Absolute Configuration Modulation Attenuated Total Reflection IR

Aug 7, 2004 - Figure 2. Principle of absolute configuration modulation spectroscopy and PSD. The principle of demodulation is shown for a signal at 17...
0 downloads 0 Views 558KB Size
Anal. Chem. 2004, 76, 5319-5330

Absolute Configuration Modulation Attenuated Total Reflection IR Spectroscopy: An in Situ Method for Probing Chiral Recognition in Liquid Chromatography Ronny Wirz,† Thomas Bu 1 rgi,*,†,‡ Wolfgang Lindner,§ and Alfons Baiker†

Institute for Chemical and Bioengineering, ETH-Ho¨nggerberg, CH-8093 Zu¨rich, Switzerland, and Institute of Analytical Chemistry, University of Vienna, Vienna, Austria

A method to selectively probe the different adsorption of enantiomers at chiral solid-liquid interfaces is applied, which combines attenuated total reflection infrared spectroscopy and modulation spectroscopy. The spectral changes on the surface are followed while the absolute configuration of the adsorbate is changed periodically. Demodulated spectra are calculated by performing a subsequent digital phase-sensitive data analysis. The method is sensitive solely to the difference of the interaction of the two enantiomers with the chiral surface, and the small spectral changes are amplified by the phasesensitive data analysis. Its potential is demonstrated by investigating an already well-studied system in liquid chromatography, namely, the enantiomer separation of N-3,5-dinitrobenzoyl-(R,S)-leucine (DNB-(R,S)-Leu) using tert-butylcarbamoyl quinine (tBuCQN) as the chiral selector immobilized on the surface of porous silica particles. The performed experiments and density functional theory calculations confirm an interaction model that was proposed earlier based on solution NMR and XRD in the solid state. It emerges that the ionic interaction is the strongest one, but the main reason for the potential for enantioseparation of the chiral stationary phase (CSP) is the distinct formation of a hydrogen bond of the (S)enantiomer with the chiral selector. This H-bond is established between the amide N-H of DNB-(S)-Leu with the carbamate CdO of the CSP. The (R)-enantiomer instead shows no specific hydrogen bonds. Only the unspecific ionic bonding between the protonated quinine part of the tBuCQN and the carboxylate of the DNB-(R)Leu (holds also for DNB-(S)-Leu) is observed. The increasing demand for enantiomerically pure compounds stimulated research in the field of asymmetric synthesis and also advanced the knowledge of enantiomer separation strategies. Methods to separate a racemic mixture are mainly found in the * To whom correspondence should be addressed. Tel.: ++41-32 718 24 12. Fax: ++41-32 718 25 11. E-mail: [email protected]. † Institute for Chemical and Bioengineering. ‡ Current address: Department of Chemistry, University of Neuchaˆtel, Switzerland. § University of Vienna. 10.1021/ac049428x CCC: $27.50 Published on Web 08/07/2004

© 2004 American Chemical Society

field of chromatography (GC, LC, CEC), where the chiral stationary phase (CSP) consists of a chiral selector usually immobilized onto a support material like silica or organic polymers. This chiral selector moiety is capable of preferentially binding one enantiomer. Consequently, the weaker bound enantiomer is eluted faster than the stronger bound one resulting in enantiomer separation. The rational design of new chiral selectors that are capable of efficiently separating one enantiomer from the other requires understanding of the specific enantiodiscriminating interactions. One efficient way to investigate such distinct interactions is to use nuclear magnetic resonance spectroscopy (NMR) and to analyze the spectra of the dissolved diastereomeric complex formed between the selector (SO) and either of the two enantiomers of the selectand (SA).1 By analyzing the intramolecular and intermolecular NOEs, one can get information about the resulting conformations of such SO-SA complexes and also as a function of the solvent used. A similar situation is given in liquid chromatography where the immobilized selector moiety is in a solvated state, but due to the geometric aspects related to the surface and immobilization chemistry, the observable overall intermolecular SO-SA interaction events may not necessarily be exactly the same as in free solution. Therefore, techniques are desired that can give such surface-specific information as, for example, by an approach published recently using HR/MAS 2D transfer NOESY.2 Another approach situated closer to the chromatographic conditions is in situ attenuated total reflection infrared (ATR-IR)3 spectroscopy in combination with phase-sensitive detection (PSD).4 Here we apply this technique to an already well-studied system in chromatography,1,5-7 namely, the enantiomer separation of N-3,5-dinitrobenzoyl-(R)-leucine (DNB-(R)-Leu) and DNB-(S)-Leu (3, se(1) Maier, N. M.; Schefzick, S.; Lombardo, G. M.; Feliz, M.; Rissanen, K.; Lindner, W.; Lipkowitz, K. B. J. Am. Chem. Soc. 2002, 124, 8611-8629. (2) Hellriegel, C.; Skogsberg, U.; Albert, K.; La¨mmerhofer, M.; Maier, N. M.; Lindner, W. J. Am. Chem. Soc. 2004, 126, 3809-3816. (3) Harrick, N. J. Internal reflection spectroscopy; Intersience: New York, 1967. (4) Fringeli, U. P.; Baurecht, D.; Siam, M.; Reiter, G.; Schwarzott, M.; Bu ¨ rgi, T.; Bru ¨ esch, P. Handbook of Thin Film Materials; Academic Press: New York, 2001; Vol. 2, p 191. (5) Maier, N. M.; Nicoletti, L.; Lammerhofer, M.; Lindner, W. Chirality 1999, 11, 522-528. (6) Lesnik, J.; Lammerhofer, M.; Lindner, W. Anal. Chim. Acta 1999, 401, 3-10. (7) Lah, J.; Maier, N. M.; Lindner, W.; Vesnaver, G. J. Phys. Chem. B 2001, 105, 1670-1678.

