Development of methane positive chemical ionization gas

Mar 1, 1986 - Robert D. Voyksner, Joan T. Bursey, Terry W. Pack, and Randy L. Porch. Anal. ... Gauthier Eppe , Alexander Schaechtele , Stefan van Leeu...
0 downloads 0 Views 763KB Size
621

Anal. Chem. 1986, 58,621-626

Development of Methane Positive Chemical Ionization Gas Chromatography/Mass Spectrometry Procedures To Determine Polychlorinated Biphenyls Robert D. Voyksner,* Joan T. Bursey,' Terry W. Pack, and Randy L. Porch' Research Triangle Institute, Analytical and Chemical Sciences, P.O. Box 12194, Research Triangle Park, North Carolina 27709

Positive Ion chemical Ionization GC/MS was evaluated for the quantltatlon of polychlorobiphenyis (PCBs) using 10 or less selected PCB standards. Posltlve Ion chemlcal lonizatlon MS conditions were found that resulted In the best sensftlvlty and least fragmentation for the PCBs. The optimal conditions were a source temperature of 110 OC and source pressure of 0.6 torr, using 200 eV of energy. Mass spectral and response factor (RF) Information was gathered for 137 pure cogeners to evaluate methodology of grouplng PCBs by level of chlorlnatlon or by clusterlng accordlng to similar responses for quantltation using 10 or less standards. Positlve methane chemlcai ionization was well-suited for their analysis, since the RF values only varied by a factor of 12 at extremes and the technique was 2-6 times more sensitive than electron Impact ionization. I t was determined that quantltation by using a standard from each level of chlorination showed an overall mean devlatlon of 21%. It proved more useful for analysis of environmental samples than the cluster scheme, which required Isomer identiflcatlon for proper classification of the PCB to a cluster. The analysls of a sediment extract splked with PCB congeners resulted In a 14% f 5 % deviation between actual concentrations and measured concentrations uslng PCB standards at each level of chlorlnatlon.

Polychlorinated biphenyls (PCBs) constitute a class of widespread environmental contaminants (1-3). Environmental persistence and toxicity or suspected carcinogenicity of PCBs are reasons for continued development of new methods of detection and measurement for PCBs. Most analytical methods for PCBs have relied on gas chromatography (GC) coupled with various selective detectors such as electron capture (EC) ( 4 ) or mass spectrometry (MS) (4-9). Most analytical methods do not attempt to distinguish PCB congeners. Difficulties are encountered in precise quantitative analysis because of variability of detector response of the PCB congeners, the lack of unique analytical standards, and interferences in the chromatogram (10-14). Typically, quantitation has been based upon the degree of similarity of an Aroclor standard to the sample (12). However, due to the difficult degradation processes, bioaccumulation in living organisms, and selective biotransformation of PCBs, as well as the presence of incidentally generated PCBs, the Aroclor comparison-based approach is not always valid (15, 16). Analytical methods that are not based on Aroclors for accurate quantitation of the 209 isomers are badly needed. The ideal way to analyze PCBs would be to obtain, individually, all 209 isomers in pure form and to achieve complete GC separation. However, pure standards for all 209 PCBs are not available, and chromatographic conditions that will Present address: Radian Corp., 900 Parameter Park, Morrisville,

NC 27560.

