Device for Raman Difference Spectroscopy - Analytical Chemistry

Jul 12, 2007 - Fast and Highly Sensitive Fiber-Enhanced Raman Spectroscopic Monitoring of Molecular H2 and CH4 for Point-of-Care Diagnosis of Malabsor...
11 downloads 28 Views 500KB Size
Anal. Chem. 2007, 79, 6159-6166

Device for Raman Difference Spectroscopy Torsten Frosch,† Tobias Meyer,‡ Michael Schmitt,† and Ju 1 rgen Popp*,†,‡

Institut fu¨r Physikalische Chemie, Friedrich-Schiller-Universita¨t Jena, Helmholtzweg 4, D-07743 Jena, Germany, and Institut fu¨r Photonische Technologien, Albert-Einstein-Strasse 9, D-07745 Jena, Germany

A new layout of a versatile Raman difference setup is presented. The new device combines the advantages of the rotating cell for exploitation of the resonance Raman enhancement and the high precision of Raman difference spectroscopy together with the multiplex advantages and very high quantum efficiency offered by a CCD detector. While Raman difference spectroscopy is the most accurate method for the detection of very small band shifts, the method requires the strict prevention of any environmental perturbation of one of the two spectra, which are used for the difference spectra calculation. The presented device satisfies this requirement by implementing a double-beam layout, where the simultaneously detected Raman signals of two sample cells are combined within a Y-fiber bundle and imaged together onto the CCD detector. The accuracy of the new apparatus in detecting frequency shifts and minor sample components is greatly increased compared to conventional Raman spectroscopy as shown by studying binary mixtures of CHCl3 and CCl4. Hereby it was possible to resolve a formerly undetected shift of 1), δ ∼ ∆ν is valid.23 On the other hand, for ∆ν/Γ < 0.2, the Raman shift ∆ν is only a function of Γ and d/I016 (where I0 is the peak intensity of the Raman line, d ) |IDmax| + |IDmin| is the sum of the magnitudes of the intensity maximum and minimum in the difference signal, and the factor F is given by the band shape of the Raman line, e.g., Lorentz, FL ) 0.3501, or Gauss, FG ) 0.3849).

∆ν ) dFΓ/I0

(2)

For a given line shape (F, Γ) and intensity I0, the Raman shift ∆ν is determined by d and, therefore, by the intensity magnitudes in the difference signal. The RDS method can also be applied for the accurate determination of band broadening and also for obtaining a higher sensitivity in detecting minor sample components (by suppressing unwanted signals, e.g., from the solvent).20,24 Obviously, this increase in precision due to the difference spectra calculation requires the strict prevention of any environmental perturbation of one of the two spectra, which are used to calculate the difference spectrum. The new device, presented within this contribution, satisfies this requirement by the simultaneous detection of the two Raman signals by means of a doublebeam layout. The device has the great advantage compared to older RDS setups20,21,25-31 in exploiting the full 2D multiplex advantages and very high quantum efficiency (QE) in a wide spectral range (UV-NIR) offered by a charge-coupled device (CCD), being the detector of choice for low-light applications like Raman spectroscopy. RESULTS AND DISCUSSION The former Raman difference setups are summarized in the following. The progress in instrumentation (CCD detector, imaging spectrograph, etc.) and the introduction of the new device are discussed, and the increased accuracy is demonstrated. Description of Former Devices and Progress in Instrumentation. Raman difference spectroscopy was introduced by (22) Gardiner, D. J.; Girling, R. B.; Hester, R. E. J. Chem. Soc., Faraday Trans. 1975, 2, 71, 4, 709-713. (23) Laane, J.; Kiefer, W. J. Chem. Phys. 1980, 72, 10, 5305-5311. (24) Laane J. J. Chem. Phys. 1981, 75, 6, 2539-2545. (25) Covington, A. K.; Thain, J. M. Appl. Spectrosc. 1975, 29, 5, 386. (26) van den Boom, H.; Breemer, R. E. Rev. Sci. Instrum. 1975, Vol. 46, No. 12, 1664. (27) Rousseau, D. L. J. Raman Spectrosc. 1981, 10, S.94-99. (28) Shelnutt, J. A. J. Phys. Chem. 1983, 87, S.605-616. (29) Laane, J.; Kiefer, W. J. Appl. Spectrosc. 1981, 35, 3, 267. (30) Laane, J.; Kiefer, W. J. Appl. Spectrosc. 1981, 35, 4, 428. (31) Moskovits, M.; Michaelian, K. J. Appl. Opt. 1977, 16, 6, 2044.

