Anal. Chem. 2006, 78, 5813-5822
Structure-Function Relationships in High-Density Docosylsilane Bonded Stationary Phases by Raman Spectroscopy and Comparison to Octadecylsilane Bonded Stationary Phases Zhaohui Liao,† Christopher J. Orendorff,†,‡ Lane C. Sander,§ and Jeanne E. Pemberton*,†
Department of Chemistry, University of Arizona, 1306 East University Boulevard, Tucson, Arizona 85721, and National Institute of Standards and Technology, Gaithersburg, Maryland 20899
Raman spectroscopy is used to investigate the effects of temperature, surface coverage, polymerization method (surface or solution polymerized), and nature of the alkylsilane precursor on alkyl chain conformational order in a series of high-density docosylsilane (C22) stationary phases at surface coverages ranging from 3.61 to 6.97 µmol/m2. The results of this study contribute to an enhanced understanding of the shape-selective retention characteristics of these phases at a molecular level. Conformational order is evaluated using the intensity ratio of the antisymmetric and symmetric ν(CH2) modes as well as the frequency at which these modes are observed. Alkyl chain order is shown to be dependent on surface coverage, alkyl chain length, and polymerization method. In general, alkyl chain order increases with surface coverage. Temperature-induced changes are observed between 250 and 350 K for the three phases with surface coverages between 3.61 and 4.89 µmol/m2. These changes occur over a broad range of temperatures characteristic of twodimensional systems, but in general adhere to the behavior predicted for a simple first-order transition. This change is not believed to be an abrupt cooperative disassociation characteristic of a phase transition in a bulk phase, but instead is thought to involve significant changes in conformational order in segments of the surfacetethered molecules, mostly those segments at the outer edge of the alkylsilane. In contrast to the changes observed in coverages below 5 µmol/m2, a first-order change is not observed for the stationary phase with coverage of 6.97 µmol/m2. A molecular picture of the temperatureinduced disorder is proposed with disorder originating at the distal carbon and propagating only a short distance toward the proximal carbon. A comparison is made between these C22 stationary phases and similar highdensity octadecylsilane (C18) bonded phases.
* To whom correspondence should be addressed: Phone: (520) 621-8245. E-mail:
[email protected]. † University of Arizona. ‡ Present address: Sandia National Laboratories, Albuquerque, NM 87185. § National Institute of Standards and Technology. 10.1021/ac060385p CCC: $33.50 Published on Web 07/18/2006
© 2006 American Chemical Society
Reversed-phase liquid chromatography (RPLC) is one of the most widely used separation modes in liquid chromatography. Stationary phases for RPLC are typically formed by covalent bonding of mono-, di-, or trichlorosilanes to silica particulate substrates. Although this technique is now relatively mature, the development of RPLC has been mainly based on empirical observations. A fundamental understanding of the molecular basis of separation processes on chemically modified silicas has yet to be achieved. Such processes include the interactions of solute and mobile phase with the stationary phase and intermolecular interactions of the stationary phase. Each of these parameters is known to affect solute retention and separation in chromatographic systems.1-4 Thus, an understanding of the molecular processes giving rise to retention is requisite to developing chromatographic systems of well-defined selectivity. A wide range of spectroscopic techniques, including nuclear magnetic resonance (NMR) spectroscopy,5-12 infrared spectroscopy,7,13-17 fluorescence spectroscopy,18,19 and Raman spectroscopy,20-31 have been employed to (1) Dorsey, J. G.; Cooper, W. T. Anal. Chem. 1994, 66, 857A-867A. (2) Seibert, D. S.; Poole, C. F. Chromatographia 1995, 41, 51-60. (3) Kiridena, W.; Pole, C. F. Chromatographia 1998, 48, 607-614. (4) Schunk, T. C.; Burke, M. F. J. Chromatogr. 1993, 656, 289-316. (5) Brindle, R.; Pursch, M.; Albert, K. Solid State Nucl. Magn. Reson. 1996, 6, 251-266. (6) Jinno, K. J. Chromatogr. Sci. 1989, 27, 729-734. (7) Jinno, K.; Wu, J.; Ichikawa, M.; Takata, I. Chromatographia 1993, 37, 627634. (8) Albert, K.; Bayer, E. J. Chromatogr. 1991, 544, 345-370. (9) Albert, K.; Brindle, R.; Martin, P.; Wilson, I. D. J Chromatogr., A 1994, 665, 253-258. (10) Sentell, K. B. J. Chromatogr., A 1993, 656, 231-263. (11) Pursch M.; Sander, L. C.; Albert, K. Anal. Chem. 1999, 71, 733A-741A. (12) Albert K. J. Sep. Sci. 2003, 26, 215-224. (13) Sander, L. C.; Callis, J. B.; Field, L. R. Anal. Chem. 1983, 55, 1068-1075. (14) Ohtake, T.; Mino, N.; Ogawa, K. Langmuir 1992, 8, 2081-2083. (15) Tripp, C. P.; Hair, M. L. Langmuir 1992, 8, 1120-1126. (16) Singh, S.; Wegmann, J.; Albert, K.; Muller, K. J. Phys. Chem. B 2002, 106, 878-888. (17) Neumann-Singh, S.; Villanueva-Garibay, J.; Muller, K. J. Phys. Chem. B 2004, 108, 1906-1917. (18) Carr, J. W.; Harris, J. M. Anal. Chem. 1987, 59, 2546-2550. (19) Wong, A. L.; Hunnicutt, M. L.; Harris, J. M. Anal. Chem. 1991, 63, 10761081. (20) Thompson, W. R.; Pemberton, J. E. Anal. Chem. 1994, 66, 3362-3370. (21) Ho, M.; Cai, M.; Pemberton, J. E. Anal. Chem. 1998, 70, 2613-2616. (22) Doyle, C. A.; Vickers, T. J.; Mann, C. K.; Dorsey, J. C. J. Chromatogr., A 1997, 779, 91-112. (23) Ho, M.; Pemberton, J. E. Anal. Chem. 1998, 70, 4915-4920.