resolve all available isomers have not been described. A more practical approach to quantitative analysis of PCBs involves the use of GC/MS to distinguish among PCBs of different levels of chlorination and the use of one representative PCB per level of chlorination as a standard for quantitation (for example, the calibration with 10 PCBs rather than all 209 PCBs). GC/MS analysis using electron ionization (EI) has proven successful in quantitating PCBs using this scheme (11). However, sensitivity of EI-GC/MS is much lower than negative chemical ionization (NCI) GC/MS or EC detection. NCI and EC methods show a wide range in response factors (RFs) to allow the use of a representative PCB congener for quantitation (4,17,18). EI-GC/MS techniques have been demonstrated by use of PCB congeners for quantitation (19,20). However, evaluation of quantitation of PCBs by levels of chlorination has not been performed under chemical ionization conditions. Positive methane chemical ionization GC/MS techniques have been accurate in quantitating PCBs using representative Aroclor standards (21-23). Positive methane CI may prove to offer good sensitivity with similar congener responses for the quantitative analysis of PCBs. This paper evaluates the determination of PCBs by positive methane CI-GC/MS using one standard per level of chlorination and by response clustering. Experiments to maximize (M H)+ ion intensity a t each level of chlorination while minimizing the (M H2 - C1)' ion current (which would interfere with quantitative analysis of the PCBs a t the next lower level of chlorination) and to minimize the variation of RFs for different isomers were performed. Source pressure, source temperature, and electron voltage were varied to determine optimal conditions. All available pure (>95% purity) PCBs (137 congeners) were analyzed at the optimal conditions to determine RFs. Statistical treatment of RFs for each set of isomers allowed determination of the best isomer to represent the PCB level of chlorination. The estimated accuracy of the technique is discussed. The accuracy of quantitative analysis of PCBs using standards at each level of chlorination is compared with a cluster analysis of the RFs (where standards are chosen to represent a group of RFs, independent of the level of chlorination).

+

+

EXPERIMENTAL SECTION PCB Standards. The PCB standards (Wellington Scientific, College Station, TX, and Ultra Scientific, Hope, RI) were dissolved by using distilled-in-glasstoluene at levels of 8,40, and 200 ng/& The 40 ng/NL (1& injected) standard was used in all cases except when noted in the text. Components of standard mixtures were selected on the basis of previously reported retention times (17). Congeners were chosen for each mixture to ensure base line separation of each PCB on a 30-m fused silica capillary column, Typically there were from 10 to 30 congeners in each standard mixture. Three internal standards (phenanthrene-&, chrysene-dlz,and octachloronaphthalene)were spiked into all samples near the 100 ng/fiL level. GC/MS. Separation of the PCBs was performedon a 30-m DB-5 (0.25 mm id., 1bm fii thickness) column (J & W Scientific, Rancho Cordova, CA) using a helium flow rate of 1.2 mL/min.

0003-2700/86/0358-0621$01.50/00 1986 American Chemical Society

622

ANALYTICAL CHEMISTRY, VOL. 58, NO. 3, MARCH 1986

IM+nl'llM+H*-CII'

69

lM*Hl+

+

Figure 1. Histogram of (M H)' relative intensity determined by averaging the integrated ion currents for the base peak in the molecular ion cluster for each level of chlorination and then normalizing to the largest average and the ratio of (M -t H)+/(M H, - GI)+ (determined in the same manner) intensities at source temperatures from 90 to 170

+

OC.

The column was programmed from 100 to 300 "C at 1.5 OC/min. The final temperature was held for 5 min, resulting in an analysis time of 133 min. The PCB solutions were injected on-column (J & W Scientific) (1 pL) for the GC/MS analysis. The GC/MS analysis was performed on a Finnigan 4500 (Finnigan MAT, San Jose, CA) interfaced with an INCOS data system (Finnigan MAT). The instrument was operated in positive methane CI mode at an emission current of 0.25 mA. The important CI parameters were varied to achieve optimum sensitivity for the PCBs. The methane, introduced coaxially around the column into the source, was varied from 0.2 to 1.0 torr (measured with a thermocouple gauge in the ion source), the source temperature from 90 to 170 "C, and the electron energy from 50 to 200 eV. The instrument was scanned from m / z 100 to m f z 500 in 1.0 s during the analysis. All reported RFs and optimization measurements were the mean of three replicate analyses from measurements of the most intense ion in the molecular ion cluster of the PCB and the (M + H)' ion of chrysene-d,,, using the formula given below:

+

Figure 2. Histogram of the (M H)' relative intensity determined by averaging the integrated ion currents for the base peak in the molecular ion cluster for each level of chlorination and then normalizing to the largest average and the ratio of (M -I-H)+/(M H, - GI)' (determined in the same manner) relative intensities at various methane source pressures.