6160

Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

Bodenheimer et al. in 1972.20 They pointed out for the first time the usefulness of this comparison technique by cancelling out signals common to both samples, thus facilitating the observation of certain effects, like small wavenumber shifts or the detection of minor components. It was also recognized that the shifts can be measured with an increased accuracy determined by the spectrometer resolution rather than its reproducibility. The layout of the setup was a double-beam arrangement where the laser beam was alternately focused on two samples by means of a chopping mirror. The Raman scattered light (only half of the laser beam) was detected with a monochromator and a photomultiplier (PM) and readout synchronized with the trigger of the chopper. The application of RDS in resonance Raman spectroscopy was suggested by Kiefer in 1973,21 who modified the layout to a singlebeam arrangement with a divided rotating cell. The detection of the Raman signals of the samples in the two compartments of the cell was triggered by the rotation frequency of the cell (>1000 rpm) and readout with a PM. This very useful setup was later slightly modified (e.g., by use a four-compartment cell) and successfully applied to study, e.g., small wavenumber shifts in binary mixtures.22,25-30 Because a PM can be read out very quickly (>1000 rpm), the time elapsing between the detection of the two Raman signals of the two rotating cell compartments is short enough to rule out environmental perturbations (thermal drifts of the apparatus, laser fluctuations, etc.) of one of the Raman signals compared to the other. The advantages of double-beam RDS setups compared to a one-beam setup like, e.g., a greater versatility in the choice of possible sample (e.g., fibers or single crystals) and intensity balancing of the two beam channels with respect to each other have been pointed out.31,32 While the application of Raman spectroscopy was pushed by the availability of powerful laser lines covering the whole spectral range from UV to NIR (e.g., Ar/Kr lasers); detection of the scattered light by means of a monochromator and a single-element detector system (PM) was time-consuming and not really practicable. Exploitation of the multiplex advantages offered by an optical multichannel analyzer has dramatically reduced the necessary acquisition time and has improved Raman spectroscopy. The application of multichannel detectors made it possible to record the whole Raman spectrum much faster compared to singlechannel detectors and therefore with a higher signal-to-noise ratio (S/N). This reduction in acquisition time also avoids sample degradation due to shorter light exposure times. In 1990, Kamogawa and Kitagawa applied a device with a one-dimensional photodiode array parallel to the entrance slit of the monochromator, however, without exploitation of the multiplex advantage.33 Nowadays the CCD detector is the multichannel detector of choice for low light conditions such as Raman spectroscopy.34 This is due to a very low detection limit of ∼5 electrons, a very small dark current of less than one electron/pixel/h (cooled with liquid nitrogen) and a two-dimensional (2D) array capability. The CCD can therefore be exposed to a signal for hours without significant contribution from dark current and is capable of detecting intensity versus wavelength and versus position within the slit height. Thinned, backilluminated CCDs have smooth quantum efficiency (QE) distributions ranging from UV to NIR, with peak QEs of (32) Martin, J. C. Rev. Sci. Instrum. 1985, 56, 12, 2217. (33) Kamogawa, K.; Kitagawa, T. J. Phys. Chem. 1990, 94, 3916. (34) Murray, C. A.; Dierker, S. B. J. Opt. Soc. Am. A 1986, 3, 12, 2151.

Figure 1. Sketch of the Raman difference setup.