Analytical Chemistry, Vol. 78, No. 16, August 15, 2006 5813
investigate the molecular behavior of the alkylsilane chains. Although the results of these investigations have provided valuable insight into the retention process, a clear understanding of the complex architecture of alkylsilane-modified stationary phases remains a significant challenge. In conventional RPLC, retention behavior for solutes is often rationalized in terms of solute/mobile-phase polarity. Separation of analytes can usually be achieved by adjusting polar and nonpolar interactions through polarity of the mobile phase. However, explanations based solely on polarity are inadequate to address the retention behavior of solutes with similar polarity. One example of such solutes is geometric isomers such as polycyclic aromatic hydrocarbons (PAHs). Isomeric PAHs are often difficult to separate with monomeric octadecylsilane (C18) columns, since the compounds exhibit similar nonpolar characteristics, yet separations are possible with shape-selective columns such as polymeric C18 columns. Separations of shape-constrained solutes, including PAHs,32-36 polycyclic aromatic heterocycles,37 polychlorinated biphenyl congeners,38 steroids,39 and carotenoids,40-42 are relatively insensitive to changes in mobile-phase composition. Instead, these shapeconstrained solutes are separated based on molecular shape. Wise and co-workers43 defined a molecular descriptor known as “lengthto-breadth” (L/B) ratio to describe solute retention for these systems. Long, narrow solutes with large L/B values elute later than square-shaped solutes with smaller L/B values. The ability of a stationary phase to discriminate between geometric isomers or related compounds on the basis of molecular shape is referred to as shape selectivity. In general, shape selectivity often improves with increasing surface coverage.32,44,45 (24) Doyle, C. A.; Vickers, T. J.; Mann, C. K.; Dorsey, J. G. J. Chromatogr., A 2000, 877, 25-39. (25) Doyle, C. A.; Vickers, T. J.; Mann, C. K.; Dorsey, J. G. J. Chromatogr., A 2000, 877, 41-59. (26) Pemberton, J. E.; Ho, M.; Orendorff, C. J.; Ducey, M. W. J. Chromatogr., A 2001, 913, 243-252. (27) Ducey, M. W.; Orendorff, C. J.; Pemberton, J. E.; Sander, L. C. Anal. Chem. 2002, 74, 5576-5584. (28) Ducey, M. W.; Orendorff, C. J.; Pemberton, J. E.; Sander, L. C. Anal. Chem. 2002, 74, 5585-5592. (29) Orendorff, C. J.; Ducey, M. W.; Pemberton, J. E.; Sander L. C. Anal. Chem. 2003, 75, 3360-3368. (30) Orendorff, C. J.; Ducey, M. W.; Pemberton, J. E.; Sander L. C. Anal. Chem. 2003, 75, 3369-3375. (31) Orendorff, C. J.; Pemberton, J. E. Anal. Bioanal. Chem. 2005, 382, 691697. (32) Sander, L. C.; Wise, S. A. J. Chromatogr. 1984, 316, 163-181. (33) Wise, S. A.; Benner, B. A., Jr.; Liu, H.; Byrd, G. D.; Colmsjo, A. Anal. Chem. 1988, 60, 630-637. (34) Sander, L. C.; Wise, S. A. Anal. Chem. 1984, 56, 504-510. (35) Sander, L. C.; Wise, S. A. Anal. Chem. 1987, 59, 2309-2313. (36) Wise, S. A.; Sander, L. C. May, W. E. J. Chromatogr. 1993, 642, 329-349. (37) Wise, S. A.; Sander, L. C. In Chromatographic Separations Based on Molecular Recognition; Jinno, K., Ed.; Wiley-VCH: New York, 1997; pp 1-64. (38) Sander, L. C.; Parris, R. M.; Wise, S. A. Garrigues, P. Anal. Chem. 1991, 63, 2589-2597. (39) Olsson, M.; Sander, L. C.; Wise, S. A. J. Chromatogr. 1991, 537, 73-83. (40) Sander, L. C.; Sharpless, K. E.; Craft, N. E.; Wise, S. A. Anal. Chem. 1994, 66, 1667-1674. (41) Bell, C. M.; Sander, L. C.; Fetzer, J. C.; Wise, S. A. J. Chromatogr., A 1996, 753, 37-45. (42) Bell, C. M.; Sander, L. C.; Wise, S. A. J. Chromatogr., A 1997, 757, 29-39. (43) Wise, S. A.; Bonnett, W. J.; May, F. R. J. Chromatogr. Sci. 1981, 19, 457465. (44) Sander, L. C.; Pursch, M.; Wise, S. A. Anal. Chem. 1999, 71, 4821-4830. (45) Wise, S. A.; Sander, L. C. J. High Resolut. Chromatogr. Chromatogr. Commun. 1985, 8, 248-255.
