Investigation of Enantiomer Bonding on a Chiral Stationary Phase by

Research Group for Analytical Chemistry, Hungarian Academy of Sciences, P.O. Box 158, H-8201 Veszprém, Hungary ... of Technology, 2 George Street, GP...
0 downloads 0 Views 121KB Size
Anal. Chem. 1998, 70, 2766-2770

Investigation of Enantiomer Bonding on a Chiral Stationary Phase by FT-Raman Spectrometry E. Horva´th* and L. Kocsis

Research Group for Analytical Chemistry, Hungarian Academy of Sciences, P.O. Box 158, H-8201 Veszpre´ m, Hungary R. L. Frost

Centre for Instrumental and Developmental Chemistry, Queensland University of Technology, 2 George Street, GPO Box 2434 Brisbane, Q 4001, Australia B. Hren

Department of Analytical Chemistry, University of Veszpre´ m, P.O. Box 158, H-8201 Veszpre´ m, Hungary L. P. Szabo´

Research Institute of Chemical Engineering, Hungarian Academy of Sciences, P.O. Box 125, H-8201 Veszpre´ m, Hungary

The bonding of serine, phenylalanine, and mandelic acid enantiomers on an N-3,5-dinitrobenzoyl-L-leucine chiral stationary phase (on zeolite A support) has been investigated by FT-Raman spectrometry. It was found that retention is due to hydrogen bonds and π-stacking interactions between the stationary phase and the analyte. The involvement of the two different amide groups (as donor and/or acceptor) in the complexation reaction can be followed based on spectral data. A correlation was found between the ratio of the amide I and the ring stretching (1532 cm-1) bands and retention data. One of the fundamental questions in chromatography is how can a chiral stationary phase (CSP) discriminate between enantiomeric analytes? To answer this problem, it is necessary to know which functional groups on the CSP take part in the enantiodiscrimination and what are the intermolecular forces holding together the transient analyte-CSP complex? These questions are of importance not only to understand the nature of chiral separations but also to improve the efficiency of the process itself. The impact of chromatographic investigations made to date on the fundamental aspects of the enantioselective behavior of CSPs is rather limited due to the indirect nature of information obtained.1 On the other hand, the wealth of commercially available stationary phases often makes it difficult to select the “best” system out of the many “good” ones. Therefore, any analytical technique capable of providing additional information on chiral separations is a potential tool in the hands of the chromatographer.2 In situ or quasi in situ studies on brush-type CSPs are of considerable importance to better understand chiral recognition in chromatography.1,3,4 Although vibrational spectroscopy is the (1) Dorsey, J. G.; Foley, J. P.; Cooper, W. T.; Barford, R. A.; Barth, H. G. Anal. Chem. 1992, 64, 353R-389R. (2) Lipkowitz, K. B. J. Chromatogr. A 1994, 666, 493-503.

2766 Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

method of choice in the study of adsorbed or chemisorbed molecules, the application of infrared and Raman spectroscopy for the investigation of chromatographic systems is rather difficult5-8 due to the strong absorption as well as the low Raman scattering and fluorescence of the stationary phase. It is possible, however, to use FT-Raman spectrometry, because in this case the disturbing effect of fluorescence can be eliminated. In the present work, the potential application of FT-Raman spectrometry for the study of enantioselective bonding on brush-type CSPs is demonstrated. EXPERIMENTAL SECTION The bonding mechanism of serine, phenylalanine, and mandelic acid enantiomers on an N-3,5-dinitrobenzoyl L-leucine chiral stationary phase connected to a zeolite A support (Figure 1) has been investigated. The preparation of the stationary phase and the way of determining the active sites of a brush-type CSP has been reported previously.9 The active sites (30 mmol/100 g of CSP) were partially coated with 5 mg/cm3 aqueous solutions of the enantiomers in a way that a series of samples representing 10, 30, 50, and 70% coverage for each enantiomer was obtained. The Raman spectra of the air-dried samples were then recorded by means of a Bio-Rad dedicated FT-Raman spectrometer (Nd: YAG laser, 500 mW laser power, 3000 scans at a resolution of 2 cm-1) and corrected to the white light background. (3) Hanson, M.; Unger, K. K.; Schmid, J.; Albert, K.; Bayer, E. Anal. Chem. 1993, 65, 2249-2253. (4) Lightfoot, E. N.; Athalye, A. M.; Coffman, J. L.; Roper, D. K.; Root, T. W. J. Chromatog. A 1995, 707, 45-55. (5) Tran, C. D. J. Chromatogr. 1984, 292, 432-438. (6) Fujimoto, C.; Morita, T.; Jinno, K. J. Chromatog. 1988, 438, 329-337. (7) Ni, F.; Thomas, L.; Cotton, T. M. Anal. Chem. 1989, 61, 888-894. (8) Soper, S. A.; Ratzlaff, K. L.; Kuwana, T. Anal. Chem. 1990, 62, 14381444. (9) Szabo´, L. P.; Kallo´, D.; Szotyory, L. Mikrochim. Acta 1996, 124, 263-271. S0003-2700(97)01241-9 CCC: $15.00

