppy disk by the system
0003-2700/86/0358-2696$0 1.50/0 0 1986 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 58, NO. 13, NOVEMBER 1986
2697
Table I. Identification of PAHs in the Extract from Diesel Engine Particulate Matter
I
0
peak
E
no. 1 2 3 4
M
5 6
0.00 min
7 8 9
0
1
10 11 12 13
4
OJ,....
U,L-,f ,
'
I
,
* 7
? 8
'
'
255 nm
2 4 6 8 10.00min Figure 1. Comparison of predicted and measured chromatograms of PAHs with supercritical COP as the mobile phase. (A) Predicted chromatogram by the microcomputer-assisted retention prediction system: mobile phase, CO,; pressure, 200 kg/cm2;temperature, 40 "C; peak assignment, (A) naphthalene, (B)fluorene, (C) phenanthrene, (D) anthracene, (E) fluoranthene, (F) pyrene, (0)benz[a ]anthracene, (H) chrysene, (J) benzo[b]fluoranthene, (K) benzo[k]fluoranthene, (L) benzo[a Ipyrene, (M) dibenz[a ,/?]anthracene. (B) Measured chromatogram of the extract from diesel engine particulate matter at the above SFC condition.
0.00
supplier, and at the analysis stage, no standard substances were used. The microcomputer-assisted retention prediction system for SFC had been constructed in our previous work (8) by the use of five standard PAHs such as naphthalene, fluorene, anthracene, pyrene, and chrysene on the PC-9801 microcomputer. The sample was the extract from diesel engine particulate matter. The dichloromethane extract from particulate matter collected from the exhaust of a test-diesel engine was separated into seven fractions by a silica column (NQ-2 column, size 29 mm i.d. X 360 mm, Wako Chemicals, Tokyo, Japan) with gradient elution mode from 100% n-hexane to 100% dichloromethane. ks some fluorescence emission from the third fraction was observed by irradiation with an UV lamp, this fraction was considered as that containing PAHs. This PAHs fraction of the concentrated extract was redissolved in acetonitrile and analyzed by the subsequent SFC separations.
RESULTS AND DISCUSSION As the basic concept of the retention prediction of PAHs in SFC has been described in our previous study (8-11). For the column used in this investigation, the retention prediction was log 12' = (-0.000109P 0 . 0 8 2 4 ) ~ ~ 0.00121P - 1.240
+
+
(1) where P is in kg/cm2 and a is a molecular polarizability. This equation means that if P and CY are known, the logarithm of capacity factor of any solute can be determined for the given chromatographic condition. The computerized system constructed for this column in SFC is similar to that of reversed-phase LC (9-11). Identification of Middle-Size PAHs in the Extract. Since most important carcinogenic or mutagenic compounds in PAHs are considered as middle-size compounds (12), primary concern in this analysis is focused on these middle-size PAHs. T o establish the separation condition for these middle-size PAHs, the microcomputer-assisted retention prediction system was used, where the desired condition was that dibenz[a,h]anthracene (five rings) should elute in less than 10 min. The prediction system responded with the most optimized condition being 200 kg/cm2 and 40 "C of carbon
approximate assignment assignment by retention prediction UV multichannel detector fluoranthene pyrene benz[a]anthracene chrysene
fluoranthene pyrene benz [a]anthracene chrysene
a
a f benzo[b]fluoranthene* r: benzolklfluoranthene
benzo[b]fluoranthene benzolklfluoranthene .. benzo[e]pyrene benzotalpyrene
.