5814
Analytical Chemistry, Vol. 78, No. 16, August 15, 2006
Although temperature often plays a minor role in the development of separations method in liquid chromatography, the use of elevated and subambient temperatures has been reported as a way of changing selectivity for the separation of PAH molecules.46 In this previous report, Sander and Wise demonstrated differences in shape selectivity for the separation of planar versus nonplanar molecules as a function of temperature between 256 and 372 K. As indicated by chromatographic studies,47 stationary-phase shape selectivity increases with increasing alkyl chain length. Therefore, longer chain stationary phases, such those prepared from docosylsilane (C22), are usually used for separation of shape-constrained compounds. Sander and co-workers have reported shape selectivity for a series of C22 stationary phases prepared from different precursors and grafting procedures, yielding stationary phases with surface coverages ranging from 3.61 to 6.97 µmol/m2.48 An empirical model called the “slot model” has been developed to describe shape-selective solute retention.45 In this model, alkyl chains are viewed as slots into which solute molecules penetrate. Thus, solute shape and slot width are important to this representation. Bulky, square molecules should be retained less than extended, planar molecules. This model also predicts that shape selectivity and absolute retention should increase with increasing surface coverage of the stationary phase. However, at very high bonding densities, solute association with the alkyl chains may be limited by the available interchain spaces, and retention should decrease. This phenomenon was first observed experimentally by Sentell and Dorsey49 for a series of monomeric octadecylsilane (C18) stationary phases with loadings of between 1.6 and 4.1 µmol/ m2. Sander and Wise50 observed a similar trend. Although numerous chromatographic observations indicate that stationary-phase shape selectivity is affected by surface coverage, polymerization method, alkyl chain length, and column temperature,44,51 the shape recognition process at the molecular level is still not completely understood. Understanding the stationary-phase structure resulting from changes in these parameters should facilitate the development of an understanding of the origin of stationary-phase shape selectivity at the molecular level. Previous studies have demonstrated that Raman spectroscopy is a powerful tool for determining subtle conformational changes in alkyl chains.23,26-31,52-61 Such information is useful in understanding the chemical environment of alkyl chains and provides (46) Sander, L. C.; Wise, S. A. Anal. Chem. 1989, 61, 1749-1754. (47) Sander, L. C.; Wise, S. A. Anal. Chem. 1987, 59, 2309-2313. (48) Pursch, M.; Sander, L. C.; Egelhaaf, H.; Raitza, M.; Wise, S. A.; Oelkrug, D.; Albert, K. J. Am. Chem. Soc. 1999, 121, 3201-3213. (49) Sentell, K. B.; Dorsey, J. G. Anal. Chem. 1989, 61, 930-934. (50) Sander, L. C.; Wise, S. A. Anal. Chem. 1995, 67, 3284-3292. (51) Sander, L. C.; Wise, S. A. J. Chromatogr., A 1993, 656, 335-351. (52) Orendorff, C. J.; Ducey, M. W.; Pemberton, J. E. J. Phys. Chem. A 2002, 106, 6991-6998. (53) Snyder, R. G.; Schachtschneider, J. H. Spectrochim. Acta 1963, 19, 117168. (54) Spiker, R. C.; Levin, I. W. Biochim. Biophys. Acta 1975, 388, 361-373. (55) Snyder, R. G.; Hsu, S. L.; Krimm, S. Spectrochim. Acta 1978, 34A, 395406. (56) Snyder, R. G.; Scherer, J. R.; Gaber, B. R. Biochim. Biophys. Acta 1980, 601, 47-53. (57) Wallach, D. F. H.; Varma, S. P.; Fookson, J. Biochim. Biophys. Acta 1979, 559, 153-208. (58) Gaber, B. P.; Peticolas, W. L. Biochim. Biophys. Acta 1977, 465, 260-274. (59) Larson, K.; Rand, R. P Biochim. Biophys. Acta 1973, 326, 245-255. (60) Snyder, R. G.; Scherer, J. R. J. Chem. Phys. 1979, 71, 3221-3228.
Table 1. Stationary-Phase Properties48
a
stationary phase
previous stationaryphase designationa
silane functionality
polymerization method
% carbon
surface coverage (µmol/m2)
shape selectivity parameter Rb
DFC22SL DFC22SF TFC22SL TFC22SF
A B C D
dichloromethyl dichloromethyl trichloro trichloro
solution surface solution surface
15.01 18.33 18.17 23.38
3.61 4.65 4.89 6.97
1.25 0.70 0.47 0.12
Used as designation in previously published work on these phases; see ref 48. b As defined in ref 47.
valuable insight into the interactions between alkyl chains and their environment. Alkyl chain order was assessed using the intensity ratio of the antisymmetric to symmetric ν(CH2) modes (I[νa(CH2)]/I[νs(CH2)]) as well as the frequency at which these modes are observed. Temperature-induced changes in alkyl chain conformational order were previously measured between 258 and 343 K for a similar homologous series of octadecylsilane (C18) stationary phases.27 Using a Raman spectral indicator of conformational alkyl order as a metric of pressure, it was found that these changes generally conformed to the Clapeyron relationship between pressure and temperature for a simple first-order phase transition. For the limited range of systems investigated, however, a simple correlation between stationary-phase preparation (surface versus solution polymerization and nature of the silane precursor) and extent of alkyl chain order was not clearly observed. Instead, alkyl chain order was generally observed to be dependent on absolute alkylsilane surface coverage. The alkyl chain structural attributes for a similar series of C22 stationary phases as developed by Sander and co-workers48 are investigated here using Raman spectroscopy. This report describes the effects of temperature, surface coverage, nature of the alkylsilane precursor (trichloroalkylsilane or dichloromethylalkysilane), and surface modification method (surface or solution modification) on alkyl chain conformational order in a series of high-density C22 stationary phases ranging in surface coverage from 3.61 to 6.97 µmol/m2 to elucidate trends in behavior that may be attributed to differences in shape selectivity. As was done previously with the C18 phases,27 the resulting behavior is analyzed using the Claperyon relationship for a simple first-order phase transition. A comparison is made between the Raman spectral results from C18 and C22 stationary phases in an attempt to better understand differences in shape selectivity and retention resulting from alkyl chain length. EXPERIMENTAL SECTION Stationary Phases. C22 stationary phases were prepared and characterized as previously described.48 The preparation method can be summarized as follows. Difunctional and trifunctional stationary phases were prepared using dichloromethyl- and trichlorodocosylsilane, respectively. Of the four stationary-phase materials investigated, two were synthesized by the “surface polymerization” method, and two were synthesized by the “solution polymerization” method. Surface-polymerized phases were (61) Synder, R. G.; Strauss, H. L.; Elliger, C. A. J. Phys. Chem. 1982, 86, 51455150.