© 1998 American Chemical Society Published on Web 05/22/1998

Table 1. Band Assignment Data of Functional Groups Taking Part in CSP-Analyte Interactions

Figure 1. Schematic and molecular graphical presentation of the CSP.

Modeling of enantiomer bonding to CSP was carried out by the Desktop Molecular Modeler version 2.0 (DTMM) program. The DTMM program is a first-generation structure estimation program based on molecular mechanics calculating conformational energy minima for valence type (bond strength, bond and torsional angles) and nonvalence type (van der Waals and simple electrostatic) interactions. With this method, qualitative comparisons can be made for the possible structures of a given enantiomerCSP complex. RESULTS AND DISCUSSION FT-Raman spectroscopic detection of changes caused by weak interactions is particularly difficult when the concentration is low and the interacting groups are of polar character (interactions and reduced symmetry result in lower scattering). Therefore, particular attention shall be paid to the selection of the measurement parameters and possibly to the optimal enantiomer concentration. It was found that a 30-50% coverage of the CSP active sites is best suited for the study of enantiomer-CSP interactions. In this case, reproducibility can be assured, and spectral changes show a definite tendency. All spectra in this work refer to a coverage of 50%. Although reliable detection of spectral changes can be enhanced by selecting low interferometer velocity and resolution, in many of the cases chemometric methods are necessary to extract spectral information. Spectral details showing the most significant information on the mechanism of bonding are given in Figures 2 and 3. The assignments of the bands are summarized in Table 1. To extract spectral information, a Savitzky-Golay filtration10 was performed in the 3375-2800 and the 1700-1480 cm-1 ranges followed by a noise minimization between the original and the filtered spectra (after Iwata et al.11-14) using the modified NIPALS (nonlinear (10) (11) (12) (13)

Savitzky, A.; Golay, M. J. E. Anal. Chem. 1964, 36, 1627. Geladi, P.; Kowalsky, B. R. Anal. Chim. Acta 1986, 185, 1-17. Iwata, T.; Koshoubu, J. Appl. Spectrosc. 1994, 48, 1443-1456. Iwata, T.; Koshoubu, J. Appl. Spectrosc. 1996, 50, 747-752.

wavenumber (cm-1)

assignment

3375-3275 3200-3100 3080-3040 1685 1623

νNH νNH νCH(aromatic) νDNB-C overtones amide I (νCdO + νN-C)

1574 1565 1532 916 884 1608 1600 1584 1602 1589

amide II (βNH + νCN) amide II (βNH + νCN) ring stretch νCsN(H)sC(dO) νC(dO)sN(H) βNH + νCdO(OH) νCdO(OH)

structure bonded sN(H)sC(dO)sC*s sN(H)sC(dO)sDNB sN(H)sC(dO)sDNB sN(H)sC(dO)sC*s sDNB R1sNHsC(dO)sR2 R1sNHsC(dO)sR2 L- and D-phenylalanine L-

