2674
Anal. Chem. 1986, 58,2674-2680
o-iodoaniline, 615-43-0; m-iodoaniline, 626-01-7; p-iodoaniline, 540-37-4; o-nitroaniline, 88-74-4; n-nitroaniline, 99-09-2; p nitroaniline, 100-01-6;m-chloroaniline, 108-42-9;p-chloroaniline, 106-47-8;o-nitrotoluene, 88-72-2; m-nitrotoluene, 99-08-1.
LITERATURE CITED Hinze. W. L. Sep. Purif. Methods 1981, 10, 195-237. SmoikovB-Keulemansova, E. J. Chromatogr. 1982, 251, 17-34. Uekama, K.; Hirayama, F.; Ikeda, K.; Inaba, K. J. Pharm. Sci. 1977, 66,706-710. Uekama, K.; Hirayama, F.; Nasu, S.; Matsuo, N.; Irie, T. Chem. Pharm. Bull. 1978. 26, 3477-3489. Armstrong, D. W. J. Li9. Chromatogr. 1980, 3 ,895-900. Hinze, W. L.; Armstrong, D. W. Anal. Lett. 1980, 13, 1093-1104. Burkert. W. G.; Owensby, C. N.; Hinze, W. L. J. Li9. Chromatogr. 1981, 4, 1065-1085. Debowski, J.; Sybilska, D.; Jurczak, J. J . Chromatogr. 1982, 237, 303-306. Sybilska. D.; Lipkowski, J.; W6ycikowski, J. J . Chromatogr. 1982, 253,95-100. Debowski, J.; Sybilska, D.; Jurczak, J. Chromatographia 1982, 16, 198-200. Nobuhara, Y.; Hirano, S.; Nakanishi, Y. J. Chromatogr. 1983, 258, 276-279. Debowski, J.; Jurczak, J.; Sybilska, D. J. Chromatogr. 1983, 282, 83-68. Armstrong, D. W.; Stine, G. Y. J. A m . Chem. SOC. 1983, 105, 2962-2964. Sybiiska, D.; Debowski. J.; Jurczak, J.; Zukowski, J. J. Chromatogr. 1984, 286, 163-170. Korpela, T. K.; Himanen, J. P. J. Chromatogr. 1984, 290,351-361
Debowski, J.; Jurczak. J.; Sibilska, D.; Zukowski, J. J. Chromatogr. 1985, 329,206-210. Tanaka, M.; Miki, T.; Shono, T. J. Chromatogr. 1985, 330,253-261. Zukowski, J.; Sybilska, D.; Jurczak, J. Anal. Chem. 1985, 57, 22 15-22 19. Debowski, J.; Grassini-Strazza, G.; Sybiiska, 0.J , Chromatogr. 1985, 349, 131-136. Gazdag, M.; Szepesi, G.; Huszar, L. J . Chromatogr. 1986, 351, 128-135. Cline Love, L. J.; Arunyanart, M. ACS Symp. Ser. 1986, N o . 297, 226-243. Matsui, Y.; Mochida, K. Bull. Chem. SOC.Jpn. 1979, 52,2808-2814. Fujimura, K.: Ueda, T.; Ando, T. Proceedings of 25th Meeting of the Liquid Chromatography Research Society; Kyoto, Japan, Feb 23-24, 1982; pp 47-50. Osa, T.; Matsue, T.; Fujihira, M. Heterocycles 1977, 6 , 1833-1839. Fujimura, K.; Ueda, T.; Ando, T. Anal. Chem. 1983, 55, 446-450. Armstrong, D. W.; DeMond, W. J. Chromatogr. Sci. 1984, 22, 411-415. Kawaguchi, Y.; Tanaka, M.: Nakae, M.; Funazo, K.; Shono, T. Anal. Chem. 1983, 55, 1852-1657. ranaka. M.; Kawaguchi, Y.; Nakae, M.; Mizobuchi, Y.; Shono, T. J. Chromatogr. 1984, 299,341-350. Tanaka, M.; Kawaguchi, Y.; Shono, T.: Uebori, M.; Kuge, Y. J. Chromatogr. 1984, 301,345-353.
RECEIVED for review March 18, 1986. Accepted July 2, 1986. Part of this work was presented at the 25th Meeting of The Liquid Chromatography Research Society, Kyoto, Japan, Feb 23-24, 1982.