prepared by adding the di- or trifunctional silane to previously humidified silica, while the solution-polymerized phases were prepared by the addition of water to the precursor/silica slurry. These stationary phases were prepared using 5-µm YMC silica (YMC, Inc.; pore size 200 Å.) The identity of the precursor, polymerization method, and resulting properties of each stationary phase are summarized in Table 1. Instrumentation. Raman spectra were collected using 100 mW of 532-nm radiation from a Coherent Verdi Nd:YVO4 laser (Coherent, Inc.) on a Spex Triplemate spectrograph. Slit settings of the Triplemate were 0.5/7.0/0.15 mm for all experiments, corresponding to a spectral band-pass of 5 cm-1. The detector in these experiments was a Princeton Instruments charged-coupled device (CCD) system based on a thinned, back-illuminated, antireflection-coated RTE-110-PB CCD of pixel format 1100 × 330, which was cooled with liquid N2 to 183 K. Samples were sealed in 5-mm-diameter NMR tubes and positioned in a 200-µm-diameter laser beam using a copper sample mount. The temperature of the copper mount was regulated by circulating through it a temperature-controlled medium (50:50 ethylene glycol/water), allowing temperature control from 258 to 343 K. Samples were allowed to equilibrate at each temperature for a minimum of 30 min prior to analysis to ensure that any temperature-induced changes in conformational order were complete. A minimum of three measurements was made on each sample at each temperature, and standard deviations were determined. Error bars are plotted for all data points in the figures. In cases wherein the error bars are not obvious, they are smaller than the size of the symbol used to indicate the average value. Integration times for each spectrum are provided in the figure captions. Low laser power and short acquisition times were used to avoid unnecessary sample heating by the laser beam. Spectral Data Analysis. Spectra were superimposed on a linear background in the ν(CH) frequency region. This background is a complex convolution of detector response, spectrometer response, and sample (alkylsilane and background) scattering characteristics. Spectral data were fit to a linear background in the frequency region between 2750 and 3050 cm-1 that was subtracted from the raw spectra to remove these contributions in a single step. Peak intensities of the νa(CH2) and νs(CH2) bands were determined as peak intensities measured from the flat baseline in the resulting spectra. Molecular Pictures. Pictures of the solvent/stationary-phase interface were constructed using Chem3D (CambridgeSoft). Alkyl chains of the stationary phase were arranged in a 6 × 3 (or 4 × 3 depending on surface coverage) matrix; the spacing of the chains Analytical Chemistry, Vol. 78, No. 16, August 15, 2006
5815
was determined from the relationship between the average SiO-Si spacing in the silica matrix (3.2 Å62) and the theoretical bonding density of close-packed alkyl chains on a silica substrate (∼8 µmol/m2). For example, a 4.00 µmol/m2 coverage material is half the coverage of the theoretical limit. Therefore, the alkylsilane spacing of a homogeneously distributed 4.00 µmol/ m2 material should be twice the silica Si-O-Si spacing at the theoretical limit, or 6.4 Å. For the spacing between difunctional alkylsilanes containing Si-bound methyl groups, the minimum distance between alkyl chains is assigned as the diameter of the methyl groups in both dimensions in order to account for all possible orientations of the methyl groups. The silica surface is depicted with surface curvature in the resulting figures to more accurately represent the surface morphology of the particulate silica support. Molecular models from molecular mechanics were energyminimized using the modified MM2 force field computation provided by Chem3D Pro. This computation employs numerous chemical parameters including bond stretch energy, angle bend energy, torsion and nonbonding constraints, π-system calculations, charge and dipole terms, and cutoff parameters for van der Waals and electrostatic interactions. During the energy minimization computation, the silicon atoms, the oxygen atoms, and the five proximal methylene groups on each alkyl chain were fixed in space, simulating a surface-confined, densely packed array of alkyl moieties. The molecular models at high temperature were energyminimized along with dynamic control of heating to 350 K. It is important to note that while the models presented here are energyminimized by the modified MM2 computation, they are used only to develop a less arbitrary picture of the chromatographic interface and to provide a framework for visualizing the relevant intermolecular interactions reported by the Raman spectral data.
Figure 1. Raman spectra in the ν(CH) region for DFC22SL, DFC22SF, TFC22SL, and TFC22SF at (a) 258, (b) 263, (c) 313, and (d) 333 K. All spectra except DFC22SL acquired with 2-min integrations, while DFC22SL spectra acquired with 5-min integration. Table 2. Raman Peak Frequencies and Assignments for C22 Stationary Phases peak frequency (cm-1)
RESULTS AND DISCUSSION Alkylsilane Order from ν(CH2) Intensities. Raman spectral studies of C18 stationary phases have been described in previous publications from this laboratory.21,26-31 As expected, C22 stationary phases exhibit similar Raman spectroscopy. Nevertheless, the behavior of the C22 phases is sufficiently different to warrant their further consideration. Here, the effect of temperature on the conformational order of a series of high-density C22 stationary phases is studied over the temperature range from 258 to 343 K by determination of the magnitudes of two Raman spectral order indicators in the ν(CH) region between 2750 and 3050 cm-1. Within this region, bands due to symmetric and antisymmetric ν(CH2) and ν(CH3) vibrations, as well as bands due to Fermi resonance of these modes with overtones of C-H bending modes, are observed. This region is highly complex for alkanes and not suitable for spectral decomposition because of the large number of methylene vibrational and Fermi resonance modes of poorly defined spectral characteristics (e.g., peak frequency, width, etc.)53,55,60,61 Thus, empirical indicators of conformational order have been extensively used in the study of alkyl-containing systems26-31.52-64 such as those of interest here. Figure 1 represents typical Raman spectra in this region for stationary phases prepared from difunctional (DFC22SL and DFC22SF) and trifunctional (TFC22SL and TFC22SF) docosylsi(62) Stevens, M. J. Langmuir 1999, 15, 2773-2778.