Analytical Chemistry, Vol. 76, No. 18, September 15, 2004 5319

Chart 1. Used Chiral Selectands (SAs) and Chiral Selector (SO) Units Immobilized on a Mercaptopropyl Silica Surface

lectand, SA) using tert-butylcarbamoyl quinine (tBuCQN) (2) as the chiral selector (SO), which is immobilized on porous silica particles (see Chart 1). The in situ ATR-IR technique is applied in conjunction with absolute configuration modulation spectroscopy, which allows us to uncover the different interaction of enantiomeric probe molecules with chiral centers on a surface. For this purpose, the stationary chiral phase (selector) is exposed to periodic changes (modulation) of the concentration of the two enantiomers of the probe molecule (selectand). The total concentration of selectand (cTot ) c(R) + c(S)) in solution stays constant during the experiment, and only signals derived from the change of the absolute configuration of the selectand bound to the surface are detected. The aim of this study is to link and verify the observed effects in chromatography and the performed ex situ investigations (FTIR6 and NMR1,2) by absolute configuration modulation (ACM) ATR-IR spectroscopy4,8 as an in situ method. Absolute Configuration Modulation Infrared Spectroscopy. Modulation spectroscopy is a sensitive technique for the investigation of reversible systems.4,9-11 The method is based on the disturbance (stimulation) of the system under investigation by a periodic alteration of an external parameter, such as, for example, temperature,9 pressure, electric field, or concentration.11 The system will then respond periodically at the frequency of the stimulation. Here we use ATR infrared spectroscopy to follow the periodic changes of the system under investigation by collecting time-resolved infrared spectra. The experimental setup is shown schematically in Figure 1. The CSP is fixed on an internal reflection element (IRE), which itself is mounted in a flow-through cell. The solvent and solutions of the (R)- and (S)-enantiomer, (8) Wirz, R.; Burgi, T.; Baiker, A. Langmuir 2003, 19, 785-792. (9) Muller, M.; Buchet, R.; Fringeli, U. P. J. Phys. Chem. 1996, 100, 1081010825. (10) Baurecht, D.; Fringeli, U. P. Rev. Sci. Instrum. 2001, 72, 3782-3792. (11) Burgi, T.; Baiker, A. J. Phys. Chem. B 2002, 106, 10649-10658.

5320 Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

respectively, are stored in separate glass bubble tanks. For a modulation experiment, two out of the three fluids are alternately pumped through the cell. Three different such experiments are possible depending on which two fluids are chosen: modulation of (R)-enantiomer against solvent, modulation of (S)-enantiomer against solvent, and modulation of (R)-enantiomer against (S)enantiomer. In the former two experiments, the concentration of either one of the enantiomers of the probe molecule is changing with time (concentration modulation). In the latter, the absolute configuration of the probe molecule is changing with time (absolute configuration modulation). During the modulation

Figure 1. Experimental setup. Three bubble tanks containing the neat solvent (acetic acid in acetonitrile), dissolved DNB-(R)-Leu, and dissolved DNB-(S)-Leu are connected via two steel valves to the flowthrough cell in a way that two of these solutions can be alternately flown over the sample (“modulated”) by switching the two computercontrolled pneumatically actuated Teflon valves. During the “modulation”, ATR-IR spectra are measured.

Figure 2. Principle of absolute configuration modulation spectroscopy and PSD. The principle of demodulation is shown for a signal at 1717 cm-1 for two different demodulation phase angles (270° and 360°). For demodulation, a sine function with the same frequency (1/T) and a phase angle, which can be chosen by the operator, is multiplied with the measured signal and integrated over the whole period according to eq 1. The result, the demodulated signal, is shown on the right side. If the relative phase angle ΦPSD rel is 0°, the absorbance in the demodulated spectrum is equal to the amplitude of the measured signal. This is called the in-phase spectrum. By changing the relative phase angle ΦPSD rel from 0° to 90°, the absorbance in the demodulated spectrum reaches 0 and is called the “out of phase” spectrum.

experiments, ATR spectra are recorded (time-resolved spectra). By a subsequent digital phase-sensitive data analysis according to eq 1 phase-resolved (demodulated) spectra are obtained from PSD

Aφk k (v˜) )

2 T

∫ A(v˜,t) sin(kωt + φ T

0

PSD k )

dt

(1)

the set of time-resolved spectra. Here k ) 1, 2, 3,... determines the demodulation frequency, i.e., fundamental, first harmonic, second harmonic, etc., T is the duration of the modulation period, φPSD is the demodulation phase angle, ω is the modulation frequency, and A(v˜,t) is the timedependent absorbance at wavenumber ˜v (response signal). Figure 2 illustrates the connection between a time-resolved signal at a certain wavenumber A(1717 cm-1,t) and the corresponding phaseresolved signal A270° and A360° (1717 cm-1) at the same wavenumber for two different demodulation phase angles (270° and 360°). In one case (Figure 2 top), the phase angle is chosen such that the sine function and the absorbance signal are in phase. This is

the phase angle, which results in the maximum signal at this wavenumber in the demodulated spectra. In the other case shown (Figure 2 bottom), the sine function and the absorbance signal are out of phase. After multiplication of the two functions and integration according to eq 1, the resulting demodulated signal is zero. In case the system response is fast compared to the stimulation, the resulting demodulated spectra represent difference spectra between two states of the system. In case the system response is in the order of the stimulation period, phase lags between stimulation and response can be measured, which are connected to the kinetics of the system. Important features of the method, which are crucial for the following application, are as follows: (i) Only periodically changing signals show up in the demodulated spectra, whereas static signals cancel (integrating a sine function over one period equals zero!). (ii) The application of phase-sensitive data analysis results in high-quality spectra, i.e., large signal-to-noise ratio, compared to conventional difference spectra. This is due to the fact that noise at a frequency different Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