and D-mandelic acid

iterative partial least squares) method. Bond resolution in the νNH range was performed by the Jandel Scientific Peakfit program package (version 3.18). As a result of interactions, a broadening of the νNH bandsas compared to that of the pure CSPscan be witnessed for all the enantiomer-CSP complexes investigated. Based on the first and second derivatives of the curves, all νNH bands can be resolved to minimum four Lorentz-type components. The degrees of the fit (r2) as well as the positions of the Lorentz bands are summarized in Table 2. Comparing the data, it can be concluded that the νNH bands show a slight shift to higher frequencies as a result of complexation. The increase of the νNH frequency for secondary amide groups is explained by the formation of a Z-isomer.15 In light of this observation, it could be concluded that the Zisomerism is favored as a result of interaction. As to the mechanism of bonding, valuable information can be obtained from Figures 2 and 3. All functional groups suitable in principle to interact with the analyte show peak broadening and a changing intensity. The decrease or increase of the Raman scattering as well as the broadening of the bands are different for different chemical groups and enantiomer-CSP complexes, depending on which -NH and -CdO groups available on the stationary phase in two different chemical environments take part with a higher probability in the interaction (see Figure 1). Figures 2 and 3 show that each band shows a smaller or larger change as referenced to pure CSP. It means that the interaction cannot be confined to a single structure, i.e., several different structures can be assumed. Figure 2 also shows that the NH group of the A-type amide is preferred over the B-type one (due to steric hindrance), since the band at 1565 cm-1 shows a little change only, while the one at 1574 cm-1 undergoes a significant decrease in intensity and shows band broadening. At the same time, the bands of the enantiomers connecting to the CSP appear in the spectra super(14) Szabo´, L. P.; Lippai, EÄ . H.; Selmeczi, K.; Sze´csi, A.; Major, H. HPLC Resolution of Chiral Propionic Acid Analogues. Proceeding of the 21th International Symposium on Chromatography, Stuttgart, September 15-20, 1996; p 286. (15) Lin-Vien, D.; Colthup, N. B.; Fateley, W. G.; Graselli, J. G. The Handbook of Infrared and Raman Characteristic Frequencies of Organic Molecules; Academic Press: San Diego, CA, 1991; pp 143-175.

Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

2767

Figure 2. FT-Raman spectra of the pure CSP (1), L-phenylalanine (2), D-phenylalanine (3), L-serine (4), D-serine (5), L-mandelic acid (6), and D-mandelic acid (7) connected to CSP.

Figure 3. νCH (aromatic), νNH (bonded), and νNH (stretching) spectral ranges of the pure CSP (1) and L-phenylalanine (2), D-phenylalanine (3), L-serine (4), D-serine (5), L-mandelic acid (6), and D-mandelic acid (7) on CSP.

imposed on the falling amide I and the rising amide II bands (see the designation in the Figure 2). The enantiomer bands show no shift, although a broadening can be observed. The formation of hydrogen bonds can be sufficiently proved by the shift of the amide II band to higher frequencies accompanied with a significant band broadening and a decrease in peak intensity. Spectral changes in the rising amide I band can be explained by the activity of the amide group to form hydrogen bonds, since the N-CdO 2768 Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

conjugation results in the increase of the N-H bond length, thereby facilitating the hydrogen bonding capability of both the CdO and the NH groups. In our view, the carbonyl group of the A-type amide plays a more significant role in the formation of hydrogen bonds, in harmony with the deformation of the overtone at 1685 cm-1 as a spectral evidence. It should be noted that the connection of L- and D-serine via hydrogen bonding can only be detected in the 1700-1480 cm-1 range. No significant difference

Table 2. Regression Data, Lorentz Peak Maxima, Band Ratios, Retention Figures (tR ) tenantiomer/t L-leucine with Water as Eluent) for Systems Studieda

regression r 2 peak maxima

L-phenylalanine

+ CSP

D-phenylalanine

L-serine

D-serine

CSP alone 0.997 3315 3322 3328 3337

0.997 3319 3326 3332 3338 3.6 0.65 6.50 8.8* 9.5 9.6** 8.4

0.997 3316 3323 3329 3337 N/A 0.69 6.38 9.0* 9.6 9.8** 4.0

0.995 3318 3325 3331 3339 1.0 0.78 5.78 4.5

0.996 3320 3326 3331 3340 1.1 0.75 5.81 5.6

4.6

6.7

tR Aν884/Aν916 Aν1623/Aν1532 AνCHar/AνNH × 100

0.78 5.72 3.4

AνNHbond/AνNH × 100

1.9

a

+ CSP

+ CSP

+ CSP

L-mandelic

+ CSP

acid

0.996 3319 3325 3330 3338 4.1 0.71 6.61 7.1* 7.4 8.2** 1.3

D-mandelic

+ CSP

acid

0.992 3322 3329 3336 3345 5.7 0.67 6.80 10.2* 10.6 11.1** 3.7

*, 30% coverage; **, 70% coverage.

