Anal. Chem. 2004, 76, 5486-5497
Retention Indexes for Temperature-Programmed Gas Chromatography of Polychlorinated Biphenyls Shaogang Chu† and Chia-Swee Hong*,†,‡
Department of Environmental Health and Toxicology, School of Public Health, State University of New York at Albany, Albany, New York 12201-0509, and Wadsworth Center, New York State Department of Health, Albany, New York 12201-0509
A noninteger retention index was defined based on a series of PCB internal standards, namely congeners 8 (2,4′-dichlorobiphenyl), 31 (2,4′,5-trichlorobiphenyl), 44 (2,2′,3,5′-tetrachlorobiphenyl), 101 (2,2′,4,5,5′-pentachlorobiphenyl), 138 (2,2′,3,4,4′,5′-hexachlorobiphenyl), 180 (2,2′,3,4,4′,5,5′-heptachlorobiphenyl), and 194 (2,2′,3,3′,4,4′,5,5′-octachlorobiphenyl). These retention index markers are common congeners present in technical mixtures and most environmental samples, and they show a linear dependence of retention time on the number of chlorine atoms, in the temperature-programmed analysis. The index values are calculated with a single regression equation instead of the Van den Dool and Kratz equation. The retention indexes of all 209 PCBs on two commonly used columns (DB-XLB and DB-5), as well as on a supplementary column of DB-17 in capillary gas chromatography, were determined using this system. The reliability of the retention index is quite good, with the average 95% confidence limits for three measurements on each PCB being (0.1 index unit under the same chromatographic conditions and (0.4 index unit under different column head pressures. The effect of heating rate of the programmed runs on the retention index was also investigated. The inversion of the elution order of some congener pairs on the DB-XLB column for different temperature heating rates was observed. Our index values were compared with those of Castello and Testini. The specification of retention in gas chromatography (GC) in terms of the retention index has found widespread application. The use of retention indexes has a great advantage over the use of relative retention times (RRTs), retention volumes, or capacity ratios, in that retention indexes are independent of the column dimensions, film thickness, and phase ratio. Large amounts of isothermal GC retention data have been published as Kova´ts retention indexes, based on the use of n-alkanes as calibration standards.1 To facilitate the retention specification with temperature-programmed operation, Van den Dool and Kratz2 developed * To whom correspondence should be addressed. Phone: 518-473-7299. Fax: 518-473-2895. E-mail:
[email protected]. † State University of New York at Albany. ‡ Wadsworth Center. (1) Kova`ts, E. Helv. Chim. Acta 1958, 41, 1915-1932. (2) Van den Dool, H.; Kratz, P. D. J. Chromatogr. 1963, 11, 463-471.
5486 Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
a temperature-programmed retention index (ITPGC), using an arithmetic relationship similar to the logarithmic one of Kova´ts. Soon after the introduction of the n-alkanes as calibration standards, alternative reference series began to be used, and this has largely continued to the present time. There are two main reasons that the alternative reference series have been widely used. The first is the introduction of specialized and element-specific detectors, for which the response of the n-alkanes is unsatisfactory. There have been a large number of alternative reference series proposed for use with the electron capture (ECD) and element-specific detectors, particularly for nitrogen-, phosphorus-, and halogen-containing materials, drugs, polychlorinated biphenyls (PCBs), and other aromatic compounds, which have become of considerable importance. The second reason for the use of alternative reference series is that the effects of sorption on retention indexes should be considered. When compounds having polarities different from those of analytes are used as retention probes, the dependence of the obtained retention index on temperature may be great, due to the different solutesolvent interactions of the analytes and of the probes. The more similar the reference standards to the compounds analyzed, the more reliably the index will be measured, especially for polar columns and in temperature-programmed operation. Lee found that the retention indexes of polycyclic aromatic hydrocarbons (PAHs) on a low-polarity capillary column (SE-52) using n-alkanes as standards showed poor statistical reliability, and a PAH reference series was introduced.3 Many alternative reference series have been proposed for other compounds, such as alcohols, ketones, esters, n-alkylbenzenes, and