+

100 r

Electron Energy

(area)PCB congener X ng of internal standard RF = (area)internal standard X ng of PCB congener GC/MS A n a l y s i s . Three to five grams of sediment was mixed with 100 mL of water. The pH was adjusted to 12-13 with 1 N sodium hydroxide and extracted with 20 mL of methylene chloride. The extract was dried over 2 g of sodium sulfate and concentrated to 200 fiL. Three internal standards and 10 PCBs (C14C18level of chlorination) in the 50-70 ng/gL range were spiked into this extract. The PCBs spiked into the extract were chosen so they could easily be resolved from each other on the 30-m, DB-5 column. One microliter of the extract was analyzed by positive methane CI-GC/MS using the optimized conditions described in the solvent standard section. Sediment E x t r a c t

RESULTS AND DISCUSSION CI O p t i m i z a t i o n . Experimental parameters such as source temperature, source pressure, and electron voltage were varied to determine the settings for maximum sensitivity and least fragmentation using a representative PCB mixture containing one PCB of each level of chlorination. The bar graphs in Figures 1-3 represent the averages of the integrated intensities of the base peak in the molecular ion cluster for one PCB of each level of chlorination. The integrated intensities were normalized to the largest value. For source temperature (Figure 1)it was obvious that low temperatures (90 OC) were needed both to achieve sensitivity and to limit fragmentation. A change in temperature to 110 O C did not reduce sensitivity drastically, but the relative intensity of the (M -t- H2- Cl)' fragment increased. This fragmentation should be minimized since the fragment ion could interfere with the next lower PCB congener.

0

IM+HI */IM+HZ-Cll*

IM+HIt

Flgure 3. Histogram of the (M

+

H)' relative intensity determined by averaging the integrated ion currents for the base peak in the molecular ion cluster for each level of chlorination and then normalizing to the largest average and the ratio of (M 4- H)'/(M f H, - CI)' (determined in the same manner) relative intensities at three different electron voltages.

The optimal source pressure for methane was 0.6 torr based on signal level (Figure 2). Fragmentation was slightly reduced if the source pressure was reduced to 0.4 torr. Based on these observations, optimal conditions were considered to be 0.5-0.6 torr. At higher pressures, losses in sensitivity and increased fragmentation were observed. Electron voltage also played a role in sensitivity and fragmentation of the PCBs analyzed (Figure 3). An electron voltage of 200 eV resulted in the best sensitivity, but the high energy promoted fragmentation. On the other hand, 50 eV minimized fragmentation but with a sacrifice of 1 5 2 0 % in sensitivity. Before the choice of optimal conditions could be finalized, experiments to determine reproducibility and RF variability at settings near optimal sensitivity conditions were performed. These experiments ensured that conditions chosen to promote sensitivity did not hinder the precision of the analysis or increase the variability of RF values among isomers. A mixture of pentachlorobiphenyls (about 40 ng of each PCB) was analyzed in triplicate at optimal conditions depicted in Figures 1-3. Table I shows the mean RF for the 15 different pentachlorobiphenyls selected to ensure chromatographic

ANALYTICAL CHEMISTRY, VOL. 58, NO. 3, MARCH 1986

Table I. Mean RF, Relative Standard Deviation, and Range in RFs for the Analysis of 15 Pentachlorobiphenyls Analyzed with the Conditions Listed conditions 0.6 torr, 110 "C, 200 eV, 40 ng 0.6 torr 110 "C, 50 eV, 50 ng 0.6 torr, 90 "C, 200 eV, 40 ng 0.6 torr, 130 "C, 200 eV, 40 ng 0.4 torr 110 "C, 200 eV, 40 ng 0.8 torr 110 "C, 200 eV, 40 ng 0.6 torr, 110 "C, 200 eV, 8 ng 0.6 torr 110 "C, 200 eV, 200 ng