in bond strength could be observed between the L- and Dantipodes, which is in agreement with the retention behavior. According to the chromatographic literature, hydrogen bonds of different strengths cannot be responsible for such a difference in the retention behavior other than the one observed in this system between the L and D forms of mandelic acid. On a Pirkle type chromatographic stationary phase (i.e., with π-stacking interactions), retention is mainly due to the interaction between the electron-rich benzene ring of the enantiomer (π-base) and the electron-deficient benzene ring of the phase (π-acid).1,2,14 A spectral evidence of this supposition can be found in the νCH(aromatic) range of the spectrum (Figure 3). The νCH(aromatic) band intensities of the stationary phase were comparable with the noise level. For the phenylalanine and mandelic acid enantiomers, however, the aromatic νCH band could be detected with certainty. The intensity was in all cases an order of magnitude lower than that of the νNH band. In the 3100-3200 cm-1 range E- and Z-bonded νNH bands could be detected. Since the intensity and area of the νNH band referred to those of the νCH(aromatic) and the νNH(bonded) ones can be considered constant, the AνCH(aromatic)/ AνNH as well as the AνNH(bonded)/AνNH ratios were calculated (see Table 2). The AνCH(aromatic)/AνNH ratios were also calculated for mandelic acid and phenylalanine at 30 and 70% coverage. Since the band ratios showed an increasing tendency with increasing coverage, it is reasonable to suppose that the enantiomer-CSP connections originating from nonspecific interactions can be ignored. With respect to the aromatic νCH vibration, it could be expected that the area ratios and the frequency shifts correlate with retention, i.e., show a decreasing tendency from D-mandelic acid to L-phenylalanine. The spectra can only be interpreted on the basis of the retention data. Of course, π-stacking interactions also require certain steric conditions, but if an enantiomer is connected to the stationary phase, the aromatic νCH vibration can appear in the spectrumsdue to the increased surface concentrationssimilarly to the βNH and ring breathing vibrations that can be detected. With the formation (and strengthening) of π-stacking interactions, however, the aromatic νCH intensity should decrease and the bond frequency should shift to lower values as a result of the decreasing change in polarization. Figure 3 indicates, however, that the νCH(aromatic) band in the enantiomer-CSP complex shows significant deviations not only in intensity but also

in peak broadening. The bands obtained for L-phenylalanine and D-mandelic acid are of diffuse character showing several local maxima. The νNH(bonded) peaks also show significant broadening in harmony with the conclusion drawn from Figure 2 regarding the possible coexistence of different structures. The split of the νNH(E-bonded) bands as well as their increased intensity and area (with respect to CSP) show the dominating donor character of the NH groups. This is especially true for L-phenylalanine, where an 8-fold intensity increase could be observed in the case of the νNH(E-bonded) band. The diffuse nature of the νCH(aromatic) band supports the coexistence of several structures. It should be noted that the analyte is connected to the phase with its carboxylic group toward the B-type amide (head-to-head interaction). When intermolecular connections form toward the A-type amide, the connection is called head-to-leg interaction. In light of the spectral changes, both head-to-head and head-to-leg connections can be considered for the L-phenylalanine-CSP complex. No π-stacking can occur with the head-to-leg connection. From the side of the analyte, the carboxylic CdO and the NH2 groups, and from the side of the phase, the B-type amide CdO and the A-type amide NH groups can take part in the formation of the hydrogen bonding. With a head-to-leg interaction the NH groups of both types of amides can take part as donors. Although the conditions for the formation of a π-stacking are also met, this is less likely considering the high AνCH(aromatic)/AνNH ratio and the low shift of the νCH(aromatic) band. In the case of the D-phenylalanine-CSP complex, the donor character of the NH group of the phase is reduced and the conditions of a π-stacking are improved. The NH2 and carboxylic OH groups of the analyte can form hydrogen bonds via two points, while hydrogen bonding is possible at a single point with the carboxylic CdO and the NH group of the phase. The decrease in band frequency and a slight broadening of the νCH(aromatic) band indicates the higher probability of formation for the head-to-head connection. Both the donor and acceptor character of the CSP NH groups dominate in the L-mandelic acid-CSP complex. The head-to-leg and head-to-head interactions are of the same probability to form, as indicated by the broadening of the νCH(aromatic) band and the appearance of local maxima. In all cases the “two-point” hydrogen bond is likely to form. In the case of a head-to-head interactionsdue to steric reasonssthe involvement of the A-type amide in the formation of Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