Study of Temperature and Mobile-Phase Effects in Reversed-Phase High-Performance Liquid Chromatography by the Use of the Solvatochromic Comparison Method Peter W. Carr* Department of Chemistry, Kolthoff and Smith Halls, University of Minnesota, 207 Pleasant Street, Minneapolis, Minnesota 55455 Ruth M. Doherty and Mortimer J. Kamlet Naval Surface Weapons Laboratory, White Oak Laboratory, Silver Spring, Maryland 20903-5000 Robert W . Taft Department o f Chemistry, University of California, Irvine, California 9271 7 Wayne Melander and Csaba Horvath Department of Chemical Engineering, Yale Unioersity, New Haven, Connecticut 06520 Retentlon of a serles of aromatic sotutes In reversed-phase HPLC has been studied as a functlon of temperature and moblle-phase composition in acetonltrlle-water mixtures. The chief factors that control retentlon are the solute slre (molar volume) and hydrogen bond basicity. Less Important factors are the polarlzaMIlty4polarltyand hydrogen bond acldlty of the solute. Through the use of the sohratochromlc comparison method and h e a r solvation energy relationships, the key system parameters are seen to decrease In magnitude as the mobile phase becomes rlcher In organlc modlfter. That Is,the endoergk cavity formation energy and the exoerglc hydrogen bond formation processes both become less signHkant as the temperature increases. RelationsMps between the logarlthmlc capacity factor and moblle-phase composltlon are predicted to be more linear for small nonpolar solutes and hydrogen bond bases than for large nonpolar solutes.
In the preceding paper in this series, the solvatochromic comparison method and linear solvation energy relationships 0003-2700/88/0358-2674$0 1.50/0
(LSER) were used to examine the chemical and physical characteristics of a solute that govern retention in reversedphase HPLC ( I ) . In that work, the effects of the type of bonded phase, including hydrocarbon vs. fluorocarbon, and the amount and nature of the organic modifier, were tested as variables. We found that the most significant factors by far were the solute size, as measured by molar volume, and hydrogen bond basicity. Solute polarizability-dipolarity was a statistically significant but less important variable. Due to the nature of the data then available to us, the solute hydrogen bond basicity could not be examined over as large a range as desired. However, in related studies concerning the factors that govern water solubility and octanol-water partition coefficients of organic nonelectrolytes, it was similarly shown that hydrogen bond basicity of the solute is significantly more important than its hydrogen bond acidity (2, 3). It is clear that temperature is an important variable in reversed-phase HPLC (4-7) and can be used to optimize a separation (8,9). To date, we have not examined the effect of temperature on solubility or phase transfer process by the solvatochromic comparison method. The data set examined 0 1986 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 58, NO. 13, NOVEMBER 1986
in one of these laboratories (Yale University) contained a sufficient body of information on materials whose solvatochromic parameters are available to enable such a study (6). The effect of temperature on any phase-transfer process is likely to be quite complex and experimentally difficult to define, since the entropy and thereby the enthalpy of the process are obtained from the slope, i.e., derivative of the free energy change vs. temperature. Recently, Gill and his coworkers have shown that an accurate measurement of the enthalpy of transfer of nonpolar gases into water could be obtained from AG measurements only by fitting the data to a fourth-order polynomial in temperature (10). This is due to the fact that the heat capacity of water is very strongly temperature dependent. In the case of liquid chromatography, the situation is further complicated by the fact that there are two different phases to be considered. The stationary phase is a complex milieu comprised of the nonpolar alkyl chains, residual silanol groups, and significant quantities of the organic modifier and water. Burke and co-workers (11,12) have measured the amount of water and modifier associated with the stationary phase via gas chromatographic analysis of displaceable substances. Their work clearly indicates that there is a great deal of water present in the stationary-phase region and its composition changes in a complex fashion with the composition of the mobile phase. Gilpin and his group have studied temperature effects on reversed-phase HPLC and observed a profound hysteresis effect in plots of In k’vs. reciprocal temperature (13). For the above reasons, we decided to examine the retention data directly rather than attempt to obtain an enthalpy of transfer by fitting the observed capacity factor to some function of temperature and then examine the enthalpy of transfer vs. the solvatochromic parameters. We will continue to use the previously established conventions and “ground rules’’ for the solvatochromic parameters (I). In the case of a chromatographic retention, SP, in the equation below, denotes a logarithmic capacity factor, the subscript 2 designates a solute property such as molar volume ( Vz/lOO), polarizability-dipolarity (rz*), hydrogen bond acidity (a2),or hydrogen bond basicity (&).