5321

from ω is efficiently filtered. We take advantage of these possibilities to investigate the different adsorption/desorption behavior of enantiomers at a chiral solid-liquid interface. The external parameter that is periodically altered is the absolute configuration of the chiral guest molecules (SAs) (absolute configuration modulation).8 Note that the concentration profile is not necessarily sinusoidal. In our case, it resembles more a square wave pattern. This does not, however, change the shape of the resulting spectra. Signals arising from dissolved species are filtered out because the spectra of enantiomers are identical and do not change during modulation, assuming equal concentration (see (i) above). Also, nonspecific interactions of the enantiomers with the surface are filtered out, for the same reason. Only signals are detected, which result directly from the different diastereomeric associates with the chiral selector generated via the alternating interactions with the (R)and (S)-enantiomers of the selectand. Hence, the received information is solely due to the changing absolute configuration of the adsorbate interacting with the chirally modified surface. The expected differences may be small, but these signals are enhanced by the phase-sensitive data analysis resulting in improved signal-to-noise ratio (see (ii) above). EXPERIMENTAL SECTION Experimental Setup and Data Acquisition. Infrared spectra were measured on a Bruker IFS 66/S FT-IR spectrometer equipped with a dedicated ATR-IR attachment (Optispec) and a liquid nitrogen-cooled MCT detector. All spectra were recorded at a resolution of 4 cm-1. The silica-based CSP-covered ZnSe prism was fixed within an in-house-built stainless steel flow cell. A scheme of the experimental setup is depicted in Figure 1.11 The design of the cell allowed concentration modulation direct on the surface of the IRE.12 The gap between the polished steel surface of the cell and the IRE was 250 µm and defined by a 30 × 1 mm viton O-ring (Johannsen AG) fit into a precision electroeroded nut of the steel cell. The total volume of the cell is 0.077 mL. The flow-through cell was cooled by means of a thermostat, and the measurements were performed at 25 °C. The flow of liquid over the sample was controlled by means of a peristaltic pump (Ismatec Reglo 100) located behind the cell. Liquid was provided from three separate glass bubble tanks, where dry nitrogen gas (Pangas) was bubbled through the solutions to remove air/moisture. Two steel valves were used to control the flow of the two solutions, which are modulated against each other, for example, (R)-enantiomer against solvent; see Figure 1. The flow of the two solutions was determined by two computer-controlled pneumatically actuated three-way Teflon valves (Parker PV-1-2324). Steel tubing was used throughout. For transmission experiments, a normal sample holder replaced the ATR-IR attachment. A variable-path length IR cell (Specac) with KBr windows was used with a path length adjusted to ∼100 µm. Materials and Chiral Film preparation. As chiral selector moiety served tBuCQN and cinchonidine (CD, Chart 1). These compounds were immobilized on porous spherical silica particles (5 µm) of a surface area of ∼300 m2/g resulting in a selector (12) Urakawa, A.; Wirz, R.; Burgi, T.; Baiker, A. J. Phys. Chem. B 2003, 107, 13061-13068.

5322

Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

loading of 340 µmol/g. Such materials were synthesized according to a standard procedure described elsewhere.5 DNB-(S)-Leu was purchased from Sigma, and DNB-(R)-Leu was synthesized according to a standard protocol.13 Although ATR-IR spectroscopy is an excellent tool to investigate the interactions at solid-liquid interfaces, it is often difficult to establish a stable solid film on the IRE, especially if one is using a flow-through cell and silica-based particles. We applied a method, proposed recently,14 that makes use of polyethylene (PE) to fix silica particles on an IRE. It was claimed later that this method did not work when extended to polymer/surfactant or DNA/ protein systems, because the larger molecules could not penetrate into the PE/silica powder matrix.15 Nevertheless, it worked perfectly for our purposes. Compared to the original work,14 we decreased the ratio of PE to silica, or CSP in our case, to a limit, so that the amount of PE is enough to fix the silica-based particles on the IRE but does not cover and block the surface for interactions with dissolved selectand species. The PE did not change the chemical behavior of the used CSP as the IR investigation shows. For film preparation, 1.5 mL of toluene was heated to 100 °C and ∼0.6 mg of PE (Fluka) was added. After stirring for ∼20 min, ∼14 mg of CSP (2 or 3) was added. After another 2 min of intense stirring, 0.7 mL of this suspension was spread on a ZnSe IRE (Komlas, 50 × 20 × 2 mm) to get a flow-stable chiral film. Toluene was allowed to evaporate. Excess chiral film was removed from the edges of the IRE, and the latter was put under vacuum to get rid of residues of toluene. Assuming refractive indices of 1.4 for the wet film and 2.4 for ZnSe, a penetration depth dp of 1.7 µm (0.6 µm) at 1000 cm-1 (3000 cm-1) was calculated. Modulation Experiments. Immobilized tBuCQN (2) and CD (3) were used as chiral selectors (SO). DNB-(R)-Leu and DNB(S)-Leu (1) were the selectands (1 mM each). Acetic acid (Fluka, puriss p.a.) in acetonitrile (Fluka, puriss p.a.) was used as solvent (210 mM). The (R)- and (S)-selectands form quite strong diastereomeric complexes with SO due to their strong ionic interaction paired with additional hydrogen-bonding and π-π-type intermolecular interactions. The acetic acid in the solvent stimulates the exchange/removal of the adsorbed DNB-(R)-Leu and DNB(S)-Leu (1), making possible to detach most of the adsorbate within half of a modulation period. As will be seen later, the acetic acid did not change the interaction of the SAs with the selector forming the diastereomeric complex. In a first try, trifluoroacetic acid (TFA) was used instead of acetic acid, but it turned out that this acid is too strong and thus blocks the cationic site of the SO, preventing adsorption of the weaker acidic SAs. A series of experiments was typically performed in the following way: Neat acetonitrile was flown first over the CSP until no variation in the spectrum could be detected. Then solvent (acetic acid in acetonitrile, 210 mM) was flown over the CSP until no variation in the spectrum could be detected. At this point, the reference for all further spectra was measured. The valves were set so that (R)-1 and solvent could reach the flow-through cell (13) Veigl, E.; Bohs, B.; Mandl, A.; Krametter, D.; Lindner, W. J. Chromatogr., A 1995, 694, 151-161. (14) Ninness, B. J.; Bousfield, D. W.; Tripp, C. P. Appl. Spectrosc. 2001, 55, 655-662. (15) Jiang, C.; Li, H.; Tripp, C. P. Appl. Spectrosc. 2003, 57, 1419-1424.