1
benzo[e]pyrene benzo [a]pyrene
a a a a
a
fluorene
indeno[1,2,3-cd]pyrene benzo[ghi]perylene a
14
a
15
phenanthrene
fluorene a phenanthrene
16 17 18
a
a
anthracene
anthracenec
a
U
"Not identified. b f , front part of the peak; r, rear part of the peak. correlation number of 0.80. dioxide supercritical state. At this condition, the predicted retention time of benz[a,h]anthracene is ca. 8.5 min. The synthesized chromatogram as a response of the system is shown in Figure lA, in which retention of 12 PAHs such as naphthalene, fluorene, phenanthrene, anthracene, fluoranthene, pyrene, benz[a]anthracene, chrysene, benzo[ blfluoaranthene, benzo[k]fluoranthene, benzo[a]pyrene, and dibenz[a,h]anthracene is predicted. Figure 1B shows the actual chromatogram of the diesel extract obtained a t 200 kg/cm2 and 40 "C. Since it is very difficult to identify some of the components that eluted in less than 1.5 min because of complexity of the chromatogram under this separation condition, the identification of the main eight peaks numbered in the measured chromatogram was performed first. Approximate assignments for six of the eight peaks by the comparison between predicted and measured retention are summarized in Table I. The more precise identification of peaks in the measured chromatogram can be accomplished by the use of several functions of the UV multichannel detector. Since one of the functions of the multichannel detector is to give the UV spectrum of the solute separated in one chromatographic run, this capability was applied to assign these peaks. In addition, the search capability of the standard UV spectrum from the stored data file was used to match the measured spectrum. The first step in the identification process concerned the component responsible for peak 1. The automatic correlation study matching the observed UV spectrum and the stored UV spectra of PAHs yielded the computer output shown in Figure 2. The results show that the UV spectrum of peak 1 is closely correlated to that of fluoranthene. Both UV spectra show excellent agreement with each other in the wavelength range >230 nm. Even though the measured UV spectrum is different from that of the standard UV spectrum of fluoranthene in the shorter wavelength range, one can conclude that peak 1contains fluoranthene because of the following two reasons: first, the retention time of peak 1 is almost consistent with the predicted retention of fluoranthene by the retention prediction system, and second, the shorter wavelength range of the UV spectrum is sometimes critically affected by background subtraction. Similar procedures were then applied to the identification of components making up peaks 2 to 8 in the measured
2698
ANALYTICAL CHEMISTRY, VOL. 58, NO. 13, NOVEMBER 1986
I
100
,f
The identification of peaks 5 and 7 in Figure 1A and one that appeared outside the 10-min range (peak no. 9) was difficult, even though the UV spectra were obtained clearly. No UV spectra stored in the data base matched well to those spectra. However, the component in peak 7 is tentatively identified as benzo[e]pyrene, because the UV spectrum is very similar to that of benzo[e]pyrene in the literature (13) and its predicted retention time (6.2 min) is in good agreement with that of peak 7 (6.4 min). Identification of Small-Size PAHs in the Extract. As seen in Figure 1,the separation condition of 200 kg/cm2 and 40 O C is not suitable to separate small-size PAHs (one to three rings). Therefore, the retention prediction system was used to examine the separation conditions for the small-size PAHs in the extract. The desired condition in this case is that anthracene elutes in less than 2 min. As the result of the retention prediction, the condition of 100 kg/cm2 and 40 O C appeared on the CRT of the microcomputer. The measured and three-dimensional chromatograms are shown in Figure 4. Seven prominent peaks appeared in the measured chromatogram. By use of the approximate retention information from the prediction system and the powerful functions of the multichannel UV detector, such as the peak deconvolution technique, it has been confirmed that two of them include fluorene and phenanthrene, respectively, although those peaks consisted of multicomponents. The deconvolution procedure was applied to that peak and the peak at 1.05 rnin should be assigned to fluorene, because the spectrum is almost the same as that of fluorene. A similar procedure has been used to the assignment of another peak. The UV spectra at 1.47 min and 1.56 min clearly indicate that at least two different components
\
0 350
195 No.
1 FLUORRNT
Figure 2. Output of the UV spectrum search process for peak 1 in the chromatogram of Figure 1B. Result of search is as follows: (1) 0.93 correlation, fluoranthene (SFC); (2) 0.89, fluoranthene (LC); (3) 0.79, benzo[b]Wanthene (SFC); (4) 0.74, benzo[b]fluoranthene (LC); (5) 0.73, phenanthrene (SFC); (6) 0.72, naphthalene (SFC); (7) 0.71, indeno[ 1,2,3-cd]pyrene (SFC); (8) 0.71, fluorene (SFC); (9) 0.68, fluorene (LC); (10) 0.60, benzo[k]fluoranthene (SFC).