SP = SP, + M(6,2 - S,2)V2/1O0 + S(a,* - ..,*)a2 + A(& + B(a, - am)& (1) Each solute property is multiplied by a term that represents the difference in complementary “solvent” properties for the two phases. In practice, the coefficients of the solute characteristics are unknown. They are obtained by carrying out multifactor simultaneous least-squaresregressions (vide infra). In chromatography these phases are designated with the subscripts m (mobile) and s (stationary). For the case of bonded phase chromatography, the above formalism should not be understood to imply that retention arises by partition as opposed to adsorption processes. To be explicit, the chemical property of the bulk or bonded phase that complements the solute’s ability to accept a hydrogen bond (p2) is the phase’s ability to donate a hydrogen bond (a, or aJ. The chemical property of the mobile phase or the stationary phase that complements the solute molar volume is its cohesive energy taken as bH2, the square of the Hildebrand solubility parameter. For those solutes that strongly selfassociate in the pure liquid state, the a and @ parameters as a solute will differ from the value for the pure substance. This is an important consideration chiefly for alcohols. The detailed procedures by which the solvatochromic parameters are assigned are given elsewhere (1);the parameters for the solutes used in this work are given in Table I. It must be understood that the same value for all solute parameters is used at each temperature. Although it would be possible to estimate the molar volume at other temperatures, the
2675
Table I. Solvatochromic Parameters for the Test Solutes” no.
solute
benzene fluorobenzene 3 chlorobenzene 4 bromobenzene 5 toluene 6 iodobenzene 7 m-dichlorobenzene 8 o-dichlorobenzene 9 p-dichlorobenzene 10 ethylbenzene 11 m-xylene 12 o-xylene 13 mesitylene 14 isopropylbenzene 15 o-diethylbenzene 16 durene 17 tert-butylbenzene 18 benzyl alcohol 19 2-phenylethanol 20 dimethylbenzyl alcohol 21 3-phenylpropanol 22 nitrobenzene 23 benzaldehyde 24 phenol 25 p-chlorophenol 26 p-nitrophenol 1 2
v2;/100 0.989 1.039 1.118 1.159 1.163 1.215 1.226 1.227 1.241 1.324 1.328 1.329 1.489 1.494 1.625 1.645 1.649 1.169 1.298 1.305 1.451 1.129 1.116 0.989 1.118 1.150
0.11 0.07 0.07 0.06 0.11 0.05 0.03 0.04 0.04 0.12 0.13 0.12 0.13 0.12 0.12 0.13 0.12 0.55 0.55 0.61 0.55 0.30 0.44 0.33 0.25 0.50
0.59 0.62 0.71 0.79 0.54 0.82 0.81 0.80 0.70 0.48 0.47 0.47 0.41 0.47 0.42 0.40 0.43 0.80 0.70 0.75 0.60 1.01 0.92 0.75 0.90 1.20
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0.33 0.33 0.33 0.33 0 0 0.61 0.69 1.00
These values were used at all temperatures.
change over the range in this study is small. Because the other solvatochromic parameters are obtained as relative measurements, their variation with temperature was also neglected. Studies by Laurance and Nicolet indicate that @ and r* are not very temperature sensitive (14). This is admittedly approximate. Values of the solvatochromic parameters are generally not available at other than ambient temperature. There are two other temperature-dependent factors that influence the capacity factor, namely, the stationary-phase volume and the mobile-phase volume. In our previous study of mobile-phase effects and stationary-phase effects by the solvatochromic comparison method (11, we examined the capacity factor per se without correction for variation in the phase ratio as a function of the mobile phase. Such changes are known to be quite significant (11, 12). The change in column dead volume ( V,), as measured by the elution time of deuterium oxide, did not vary with temperature. To our knowledge, there have no direct studies of the effect of temperature on the volume occupied by the bonded phase. Thus we did not correct the experimental capacity factor for variations in the phase ratio. This maintains the consistency of data analysis with our preceding work and in any case such a correction would have to presuppose a “mechanism” for the solute sorption process. Finally, it is likely that the variations in k’ on a single column are due primarily to the chemical factors and not to variations in the phase ratio.