5816 Analytical Chemistry, Vol. 78, No. 16, August 15, 2006
DFC22SL
DFC22SF
TFC22SL
TFC22SF
assignment
2964 2928 2908 2896 2884 2851
2964 2926 2906 2896 2880 2848
2964 2925 npa 2897 2880 2848
2963 2930 np 2898 2880 2847
νa(CH3) νs(CH3)FR ν(CH3)Si νs(CH3) νa(CH2) νs(CH2)
a
np, methyl group not present in stationary phase.
lane precursors with surface coverages of 3.61, 4.65, 4.89, and 6.97 µmol/m2, respectively. Vibrational mode assignments for these systems are well-established and are made based on those previously reported for alkyl chain systems.52-61 Table 2 provides the assignments for each of the stationary phases. In general, the prominent modes observed in each spectrum correspond to the νs(CH2) at ∼2850 cm-1, the νa(CH2) near 2880 cm-1, the νs(CH3) near 2896 cm-1, and the νs(CH3)FR near 2928 cm-1. The νs(CH3)Si mode for the methyl groups attached to the proximal silicon atom is present in the spectra of the difunctional silanes, DFC22SL and DFC22SF. This mode is located at 2906 cm-1 and is superimposed on the νs(CH3)FR near 2928 cm-1 and the νs(CH3) near 2896 cm-1. As a result of its position, it may also contribute slightly to the intensity of the νa(CH2). It was previously demonstrated that Raman spectra provide detailed information regarding alkyl chain conformation, especially
in the ν(CH) region.20,21,23,26-31,52-61 Various perturbations (e.g., temperature, solvent, etc.) cause changes in certain ν(CH) modes buried under the complex envelope. As a result, the relative peak intensity of the two prominent modes, νa(CH2) and νs(CH2), varies in response to these perturbations. In a recent study from this laboratory in which various Raman spectral indicators were quantitatively correlated,52 it was shown that, even in the crystalline state of bulk alkanes in which either no or few gauche conformers can be detected, a decrease in the I[νa(CH2)]/I[νs(CH2)] value from 2.1 to 1.2 is observed. For values from 2.1 to ∼1.5, this decrease is due to a decrease in lateral coupling between chains;58 however, from 1.5 to 1.2, this change is known to be due to the formation of end gauche and kink conformers near the ends of the alkyl chains just prior to melting.63,64 Further into the biphasic region, the average number of gauche conformers increases, starting at a I[νa(CH2)]/ I[νs(CH2)] value near 1.2 and decreasing to 0.75. In the liquid state, I[νa(CH2)]/I[νs(CH2)] remains constant while the number of gauche conformers increases with increasing temperature. Thus, I[νa(CH2)]/I[νs(CH2)] reflects the total fraction of alkyl chains that deviate in any way (i.e., conformational or gauche disorder) from an all-trans C-C bond configuration. The frequencies at which the νs(CH2) and νa(CH2) modes are observed also reflect alkane order. These frequencies are observed at 2856 and 2888 cm-1, respectively, for alkanes in the liquid state but decrease by 6-8 cm-1 for crystalline alkanes.57,58 Changes in the frequencies of these bands are largely related to the extent of coupling between neighboring alkyl chains. A higher degree of chain coupling results in a decrease in frequency for both modes.52,61 Qualitative examination of the spectra in Figure 1 reveals a wide range of rotational/conformational order within this series of stationary phases at all temperatures as reflected by the I[νa(CH2)]/I[νs(CH2)]values. Particularly pronounced is the difference in alkyl order between DFC22SL (3.61 µmol/m2) and TFC22SF (6.97 µmol/m2). This striking difference indicates that the alkyl order is affected by surface coverage of the stationary phases at a given temperature. Furthermore, when the temperature dependence of the Raman spectral response between 258 (Figure 1a) and 333 K (Figure 1d) is considered, a significant difference in alkyl chain order is especially apparent for the lowest three surface coverage materials (i.e., DFC22SL, DFC22SF, and TFC22SL). When I[νa(CH2)]/I[νs(CH2)] is investigated from 258 to 343 K, the dependence on temperature becomes even more apparent for each of the stationary-phase materials examined. Figure 2a shows plots of I[νa(CH2)]/I[νs(CH2)] as a function of temperature for the four stationary phases studied here. First, for a material of a given surface coverage, I[νa(CH2)]/I[νs(CH2)] decreases with increasing temperature, indicating that disorder is introduced. Also, at a given temperature, the magnitude of I[νa(CH2)]/ I[νs(CH2)] generally increases with surface coverage, indicating an increase in alkyl order. Interestingly, however, DFC22SF (4.65 µmol/m2) and TFC22SL (4.89 µmol/m2) exhibit a reversal of this trend; DFC22SF with lower surface coverage has a higher (63) Marconcelli, M.; Qi, S. P.; Strauss, H. L.; Snyder, R. G. J. Am. Chem. Soc. 1982, 104, 6237-6247. (64) Kim, Y.; Strauss, H. L.; Snyder, R. G. J. Phys. Chem. 1989, 93, 7520-7526.
Figure 2. (a) I[νa(CH2)]/I[νs(CH2)] as a function of temperature for TFC22SF (b), DFC22SF (4), TFC22SL (9), and DFC22SL (1). Error bars represent one standard deviation. Best fits for I[νa(CH2)]/ I[νs(CH2)] ) 1/(a ln T + b) shown as lines: TFC22SF (- ‚ - ‚ -), DFC22SF (‚ ‚ ‚), TFC22SL (-), DFC22SL (- - -). (b) dR/dT from bestfit lines in (a) as a function of 1/T. TFC22SL (9), DFC22SF (4), DFC22SL (1), and TFC22SF (b).