Table 11. Comparison of Base Peak Intensities for E1 and Positive Methane CI-GC/MS Analysis for One Congener at Each Level of Chlorination

mean RF re1 std dev RF range 0.49 0.52 0.49 0.47 0.48 0.74 0.71 0.92

16 13 16 28 27 18 15 27

0.26 0.24 0.23 0.41 0.40 0.41 0.40 0.88

resolution, along with the standard deviation and differences between the high and low R F values for the PCBs. Table I shows the RF value range is minimized at the pressure of 0.6 torr, with the source temperature at 90 or 110 "C and at either 50 or 200 eV electron energy. The standard deviation for the analysis (run-to-run reproducibility) is also minimized under these conditions. Changes in pressure and concentration have a more pronounced effect on reproducibility than temperature or electron voltage and result in an increase in the R F value range. Analysis of variance was performed for each parameter shown in Table I to determine the R F value dependence of each variable. RF values are statistically independent of source temperature and electron voltage. This independence was demonstrated in Table I, since changes in temperature or electron voltage produced no drastic change in the mean RF or RF range. However, methane source pressure and PCB concentration play a significant role in R F determination. Pressure must be controlled carefully during the evaluation so there will not be a shift in R F values. The optimal conditions for maximum sensitivity and minimal variance in R F values are 0.6 torr methane at 110 OC source temperature and 200 eV electron energy. Measurement of RFs of PCB Congeners. Response factors were determined for all pure PCB congeners in stock (137) at previously stated optimal conditions. The RFs calculated for each PCB were based upon the most intense ion in the molecular ion cluster. The RF values over all congeners ranged from 0.14 to 1.79. This range is comparable to that encountered in EI-GC/MS measurement of RF values, which ranged from 0.22 to 4.08 (17). The PI-GC/MS RFs calculated for 77 congeners used the average of four measuremen$s a t one concentration (about 50 ng) relative to 3,3',4,4'-tetrachlorobiphenyl-d6, using standard E1 conditions (70 eV) and the same type of column as in the positive methane CI analysis (17). However, positive methane CI resulted in a 2-6 times more intense (M + H)+ ion than did E1 for (M)+ ions of selected isomers representing the entire degree of chlorination range (Table 11) when analyzed under the same conditions pn the same instrument. The RF value range (a factor of 12) in positive CI was less than that for studies reported by ECD techniques (17). GC/ECD results have shown RF values to range by a factor of 165 between two extreme isomers for the 209 isomers measured (17).(For example, isomer 3 has an R F value 165 times larger than isomer 43.) However, within a level of chlorination the maximum RF value range observed was for the tetrachlorobiphenyls, which showed an RF range of a factor of 10 (17). However, this variability is compensated for by the superior sensitivity of ECD detection over positive methane CI or EI/MS techniques. The positive ion chemical ionization mass spectra for the PCB analyzed exhibited a number of common features. The base peak in all cases was in the molecular ion cluster (M 1, M + 3, or M + 5 depending on the number of chlorines present). The spectra consisted of two cluster ions, one was

+

623

congener level of chlorination

no.

E1

1 2 3 4 5 6 7 8 9 10

1 5 29 50 87 154 183 200 207 209

1 1 1 1 1 1 1 1 1

relative intensitf positive methane CI 2.1 3.3 2.5 6.2 5.5 5.5 4.0 2.4 2.5 2.2

1

All intensities are normalized to E1 base peak in the molecular ion cluster for each congener.

-

-

20 UI 18-

0

i:::

C .

Mean RF values lor 3 determinations a1 40 ng Mean RF value per level of chlorination

15-

":la-

1.3-

:::1

I 10

11.00.9

5

8 0.8 f 0.70.6 0.5-

a.