RESULTS AND DISCUSSION The solvatochromic comparison method could be applied to the retention data examined in this work in three distinct ways. First, as mentioned above, the logarithmic capacity factor for each solute could be fitted to a function of temperature and an approximate enthalpy and entropy of transfer obtained. Although the capacity factors decrease monotonically with temperature for all available solutes, it would require quite a few measurements over a reasonable range in temperature in order to derive statistically and thermodynamically valid results. Thus this approach was not pursued. A second approach would be to examine the dependence of the retention on mobile-phase composition. As shown in this study, the data of Chen, Melander and Horvath (6) can be
2676
ANALYTICAL CHEMISTRY, VOL. 58, NO. 13, NOVEMBER 1986
fit with reasonable precision to a linear function of volume fraction (4) of organic modifier. It should be noted that although certain solutes are well fit by a linear relationship between log k’and 4, the more hydrophobic solutes appear to deviate rather more substantially at the low modifier end of the data set. We found that the overall sum of residuals squared was not substantially improved by using a quadratic least-squares fit, nor was it improved by use of log k’vs. log [organic modifier] plots, which is suggested by the displacement model developed by Xindu and Regnier (15). In many cases log-log plots are less linear than the more conventional plot (results obtained a t Yale University). Furthermore, we found that the residual variance generally exceeded that obtained in statistical simulation experiments in which the coefficient of variation in the K’ values was as large as 5 % . Moreover, the standard deviation in the slope was much poorer than that in the intercept. Thus we believe that the parameters of the fit, i.e., slope and intercept of plots of log h’vs. 4, would not be sufficiently unbiased to be useful as input data for the solvatochromic comparison studies (see below). Based on the above comments and preliminary studies, the temperature and mobile-phase dependence of the solute capacity factors were examined by a third approach that involved regressing the data against the solvatochromic equation (see eq 2 ) , which is the correlation equation equivalent to eq
SP = SPO + mV2/100
+ sxp* + ua2 + b&
(2)
1. We found it necessary to exclude from consideration only a few solutes. First, although results using benzaldehyde are available under many conditions, this species is involved in a specific hydration reaction, [CGHS-CH=O + H 2 0 * C6H6-CH(OH)2], and further, it was found to be highly influential and quite often an outlier by Cook’s statistic and a t test (16). Second, unlike the alcohols in Table I, the three phenols are stronger hydrogen bond donors than monomer water. It is not as yet clear to us whether the proper parameter to use for very strong hydrogen bond acids such as phenols is a, or Aam of the acid with respect to monomer water. In essence, water is such a strong hydrogen bond acid that there are few species that can compete with it, not even alcohols, for formation of a hydrogen bond with an acceptor. Phenols are such strong donors that it is likely they can compete with water and therefore not conform to the relationship (eq 2 ) used to model other solutes. We have, therefore, for the present, restricted the study to non-hydrogen-bond donors and weak hydrogen bond donor solutes. This is similar in spirit to the exclusion of amine solutes in our previous study due to the well-known silanophilic interaction (16, 17), which cannot be studied solvatochromically as yet. It should be noted that Jinno and his co-workers found a similar problem in correlating the reversed-phase retention of phenols (18). In order to accommodate phenols in their predictive model, an additional term related to the pK, of the phenol was added to their correlation equation. Thus this lack of fitting ability is real and not simply a consequence of 5xperimental error. We found that correlations vs. V2/100, &, and either x2* or a2 were virtually equally good. No improvement in the correlation coefficient or average residual was obtained when all four were used, and there was considerable variance inflation in the standard deviations of the parameters of the fit when all four were retained. Since previous work employed rz*,we opted for this parameter to maintain consistency and ready comparison; consequently, all correlations were by eq 2 with a2 deleted. The results are summarized in Table 11. As before the two most important solvatochromic parameters are the solute size and hydrogen bond basicity, as shown by the magnitude and precision of the coefficients of v , / l O O and Pr (see m and b in Table 11). The correlation coefficients
1.7AI1 S o l u t e D a t a
1.5 -
1.31
1.1
1
#a
0.7
0 0
A
0.1
308°K 323’K 338°K
t
1 40
50
60
70
80
Voi u m e Fraction Acetonitrile Figure 1. Dependence of m on composition and temperature. All
solutes described in text were employed. are close to unity and the standard deviation of the fit is as good as in our previous liquid chromatographic study. Again, we note that the dependence on r2*is weak. The previous work indicated that when methanol was used as the modifier, the coefficient of x q * , Le., s, became more positive as water was added; the same trend is observed here. Because the dependence on x2* is so slight, we will not examine it in detail. The trends in m and b as a function of composition and temperature are shown in Figures 1 and 2, respectively. In general, both coefficients decrease in magnitude as acetontrile is added to the mobile phase. Thus both the endoergic cavity forming process, as measured by m, and the exoergic hydrogen bond effect, as seen in b of the mobile phase, diminish upon replacement of water with acetonitrile. These are the expected trends since water is far more cohesive than acetonitrile and is a much stronger hydrogen bond acid. The plots in Figures 1 and 2 also show that the magnitudes of m and b generally decrease with an increase in system temperature. This is the anticipated result, particularly for the cavity term (m). There are a limited number of cases, however, where the reasons for the progressions are not immediately obvious. These results may be due to experimental error, or may reflect temperature effects on competitive hydrogen bond acceptance between the solute and acetonitrile for the weaker hydrogen bond donor power of water. They may also be a consequence of our ignorance of temperature effects on the phase ratio. In order to eliminate this possibility, all regressions were repeated using benzene as a reference solute; i.e., SP was redefined as log (K’/kkmne). This should cancel the phase ratio and, therefore, any effect of temperature or mobile phase composition on m, s, or b via the ratio. Further, we did not include the reference solute in the second series of regressions in order to eliminate the systematic bias which this would introduce. In general, m, s , and b were not altered by more than 0.05 units and seldom by more than 0.02. Of course, the term SP, is altered due to the use of the reference; Le., the intercept shifts. As
ANALYTICAL CHEMISTRY, VOL. 58, NO. 13, NOVEMBER 1986 2.4
-
2.2
-
2.0 -
2677
1.7,
1.8-
a c
a
1.3’
1.6-
1.4 -
-b 1.2 -
1.0-
1.05 0 298°K 0 308°K 0 323°K
0.8-
1.85
Flgure 3. Correlation of mW vs. m. See eq 4 for computation of mnp.