>90% in the range 500-650 nm. Altogether, a CCD camera is the perfect multichannel detector for Raman spectroscopy. Deckert et al. suggested a single-beam layout for a RDS device applying a CCD camera.35 However, they did not fully consider the new 2D array capabilities of this versatile detector because their setup was lacking the important improvement of an imaging (singlestage) spectrograph. Since they did not detect the signals simultaneously, an uncertainty due to environmental perturbations of the signals with respect to each other was introduced with this layout. Otherwise a rather fast readout of the CCD would also be a significant drawback, since a CCD detector is read noise limited in case of low light conditions (One long signal exposure (onchip integration) is better than many short exposures (frameaveraging)). This drawback is a true limitation, since the precision in detection of very small wavenumber shifts ∆ν is determined by the accurate measurement of the intensities of the maximum and the minimum of the difference signal (see eq 2) and is therefore noise limited. Another disadvantage by just exchanging a PM21 with a CCD camera is that the CCD camera cannot be triggered by the high frequency of the rotating cell. Deckert et al. therfore masked three of the four compartments of the rotating cell by means of a blind, thus using only 25% of the laser exposure time, while noise and background is accumulated all the time. They obviously needed at least a factor of 4 longer measurement time. Because RDS is the most accurate method detecting very small wavenumber shifts (e.g., occurring in interaction studies of (35) Deckert, V.; Liebler, W.; Eck, R.; Kiefer, W. J. Appl. Spectrosc. 1997, 51, 7, 939.

biomolecules) and no Raman difference device is known exploiting the full capabilities of a CCD camerasa new, versatile 2D multichannel layout for the true simultaneous detection of two (or multiple) signals is described in the following. Description of the New Device. The basic layout of the setup is shown in Figures 1 and 2 and is explained in the following. Laser Excitation Source. An argon ion laser (LA1: Innova300MotoFreD, Coherent Inc.) with lines in the visible (528.7, 514.5, 501.7, 496.5, 488.0, 476.6, 457.9, 454.5 nm) as well as the fundamental (363.8 and 351.1 nm) and the strongest frequency doubled (257.2, 244.0, and 228.9 nm) lines in the UV is used as Raman laser excitation source. The resonator cavity runs in the TEM00 mode, and plasma lines are removed by a quartz PellinBroca prism (PBP), mounted in least deviation configuration, and an adjustable slit (SL1). A HeNe laser (LA2: 632.8 nm, 35 mW, Coherent Inc.) as a second, independent laser source can be easily coupled into the beam line by just moving the PBP along an adjusted rail. Raman Difference Setup. The laser beam is separated into two uniform beams by a nonpolarizing 50% beam splitter (BS). The two laser beams are focused by quartz lenses (L1: d ) 25.4 mm, f ) 100 mm) into two identical, cylindrical, rotating, quartz cells (C), while the laser beam waist is 40 µm and the Rayleigh length 2 mm. Employing a 90° geometry, the scattered light is collected and parallelized by the quartz lens L2 (d ) 25.4 mm, f ) 35 mm) and focused onto the head (FHL1: see Figure 2A) of a customdesigned Y-fiber (UV/vis, NA ) 0.22) by the quartz lens L3 (d ) 25.4 mm, f ) 50 mm). The lenses are optimized via ray tracing to yield a good compromise between efficiently collecting the Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

6161

Figure 2. Layout of the sample excitation, Raman light collection in 90° geometry, and coupling into the fiber head. Two custom-designed fiber heads are used for the rotating cell (A, liquid samples FHL) and the rotating sample holder (B, solid samples FHS).

scattered light, matching the NA of the fiber, minimizing spherical and chromatic aberration, and creating a parallel beam path with 15-mm transversal dimension (the reason for using a parallel beam path to include a notch filter36-38 and a polarizer is explained later). 6162

Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

The two identical front fiber heads (FHL1) within the two beam lines consist of 25 linearly aligned 100-µm UV/vis multimode (36) Yang, B.; Morris, M. D.; Owen, H. J. Appl. Spectrosc. 1991, 45, 9, 15331536.