PCBs.4 High-resolution GC/ECD or GC/mass selective detector (MSD) in selected ion monitoring (SIM) mode is the most suitable method for congener-specific PCB analyses.5 GC/ECD is generally the method of choice for analysis of PCBs in environmental samples, because it is easier to perform, provides better sensitivity and precision, and is preferred when a large number of samples have to be analyzed, in order to monitor time or spatial distributions. Although all 209 PCB standards are now available from commercial distributors, it is still an expensive and time-consuming process to acquire retention data for all of these on a given chromatographic system. For this reason, reference retention data are very useful for identification of individual PCB congeners. A (3) Lee, M. L.; Vassilaros, D. L. Anal. Chem. 1979, 51, 768-773. (4) Evans, M. B.; Haken, J. K. J. Chromatogr. 1989, 472, 93-127. (5) Larsen, B. R. J. High Resolut. Chromatogr. 1995, 18, 141-151. 10.1021/ac049526i CCC: $27.50
© 2004 American Chemical Society Published on Web 08/18/2004
large number of retention data, including RRTs, retention orders, and retention indexes, have been published and are widely used in PCB identification.6-13 A uniform parameter for the comparison of chromatographic retention data is the retention index system. The extremely large number of possible isomers, coupled with the unavailability of reference standards in many laboratories, makes standardization of retention data essential. The use of the retention index may enable a chromatographer to perform qualitative PCB analysis without a large number of congener standards and is therefore an attractive option. The classical Kova´ts index system, based on n-alkanes as retention markers, is impractical for determining PCB indexes, because the ECD used for PCB analysis does not respond well to the alkanes. In early reports, the use of n-alkyl iodides or n-alkyl trichloroacetates as reference standards with ECD was described.14,15 Morosini and Ballschmiter used a homologous series of eight 2,4,6-trichlorophenyl alkyl ethers as retention index markers, to determine retention indexes of 28 PCBs with GC/ECD on a DB-5 column.16 Chromatographic theory predicts that if the reference series compounds have polarities different from those of the analytes, the retention indexes measured during temperature-programmed operation will vary with the program, since ∆I/∆T differs for compounds of differing polarities. For this reason, reference series based on PCB congeners were suggested.10-12 In our previous work, retention indexes for 59 PCB congeners on an SE-54 column were reported, through the use of seven PCB congeners as internal standards.12 In the present study, retention indexes of all 209 PCBs on two commonly used columns (DB-XLB and DB-5), as well as on a supplementary column of DB-17 in capillary GC, were determined using this system. The effects on the retention indexes of variations in head pressure and temperature-programming rate are examined. The detailed behavior of the new retention index and its relationship with the traditional index are also discussed. EXPERIMENTAL SECTION All individual PCB congeners and PCB calibration mixtures were purchased from AccuStandard Inc. (New Haven, CT). The nine PCB calibration mixtures, which provide all 209 congeners, were diluted with isooctane to the concentration of 20 ng/mL for each congener. All solvents used were nanograde from Burdick & Jackson (Muskegon, MI). Gas chromatography was performed on a Hewlett-Packard 6890 GC equipped with an ECD. Three columns were used to generate the retention data in this study. Column DB-XLB was (6) Mullins, M. D.; Pochini, C. M.; McCrindle, S.; Romles, M.; Safe, S. H.; Safe, L. M. Environ. Sci. Technol. 1984, 18, 468476. (7) Fischer, R.; Ballschmiter, K. Fresenius’ Z. Anal. Chem. 1988, 332, 441446. (8) Vetter, W. E.; Luckas, B. J. Chromatogr., A 1995, 699,173-182. (9) Bolgar, M.; Cunningham, J.; Cooper, R.; Kozloski, R.; Hubbel, J.; Miller, D. P.; Crone, T.; Kimball, H.; Janooby, A.; Miller, B.; Faireless, B. Chemosphere 1995, 31, 2687-2705. (10) Castello, G.; Testini, G. J. Chromatogr., A 1996, 741, 241-249. (11) Castello, G.; Testini, G. J. Chromatogr., A 1997, 787, 215-225. (12) Chu, S.; Miao, X.; Xu, X. J. Chromatogr., A 1996, 724, 392-397. (13) Frame, G. M. Fresenius’ J. Anal. Chem. 1997, 357, 701-713. (14) Vetter, W. E.; Luckas, B.; Buijten, J. J. Chromatogr., A 1998, 799, 249258. (15) Morosini, M.; Ballschmiter, K. Anal. Chim. Acta 1994, 286, 451-456. (16) Morosini, M.; Ballschmiter, K. Fresenius’ J. Anal. Chem. 1994, 348, 595597.