0.6 -
I
1.45 IIC
I
I
I
I
I
40
50
60
70
80
recovered by the solvatochromic comparison method using the complete data set. We note that extrapolation of the values of m and b to pure acetonitrile does not lead to such small quantities as would be expected on the basis that acetonitrile is much less cohesive than water (aH2, acetonitrile = 138 cal/cm3, SH2, water = 549 cal/cm3) and acetonitrile is a much weaker hydrogen bond donor than water (a2= 0.19 for acetonitrile vs. a2 = 1.17 for water). The significant m and b values obtained by extrapolation are probably due to the fact that the local solvent environment around hydrogen bond bases, Le., the cybotactic region, is highly enriched-perhaps saturated-with water, even with 80% acetonitrile (by volume) in the mobile phase. Because of this “solvent sorting” by the HBA solute, the extrapolation underestimates the change in b. Reichardt and his co-workers have measured the solvatochromism of the indicator ET(30) in both acetonitrile and methanol-water mixtures (20,21). This indicator is quite sensitive to hydrogen bond effects and shows a very nonlinear dependence of the spectroscopic shift vs. volume percent as water is added to the mixture. The data of Reichardt have been replotted in Figure 4 vs. the conventional volume percent of strong solvent, Le., the fraction by volume taken. It is evident that methanol-water and acetonitrile-water mixtures have similar hydrogen bond acidities up to about 60% by volume of the organic solvent; thereafter, the acetonitrile-water mixtures are far weaker hydrogen bond donor media than are methanol-water mixtures. It is evident that extrapolation of data obtained in the 40-80% volume fraction range would significantly overestimate the apparent hydrogen bond acidity of pure acetonitrile. This is consistent with the above observation concerning the magnitude of the solvatochromic b value when extrapolated to pure acetonitrile. At this point, we also speculate that due to the smaller difference between ETNfor methanol and water vs. the same difference for acetonitrile and water, mobile-phase effects in RP-LC will be more complex with acetonitrile than with methanol as the solvent modifier. The nonpolar part of an organic solute will try to surround itself with the organic modifier and avoid water, whereas the polar, and particularly the hydrogen bond base groups, will be more comfortable in an aqueous environment. The molecule as a whole will be forced to compromise to achieve an overall minimum free energy state. With acetonitrile present, this balance will be
2678
ANALYTICAL CHEMISTRY, VOL. 58, NO. 13, NOVEMBER 1986
Table 11. Correlation Equations for the Effect of Temperature and Composition on the Solvatochromic Comparison log k ’ = SP,
+ mV2/100 + sa*:!+ bP2
values at the following vol 70 acetonitrile 40
50
60
temperature: -SP0 m S
-b slm -b/m r
std devb nc
0.85 (0.16) 1.84 (0.10) 0.31 (010) 2.35 (0.07) 0.17 1.28 0.996 0.04 16
298
0.51 (0.14) 1.34 (0.08) 1.88 (0.06) 0.07 1.40 0.995 0.05
m S
-b slm -b/m ra
std devb ne
0.41 (0.32) 1.48 (0.19) 0.19 (0.19) 2.17 (0.13) 0.13 1.47 0.983 0.10 17
0.44 (0.18) 1.27 (0.09) 0.05 (0.11) 1.82 (0.07) 0.04 1.43 0.992 0.06 19
m S
-b slm -b/m P
std devb ne
0.15 1.25 -0.01 1.92
(0.36) (0.21) (0.21) (0.15)
-0.01 1.54 0.974
0.11 17
0.27 (0.21) 1.11 (0.13) -0.02 (0.12) 1.73 (0.08) -0.02 1.56 0.986 0.07 20
m S
-b slm -b/m ra
std devb nc
0.51 (0.23) 1.53 (0.13) -0.04 (0.15) 1.98 (0.09) -0.03 1.29 0.993 0.06 15
0.23 (0.31) 1.03 (0.17) -0.04 (0.22) 1.69 (0.13) -0.04 1.64 0.980
0.08 16
323
(0.12)
(0.06) (0.07)
(0.05)
-0.02
0.67 (0.12) 0.82 (0.07) 0.00 (0.07) 1.05 (0.05)
0.00
1.36 0.993 0.04 21
1.28 0.988 0.04 21
0.44 0.85 -0.05 1.28
0.66 (0.13) 0.82 (0.08) -0.04 (0.07) 1.25 (0.17) -0.05 1.52 0.974 0.04 16
(0.15)
(0.09) (0.08) (0.06)