."$

i"

.I

04-

a

I

030 O 12 t

I 1

01

I

I

I

2

3

4

1 5

I

1

I

I

1

8

I

B

9

10

Chlorine Number

Figure 4. RF values plotted against the level of chlorination in the PCB congeners analyzed by methane positive CI-GCIMS. The square point represents the mean RF for each group of isomers of a given level of chlorination.

+

an (M C2H# ion and the other was an (M + C3HS)+ion from interactions with the methane reagent gas. (M CZH$ was always more intense than (M + C3Hs)+and usually was from 20 to 60% of the base peak intensity. The only fragment ion consistently detected was the (M Hz - C1)' ion, which ranged from 1 to 25% of the base peak intensity. Quantitation of PCB Using Representative Standards. Two approaches were evaluated for quantitation of individual congeners using a limited number of PCB standards: use of representative congeners at each level of chlorination and use of representative congeners from clustering R F values. Quantitation using representative PCBs a t each level of chlorination is conceptually easy to execute. Figure 4 graphically displays the mean RF values (average of a t least three determinations a t one concentration) vs. the number of chlorines in the molecule. The square point represents the average RF value for the set of PCBs for a given level of chlorination. The isomer closest to this mean is chosen to best represent the group of isomers. Table 111summarizes the data given in Figure 4. The PCB isomers closest to the square point value are listed along with the maximum and mean percent deviation from a measured RF value. Quantitation by level of chlorination exhibited a maximum percent deviation of 87% (isomer 75). The average percent deviation in a measurement by this method for the 137 PCB congeners analyzed was 21 f 3%. Since the deviation is significantly larger than the coefficient of variation (CV) for the measurement (6 i 5.5%), another method of quantitation was evaluated. Clustering PCBs by RF values enables forming tighter (less deviation in RF values) groups that are represented by one

+

+

624

ANALYTICAL CHEMISTRY, VOL. 58, NO. 3, MARCH 1986

-

1.8 1.7

.

1.6

.

1.5

.

1.4

-

1

.

1.3

7 Isomers

6 1.2 -

t 5

2

0

f

. 1.0 . 1.1

8 Isomers

J

-g 0.9 E 0.8 -

1

42 Isomers

0.6

8

0.5 07! 0.4

1

i

T

0,31 t

Mean R F value for PCBs in cluster

62

0.2

01

18 Isomers 1

3

2

5

4

Cluster Number

Flgure 5. Range of RF values encompassing a cluster for a cluster analysis scheme. The mean RF value is determined by averaging the individual RFs in a cluster. The number of PCB congeners in a cluster is given for each of the five clusters.

Table IV. Comparison of PCB Standards (Listed by Congener No.) and Mean Deviations for Quantitation by Level of Chlorination or with 5- or 10-Group Cluster Schemes

level of chlorination 2 7 26

69 84

type of analysis 5-group 10-group cluster cluster 162 207 184 or 140 194 154 7 or 33 5

165 184

195 206 207 mean deviation for all 137 isomers, %

21 f 3

61 81, 82, or 190 28, 47, 86, or 30 168 or 49 24, 38 or 6

33 3

13.2 f 8

6.2 f 7

standard. Clustering PCBs in groups of 5 and 10 was evaluated, and the percent deviation was compared to measurements made by representative standards for each chlorine number. Clusters were formed to minimize the percent deviation from a mean R F value. A graphic representation of PCB congeners divided into five clusters is presented in Figure 5 . Even with the use of half the number of standards proposed for quantitation by level of chlorination, the five-cluster method exhibited nearly half the percent deviation. The maximum deviation was 56% (isomer 202), but the mean percent deviation was reduced to 13 8%. This is 8% lower than observed when quantitation was performed by using chlorine numbers (using 10 representative standards). The use of 10 clusters (each cluster in the five-cluster method is divided in half) further reduced the average deviation by subdividing the range of RF values. The maximum percent deviation was reduced to 6 f 7 % , which is within the precision of the RF measurement. Any additional gain in