Figure 3. Conventional Raman spectrum (A) and Raman difference spectrum (C) of a mixture of 5 µL of CHCl3 in 1 mL of CCl4. The intensity scale of the section of the conventional Raman spectrum (B) is stretched, for better illustration.

fibers. The fibers of the two bundles are vertically arranged in a line and separated by 10 blind fibers. The combined back fiber head (FHL2), consisting of 60 fibers, therefore has a vertical dimension of ∼7.2 mm, while the two channels are arranged above each other and separated by ∼1.2 mm. This custom-designed back fiber head (FHL2) is placed in the plane of the spectrometer slit without F# matching (F# matching has also been considered, but not applied because: the stray light suppression of the spectrograph is very good, the associated magnification of FHL2 is unwanted, and the throughput cannot be improved anyway.). The vertical dimension of FHL2 is designed to match the height of the chosen CCD chip (8 mm), while the vertical magnification of the imaging spectrograph is M ) 1. The spectrograph (SG: Acton SP2758i, f ) 750 mm, Czerny-Turner aspheric imaging design, equipped with a stabilized triple grating turret (2400, 1800, 600 g/mm)) images FHL2 onto the CCD chip (Princeton Instruments, Spec-10 400B/LN). The back-illuminated CCD camera having a peak QE > 90% (500-650 nm) consists of 1340 × 400 pixels of 20 µm × 20 µm size, i.e., a dimension of 26.8 mm × 8 mm. Two ranges of interest (ROI) are defined symmetrically to the center of the chip. A sophisticated holder allows for precise 3D translation, tilting, and rotation of the fiber head with respect to the plane of the spectrograph slit. The signals of the two ROI (originating from the two rotating cells) are vertically binned, read out from the detector, and handled with a homemade software routine (Labview, National Instruments). An online visualization of the difference signal between the two ROI can be used to control whether an artificial difference signal is present due to a misalign(37) Tedesco, J. M.; Owen, H.; Pallister, D. M.; Morris, M. D. Anal. Chem. 1993, 65, 9, 441A-449A. (38) Schoen, C. L.; Sharma, S. K.; Helsley, C. E.; Owen, H. J. Appl. Spectrosc. 1993, 47, 3, 305-308.

ment. This is done by means of an appropriate reference substance (e.g., cyclohexane, with sharp and intense Raman lines) in both cells, which allows for a very precise adjustment of the setup. Control measurements with the reference substance (same substance in both cells) have always been performed before and after the Raman difference experiments to exclude any artificial difference signal and can be used to calculate the possible accuracy (error bar) of these experiments. Versatility of the Setup. All optical components are chosen to allow for a broadband application of the device from the UV to the NIR range with no or only minor changes in the optical setup. Notch filter (N: two similar super notch plus filter from one charge, Kaiser Optical Systems, aperture 15 mm) can be included into the two parallel beam paths to measure Raman spectra very close to the Rayleigh line. Measurements of polarized Raman spectra can be performed by including a calcite Glan laser polarizer (P) and a λ/2 double Fresnel rhomb (FR) for purification and rotation of the polarization of the exciting laser light, as well as two identical calcite Glan laser analyzers (A: clear aperture 15 mm) in the two parallel beam lines between L2 and L3. The experiments have been performed by rotation of the Fresnel rhomb and fixed analyzer, to avoid ambiguities because of differences in the grating reflectivity. The quartz cells can be easily exchanged by rotating sample holders, which allow for resonance Raman difference spectroscopy of solid samples, e.g., mixtures embedded in a KBr matrix. For such a case, another custom-designed Y-fiber (FHS, see Figure 2B) is applied, counting for the more spherical scattering geometry. The front head (FHS1) consists of 19 circular closepacked 100-µm multimode fibers, approximating the scattering geometry of the solid sample. Both fibers are again vertically Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

6163

Figure 4. Integral intensities of the Raman lines of CCl4 as a function of the volume fraction VCCl4/VCHCl3 (A) and of the Raman lines of CHCl3 as a function of the volume fraction VCHCl3/VCCl4 (B), respectively.