-0.06 1.50 0.988 0.05 20
K
0.05 20 338
0.67 0.96 -0.02 1.31
K
0.39 (0.15) 0.97 (0.09) -0.06 (0.09) 1.42 (0.06) -0.06 1.46 0.990
Temperature: -SP,
308
0.42 (0.15) 1.04 (0.09) -0.03 (0.09) 1.51 (0.06) -0.03 1.45 0.9905 0.05 21
Temperature: -SPo
(0.08) (0.08) (0.05)
1.38 0.993 0.04 21
Temperature: -SPo
(0.13)
0.06
21
80
K
0.58 1.15 0.07 1.59
0.10 (0.08)
70
0.42 (0.13) 0.82 (0.07) -0.10 (0.08) 1.27 (0.05) -0.12 1.54 0.990 0.04
21
0.40 (0.12) 0.63 (0.06) -0.21 (0.08) 0.98 (0.04) -0.33 1.55 0.993 0.03 16
0.46 (0.15) 0.77 (0.08) -0.02 (0.11) 1.25 (0.06) -0.03 1.62 0.992 0.04 16
0.69 (0.16) 0.73 (0.09) 0.07 (0.11) 1.10 (0.06) 0.10 1.51 0.987 0.04 16
K
0.33 (0.20) 0.93 (0.11) -0.17 (0.15) 1.38 (0.08) -0.18 1.48 0.989 0.06 1:
Correlation coefficient. *Root mean square residual. Number of test solutes. more difficult to achieve, particularly in mobile phases that are relatively rich in acetonitrile (>50% volume, see Figure 4). This hypothesis may well explain why plots of In k’vs. volume fraction are less linear when acetonitrile vs. methanol is used as the organic modifier. We wish to thank H. Colin for calling this point to our attention. The ratio -blm, rather than the absolute value of either, is more consistent from column to column as the amount of mobile-phase modifier is varied. As in our preceding study, with a considerably different group of solutes, -b/m is about +1.3 to +1.4 in acetonitrile-water mixtures. This should be compared to +1.3 for methanol-water and +1.75 for THFwater. The results in Table I1 indicate a slight upward trend in the ratio with temperature in accord with our statement that m decreases more than -b. In our previous study, we found that b and m were very strongly correlated when both the mobile and stationary phases were changed. As shown in Figure 5 and its caption, when the result at 40% acetonitrile is excluded, the remaining data are very highly correlated. Indeed, the intercept of plots of b vs. m is small-less than 0.2 except a t 338 K-and the slopes are significantly different from unity, except a t the
highest temperature. It appears that the system is dominated almost entirely by the variation in water content. The strong correlation between b and m as the mobile phase and temperature are varied lends support to our previous hypothesis that Snyder’s concept of a single master parameter (22) for reversed-phase HPLC represents some linear combination of the rn and b coefficients which varies as the water content is manipulated. The ability of the LSER to represent solute retention in terms of the solvatochromic parameters is demonstrated in Figure 6, which is a plot of the residual between the calculated and experimental logarithmic capacity factors vs. the experimental value. It can be seen that the distribution about the mean of both axes, indicated by the solid lines, is fairly uniform. The alkylbenzenes tend to lie above the mean and the alcohols below the mean, but there are some of each group on either side of the average residual. The various contributions to retention for an alkylbenzene can be clearly discerned in Table 111. It is surprising that the hydrogen bond basicity of alkylbenzenes contributes so strongly to the changes in retention. This result suggests that linearity in plots of log k’ and volume fraction of modifier
ANALYTICAL CHEMISTRY, VOL. 58, NO. 13, NOVEMBER 1986
2679
0
1
b Holooromolic Dichlorooromotic
10 Nitrobenzene
Methanol 0
Condilions 298'K. 50% Acetonitrile
+
, 0
Acetonitrite
0.4
1
j 1.2
0.8
1.6
log10 G x p
Figure 6. Residual plot. All results are for solutes at 298 K and 50% acetonitrile. 1.9
0
025-
9
1.7
Retention a t 298'K
1.5
0
\ 20 40 60 80 100
1.3
0
VOLUME FRACTION ORGANIC MODIFIER
Flgure 4. Plot of ETNvs. volume fraction of solvent in water. Data from ref 21.