*

625

ANALYTICAL CHEMISTRY, VOL. 58, NO. 3, MARCH 1986

Table V. Comparison of Chlorine Number and Cluster Analysis for Quantitating PCBs in a Sediment Sample PCB congeners spiked into sample (level of

actual concn spiked,

chlorination)

ng/&

measd quantities using level of chlorination scheme, ng/rL

69 (4) 86 (5) 121 (5) 136 (6) 139 (6) 166 (6) 168 (6) 173 (7) 193 (7) 196 (8)

47 53 57 63 56 66 62 63 53 73

39 58 69 58 63 73 73 76 63 84

measd quantities using

10-cluster scheme, exptl error, %

exptl error, %

ng/dJ

-11 +6 +2 +14 +2 -2 -6 +3 -8 +19

42 50 58 72 57 65 58 65 49 87

-17 +9 +21 -8 +13 +11 +8 +21 +19 +15

7.3 f 6%"

14 f 5%"

Absolute mean error. precision would be fruitless unless the run-to-runmeasurement reproducibility is increased. Table IV summarizes the choice in representative PCB congener numbers (24) for quantitation by the discussed schemes. In most cases, other commercially available PCB congeners with RF values close to the target RF value can be substituted for the representative quantitation standard, as long as chromatographic separation of representative isomers is achieved. For example, two different congeners chosen to represent their respective clusters cannot exhibit the same retention time. The clustering approach, while offering better accuracy, requires PCB isomer identification and analysis of all 209 PCB isomers so that each isomer can be classified in a cluster. Part of the disadvantage of clustering is that the PCB to be quantitated must be identified so that the proper standard for quantitation can be chosen. Identification of a PCB can be difficult with only a partial set of standards due to the similarity of GC retention time and mass spectral information for many isomers. If the wrong representative standard is chosen or if the PCB isomer is not one of the 137 PCBs classified in various clusters then the accuracy of the analysis could be poorer than quantitation by chlorine number. This method has very little practical value in real-world complex extracts. Quantitation by chlorine number may offer advantages in real-life unknown samples since exact identification is not needed for quantitation. Mass spectral information will determine chlorine numbers of PCBs found in the sample allowing for quantitation by the appropriate standard. If PCBs are found that have not been analyzed in this study, quantitation still can be performed by chlorine number with a realization that the maximum error limits could increase. For these reasons, quantitation by chlorine number might be preferred in a complex unknown, while quantitation by clustering would be preferred for samples of less complexity in which more information about possible PCB isomers is known. However, the clustering scheme could prove to be a useful approach if clusters were based on RF and retention time. Now the PCBs do not have to be identified to be classified to a cluster. Also, since all retention times for PCBs are known, there is no need to acquire all PCB standards. Clustering by this technique and its usefulness in environmental samples are still under investigation. Quantitation of PCBs Spiked into a Sediment Extract. A sediment extract spiked with known PCBs was analyzed to compare the quantitative results obtained by level of chlorination and cluster analysis schemes. Figure 6 shows the total ion current chromatogram for the sediment extract. The PCBs and internal standard peaks are labeled on the chromatogram. Table V compares the quantitative results of the two schemes with the known spiked concentration. Clearly

Octachlomnaphlhalene

I

50W

5620

6640

7500

8320

9140

Time (min)

Figure 6. Total ion current chromatogram for the GClMS analysis of a Sediment extract spiked with PCBs. The PCBs (listed by congener number) and internal standards are listed on the chromatogram.