magnitude of I[νa(CH2)]/I[νs(CH2)], indicating that either precursor or grafting method plays a significant role in alkyl chain order. Three important regions of alkyl order as indicated by the I[νa(CH2)]/I[νs(CH2)] magnitude are shown by different shading in Figure 2a. The highest surface coverage stationary phase, TFC22SF (6.97 µmol/m2), exhibits I[νa(CH2)]/I[νs(CH2)] values ranging from 1.75 down to 1.25. This range of values extends from crystalline-like order at the lowest temperature to a state in which the onset of gauche conformer formation occurs at higher temperature. In other words, TFC22SF exhibits considerable crystalline character52 even at the highest temperature (343 K) examined here. Both TFC22SL and DFC22SF phases, with similar surface coverage but prepared from different precursors, exhibit I[νa(CH2)]/I[νs(CH2)] values ranging from 1.55 to 1.00. These values correspond to crystalline-like order at the lowest temperature to an almost liquidlike state at higher temperatures (>313 K). Finally, the stationary phase with the lowest surface coverage, DFC22SL (3.61 µmol/m2), exhibits I[νa(CH2)]/I[νs(CH2)] values going from 1.35 to 0.95. Thus, even at the lowest temperature, this phase is in a state in which some gauche conformers exist and becomes increasingly disordered with increasing temperature. In other words, based on the I[va(CH2)]/I[νs(CH2)] values observed, the lowest surface coverage stationary phase exhibits Analytical Chemistry, Vol. 78, No. 16, August 15, 2006
5817
a level of disorder almost comparable to a true liquid despite the fact that one end of all alkyl chains is tethered to the silica surface. It is interesting to note that the three stationary phases with surface coverage greater than 4 µmol/m2 investigated here exist in a crystalline-like state (i.e., few gauche conformers) at room temperature. The conformational order of this series of C22 stationary phases has also been studied using solid-state 13C CP/MAS NMR by Pursch and co-workers48 at 295 K. The results from these studies indicate that TFC22SF alkyl chains exist in an almost all-trans configuration and that the fraction of gauche defects increases with decreasing surface coverage, with DFC22SL having the greatest number of gauche conformers in agreement with the results obtained here. For each stationary phase, systematic trends in I[νa(CH2)]/ I[νs(CH2)] values as a function of temperature are indicative of significant differences in alkyl order that might reflect phase changes or partial changes in alkyl order that are similar to phase changes. Previous studies from this laboratory27 have shown that the temperature-induced changes in I[νa(CH2)]/I[νs(CH2)] for C18 stationary phases reflect what can roughly be characterized as first-order phase changes. Under these conditions, assuming a direct relationship of I[νa(CH2)]/I[νs(CH2)] and pressure, P,27 and defining R to be the inverse of I[νa(CH2)]/I[νs(CH2)], i.e.
R ) I[νs(CH2)]/I[νa(CH2)]
(1)
then
P ∝ R ) KPref + K
( )
( )
∆Hfus ∆Hfus ln T - K ln Tref ∆Vfus ∆Vfus
(2)
where K is an unknown proportionality constant, P is pressure, Pref and Tref represent arbitrary starting reference pressure and temperature, respectively, ∆Hfus represents the molar enthalpy change of fusion, ∆Vfus represents the molar volume change of fusion, and T is temperature. Thus,
( )( )
∆Hfus 1 dR )K dT ∆Vfus T
(3)
As expected, the shapes and relative positions of the resulting plots resemble those of the phase diagrams for C18 phases documented in the previous publication.27 Figure 2a also shows the best-fit lines for the function I[νa(CH2)]/I[νs(CH2)] ) [1/(a ln T + b)] with a ) [K(∆Hfus/∆Vfus)] and b ) {KPref - [K(∆Hfus/ ∆Vfus) ln Tref]} as variable parameters. The R2 values for these fits are g0.97 in all cases. The inverse of the slopes of the best-fit lines in Figure 2a, dR/ dT, were determined and are plotted as a function of inverse temperature in Figure 2b according to eq 3. The slopes of these Clayperon plots reflect the magnitude of [K(∆Hfus/∆Vfus)] for a transition that generally mimics a crystalline-like to liquidlike phase change. It should be noted here that the term “phase change” is used loosely, since a cooperative disassociation of alkyl chains at a particular transition temperature is not envisioned for the process that might be occurring within these tethered 5818 Analytical Chemistry, Vol. 78, No. 16, August 15, 2006
Table 3. Slope of Clapeyron Plots for C22 Stationary Phases stationary phase
surface coverage (µmol/m2)
slope of Clapeyron (K)
DFC22SL DFC22SF TFC22SL TFC22SF
3.61 4.65 4.89 6.97
1.07 ( 0.07 1.14 ( 0.07 1.27 ( 0.07 0.71 ( 0.04
alkylsilanes. Rather, the processes envisioned involve disordering of only a small portion of the total alkane chain of each alkylsilane molecule, most likely that portion toward the outer edge of the alkylsilane submonolayer. The Clayperon slopes determined from the Raman spectral data are given in Table 3. As discussed previously,27 for a homologous series of surface-confined alkanes, the slope of the Clayperon plot is expected to increase and then level off with surface coverage for alkylsilane systems that can undergo crystalline-like to liquidlike disordering, assuming that K is constant for all stationary phases. The data in Figure 2b suggest that only the lowest three surface coverage stationary-phase materials in this study can access order parameter values that span the full range from crystalline-like to liquidlike. In contrast, the highest surface coverage material, TFC22SF (6.97 µmol/m2), remains crystallinelike over the entire range of temperatures investigated in this study. This behavior invalidates the assumption of a constant value of K for this stationary phase compared to the other three. This analysis leads to the prediction that, for the four stationary phases studied here, the slopes of the Clayperon plots should increase with C22 surface coverage in the order DFC22SL < DFC22SF < TFC22SL. As shown in Table 3 for these three stationary phases, this trend is exactly what is observed. In contrast, the slope of the Clayperon plot for TFC22SF deviates substantially from this trend as expected. The fact that the value for TFC22SF is lower than those of the other three phases indicates that its value of either K or ∆Hfus (or both) is (are) lower, since at this coverage structural changes induced thermally are quite minor (hence restricting the values for ∆Vfus.) The existence of phase changes in alkylsilane stationary phases has been controversial for several decades.65 For chemically bonded C22 phases specifically, numerous researchers have reported direct and indirect evidence for phase changes. Indirect evidence in the form of nonlinear Van’t Hoff plots for C18 and C22 phase changes was provided by Morel and Serpinet in the early 1980s.66 For example, these researchers demonstrated nonlinear variations of the logarithm of the capacity factor, k′, as a function of inverse temperature that suggested phase transitions of densely grafted (g4 µmol/m2) C22 phase. Evidence for phase changes in dense C18 and C22 stationary-phase materials on the basis of differential scanning calorimetry studies was also reported by Hansen and Callis.67 In contrast to densely covered stationary phases, these researchers reported that no phase transition was observed for C22 materials with coverages less than 2.0-2.5 µmol/ m2. The conclusions presented here from the Raman spectral (65) Wheeler, J. F.; Beck, T. L.; Klatte, S. J.; Cole, L. A.; Dorsey, J. G. J. Chromatogr., A 1993, 656, 317-333. (66) Morel, D.; Serpinet, J. J. Chromatogr. 1982, 248, 231-240. (67) Hansen, S. J.; Callis, J. B. J. Chromatogr. Sci. 1983, 21, 560-563.