1.1
4
-g
0
0.9
0.7
0.5
0.z
0.1
1
-0.1
298°K
0 308"K D
vol %
I
4
0.7
1.1
as
1
1
9-0
Table 111. Retention of Ethylbenzene as a Function of the Solvatochromic Parameters"
323°K
0 338'K
I
1
50 60 70 80 Volume Fraction Acetonitrile
Figure 7. Plot of log k' vs. volume fraction acetonitrile for alkylbenzenes. All data are at 298 K.
n 0
1
40
I
I
1.5
1.9
+
1.55m: Figure 5. Correlation of b vs. m: (a) 298 K, -b = -0.19 r = 0.9994; (b) 308 K, -b = 0.19 -t 1.28m: r = 0.999: (c) 323 K, -b = 0.02 -I- 1.53m; r = 0.9997; (d) 338 K, -b = 0.40 4- 1.03m; r = 0.993.
arises in part from a compensation in cavity formation and hydrogen bond acidity of the mobile phase. Further, it sug-
40 50 60 70 80
*
SPo -0.85 -0.51 -0.58 -0.67 -0.67
m~z/lOO 2.436 1.774 1.523 1.271 1.086
SQ*
0.149 0.048 0.034 -0.009 0.00
bP2
log k'
log k', calcd
0.305 0.244 0.207 0.17 0.136
1.430 1.068 0.77 0.422 0.28
1.476 1.098 0.772 0.539 0.281
@Resultsare for data at 298 K. *Acetonitrile-water mixtures.
gests that plots of log k'vs. volume fraction for large nonpolar, nonhydrogen bond accepting species should be less linear than for small nonpolar species. This hypothesis is explored in the data shown in Figure 7, which is a plot of the logarithmic
2680
Anal. Chem. 1986,58,2680-2683
capacity factor for a series of alkylbenzenes vs. volume fraction of acetonitrile. It is clear that the larger solutes deviate from the line defined by the results from 50 to 90% modifier to a greater extent as the test solute increases in size.
ACKNOWLEDGMENT The authors also thank C. Reichardt for sharing his data on the ET^ values in mixed aqueous solvents. Registry No. H20,7732-18-5;benzene, 71-43-2;fluorobenzene, 462-06-6; chlorobenzene, 108-90-7; bromobenzene, 108-86-1; toluene, 108-88-3; iodobenzene, 591-50-4; rn-dichlorobenzene, 541-73-1;o-dichlorobenzene,95-50-1;p-dichlorobenzene, 106-46-7; ethylbenzene, 100-41-4; m-xylene, 108-38-3; o-xylene, 95-47-6; mesitylene, 108-67-8;isopropylbenzene,98-82-8; o-diethylbenzene, 135-01-3; durene, 95-93-2; tert-butylbenzene, 98-06-6; benzyl alcohol, 100-51-6;2-phenylethanol, 60-12-8; dimethylbenzyl alcohol, 29718-36-3; 3-phenylpropanol, 122-97-4; nitrobenzene, 98-95-3;benzaldehyde, 100-52-7;phenol, 108-95-2;p-chlorophenol, 106-48-9;p-nitrophenol, 100-02-7;acetonitrile, 75-05-8.
(6)Melander. W. R.; Chen, B.-K.; Horvath. Cs. J. Chromatogr. 1985, 378, 1. (7) Melander, W. R.; Horvath, Cs. Chromatographia 1984, 353, 18. (8) Snyder, L. R. J. Chromatogr. 1979, 179,167. (9) Gant, J. R.; Dolan, J. W.; Snyder, L. R. J. Chromatogr. 1979, 785, 153. (10) Dec, S. F.; Gill, S. J. J. Chem. f d u c . 1985, 62,879. (11) Yonker, C. R.; Zwier, T. A.; Burke, M. F. J. Chromatogr. 1982, 2 4 4 , 257. (12) Yonker. C. R.; Zwier, T. A,; Burke, M. F. J . Chromatogr. 1982, 244, 269. (13) Gilpin, R . K.; Squires, J. A. J. Chromatogr. Sci. 1981, 79,195. (14) Laurance, C.; Nicolet, P. J. Chem. Soc., Perkin Trans. 2 ,in press. (15) Xindu, G.; Regnier, F. E. J. Chromatogr. 1985, 332, 147. (16) Weisberg, S. Applied Linear Regression; Wiley: New York, 1980. (17) Nahum, A.; Horvath, Cs. J. Chromatogr. 1981. 203,33. (18) Jinno, K.; Kawasaki, K. J. Chromatogr. 1984, 376, 1. (19) Sadek, P. C.; Carr, P. W.; Bowers, L. D. J. Li9. Chromatogr. 1985, 8 , 2359. (20) Krygowski, T. M.; Reichardt, C.; Wronka, P. K.; Wyszomirska, C.; Zielkowska, U. J. Chem. Res. 1983, 116. (21) Krygowski, T. M.; Wronka, P. K.; Zielkowska, U.; Reichardt, C. Tetrahedron 1985, 47,4519. (22) Antle, P. E.; Goldberg, A. P.; Snyder, L. R. J. Chromatogr. 1985, 321, 1.