60 m × 0.25 mm i.d., with 0.25-µm film thickness; column DB-5 was 60 m × 0.25 mm i.d., with 0.10-µm film thickness; and column DB-17ms was 30 m × 0.25 mm i.d., with 0.25-µm film thickness. All three columns were obtained from J & W Scientific (Folsom, CA). Unless specifically mentioned, the oven temperature profile was initial temperature 50 °C, hold for 2 min, and then ramp at 2 °C/min to 300 °C. All of the analyzed compounds were eluted during the linear increase of temperature. The carrier gas was helium with constant pressure control, and the makeup gas was nitrogen. The injector and detector temperatures were 280 and 320 °C, respectively. An HP 7673 autosampler was used for splitless injection (1 µL injected), with the split valve being closed for 1 min. The peak assignment was achieved by the injections of nine calibration mixtures and comparison with the published data in the retention and coelution database.13 Congeners of uncertain identity were identified by injection of individual authentic standards or were confirmed by GC/MS. RESULTS AND DISCUSSION In temperature-programmed gas chromatography, the traditional retention index was calculated by the equation
I ) 100n + 100(ti - tn)/(tn+1 - tn)
(1)
where ti is the retention time of the substance for which the retention index is to be determined, tn and tn+1 are the retention times of the n-alkane standards which bracket the substance of interest, and n is the number of carbon atoms in the n-alkane standard that elutes just prior to the substance of interest. For PCBs, n is the number of chlorine atoms on the molecule. Equation 1 can be rewritten as
I ) [100n -100tn/(tn+1 - tn)] + [100/(tn+1 - tn)]ti
(2)
If the retention time has a linear relation to n, then [100n 100tn/(tn+1 - tn)] and [100/(tn+1 - tn)] are constants
[100n - 100tn/(tn+1 - tn)] ) A
(3)
[100/(tn+1 - tn)] ) B
(4)
and
If the linear relationship does not hold true, then A and B have different values for different intervals, and the indexes of an analyte that elutes in a different interval should be calculated separately for each interval. In contrast, if we assume that the linear behavior actually exists, n can be calculated from eqs 3 and 4, but the values of n are not integers in most situations. On the basis of this concept, the new index system is established. The internal standards employed in this index system are PCBs 8 (2 Cl), 31 (3 Cl), 44 (4 Cl), 101 (5 Cl), 138 (6 Cl), 180 (7 Cl), and 194 (8 Cl) as shown in Table 1. These seven congeners are present in high concentrations in technical mixtures and in most environmental samples. Although the system has no rigorous foundation in chromatographic thermodynamics and does not assume or require a linear relationship between the retention time Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
5487
Table 1. Internal Standards and Their Corresponding Retention Indexes on a DB-XLB Column in Our Index System and in Castello and Testini’s System Castello and Testini’s systema
our index system
a
b
congener
structure
I
8 31 44 101 138 180 194
2,4′ 2,4′,5 2,2′,3,5′ 2,2′,4,5,5′ 2,2′,3,4,4′,5′ 2,2′,3,4,4′,5,5′ 2,2′,3,3′,4,4′,5,5′
196.51 324.91 389.28 476.32 615.64 694.85 802.48
congener
structure
I
1 9 27 69 121 151 178 200 207 209
2 2,5 2,3′,6 2,3′,4,6 2,3′,4,5′,6 2,2′,3,5,5′,6 2,2′,3,3′,5,5′,6 2,2′,3,3′,4,5,6,6′ 2,2′,3,3′,4,4′,5,6,6′ 2,2′,3,3′,4,4′,5,5′,6,6′
100 200 300 400 500 600 700 800 900 1000
Refs 10 and 11. b Calculated from the regression equation, I ) -709.99 + 12.247tR listed in Table 2.