Figure 3. Time-dependent absorbance of the νs(NO2) during one period. Each of the enantiomers was modulated against the solvent. One enantiomer and neat solvent were periodically admitted to the flow-through cell. where an IRE containing the SO was fixed. The absorbance was averaged during three full modulation periods. The absorbance of the νs(NO2) at 1346 cm-1 (1348 cm-1) coming from the DNB-(R)-Leu and DNB-(S)-Leu, respectively, is shown during the modulation period of T ) 299 s. SO was in one case tBuCQN and in the other CD. (a) DNB-(R)-Leu was modulated vs solvent, first half-period stimulation. (b) Solvent was modulated vs DNB-(S)-Leu, second half-period stimulation. (c) Solvent was modulated vs DNB-(R)-Leu, second half-period stimulation. (d) DNB-(S)-Leu was modulated vs solvent, first half-period stimulation. It can be seen that the weaker bound DNB-(R)-Leu can be released within about 45-50 s from the immobilized tBuCQN on silica, whereas it took about 150 s to release the DNB-(S)-Leu. Both enantiomers could be desorbed within 50 s of the immobilized CD on silica (solvent: acetic acid in acetonitrile, 210 mM and a flow rate of 0.9 mL min-1).

(Figure 1). Then a first modulation experiment was started (Mod1). This type of experiment provided information about the adsorption and desorption of DNB-(R)-Leu on the CSP (tBuCQN or CD immobilized on silica). For the modulation experiments, a flow rate of 0.9 mL/min was used. Liquid from the two bubble tanks was alternately admitted to the cell by switching the computer-controlled Teflon valves within the data acquisition loop of the measurement program. The system was allowed to reach a quasi-stationary state during three full modulation periods. Data were then averaged over another three modulation periods. Within one modulation period T of 299 s (corresponds to a frequency of 3.3 × 10-3 Hz), 60 spectra were recorded by coadding 40 scans/ spectrum. After this modulation experiment, the steel valves were set in a way that (S)-1 and solvent could reach the flow-through cell and solvent was flown over the surface until no signals from the (R)-1 were detected. Another modulation experiment was performed in the same way as described above but now by modulating between (S)-1 and solvent (Mod3). After flowing solvent again until no signals of (S)-1 could be detected, the steel valves were set so that (S)-1 and (R)-1 could reach the flowthrough cell. A similar experiment was performed by modulating between (S)-1 and (R)-1 (Mod5). As described in detail above, this type of experiment selectively highlights differences in the interaction of the two enantiomers with the CSP. The modulation of (R)-1 versus (S)-1 was repeated by coadding 70 scans/ spectrum (Mod7) resulting in a modulation period T of 523 s (corresponds to a frequency of 1.9 × 10-3 Hz). Demodulation was performed according to eq 1 and Figure 2. Only spectra demodulated at the fundamental (k ) 1) are reported here. Before demodulation, the single-beam spectra were transformed into absorbance spectra using the average of all single-beam spectra

as the background. It should, however, be noted that the result, i.e., the demodulated spectra, are independent of the choice of the reference spectrum. As mentioned already, these series of experiment were performed once with tBuCQN (2) and once with CD (3) as the SO. In the following the solution, which was admitted first during modulation, will be mentioned first. RESULTS ATR-FT-IR Experiments of DNB-(R)-Leu and DNB-(S)Leu Adsorption on Immobilized tBuCQN. The result of a modulation experiment is a three-dimensional array of infrared absorbance as a function of wavenumber (800-4000 cm-1) and time (0-T). Representative, time-resolved absorbances of the symmetric NO2 stretching vibration of N-3,5-dinitrobenzoylleucine at 1346 (for tBuCQN CSP) and at 1348 cm-1 (for CD CSP) are shown in Figure 3. Each enantiomer was modulated against the solvent, respectively. One can easily see in Figure 3a + b that the weaker bound (R)-1 desorbs much faster (∼45-50 s) than (S)-1 (∼150 s). In the case where CD was used as CSP, it took ∼50 s for both enantiomers to desorb (Figure 3c + d). In Figure 4 the time-resolved absorbances of the symmetric NO2 stretching and the amide II at δ(S)-complex(N-H) ) 1585 cm-1 over the whole modulation period is depicted where the (R)-1 was modulated against the (S)-1 on immobilized tBuCQN. The modulation period was increased from 299 s (Figure 4a and b) as it was the case in Figure 3, to 523 s (Figure 4c and d; Mod7) to ensure the system could reach equilibrium during half a period. Figure 5 shows the demodulated spectra, calculated from the time-resolved data, where (R)-1 was modulated against the solvent (Figure 5a) and where (S)-1 was modulated against the solvent Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

5323

Figure 4. Time-dependent absorbance of the νs(NO2) and the amide II vibration during one period. The enantiomers were modulated against each other The (R)-enantiomer and the (S)-enantiomer were periodically admitted to the flow-through cell, where an IRE containing the SO was fixed. The absorbance was averaged during three full modulation periods. The signals of the νs(NO2) at 1344 cm-1 and the amide II vibration at 1585 cm-1 are shown for a modulation period when (R)-1 was modulated against (S)-1. The modulation period was in one case 299 s (a, b) and in the other 523 s (c, d). The total concentration of selectands (SAs) in solution remains constant during the experiment. SO was tBuCQN.

(Figure 5b). Immobilized tBuCQN served as the CSP. The depicted spectra were calculated with five different demodulation phase angles (steps of 10°) ending with the so-called “in-phase” spectra for the 1346-cm-1 band (boldface), which shows the maximum amplitude at this certain phase angle. There were no phase shifts observed between the several IR bands either in the spectra of Figure 5a or in Figure 5b. All bands have their maximum absorbance at the same phase angle (ΦPSD ) 340° and 140°, respectively). However, the relative phase angle between the two enantiomers in the two experiments has a value different from zero, namely, 20° (20° ) 340° - 180° - 140°). Here 180 deg has to be subtracted from the ΦPSD Mod1 because it was a first half-period stimulation in the case of Figure 5a, whereas a second half-period stimulation was applied in the case of Figure 5b (see also Figure 3a + b). The 20° corresponds to a time difference of 16.6 s. This value can be seen as a sort of difference in retention time. The major differences in adsorption of either DNB-(S)-Leu or DNB-(R)-Leu on immobilized tBuCQN can be derived by comparing the in-phase spectra shown in Figure 5a + b (bold line). The bands that significantly differ only in intensity coming from the complexed DNB-(R,S)-Leu are the symmetrical stretching νs(NO2) at 1346 cm-1 (1345 cm-1) and the amide I band at 1670 cm-1 (1668 cm-1). The frequency of these vibrations is not significantly changing during adsorption of the SAs to the SO. The reason for the different absolute intensity is likely the higher amount of adsorbed (S)-1 compared to (R)-1. The band that shifts quite remarkably from ∼1725 cm-1 up to 1745 cm-1 in the case of adsorption of (R)-1 is the amide I of the SO. This shift was also observed after flowing dissolved acetic acid or TFA over the surface, taking the spectrum recorded in neat acetonitrile as the reference. This indicates the loss of an H-bond or the formation of a weaker H-bond of the SO’s CdO when an acid is present. In the case of adsorption of (S)-1, more band shifts could be observed. As shown in Figure 5b, the amide III band of the SO 5324 Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