Figure 3. νs(CH2) peak frequency as a function of temperature for DFC22SL (1), DFC22SF (4), TFC22SL (9), and TFC22SF (b). Error bars represent one standard deviation.
investigation of high-coverage C22 stationary phases are generally in agreement with the results of these previous studies, with the caveat that we envision the change in order occurring to involve only a portion of each C22 chain. As noted above, despite the similar surface coverage of DFC22SF (4.65 µmol/m2) and TFC22SL (4.89 µmol/m2), the magnitudes of I[νa(CH2)]/I[νs(CH2)] for DFC22SF are larger than those for TFC22SL at all temperatures studied here. This observation suggests that surface coverage is not the only factor that dictates alkylsilane order at a given temperature. For stationary phases with similar surface coverage such as TFC22SL and DFC22SF, other factors including stationary-phase precursor, polymerization method, and alkylsilane surface homogeneity may also affect the order as well. As will be discussed in more detail below, in this case, it is likely that polymerization method plays a major role in determining alkyl order. Despite relatively minor variations from the expected trend for stationary phases of similar coverage, surface coverage is the major determinant of alkyl order in these systems. Alkylsilane Order from νs(CH2) Frequency. Insight into alksylsilane order can also be ascertained from the frequencies at which the two ν(CH2) modes are observed. These frequencies not only are sensitive to conformational disorder but also are sensitive to the degree of interchain coupling between neighboring alkyl chains.52,61 Although interchain coupling can, in some cases, be directly correlated with trans/gauche conformational state of an alkane, in other cases, interchain coupling may not change monotonically with conformation.58 The temperature dependence of the νs(CH2) frequency is shown in Figure 3. The different shading in Figure 3 indicates the three important regions of alkyl chain coupling indicated by these values: a region of change dominated predominantly by loss of lateral chain interaction, a biphasic transition region, and a region in which continued changes occur due to an increase in gauche conformers. For a given material, peak frequency increases with increasing temperature, reflecting a higher degree of interchain decoupling at higher temperature. Moreover, at a given temperature, peak frequency generally decreases with increasing surface coverage, reflecting a greater extent of interchain coupling in the higher surface coverage materials. However,
for the two phases with similar surface coverage, TFC22SL and DFC22SF, the peak frequencies are not as different as their I[νa(CH2)]/I[νs(CH2)] values; both phases exhibit an almost identical trend in νs(CH2) frequency as a function of temperature, suggesting that surface coverage dictates the available interchain space, and hence, degree of interchain coupling. This dependence is further exemplified by the behavior of DFC22SL, the phase with the lowest surface coverage, in that its frequency indicates the smallest interchain coupling consistent with its largest interchain space. The data in Figure 3 further indicate that differences in interchain coupling among these stationary phases increase with increasing temperature. TFC22SF, the phase with the highest surface coverage, exhibits the greatest interchain coupling. Differences in interchain coupling at low temperature are most likely due to the differences in the available interchain space based on surface coverage as noted above. Larger differences in the frequencies for these stationary phases at higher temperature are a further manifestation of surface coverage superimposed on the effects of thermally induced disorder on chain coupling, most likely resulting from the extent of interchain decoupling. For example, since the surface coverage of TFC22SF (6.97 µmol/m2) approaches the maximum bonding density of 8 µmol/m2, decoupling is almost impossible due to the lack of available interchain space even at higher temperature. As a result, the νs(CH2) frequency increases only slightly at higher temperature and then levels off. In contrast, for DFC22SL with a quite small surface coverage (3.61 µmol/m2), the interchain space allows more decoupling upon thermal disordering such that its frequency continues to increase with temperature. Combining the results of Figures 2 and 3, a more complete molecular picture of the temperature-dependent disorder introduced into these systems is proposed, with conformational disorder originating at the distal carbon and propagating inward slightly toward the proximal carbon. Given the length of the C22 chains used for these stationary phases and their high surface coverage, this disordering is envisioned to involve only a small alkyl segment of each chain and is not believed to be a cooperative disassociation of chains over a narrow temperature range as in a bulk material. Figure 4 shows molecular pictures resulting from energyminimized molecular mechanics computations for TFC22SF and DFC22SL, the two most extreme phases based on coverage. The TFC22SF phase exhibits the highest degree of alkane order with almost all conformers in the trans state at the lowest temperature. As the temperature rises, rotational disorder is introduced into the chains at the distal end. However, a significant population of gauche conformers cannot develop, since alkyl chain motion is sterically limited by the available interchain space. The relatively minor rotational disorder introduced thermally in this case does not significantly alter the extent of alkyl chain coupling as shown in Figure 4a. As shown in Figure 4b, for the lowest surface coverage DFC22SL phase, a large population of gauche conformers exists even at room temperature. These gauche conformers decrease chain coupling and hence raise the frequency of the νs(CH2) mode relative to the value observed in the higher surface coverage phases. As temperature increases, the number of gauche conformAnalytical Chemistry, Vol. 78, No. 16, August 15, 2006
5819
Figure 4. Proposed temperature-dependent disordering process for (a) TFC22SF and (b) DFC22SL.