Table 2. Linear Regression Coefficients (I ) A + BtR) and Values of Correlation Coefficient r2 Obtained from Plotting tR versus NCl (Chlorine Number) × 100 for Seven Congeners Used as Internal Standards GC column
A
B
r2
DB-XLB DB-5 DB-17
-709.99 -556.06 -546.26
12.247 13.018 13.191
0.9944 0.9959 0.9936
and the number of chlorines, a linear regression analysis of the retention times versus the chlorine numbers of these standards on columns DB-XLB, DB-5, and DB-17 gave r2 ) 0.9944, 0.9959, and 0.9936, respectively. A reference scale of retention indexes was set that used a multiple of the chlorine number by 100 (e.g., IPCB8 ) 200, IPCB31 ) 300). When the assigned I (100n) of retention index markers is plotted against the congeners’ retention times, a linear relationship is observed. The linear regression equation obtained is used to calculate the retention indexes of all 209 PCBs, including the seven internal standards, as
I ) A + BtR
(5)
where A and B are respectively the intercept and the slope of the plot, and tR is the retention time. Table 2 lists three pairs of A, B values obtained on DB-XLB, DB-5, and DB-17 columns, respectively, for which the regression coefficients (r2) are higher than 0.9936. As a result, the retention indexes can be predicted from the A and B values with eq 5 under linear programmedtemperature conditions. The retention indexes are generated by a single regression equation, instead of several equations with a different equation for each interval. It is simply more convenient to use a single equation to calculate all of the indexes. The internal standards in this index system include dichlorinated to octachlorinated PCB; these are the predominant congeners in most environmental samples and therefore are most suitable for general analysis. Because only three monochlorinated biphenyls exist that have the lowest response on ECD, a monochlorinated biphenyl is not included in the reference series. If we note that only three minor congeners, PCBs 205 (8 Cl), 206 (9 Cl), and 209 (10 Cl), elute after PCB 194 in order with well5488
Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
resolved peaks on most nonpolar or low-polar columns, then the use of retention indexes to identify these three congeners is not necessary since they will otherwise be identified for their unusual elution pattern. Therefore, it is not necessary to include PCB 206 (the one nonachlorobiphenyl) and the 10 chlorine-containing PCB 209 as reference standards. However, we did evaluate the use of 8, 9, or 10 internal standards as index markers, through inclusion of the additional congeners PCB 3 (1 Cl), PCB 3 + PCB 206, or PCB 3 + PCB 206 + PCB 209, respectively. No significant improvement in precision or accuracy was seen when we used the additional internal standards. Castello and Testini proposed PCB congeners 1, 9, 27, 69, 121, 151, 178, 200, 207, and 209 (refer to Table 1 for PCB structures) as a retention reference series, based on the linear relationship between the retention time and the number of chlorines.10,11 In their index system, retention indexes were defined to range from 100 to 1000 units. The authors used a modified Van den Dool and Kratz equation to calculate retention indexes. A drawback to their proposed standards is that most of the congeners are generally absent from technical mixtures and environmental samples, requiring the spiking of a reference mix into samples. A large number of stationary phases in high-resolution GC have been used for separation of complex mixtures of PCBs.5,13,17 Despite the advances in capillary column technology, no single column can separate all 209 PCB congeners. The column used for PCB analysis depends on the purpose of the analysis and on the congeners being analyzed. In most conditions, a nonpolar or weakly polar stationary phase is chosen for PCB analysis. In the early days, the most popular GC stationary phase for PCB separation was SE-54.6 Frame has studied 209 PCB congeners and six Aroclors on 20 different HRGC columns.13 On the basis of published data and other information, the low-bleeding column DB-XLB (12% phenyl on siloxane backbone) seems to be the most suitable one for congener-specific PCB analysis.18,19 In the present work, three capillary columns were selected for determination of the indexes of all 209 congeners. Although the GC parameters are somewhat different from those in Frame’s work,13 the separa(17) Cochran, J. W.; Frame, G. M. J. Chromatogr., A 1999, 843, 323-368. (18) Poletti, G. Analusis 1998, 26, M41. (19) Frame, G. M.; Cochran, J. W.; Bowadt, S. S. J. High Resolut. Chromatogr. 1996, 19, 657-668.