combined with 15 separating blind fibers into the back fiber head (FHS2). Obviously, a variety of different samples (temperaturecontrolled cells, capillaries, single crystals, etc.) can be studied with the double-beam layout, while a variable, neutral density filter (DF) is used for sensitive intensity balancing between the two samples. The improved performance of the new device in detection of minor components by cancelling out unwanted solvent signals and in detection of very small wavenumber shifts due to solvent-solute interactions is exemplarily shown in the following by investigating binary mixtures of CHCl3 and CCl4, in different molar ratios that had also been used for testing former RDS devices21 (λexc ) 488 nm; 250 mW at the sample; spectral slit width, 1.5 cm-1 (slit, 100 µm; grating, 2400 g/mm)). The application of the setup to solid samples and measurement of polarized Raman spectra is also briefly demonstrated. Examination of Binary Mixtures of CHCl3 and CCl4. Detection of Minor Components. The lower limit of the device in detecting CHCl3 in CCl4 in a molar fraction of only 0.005 is demonstrated in Figure 3. The spectrum of the mixture of only 5 µL of CHCl3 in 1 mL ofCCl4 (Figure 3A) looks very similar to the spectrum of the pure solvent CCl4 (see for example ref 21). The appearance of two shoulders at 262 and 366 cm-1 (see Figure 3B) might indicate the presence of CHCl3. However, the simultaneous detection of both, i.e., the Raman spectrum of the mixture as well as the spectrum of the pure solvent CCl4, allows for subtraction of the solvent spectrum. The difference spectrum (Figure 3C) clearly shows Raman bands at wavenumber positions 366 and 262 cm-1, while the CCl4 bands at 218 , 314, and 459 cm-1 are not completely removed due to solute-solvent interaction and resulting wavenumber shifts. While the Raman spectra of the individual bands must have the same peak intensities for application of the difference calculation in the case of pure band shifts, the detected signals will differ slightly in absolute intensities as explained in the following. The scaling of the individual Raman bands to the band of interest before subtraction of each other might be performed peakwise as pointed out, e.g., in ref 40. The ray tracing of the lens system (see Figure 1 and Figure 2), however, shows that the main 6164 Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

Figure 5. Wavenumber shifts of the Raman band of CCl4 (at 218 cm-1 and 459 cm-1) as a function of the mole fraction of CHCl3 (A) and shifts of the Raman band of CHCl3 (at 262 and 366 cm-1) as a function of the mole fraction of CCl4 (B).

influence is the longitudinal chromatic aberration (a Raman spectrum of 4000 cm-1 in the visible corresponds to ∼100 nm). This wavelength dependency has been considered by fitting the quotients of the normalized intensities of the bands at 218 , 314, and 459 cm-1 of the pure solvent CCl4 with a linear function and scale all further spectra pixelwise with the related factor. In Figure 3C, it is seen that a slight second-order aberration might be present and the error bars of the detected CHCl3 signals are in the same range as the absolute values. However, a quadratic fit is not necessary because a possible band shift error that might be introduced is below 0.01 cm-1. An increasing concentration of CHCl3 results in a linear increase of the integrated intensities of the Raman lines of CHCl3 as a function of the volume fraction VCHCl3/VCCl4 (Figure 4B), since the spectra are normalized to Isolvent. The same linear dependency can be seen for the Raman lines of CCl4 (at wavenumber positions 218 , 314 , and 459 cm-1; see Figure 4A), if CCl4 is considered as solute and CHCl3 as the solvent. Hence, from measuring the Raman intensity of a solution of known concentration, it is possible to calculate the fraction of the solute in a mixture of unknown concentration by just measuring the same Raman line. Detection of Very Small Wavenumber Shifts. The quantification of wavenumber shifts from experimental Raman difference spectra has been demonstrated by Laane and Kiefer23,24,29 as discussed briefly in the introduction. They investigated line profiles of pure Gaussian and Lorentzian shapes and derived equations for these cases, while Voigt profiles can sufficiently be approximated by linear interpolation. The deconvolution of the slit function (e.g.,

Figure 6. Comparison of the difference Raman spectrum of a mixture of 0.2 mol fraction CHCl3 in CCl4 (B) and a random (thermal) drift of the apparatus within 45 min onto the spectrum of the pure solvent CCl4 (A).