shifts from 1266 cm-1 up to 1282 cm-1, whereas the amide I band of the SO was observed to shift down from 1735 to 1715 cm-1. Further, we observe a weak band shift from 852 cm-1 up to 869 cm-1, which seems to be significant but could not be assigned. Also the asymmetric stretching νas(NO2) is slightly downshifted to 1541 cm-1 in the case of (S)-1 adsorption compared to (R)-1 adsorption (1545 cm-1). Two additional bands were detected with the adsorption of (S)1, which were missing in the case of (R)-1. These were the relatively broad ν(S)-1(NH) stretching band at 3341 cm-1, which is the result of a band shift from 3350 cm-1, and the formation of another broad band at 1585 cm-1. We assign this band to the amide II δ(S)-complex(N-H) of the SA within the complex. It is shifted up from the region 1500-1550 to 1585 cm-1. The two enantiomers were also modulated against each other to mimic mutual displacement important in chromatographic separation. Experiments were performed once with the same modulation period as in Figure 5 (T ) 299 s) and once with a longer period of 523 s. The demodulated spectra are depicted in Figure 6a + b, respectively. They look quite similar to the spectra received from the modulation of (S)-1 versus solvent (Figure 5b). To check the consistency of the modulation technique and the derived spectra, the in-phase spectra of the experiment (R)-1 versus (S)-1 (Mod5, Figure 6a) was compared with the difference of the in-phase spectra of the experiment solvent versus (S)-1 (Mod3, Figure 5b) minus the in-phase spectra of experiment (R)-1 versus solvent (Mod1, Figure 5a). The modulation period for these experiments was 299 s. The result is shown in Figure 7c and d. The two spectra match each other in an excellent way. Even the smallest IR bands could be reproduced! Therefore, we can conclude that the performed experiments are consistent. Transmission FT-IR Experiments of DNB-(R)-Leu and DNB-(S)-Leu and tBuCQN. Several FT-IR measurements in transmission mode were done to compare the spectra received at the solid-liquid interface with spectra of species in solution.

Figure 5. Demodulated spectra of DNB-(R)-Leu vs solvent (a) and solvent vs DNB-(S)-Leu (b) on tBuCQN immobilized on silica. The six demodulated spectra were calculated according to eq 1 using six different phase angles (∆Φ ) 10°) ending with the in-phase spectrum (same experiment as in Figure 3a and b). Only the signals that change periodically at the same frequency as the stimulation show up in these spectra. The boldface spectra are the so-called in-phase spectra where ΦPSD ) 340° for the (R)-1 and at ΦPSD ) 140° for the (S)-1.

Thus, spectra of 5 and 50 mM tBuCQN and DNB-(S)-Leu in acetonitrile, respectively, were collected. Neither changes in the spectra of (S)-1 nor of tBuCQN were observed by increasing the concentration from 5 to 50 mM. Also, the spectra of the two diastereomeric complexes were measured. The concentration of DNB-(R)-Leu, DNB-(S)-Leu, and free (not silica grafted) tBuCQN was 10 mM each. The corresponding spectra are shown in Figures 8 and 9 patterns c and d. The spectra of the (S)-complex in solution look quite different from those of the (R)-complex. This becomes even more apparent when the difference spectrum of the (S)complex minus the (R)-complex is compared with the in-phase spectrum of the experiment (R)-1 versus (S)-1 (Mod7, T ) 523 s, ΦPSD ) 150°) as depicted in Figure 9. They look quite similar although the quality of the in-phase spectra is much higher. The similarities are in detail bands at 1344, 1540, and 1666 cm-1 and shifts of the corresponding bands from 849 cm-1 up to 867 cm-1, from 1265 cm-1 up to 1280 cm-1, from 1735 cm-1 down to 1716 cm-1, and from 3368 cm-1 down to 3354 cm-1. This indicates a similar conformation of the corresponding complexes on the surface and in the dissolved state.

Figure 6. Demodulated spectra of DNB-(R)-Leu vs DNB-(S)-Leu on tBuCQN immobilized on silica. The six demodulated spectra were calculated based on the collected data shown in Figure 4 according to eq 1 using six different phase angles (∆Φ ) 10°) ending with the in-phase spectrum (same experiment as in Figure 4). Both enantiomers were modulated against each other. All appearing signals are a consequence of the changing absolute configuration of the selectand. The modulation period was increased from T ) 299 (a) to 523 s (b). The boldface spectra are the so-called in-phase spectra, which are at ΦPSD ) 130° and 150°.