tion behavior and elution order are similar. Method translation20 in GC is a variation of components (columns, carrier gases, detectors, etc.) and parameters (pressures, temperature programs, etc.) of a method in a way that maintains the peak elution pattern. Method translation can be used to reduce analysis time, improve resolution, and adopt one method to the others that have different parameters. Theoretically, method translation is valid only under constant pressure for a single type of stationary-phase column.20 For this reason, constant pressure control, instead of constant mass flow controls,10,11 was used by us. By combining the retention data of the 209 PCB congeners on different columns and our proposed PCB reference series, we obtained the retention indexes of all 209 PCB congeners. Table 3 lists the retention indexes of all 209 PCBs on three capillary columns, acquired according to our index system, with a temperature-programmed rate of 2 °C/ min and a column head pressure of 30 psi, and calculated according to equations listed in Table 2 and the retention data from this study. These congeners are listed according to elution order, and IUPAC congener numbers21,22 are used. The numbers for congeners 107, 108, 109, 199, 200, and 201 in this paper are derived according to Guitart et al.,21 and they differ from the corresponding numbers assigned by Ballschmiter and Zell23 as 108, 109, 107, 201, 199, and 200, respectively. Retention indexes were reproduced (three replicates) with average standard deviations of 0.10, 0.16, and 0.05 index units for the DB-XLB, DB-5, and DB-17 columns, respectively. To compare our index system with the traditional index system, we have also calculated the index of each PCB by the method of Castello and Testini,10,11 using the same retention-time data as described above. The calculation of the retention index values on the DB-XLB column, under temperature-programmed operation with 30, 40, 50, and 60 psi of head pressure and using our method and the proposed internal standards (Table 1), was compared with the use of the classical Van den Dool formula and Castello and Testini’s internal standards (Table 1). In our index system, the average standard deviation of the indexes for 209 congeners is 0.36; each standard deviation is less than 1 index unit except for four congeners under different column head pressures (30, 40, 50, and 60 psi). However, use of Castello and Testini’s method results in 23 congeners having a standard deviation of the index larger than 1 unit, again under different column head pressures (30, 40, 50, and 60 psi). It is clear that our noninteger retention index system is superior for complex PCB analysis. Table 4 shows the indexes and standard deviations for selected coplanar PCBs, calculated both by our method and by Castello and Testini’s method, as a function of column head pressure. The good reproducibility of the retention index in our system results mainly from the similarity of the internal standards with the analytes and the new noninteger index system used. In the traditional index system, if the retention time for one of the retention reference compounds changes when the chromatographic parameters are changed, the indexes of other congeners that elute before or after it will exhibit a large change. This is because the indexes of the internal standards are defined as a 100-multiple of the number of chlorines, and they are unaffected (20) Blumberg, L. M.; Klee, M. S. Anal. Chem. 1998, 70, 3828-3839. (21) Guitart, R.; Puig, P.; Go´mez-Catala´n, J. Chemosphere 1993, 27, 1451-1459. (22) http://www.epa.gov/toxteam/pcbid/table.htm. (23) Ballschmiter, K.; Zell, M. Fresenius’ Z. Anal. Chem. 1980, 302, 20-31.