chromatogram, and the results are listed in Table I. The spectra of individual peaks and the standard spectrum assigned to them by the UV multichannel detector are summarized in Figure 3. Three major peaks appeared in the retention range beyond the 10 min shown in Figure 1B (peak no. 9,10, and 11). By use of the same procedures, two of those were identified as indeno[ 1,2,3-cd]pyreneand benzo[ghi]perylene, respectively. 2
3
I95
350 No.
1
6-1
350
195
PYRENE
195
No. 1 benz(a)an
350 No. 1 CHRYSENE
100 1
I
-
0
D
'oOm 350
1Y5
No
0
1 benzo(k)f
E
195
350
No. I
benzo(a)p
I o F : m
Flgure 3. Comparison of the measured UV spectra of the separated components in Figure 1B and the most correlated standard UV spectra: (A) peak no. 2, assigned to pyrene; (B) peak no. 3, assigned to benz[a]anthracene; (C)peak no. 4, assigned to chrysene; (D) peak no. G(front part), assigned to benzo[b]fluoranthene;(E) peak no. qrear part), assigned to benzo[k]fluoranthene;(F) peak no. 8, assigned to benzo[a]pyrene.
2699
Anal. Chem. 1980, 58. 2699-2704
this small PAHs region were observed with the use of this column. Therefore, to improve resolution in the analysis for small PAHs, one should use longer and higher efficiency columns with more severe separation conditions such as temperatures over 60 'C and pressures of 80 kg/cm2, although that approach had not been attempted in this study. In conclusion, 11 priority pollutants, PAHs and henzo[e]pyrene, were identified in the extract sample as shown in Table I by SFC coupled with a UV multichannel detector.
ACKNOWLEDGMENT The authors wish to thank Y. Hirata and H. Ohta of Toyohashi University of Technology for their assistance in the sample preparation. Registry No. Fluoranthene, 206-44-0:pyrene, 129-00-0: benz[a]anthracene, 56-553;chrysene, 218-01-9:fluorene, 86-73-7; henzo[b]fluoranthene, 205-99-2;benzo[k]fluoranthene, 207-089; henzo[e]pyrene, 192-97-2;benzo[a]pyrene, 50-32-8;indeno[1,2,3-cd]pyrene, 193-39-5; benzo~hi]perylene.191-24-2; phenanthrene, 85-01-8;anthracene, 120-12-7;naphthalene, 91-20-3; dibenz(a,h]anthracene,53-70-3.
d.0,
1
LITERATURE CITED I,*
Imln
Flgun 4. Measured chromatograms of PAHs with superdticsl CO, as the mobile phase. (A) Measured chromatograms of tlm extract: mobile phase. COS;presswe. 100 kglcm'; temperahre. 40 OC: peak assignments, 1 and 2. are the same as in Figure i A . (6) Three-dC mensionai chromatogram.
elute a t a similar retention time. And the spectrum a t 1.56 min can be assigned as phenanthrene, because the agreement between the filed and measured UV spectra was found. However, identification of other peaks is very difticult because of the complexity of the chromatogram and nonreliahility of the obtained UV spectra. I t is assumed that only peak 17 might be caused from anthracene, hecause the correlation number of 0.80 was obtained in the UV spectra matching process. To know the complexity of the chromatogram in this region more clearly, a contour plot was called from the data filed in the microeomputer. About 15 components can be found in this region. The separation condition of higher column temperature and lower column pressure was then examined, hut no remarkable improvements in resolution in
(1) Jones. D. G. Anal. Chem. 1985, 57. 1207A-1214A. (2) DBOSY.R. E.; Reynolds. W. D.; Nunn. W. G.: 'mu% C. A.; M o k . G. F. J . Chmmatogr. 1910. 728. 347-368. (3) Drouen. A. C. J. H.; Bliiiet. H. A. H.;DeOalan, L. Anal. m m . 1985,