2769

a hydrogen bonding (analyte carboxylic CdO S phase NH, analyte OH S phase CdO) results in a stronger π-stacking as compared to the situation when both A- and B-type amide groups are present together (analyte carboxylic CdO S phase B-type amide NH, analyte OH S phase A-type amide NH). The donor character of the NH groups of the phase in the D-mandelic acidCSP complex is increased as compared to that in the CSP. This means that a “one-point” hydrogen bonding andsconsidering the retention datasa strong π-stacking interaction can be supposed. Since the νCH(aromatic) band is broad, other (e.g., head-to-leg) connections can occur, although with lower probability. Despite its limitations, the DTMM molecular modeling method contributed substantially to the interpretation of spectral data. With this method a lower conformational energy was calculated for the L-enantiomer-CSP complexes independently of the nature of analytes used. This contradicts both chromatographic and spectral data. If we consider, however, that the DTMM method refers to the final state of the thermodynamic potential barrier and that only the hydrogen bonding connections could be taken into account, this contradiction is understandable. The calculations gave equal probabilities for the head-to-head and head-toleg interactions, and the two-point connections represented lower conformational energies than the three-point or one-point interactions. As a result of interactions, 9-12-membered rings can form with the involvement of the chiral carbon atoms. Although the DTMM method allows qualitative conclusions only, it can be used to confirm conclusions drawn from Raman spectroscopic data on the strength of enantiomer-CSP interactions and on the identification of groups that interact most likely with each other in the separation process. At the same time, the DTMM calculations can be considered as a preliminary basis for a more sophisticated quantum mechanical approach. Although an in situ or quasi in situ spectroscopic investigation of a chromatographic system is a challenging task, it can only be useful for the chromatographer if the spectral information obtained can directly be used for the design of separation processes. Since the intensity of the Raman scattering dependssin addition to chemical reasonsson the cross section (e.g., sample packing, focusing errors) and that the interaction is made via the amide and aromatic groups, correlations between spectral and chromatographic data can be expected with a band ratio analysis only. For this reason, the ratios of the νC(dO)-N(H) (884 cm-1) and the νCsN(H)sC(dO) (916 cm-1) bands were calculated and listed in Table 2. Although the A884 cm-1/A916 cm-1 ratio can predict the elution

2770 Analytical Chemistry, Vol. 70, No. 13, July 1, 1998

sequence of a given enantiomer pair, it does not correlate with relative retention. The reasons for the deviation can include a drifting baseline as a result of fluorescence, the low band intensities, and the significant band broadening. Therefore, a relatively high-intensity band was selected for the analysis whichsalthough changes with the strength of interactionsdoes not influence significantly the integration limits as a result of band broadening. Comparing the area ratios of the amide I (1623 cm-1) and ring stretching (1532 cm-1) bands to the retention times (tR), it can be seen that surprisingly good predictions can be made for the relative retentions as well. CONCLUSION FT-Raman spectroscopy can successfully be used to determine the type(s) of connections and the nature of groups that most likely take part in the interactions when an enantiomer-CSP complex is formed. Some questions (e.g., the number of possible structures at very low coverage) still have to be answered. The most important question is, of course, what is the situation in aqueous or buffered systems? It seems, however, that even these first steps can convincingly justify the use of FT-Raman spectrometry in the study of chromatographic systems. Changes associated with the enantiomer-CSP complex can be followed in the Raman spectra allowing the analyst to draw conclusions on the nature and strength of interactions. It was proved experimentally that hydrogen bonds and π-stacking interactions are responsible for retention. There are band ratios that show satisfactory correlations with chromatographic retention data. Although Raman spectrometry cannot answer all the problems of interest, it can be used as a reliable tool to select a stationary phase best suited for a given system and can contribute substantially to a better understanding of the nature of chiral separations. ACKNOWLEDGMENT This work was supported by the Hungarian Ministry of Culture and Education under Grant FPF-4004/1997. The Centre for Instrumental and Developmental Chemistry of the Queensland University of Technology is thanked for the use of the FT-Raman spectrometer. The authors thank L. Bencze and R. Szila´gyi for providing the DTMM parameter list. Received for review November 12, 1997. Accepted April 1, 1998. AC9712411