measured with a sharp laser plasma line) has also been considered,39,40 while several additional broadening effects are hardly to be approximated very accurately. Hence, it might be very difficult to derive the pure natural line width (which is a Lorentzian one) in many cases. The procedure becomes even more complex, if several Raman lines overlap with each other, what is often the case for important biomolecules. Another useful method is the simulation of difference Raman spectra via the experimental Raman spectrum. In doing so, the measured Raman line is numerically interpolated with a finer grid size than the minimal resolution and slightly shifted with respect to itself. For every shift ∆ν, a simulated Raman difference spectrum and, therefore, the value d of the difference intensity are obtained. Linear regression yields the constant A in ∆ν ) Ad (see eq 2). The determined wavenumber shifts are shown for the Raman lines of CCl4 (at 218 and 459 cm-1; see Figure 5A) as a function of the mole fraction of CHCl3 and for the bands of CHCl3 (at 262 and 366 cm-1; see Figure 5B) as a function of the mole fraction of CCl4, respectively. The simulation of the difference spectra has been performed via the measured Raman spectrum of the pure solvent and the highest concentrated 1:1 mixture, resulting in the error bars depicted in Figure 5. The observed wavenumber shifts of the binary mixtures of CHCl3 and CCl4 are in agreement too.21 However, the shift of the 218-cm-1 Raman line of CCl4 has been detected for the first time, to the best of our knowledge, accounting for the improved accuracy of the new device. The advantages and the necessity of the simultaneous detection of the two signals, which are used to calculate the difference spectrum, are demonstrated in Figures 6 and 7. In Figure 6, the difference Raman spectrum of a mixture of 0.2 mol fraction CHCl3 in CCl4 (Figure 6B) is compared to a random (thermal) drift of the apparatus within 45 min onto the spectrum of the pure solvent CCl4 (Figure 6A). The error in determining the wavenumber shifts of the 218- and 459-cm-1 Raman lines of CCl4 due to this environmental perturbation is (39) Asthana, B. P.; Kiefer, W. J. Appl. Spectrosc. 1983, 37, 4, 334-340. (40) Asthana, B. P.; Kiefer, W. In Vibrational Spectra and Structure; Durig, J. R., Ed.; M. Dekker: New York, 1992; Chapter 2.

Figure 7. Comparison of the observed Raman band shift of the lines at 218 and 459 cm-1 of CCl4 (0.2 mol fraction of CHCl3 in CCl4) in a simultaneous and a delayed measurement as a function of mole fraction of CHCl3. The real band shift in the simultaneous measurement is adulterated in the delayed measurement by the thermal drift of the apparatus.

demonstrated in Figure 7. While the drift of the apparatus is in the same range as ∆ν for small molar fractions, the observed false shift (band shift + drift) would be too large, and the dependency is no more a linear one. Random drifts of the apparatus into the opposite side would result in the observation of too small shifts. One can clearly recognize the ambiguities introduced by serial, nonsimultaneous detection of Raman signals,35 which are intended to be used for RDS calculations. Hence, the presented RDS setup using a CCD detector and simultaneous detection of the two signals leads to a much improved performance compared to formerly reported RDS devices. Versatility of the Setup. The application of the setup for rotating solid samples is briefly demonstrated for sulfur. Even Raman bands as close as 27 cm-1 to the Rayleigh line can be detected (see Figure S1 Supporting Information). The measurement of polarized spectra of CCl4 proves the high purity of the polarization optics, in observing depolarization ratios of F ∼ 0.75 and F ∼ 0 for the depolarized Raman line at 314 cm-1, as well as for the Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

6165

An interpretation of these shifts on a molecular basis is currently underway.