LITERATURE CITED Sadek, P. C.; Carr, P. W.; Doherty, R. M.; Kamlet, M. J.; Tan R. W.; and Abraham, M. H. Anal. Chem. 1985, 57, 2971-2978. Taft, R. W.; Abraham M. H.; Doherty R. M.; Kamlet, M. J. Nature (London) 1985, 313,384. Taft, R. W.; Abraham, M. H.; Famini, G. R.; Doherty, R. M.; Abboud, J. L.; Kamlet, M. J. J. Pharm. Sci. 1985, 74,807. Melander, W. R.; Campbell, D. E.; Horvath, Cs. J . Chromatogr. 1978, 158. 215. Melander, W R.; Chen, B.-K.; Horvath. Cs. J . Chromatogr. 1979, 185, 99.
RECEIVED for review March 18, 1986. Accepted June 9, 1986. Work at the University of Minnesota was supported by a grant from the National Science Foundation. The work by R. W. Taft was supported in part by a grant from the Public Health Service as was the work at Yale University. The work by M. J. Kamlet and R. M. Doherty was done under Naval Surface Weapons Center Task IR-060.
Liquid Chromatography-Gas Chromatography Interfacing Using Microbore High-Performance Liquid Chromatography with a Bundled Capillary Stream Splitter Thomas V. Raglione and Richard A. Hartwick* Department of Chemistry, Busch Campus, Rutgers University, Piscataway, New Jersey 08854
An on-line LC-GC system Is presented. HPLC peak volume reductlon was accompllshed through the use of mlcrobore (1.0 mm) columns. Further volume reduction was accomplished by use of a novel bumlled multlcaplllarystream spWtter between the LC detector and the GC. Appllcatlons of the system for the separatkn of solvent-refined coal samples are presented. ReproduclblWtles of the dlrect and split systems were 13 % and 9 % relathre standard deviatlon, respectlvely.
True multidimensional separations offer fundamental advantages over single-mode techniques in the separation of complex mixtures (1,2). Many GC analyses of environmental and biomedical samples directly or indirectly depend upon off-line LC for sample cleanup and/or class fractionation. The coupling of HPLC to high-resolution GC is thus an attractive instrumental design that should find applications in many areas of analysis. Interest in developing efficient HPLC-GC interfaces is not new. Majors (3) and Apffel and McNair ( 4 ) were among the first to attempt on-line coupling by employing conventional HPLC systems coupled to capillary GC systems. However, due to the large eluted peak volumes associated with conventional (4.6 mm i.d.) HPLC systems, it was only possible to heart-cut a small fraction of the LC peak to the GC, making 0003-2700/86/0358-2680$0 1.50/0
quantitation difficult. Nevertheless, this early work clearly demonstrated the feasibility and power of the technique, as well as the areas that needed improvement. Grob et al. (5, 6) employed an on-column concentration technique coined “retention gap” as a means of improving quantitation by permitting the transfer of larger peak volumes to the GC. Although it was possible to transfer up to 1mL of eluent, the length of retention gap required (approximately 50 m) resulted in solvent evaporation times of up to 1 h. Munari et al. (7) noted the drawbacks of such long retention gaps and investigated the use of shorter (10 m) gap lengths. The use of these shorter retention gaps places restrictions on the solutes of interest as well as the LC mobile phase. Concurrent solvent evaporation requires that the solutes of interest elute a t least 50 OC above the boiling point temperature of the mobile phase. Reversed-phase separations would have limited applicability due to the high-boiling aqueous mobile phases employed (8). Though the retention gap technique is time-consuming and has limited compatibility with reversed-phase LC, it is a valuable technique especially when applied to trace analysis. More recently, in an attempt to more closely match the volume requirements of liquid and gas chromatography, Cortes et al. (9, IO) have successfully interfaced packed capillary (300 pm i.d.) HPLC with capillary GC using the retention gap method, However, miniaturization of the HPLC to this level Q 1986 American Chemical Society