ATR-FT-IR Experiments of DNB-(R)-Leu and DNB-(S)Leu on Immobilized CD. The same experiments, which were done with tBuCQN (2) as the SO, were also performed with CD (3). Therefore, a new chiral surface was prepared and the modulation experiments were repeated. Because cinchonidine has no carbamate group, comparison with the tBuCQN experiments should help to assign the amide bands to the DNB-(R)-Leu and DNB-(S)-Leu, respectively, or to the tBuCQN moiety of the complex. Furthermore, comparison between the two stationary phases should clarify the role of the carbamate branch for enantiomer differentiation. The time-dependent absorbances at 1348 cm-1, where each of the two enantiomers were modulated against neat solvent, are depicted in Figure 3c and d. One can see that the adsorption and desorption behavior of the two enantiomers is the same. It takes about 45-50 s for either of the two adsorbates to desorb again when solvent is flown over the surface. Therefore, the enantioAnalytical Chemistry, Vol. 76, No. 18, September 15, 2004

5325

Figure 7. Consistency check of the modulation technique. The inphase spectra (a, b) of Figure 5b + a are shown and the difference was calculated (d). This difference spectrum was compared with the in-phase spectrum (c) of Figure 6a. As can be seen easily, they match each other in an excellent way. Even the smallest bands are reproducible!

differentiating power of 3 is expected to be much lower than that of 2. This is verified by comparing the corresponding demodulated spectra shown in Figure 10a and b, which look the same for both enantiomers. Furthermore, no bands were detected in the demodulated spectra where the two enantiomers were modulated against each other (not shown). Also, chromatographically, the CD-type CSP expresses almost no enantioselective discrimination between (R)-1 and (S)-1, corroborating the validity of the presented ATR-IR technique or vice versa. Density Functional Theory (DFT) Calculations. To confirm the experimentally obtained ATR FT-IR spectra, DFT calculations were accomplished. A normal-mode analysis of DNB-(S)-Leu, tBuCQN, and the corresponding complex was performed. The conformation of the complex was based on the X-ray crystal structure proposed elsewhere.1 (S)-1 was supposed to exist as a monomer with a hydrogen bond between the acid proton and the CdO of the amide. The conformation of the SO was chosen as in the case of the complex. Note that such calculations neglect the effect of the solid surface. Also, calculating a limited number of distinct structures for such a complex system cannot give a comprehensive picture. Nevertheless, the calculations could reproduce the experimentally observed vibrational bands, especially of those functional groups that were mainly involved in these strong enantiodiscriminating interactions. Before the normal mode analysis, the structures were fully optimized to minimum energy. We used Gaussian0316 with the BLYP DFT method and the 6-31G(d,p) basis set. The conformations of the species are shown in (16) Frisch, M. J.; Schlegel, G. W. T., H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A., Jr.; Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; Daniels, A. D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; Barone, V.; Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.; Adamo, C.; Clifford, S.; Ochterski, J.; Petersson, G. A.; Ayala, P. Y.; Cui, Q.; Morokuma, K.; Rega, N.; Salvador, P.; Dannenberg, J. J.; Malick, D. K.; Rabuck, A. D.; Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.; Baboul, A. G.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, M. W.; Andres, J. L.; Gonzalez, C.; Head-Gordon, M.; Replogle, E. S.; Pople, J. A. Gaussian 03, Revision A.11.4 Gaussian, Inc., Pittsburgh, PA, 2002.

5326 Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

Figure 8. Transmission spectra of the SO, the SA, and the two formed diastereomeric complexes in solution. Concentrations of the SO and the SAs forming the complexes were 10 mM each. The concentration of single SO and single SA was 50 mM; the corresponding spectra were multiplied by a factor of 0.2. The thickness of the used transmission IR cell was ∼100 µm. Differences in the spectra of the two complexes appear at ∼1716-1735, 1666, and ∼12651280 cm-1. The spectra of single SO and single SA at concentrations of 50 and 5 mM (not shown) exhibit no significant differences. Therefore, a change in monomer/dimer aggregation in this concentration range and a corresponding band shift in the ATR-IR spectra is not expected. Pure acetonitrile was used as the solvent. Some regions in the spectra are covered due to strong solvent absorption (/).

Figure 11. Subtracting the calculated energy of the free species from the one of the complex, a total binding energy ∆E of -36.7 kJ mol-1 was calculated. The resulting IR spectra are shown in Figure 12. None of the normal modes had imaginary frequency. The calculated frequencies are scaled by a factor of 1.04 so that the calculated spectra are easier compared with the experimental ones. The factor of 1.04 agrees with the proposed factor of 1.0352 for BLYP below 1800 cm-1.17 Although the frequency of some bands is still overestimated (specially the H-bonding modes), the shape and the band shifts of the spectra reproduce the experimental observations quite well. One should consider that calcu(17) Halls, M. D.; Velkovski, J.; Schlegel, H. B. Theor. Chem. Acc. 2001, 105, 413-421.

Figure 9. Comparison of the difference spectra of the two complexes and the demodulated spectrum. The difference spectrum (b) of the two formed complexes (c) DNB-(R)-Leu/tBuCQN and (d) DNB(S)-Leu/tBuCQN was compared to (a) the in-phase spectrum of the modulation experiment shown in Figure 6b. Similarities were observed at 1265-1280, 1666, 1716-1735, and ∼3350 cm-1. This indicates a similar interaction of the two enantiomers with the dissolved tBuCQN and the immobilized tBuCQN. Further, a change in the interaction mode of the enantiomers with the tBuCQN due to acetic acid in the solution (in the modulation experiments) can be excluded. The acetic acid only helps to release the adsorbed enantiomer faster from the SO. Some regions in the spectra are covered due to strong solvent absorption (/).

lated spectra hold for single molecules in a vacuum, whereas in the experiment, a much more complex surrounding prevails. The main band shifts, when (S)-1 is interacting with tBuCQN, are highlighted by arrows. The acid bending mode δCOOH(SA) shifts from 1451 to 1579 cm-1. The amide bending δN-H, Amide II(SA) shifts up from 1544 to 1568 cm-1. As in the experiment, the asymmetric stretching of the nitro group νNO2,asym.(SA) is downshifted from 1584 to 1572 cm-1. The CdO vibrations also show behavior similar to that seen in the experiments, namely, a upshift of νCdO,amide I(SA) from 1690 to 1713 cm-1, a downshift of νCdO,acid(SA) from 1848 to 1742 cm-1, and a downshift of νCdO,amide I(SO) from 1799 to 1776 cm-1. The νN-H(SA) is downshifted from 3657 to 3580 cm-1 due to hydrogen bonding.

Figure 10. Demodulated spectra of solvent vs DNB-(R)-Leu and DNB-(S)-Leu vs solvent on CD immobilized on silica. The experimental conditions were the same as where tBuCQN was the SO. The boldface patterns are the so-called in-phase spectra, which are at ΦPSD ) 160° for (R)-1 and at ΦPSD ) 340° for (S)-1. Both spectra look the same. Therefore, no enantioselective discrimination is expected for this system.