Figure 8. Comparison of the resonance Raman spectrum of a 20 mM solution of hematin (pH 7.5) with excitation wavelength 488 nm (black) and the spectrum of a 1:2 (hematin/chloroquine) mixture (red) is shown in (A) in the region from 1600 to 1650 cm-1. The Raman difference signal of the spectra of hematin and the hematinchloroquine mixture is displayed in (B).

totally symmetric band at 459 cm-1, respectively, in single-channel detection (see Figure S2 Supporting Information). Examination of the Interaction of Chloroquine with Hematin. While malaria is one the most important infectious diseases with tremendous impact onto the economical development of sub-Saharan countries,41,42 the molecular mode of action of key drugs like chloroquine is not well understood.9,43,44 The understanding of the binding affinity of chloroquine to hematin in a hydrous environment is therefore of high importance. A monitoring of the sensitive Raman peaks during drug-hematin interactions will give insight into the molecular details of the docking process and might help to design new drugs. The new Raman difference setup will be of high importance to detect even small wavenumber shift according to weak drug-target interactions. Preliminary Raman spectroscopic results concerning the interaction of chloroquine with hematin in a hydrous environment are shown in Figure 8. The resonance Raman spectrum of a 20 mM solution of hematin (pH 7.5) with excitation wavelength of 488 nm is shown in Figure 8A (black) in the region from 1600 to 1650 cm-1. This Raman peak shifts subsequently to lower wavenumbers after addition of chloroquine (solution pH 7.5). The Raman spectrum of a 1:2 (hematin/chloroquine) mixture is also shown in Figure 8A (red). The difference signal of the spectra of hematin and the hematin-chloroquine mixture is displayed in Figure 8B. (41) http://rbm.who.int/wmr2005. (42) Sachs, J.; Malaney, P. Nature 2002, 680-685 (43) Hastings, I. M.; Bray, P. G.; Ward, S. A. Science 2002, 298, 74-74. (44) Wellems, T. E. Science 2002, 298, 124-126. (45) Frosch, T.; Tarcea, N.; Schmitt, M.; Thiele, H.; Langenhorst, F.; Popp, J. Anal. Chem. 2007, 79 (3), 1101-1108.

6166 Analytical Chemistry, Vol. 79, No. 16, August 15, 2007

CONCLUSION AND OUTLOOK A new Raman difference setup has been presented. The new device has superior performance compared to former ones and full versatility in the kind of sample (solutions in a rotating cell, solutions in temperature-controlled cell, solids in a rotating holder, single crystals, etc.) under investigation. Raman excitation wavelengths from deep UV to NIR can be applied. Therefore, the selectivity and sensitivity of the resonance Raman scattering can be exploited over the full spectral range in applying the rotating sample technique. An intrinsic enhancement of the scattering (ω4 dependency) and a separation of disturbing fluorescence signals from the Raman spectra are further advantages in case of deep UV excitation.10,45 Even highly nonuniform excitation sources, like pulsed lasers, can be taken into consideration. The fundamental improvements of the CCD multichannel detector (multiplex advantage, sensitivity, versatility) have been fully exploited with a double-beam layout and simultaneous detection of two signals. These two signals are used to observe wavenumber shifts down to 0.02 cm-1. The RDS setup is therefore the most accurate technique for the detection of very small Raman band shifts, which are often present due to weak molecular interactions. Because of these extraordinary advantages and versatilities, the new device is suggested as a very promising tool for the investigation of biological molecules and their interactions. Hence, these combined advantages of the new device will allow for a fast, sensitive, selective, gentle, and precise investigation of many important questions in the field of biochemistry and pharmacy. One exemplarily topic is the study of the interaction of anti-malariaactive agents like chloroquine,7,9 mefloquine,6 quinine,11 and dioncophylline A8,10 with their biological target hematin. First results show a weak interaction of chloroquine and hematin in a hydrous environment. A molecular understanding of these interactions will inform drug chemistry and help to fight back against malaria. ACKNOWLEDGMENT The authors thank W. Fa¨hndrich, C. Jabob, and H. Fischer from our workshop for help in the construction of the mechanical parts of the setup. We grateful acknowledge support from the Fonds der Chemischen Industrie and the Deutsche Forschungsgemeinschaft (Sonderforschungsbereich 630 “Recognition, Preparation, and Functional Analysis of Agents Against Infectious Diseases“, project C1). SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review March 5, 2007. Accepted June 1, 2007. AC070440+