Table VI. PCB Congeners with the Same Level of Chlorination and Very Close Retention Times (50.8 min) to the Target Congeners" target PCB congener no. 69 86 121 136 139 166

168 173 193 196

max dev for misidentificacongener noeC

tion, %

43,b 52 82, 87, 97, 115, 116b 95 148,* 154 140, 14gb 159 153, 132* 171, 177, 189 191, 180 203, 201b

213 28 13 t60 213 28 213 45 38

>O

"These congeners could easily be mistaken for the target PCBs. Such a mistake would result in assignment of the PCB to the wrong cluster for quantitation, increasing the maximum percent deviation by the values listed. bPCB not assigned to a cluster but retention time is known (17). With retention times 50.8 min of target PCB and same level of chlorination as the target PCB. the cluster scheme showed a lower percentage error (7%) than quantitation by level of chlorination (14%). However, Table VI shows possible PCB interferences that are not distinguishable by mass spectral information and have retention times very close to the target PCB. If the target PCB was misidentified as one of the interfering PCBs listed, the error

626

Anal. Chem. 1986, 58,626-631

would increase by at least the percentage listed, making quantitation by level of chlorination equivalent or better in accuracy and much easier to perform.

ACKNOWLEDGMENT We thank William L. Budde and Ann L. Alford-Stevens for their conception and insight on the project. Registry No. Methane, 74-82-8; decachlorobiphenyl, 2051-24-3.

LITERATURE CITED (1) Kutz, F. W.; Strassmdn, S. C. National Conference on PCBs, Chicago, IL, Nov 19-21, 1975 (2) Gustafson, C. G. Environ. Sci. Technol. 1970, 4 , 814. (3) Nisbet, I.C.; Sarofim, A. E. EHP, Environ. Health Perspect. 1972, 1 , 21. (4) Oswald, E. 0.; Ley, L.; Corbett, B. J.; Walker, M. P. J . Chromatogr. 1974, 93,63. (5) Duinker, J. C.; Hellebrad, M. T. Environ. Sci. Technol. 1983, 17, 449. (6) Tindall, W.; Winlnger, P. E. J . Chromatogr. 1980, 109, 119. (7) Castelli, M. G.; Martelli, G. P.; Spargone, C.; Capellini, L.; Fanelli, R. Chemosphere 1983, 12, 291. (8) Mullin, M.; Pochinl, C.; McCrindle, S.;Romkes, M.; Safe, S.; Safe, L Envlron. Sci. Technol. 1954, 18, 468. (9) Krupcik, J.; Leclercq, P.;Simovi, A.; Colgk, M.; Hrivnak, J. J . Chromatogr. 1976, 119,271. (10) Llu, R. H.; Ramesh, S.; Liu, J. H.; Kim, S. Anal. Chem. 1984, 56, 1808. (11) Martelli, G. P.; Castelli, M. G.; Fanelii, R. Biomed. Mass Spectrom. 1981, 8, 347. (12) Webb, R . G.; McCall, A. C. J . Chromatogr. Sci. 1973, 1 1 , 366. (13) Boe, B.; Egaas, E. J . Chromatogr. 1979, 180. 127.

(14) Eicheiberger, J.; Harris, L. E.; Budde, W. L. Anal. Chem. 1974, 46, 227. (15) "PCBs: Biological Criteria for an Assessment of Their Effects on Environmental Quality"; National Research Council, Canada, NRCC No. 16077, 1978. (16) McNulty, W. P.; Becker, G. M.; Cory, H. T. Toxlcol. Appl. fharmacol. 1980, 56, 182. (17) Cooper, S. D.; Moseley, M. A.; Pelllzzari, E. D. "Development and Standardization of Methods for Analysis of Biological Tissue for PCBs"; Final Report to EPA, Las Vegas, NV, Contract 68-03-3099, 1982. (18) Boe, B.; Egaas, E. J . Chromatogr. 1979, 180, 127. (19) Slivon, L. D.; Gebhart, J. E.; Hayes, T. L.; Alford-Stevens, A. L.;Budde, W. L. Anal. Chem. 1985, 57, 2464. (20) Gebhart, J. E.; Hayes, T. L.; Alford-Stevens, A. L.; Budde, W. L. Anal. Chem. 1985, 57, 2458. (21) Harrison, A. G.; Onuska, F. I.; Tsang, C. W. Anal. Chem. 1981, 53, 1183. (22) Cairns, T.; Stegmund, E. G. Anal. Chem. 1981, 53, 1599. (23) Oswaid, E. 0.; Aibro, P. W.;McKinney, J. D. J . Chromatogr. 1974, 98,363. (24) Bailschmiter, K.; Zell, M. Fresenius' Z.Anal. Chem. 1980, 302,20.