DISCUSSION A list of the mentioned vibrational modes and the corresponding wavenumber is given in Table 1. The main binding force between the two enantiomers and the chiral selector is due to the ionic interaction. In both cases where (R)-1 and (S)-1 were adsorbed, no band at 1751 cm-1 could be detected, indicating that the amino acid derivative is fully deprotonated. The expected carboxylate band at ∼1550 cm-1 overlaps with other characteristic bands of the formed complex. It is also expected to be rather broad due to strong interaction with the protonated quinuclidine part. Moreover, it was difficult to desorb either of the two enantiomers from the CSP (tBuCQN/CD) without using the competing acetic acid, which also indicates a strong ionic interaction. Acetic acid was necessary to enhance the exchange/removal of the SAs. An important issue is the role of the acetic acid for the enantiodifferentiation of the SO-SA associates. Because the difference spectra of the two diastereomeric complexes on the surface (measured with acetic acid) and in solution (measured without acetic acid) look quite similar (Figure 9), we can exclude a significant sterical influence of acetic acid on the SA-SO interaction. In other words, acetic acid works only as a displacer with Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

5327

Figure 12. Spectra of DFT frequency calculations using BLYP/631G(d,p). (a) DNB-(S)-Leu interacting with tBuCQN. (b) Monomer of DNB-(S)-Leu. (c) tBuCQN. The corresponding conformations are depicted in Figure 11. The spectra were scaled by a factor of 1.04. The calculated spectra fit quite well to the ones measured in solution (Figure 8), and the occurring band shifts agree also qualitatively to those calculated. The main shifts of the functional groups of the participating species forming the complex are marked by arrows. These are in detail: δCOOH(SA) ) 1451 f 1579 cm-1, δN-H,amide II(SA) ) 1544 f 1568 cm-1, νNO2,asym.(SA) ) 1584 f 1572 cm-1, νCdO,amide I(SA) ) 1690 f 1713 cm-1, νCdO,acid(SA) ) 1848 f 1742 cm-1, νN-H(SA) ) 3657 f 3580 cm-1, and νCdO,amide I(SO) ) 1799 f 1776 cm-1. Figure 11. Conformations of DNB-(S)-Leu, tBuCQN, and the corresponding complex. Forming the DNB-(S)-Leu/tBuCQN complex, a total bonding energy of ∆Ebonding ) -36.7 kJ mol-1 was calculated.

the consequence that this phenomenon may also be reached by the use of other acids; however, the pKa differences of the acids compared to the SAs need to be taken into consideration due to their role in the ion exchange processes. The different intensities of certain bands such as the symmetrical stretching νs(NO2) at 1346 cm-1 (1345 cm-1), the asymmetrical stretching νas(NO2) at 1545 cm-1 (1541 cm-1), and the amide I band at 1670 cm-1 (1668 cm-1) (Figure 5) depending on the enantiomer, show that a higher amount of (S)-1 can adsorb on the tBuCQN-modified surface compared to (R)-1. Upon adsorption, interactions within the CSP are broken as revealed by the ATR spectra (see also later). The (R)-1 is able to break only a smaller part of the inter- or intramolecular interactions of the tBuCQN surface, namely, around 1/3-1/2 of the amount that the (S)-1 is able to release. Amide and acid functional groups involved in H-bonds and ionic bonds are mainly responsible for the interaction between SAs and SO. The shift of the asymmetrical stretching νas(NO2) to 1541 cm-1 in (S)-1 adsorption instead of 1545 cm-1 in (R)-1 adsorption might be due to π-π interactions. Van der Waals interactions are usually difficult to see in vibrational spectroscopy. The following discussion focuses therefore on the behavior of the groups building up H-bonds. Figure 5a and Figure 8b reveal a shift of the SO amide I band from 1726 to 1745 cm-1. This shift was also observed when small concentrations of acetic acid, TFA, or (S)-1 without acetic acid in the solvent were flown over the CSP. It indicates the cleavage of a stronger intramolecular hydrogen bond of the amide CdO with 5328 Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

an H-donor group and the formation of a weaker one as soon as an acidic compound reaches the surface. The (S)-enantiomer leads to stronger changes of the surface during adsorption. (S)-1 shows specific interactions with its functional groups, whereas the (R)-1 seems to adsorb mainly via an ionic interaction. No other interactions could be detected with IR spectroscopy. The specific interactions between the (S)-enantiomer of the SA and the SO come mainly from the amide groups of the corresponding species. In case of the SA, the amide III (1266 f 1282 cm-1) and amide II (∼1530 f 1585 cm-1) bands shift up and the amide I of the SO (1726 f ∼1717 cm-1) shifts down. This indicates a strong H-bonding of the SO with an H-donor. And indeed, a shift and broadening of the ν(NH) stretching band of the (S)-1 at ∼3350 cm-1 could be observed (Figures 8 and 9). The ν(NH) stretching band of the SO is situated a bit higher in wavenumber at ∼3371 cm-1. The demodulated spectra where (R)-1 is adsorbed on tBuCQN (Figure 5a) look similar to the spectra where each of the two enantiomers is adsorbed on CD (Figure 10a,b). Furthermore, it even takes the same time to desorb (S)-1 and (R)-1 from CD and (R)-1 from tBuCQN as the SO, namely, 45-50 s (Figure 3). Therefore, one can speculate that in the case of (R)-1 adsorption on tBuCQN, the carbamate side chain of the CSP has no significant influence on the overall adsorption behavior. This assumption is confirmed by the fact that no band shifts could be detected in the amide regions (amides I, II, III). The lack of such shifts indicates the absence of H-bonds. It is known that DNB-(R)-Leu and DNB-(S)-Leu bind to quinine, quinidine, and t-BuCQN in a 1:1 association mode.1 Further, it was shown that the SA exhibits higher affinity for quinine and quinidine than for t-BuCQN.7 The reason for this may be a different type of interaction. As shown elsewhere,18 the O-H

Table 1. Assignments of the (ATR)-FT-IR Spectraa dissolved vibrational mode

DNB-(R,S)-Leu

δ(OCN-H), amide III νs(NO2) δ(COOH) δ(OCN-H), amide II νas(NO2) χ(aromatic vib) χ(aromatic vib) χ(aromatic vib) ν(HNCdO), amide I ν(HOCdO), acid ν(N-H)