ers increases rapidly leading to further decoupling of the alkyl chains. The molecular pictures for DFC22SF and TFC22SL suggest alkyl chain coupling behavior intermediate between these two extremes. Information about alkyl chain conformational order and coupling obtained from these four stationary phases provides insight into the shape-selective behavior of these stationary phases as a function of surface coverage, polymerization method, and column temperature. As has been proposed previously, it is likely that alkyl chain conformational order and interchain coupling is the origin of the shape selectivity of these stationary phases.48 In general, conformational order and interchain coupling increase with decreasing column temperature and increasing stationaryphase surface coverage, leading to an increase in shape selectivity. As shown in Figure 4a, at low temperature, TFC22SF, the phase with highest surface coverage, is highly ordered and coupled. Thus, the nonplanar PAHs or PAHs with smaller L/B values do not fit into the slots in the alkyl chains, and as a result, the shape selectivity is relatively good. In contrast, DFC22SL, the phase with the lowest surface coverage, is less ordered and less coupled and possesses sufficient interchain space to accommodate both types of solutes. As a result, the shape selectivity of this phase is relatively poor. As temperature increases, alkyl chain conformational order and interchain coupling for both TFC22SF and DFC22SL phases decrease, leading to a decrease in their shape selectivity. 5820
Analytical Chemistry, Vol. 78, No. 16, August 15, 2006
Previous chromatographic48 and NMR spectroscopic68 work also suggests that, for phases with similar surface coverages, solution-polymerized phases usually offer better shape selectivity than comparable phases prepared using surface polymerization. This behavior is only thought to be possible if the alkylsilanes in solution-polymerized phases are attached to the silica surface in islands as the result of preassembly, oligomerization, or both in solution, whereas alkylsilanes in surface-polymerized phases are more uniformly distributed across the surface. Silica surfaces modified by surface polymerization are proposed to be more homogeneous, leading to a higher degree of conformational order within the alkyl chains, while the alkyl chains on surfaces modified by solution polymerization are more heterogeneously distributed. Although the average extent of interchain coupling in solutionpolymerized stationary phases is comparable to that in surfacepolymerized stationary phases of similar coverage based on the peak frequency of the νs(CH2), in local regions of the solutionpolymerized phase, the interchain spacing is smaller due to the presence of true siloxane linkages that form in solution prior to surface bonding. This architecture of islands of heterogeneously distributed but closely coupled alkyl chains leads to the better shape selectivity of solution-polymerized phases. Comparison of High-Density C22 Stationary-Phase Structure with High-Density C18 Stationary-Phase Structure. It has been predicted previously that alkyl chain length should have an (68) Pursch, M.; Sander, L. C.; Albert, K. Anal. Chem. 1996, 68, 4107-4113.
Figure 5. (a) I[νa(CH2)]/I[νs(CH2)] as a function of temperature for surface-polymerized phases. Left axis: TFC22SF (b) and TFC18SF (O). Right axis: DFC22SF (2) and DFC18SF (4). Note that left and right axes offset by 0.2 unit for clarity. (b) I[νa(CH2)]/I[νs(CH2)] as a function of temperature for solution-polymerized phases. Left axis: TFC22SL (9) and TFC18SL (0). Right axis: DFC22SL (1) and DFC18SL (3). Note that left and right axes offset by 0.1 unit for clarity. Error bars represent one standard deviation.
effect on alkyl chain order for systems of comparable surface coverage.35 In assessing this chain length effect, we are fortunate to have available a series of C18 stationary phases with surface coverages comparable to those of the C22 phases investigated here. Moreover, these phases were prepared from the identical functional precursor using identical polymerization methods,48 and thus, this series of C18 phases allows unambiguous determination of the effect of chain length on Raman spectral indicators of order. I[νa(CH2)]/I[νs(CH2)] values as a function of temperature for C18 and C22 phases of similar surface coverage are shown in Figure 5. Figure 5a shows the comparisons for the surface-polymerized phases prepared from trifunctional and difunctional precursors, TFC22SF versus TFC18SF and DFC22SF versus DFC18SF, respectively. Figure 5b shows the comparisons for the solutionpolymerized phases prepared from trifunctional and difunctional precursors, TFC22SL versus TFC18SL and DFC22SL versus DFC18SL, respectively. Figure 5a indicates that I[νa(CH2)]/I[νs(CH2)] for TFC22SF (C22, 6.97 µmol/m2) is greater than that for TFC18SF (C18, 6.45 µmol/m2) over the entire temperature range studied, reflecting a
higher degree of alkane order in TFC22SF. The magnitude of I[νa(CH2)]/I[νs(CH2)] for TFC22SF at room temperature indicates that it exists in a state in which almost all C-C bonds are in trans conformers. In contrast, TFC18SF possesses considerably more disorder, probably in the form end gauche and kink conformers, under the same conditions. Thus, although the C22 chains are only 22% longer than the C18 chains, this extension causes a dramatic increase in alkyl chain order. Similarly, I[νa(CH2)]/I[νs(CH2)] for DFC22SF is greater than that for DFC18SF at temperatures less than 325 K. At higher temperatures, I[νa(CH2)]/I[νs(CH2)] for DFC22SF is essentially equivalent to that for DFC18SF. The difference in I[νa(CH2)]/ I[νs(CH2)] between DFC22SF and DFC18SF at temperatures of