RECEIVEDfor review June 6, 1985. Resubmitted October 28, 1985. Accepted October 28, 1985. Although the research described in this article has been funded wholly or in part by the U S . Environmental Protection Agency, Environmental Monitoring and Support Laboratory, Cincinnati, OH, through Contract 68-03-3122, Work Assignment No. 9, directed by Ann Alford-Stevens, to Research Triangle Institute, it has not been subjected to the Agency's required peer and administrative review and therefore does not necessarily reflect the views of the Agency, and no official endorsement should be inferred.

Fluorescence Studies of the Stationary-Phase Chemical Environment in Reversed-Phase Liquid Chromatography J. W. Carr and J. M. Harris* Department of Chemistry, University of Utah, Salt Lake City, Utah 84112

The polarity of both polymerlc and monomerlc C18 statlonary phases Is determlned from the fluorescence vlbronlc flne structure of pyrene sorbed to the surface. A mlcrobore fluorescence sample cell, slurry packed wlth chromatographlc silica using a high flow rate of solvent ylelds stationary phase to mobile phase volume ratios comparable to chromatographic condltlons. The results of the study indicate that lntercalatlon of organlc modlflei Into the stationary phase produces an Inverse relatlonshlp between the polarlty of the C18 surface and that of the mobile phase over a wlde range of solution composltlons. Wlth highest water content moblle phase condltlons, the sorbed probe becomes partlally exposed to the moblle phase and/or surface sllanols due to a reductlon In statlonary phase volume and structural dlfferences In the organlratlon of monomerlc and polymeric C18 Ilgands.

Reversed-phase liquid chromatography has grown to be a widely used method of chemical separation applicable to a diverse mix of analytical problems., As a result of its widespread use, the practice of reversed-phase chromatography has advanced faster than a fundamental understanding of its physical basis. While much of observed retention behavior can be explained in terms of solvophobic or hydrophobic interactions (1-3), this model does not consider contributions

to selectivity from the stationary phase. These contributions can arise from the organization of the hydrocarbon ligands on the silica surface, intercalation of solvent into the bonded layer, activity of residual silanols on the solid support, or changes in these factors as a function of mobile phase composition. A growing body of chromatographic evidence (4-1 I) supports the importance of stationary phase contributions to selectivity and retention behavior. Before a more detailed model of solute retention can be caveloped to include these factors, the physical and chemical nature of the derivatized silica surface must first be understood. A number of spectroscopic studies of alkylated silica have been carried out to study the extent of derivatization, the chemistry of bonding, and the structure and mobility of the bound ligands using infrared (12-16) and NMR (17-23) spectroscopies. Lochmuller and co-workers have pointed out the utility of measuring fluorescence from probe molecules covalently attached to silica for studying the polarity of a reversed-phase surface environment (24), the degree of microheterogeneity (25), and the organization and distribution of the bound ligands (26, 27). A fluorescence investigation of ion pair interactions on alkylated silica surfaces has been carried out using aniline-naphthalene-sulfonic acid as a probe (28). The microviscosity of a C18 stationary phase has been estimated from the fluorescence of excited-state dimers or excimers of pyrene (29). The local density of pyrene adsorbed onto polypropylene was also studied using the yield of excimer

0003-2700/86/0358-0626$01.50/0 0 1986 American Chemical Society