1266 (1273, 1291) 1348 (1333) cov. (1451) cov. (1544) 1549 (1584)

a

1595 (1626) 1632 (1666) 1674 (1690) 1751 (1848) 3350 (3657)

complexed tBuCQN

DNB-(R)-Leu

tBuCQN

(1218)

1282 (1305) 1345 (1333) cov. (1579) 1585 (1568) 1541 (1572)

1346 1545 cov. (1608) 1593 (1634) 1622 (1672) 1726 (1799)

1629 1670

3371 (3672)

3367

DNB-(S)-Leu

∼1745

(1626) 1629 (1662) 1668 (1713) cov. (1742) ∼3341 (3580)

tBuCQN (1226, 1295)

cov. (1604) (1628) (1671) ∼1717 (1776) 3371 (3666)

(cm-1)

The values are given in wavenumbers obtained by ATR-IR. Values written in italic type were measured by transmission FT-IR. Values in parentheses were calculated by a normal coordinate analysis using BLYP 6-31G(d,p). Bands covered by more intense bands are marked with cov.

at the C9 of cinchonidine (H-donor) and the basic nitrogen of the quinuclidine part (H-acceptor) fit quite well for interacting with an acid group where the carbonyl group acts as the H-acceptor and the acid proton is transferred to the base. We explain the opposite behavior in our experiments, faster release of (R)-1 and (S)-1 bound on immobilized CD compared to (S)-1 adsorbed on immobilized t-BuCQN (Figure 3), by the presence of acetic acid in the solvent as a competitor to the SA concerning the ionic type of interaction but not for the hydrogen bond. Because (R)-1 seems not to form a specific hydrogen bond with the SO, it shows absorption/desorption behavior similar to that of both SAs on immobilized CD. Taking the proposed interaction model as a starting point for a structure optimization and performing a normal mode analysis using DFT methods, we could reproduce the experimentally obtained spectra, which corroborate the proposed model. This also allows a more accurate assignment of the IR bands and confirms the general behavior of the observed band shifts. Further, a total binding energy ∆E of -36.7 kJ mol-1 was calculated. Microcalorimetric titrations7 revealed a ∆Hb° of -33 kJ mol-1. Finally, from the intensity ratio of the νs(NO2) band at 1345 cm-1, we estimated a ∆∆G value of 2.2 kJ/mol. For comparison in chromatography, a ∆∆G of 5.11 kJ/mol was calculated from the enantioselectivity value R for the (R,S)-1/ tBuCQN system.6 It should be noted that the value 5.11 kJ/mol was calculated using methanol/ammonia acetate as the mobile phase, whereas we used acetic acid in acetonitrile. CONCLUSIONS The present study shows that ACM ATR-IR spectroscopy in combination with a phase-sensitive data analysis is a valuable tool to get specific information from the solid-liquid interface of a chiral stationary phase used in chromatography for enantiomer separation. The most powerful advantage of this analytical method is the exclusive sensitivity for diastereomeric interaction of either of the two enantiomers adsorbed on a chiral surface. The reproducibility of even the smallest IR bands is excellent. The method was applied to study a well-known system in liquid (18) Burgi, T.; Vargas, A.; Baiker, A. J. Chem. Soc., Perkin Trans. 2 2002, 15961601.

chromatography, namely, the enantiomer separation of DNB-(R,S)Leu with a silica grafted tBuCQN-type CSP. For the investigated systems, the ionic interaction is the strongest one as expected. The main force causing the enantiodifferentiation is however the specific H-bonding of the DNB-(S)Leu with the carbamate branch of the immobilized tBuCQN. The ATR-IR experiments lead to the conclusion that (R)-1 does not form such H-bond interactions and therefore is not adsorbed as strongly as the (S)-1 enantiomer. This was also observed in HPLC separation mode, where the tBuCQN showed a preferential binding of the (S)-enantiomer.5,19 The indications for the formation of this H-bonding are in particular the amide III (SA) shift from 1266 cm-1 up to 1282 cm-1, the amide II shift from around 1500-1550 cm-1 up to 1585 cm-1 (SA), and the shift of the amide I from 1727 cm-1 down to 1717 cm-1 (SO on CSP). The discriminative interaction of DNB-(R)Leu and DNB-(S)-Leu with immobilized tBuCQN (CSP) is similar to the interaction observed in solution. Therefore, conclusions of NMR investigations concerning the H-bond interaction of DNB(R)-Leu and DNB-(S)-Leu with tBuCQN made in solution can also be applied to the immobilized tBuCQN. The same interaction model was used to perform a normal mode analysis, which resulted in similar spectra as they were obtained experimentally. The tendential behavior of the observed band shifts could also be reproduced, confirming the proposed model. The ATR-IR MES spectroscopy results are in accordance with the findings of earlier spectroscopic and chromatographic studies on this system. Absolute configuration modulation ATR-IR spectroscopy together with DFT calculations afforded detailed information on the relevant structures responsible for chiral recognition in an experimentally fast and convenient way. It may be used as a screening technique and as a tool for developing new chiral selectors that are capable of separating other mixtures of enantiomers. ACKNOWLEDGMENT Financial support of the Swiss National Science Foundation and the Foundation Claude and Giuliana is gratefully acknowledged. We thank Atsushi Urakawa for helpful advice concerning (19) Lammerhofer, M.; Lindner, W. J. Chromatogr., A 1996, 741, 33-48.

Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

5329

DFT calculations and Prof. U. P. Fringeli, Institute of Physical Chemistry, University of Vienna, for valuable discussions on modulation spectroscopy. Dr. Dieter Baurecht, Institute of Physical Chemistry, University of Vienna, is acknowledged for making his demodulation program available to us. Dr. Norbert M. Maier is gratefully acknowledged for providing the various CSP and

5330

Analytical Chemistry, Vol. 76, No. 18, September 15, 2004

selectand samples. CSCS in Manno and ETH-Zurich are acknowledged for providing computing time. Received for review April 15, 2004. Accepted June 18, 2004. AC049428X