by the change in parameters. In the proposed index system, the retention behavior of some reference congeners may also differ somewhat from the behavior of other analytes when the chromatographic conditions are changed. However, there is only a small effect on the equation (the value of A or B). Therefore, the change of the index in our index system will be less than that in the traditional index system. For the same reason, Fischer and Ballschmiter used the sum of the retention times of two congeners (PCB 52 and 180), instead of a single retention time, as the standard reference to calculate the relative retention times.7 With the ubiquity of computers, it is very easy to produce a mathematical relationship or equation to interrelate a number of experimentally determined data points. In our index system, the indexes of reference compounds, which depend on the chromatographic parameters, are not defined as 100-multiples of integers. According to this concept, the retention index of the standard reference for differing columns or under differing operation conditions will vary. This may seem inconvenient to an analyst who is familiar with the traditional index system. However, if our system is compared with the traditional index system in which different equations are required for different intervals, it is clear that the calculation using a single equation in our index system is more convenient, especially when a very large number of data need to be treated. Although the PCB reference standards show a linear dependence of their retention times on the numbers of chlorine atoms on molecules under temperature-programmed operation, this does not mean that the retention order of all PCB congeners is independent of the temperature heating rate. Reversed elution order occurs for some pairs of congeners when the temperature heating rate changes. Congeners 88 and 121 exhibit the most obvious reversal. Congener 88 eluted before 121 at the rate of 2 °C/min; the two congeners coeluted at the rate of 4 °C/min, and the original elution order was reversed at the rate of 6 and 8 °C/ min. Figure 1 shows the chromatogram with the separation of congeners 88 and 121 at different temperature-programmed rates. The others, including pairs of congeners 13 and 24, congeners 152 and 112, and congeners 136 and 154, show similar elution behaviors with the change of heating rate. The elution order inversion with different heating rate has also been observed on the DB-5 column. In order to demonstrate the applicability of the proposed method, the IPCB values of 209 PCB congeners on three columns under programmed-temperature analysis conditions are listed in Table 3 according to the elution order. The stationary phases tested, all having low polarities, elute all of the PCB congeners in a similar way. The differences in IPCB values are great enough between DB-XLB and DB-17, or between DB-5 and DB-17, to confirm the identification of congeners through comparison of the results obtained on columns of differing polarity. The peak width at half-height averages over the range of PCBs is approximately 0.70 index unit. This means that two PCBs whose retention indexes differ by 1 index unit should have better than 50% resolution. The ∆I is obtained by subtracting the retention index value of each congener from the value of the following one; in this way it is possible to evaluate that, for all three columns considered, the I values of about 75% of the congeners each differ from the value of the contiguous peak by more than 1 index unit and can therefore permit the identification of the compound Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
5489
Table 3. Retention Indexes of 209 PCB Congeners on Three Types of Capillary Columns
5490
Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
Table 3 (Continued)
Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
5491
Table 3 (Continued)
5492
Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
Table 3 (Continued)
Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
5493
Table 3 (Continued)
5494
Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
Table 3 (Continued)
Analytical Chemistry, Vol. 76, No. 18, September 15, 2004
5495
Table 3 (Continued)
a IUPAC numbers21,22 according to GC retention index order. b ∆I ) retention index value subtracted from values of the following congener. 4I values were calculated before rounding. c IUPAC numbers of congeners in boldface are the predominant PCBs present in Aroclors. d Numbers with underlining indicate ∆I
