Anal. Chem. 2000, 72, 5718-5724
Surface-Enhanced Resonance Raman Spectroscopy as an Identification Tool in Column Liquid Chromatography Reza M. Seifar, Maarten A. F. Altelaar, Reyer J. Dijkstra, Freek Ariese, Udo A. Th. Brinkman, and Cees Gooijer*
Department of Analytical Chemistry and Applied Spectroscopy, Free University, De Boelelaan 1083, 1081 HV Amsterdam, The Netherlands
The compatibility of ion-pair reversed-phase column liquid chromatography and surface-enhanced resonance Raman spectroscopy (SERRS) for separation and identification of anionic dyes has been investigated, with emphasis on the at-line coupling via a thin-layer chromatography (TLC) plate. SERR spectra using silver sols were recorded both for aqueous solutions and for samples deposited on aluminum oxide and silica TLC plates at 514.5- and 457.9-nm laser excitation. For some dyes, the shorter wavelength was needed to diminish the fluorescence background. For aqueous solutions and for samples deposited on aluminum oxide, clear SERR spectra were obtained upon addition of poly(L-lysine); for the silica plates, the addition of nitric acid was required. Upon drying the plates, the SERRS signals decreased in intensity; simply adding a drop of water could largely restore them. At-line coupling of LC and SERRS was successfully achieved when using silica, but not aluminum oxide, plates. The application of a gradient, a high water content, and the presence of ion-pair reagents needed for the separation did not adversely affect the deposition and the recording of SERR spectra. The identification limits were 10-20 ng of deposited material, depending on the dye selected, which corresponded to injected concentrations of 5-10 µg mL-1. The applicability of Raman spectroscopy (RS) as an identification tool in column liquid chromatography (LC) is hampered by the extremely low Raman cross sections of most molecular systems. As a result, the detection and identification limits in RS are very poor. One recent approach to improve the sensitivity of LC-RS is directed at the increase of the optical path length of detector cells by using liquid core waveguides (LCWs) that can be used for aqueous solutions.1,2 Though the results are promising and open perspectives for identification purposes in the bioanalytical field, provided that the analyte concentrations are high enough, it is obvious that RS will never become suitable for traceanalytical problems. (1) Holtz, M.; Dasgupta, P. K.; Zhang, G. Anal. Chem. 1999, 71, 2934-2938. (2) Dijkstra, R. J.; Bader, A. N.; Hoornweg, G. P.; Brinkman, U. A. Th.; Gooijer, C. Anal. Chem. 1999, 71, 4575-4579.
5718 Analytical Chemistry, Vol. 72, No. 22, November 15, 2000
Other approaches to improve the sensitivity make use of resonance phenomena (resonance Raman spectroscopy, RRS) or surface-enhancement effects (surface-enhanced Raman spectroscopy, SERS).3-5 If the laser excitation wavelength matches the absorption of the analytes, the two techniques can be combined to surface-enhanced resonance Raman spectroscopy (SERRS).6 In principle, RRS can be coupled on-line to LC. Unfortunately, the resonance enhancement of Raman signals is often accompanied by an increased fluorescence background, so that such a coupling shows little perspective, except maybe for the deep ultraviolet region of the electromagnetic spectrum where fluorescence seems to play only a minor role.7-9 SERS and SERRS are most easily coupled to LC in the at-line mode, although on-line approaches have been described in the literature.10-12 In the present paper, the at-line approach is followed: the LC effluent is deposited onto a suitable substrate such as a thin-layer chromatography (TLC) plate by means of a jet-spray interface, followed by the deposition of silver colloids.13,14 Emphasis will be on the identification of dyes, a class of analytes that is highly relevant especially in the field of food, beverages, cosmetics, inks, and forensic science. Dye samples are usually analyzed by means of ion-pair reversed-phase LC.15 In a previous preliminary publication, we have shown the applicability of at-line ion-pair LC-SERRS for the identification of cationic, i.e., positively charged, dyes.14 Interestingly, there was no deposition problem despite the presence of ion-pair reagents; furthermore, these (3) Fleischmann, M.; Hendra, P. J.; McQuilla, A. J. Chem. Phys. Lett. 1974, 26, 163-166. (4) Jeanmaire, D. C.; Van Duyne, R. P. J. Electroanal. Chem. 1977, 84, 1-20. (5) Creighton, J. A. Blatchford, C. G.; Albrecht, M. G. J. Chem. Soc., Faraday Trans. 1979, 75, 790-798. (6) Laserna, J. J. Modern Techniques in Raman Spectroscopy; J. Wiley: Chichester, U.K., 1996. (7) Asher, S. A.; Johnson, C. R. Science 1984, 225, 311-313. (8) Asher, S. A. Anal. Chem. 1993, 65, 59A-66A. (9) Asher, S. A. Anal. Chem. 1993, 65, 201A-210A. (10) Force, R. K. Anal. Chem. 1988, 60, 1987-1989. (11) Pothier, N. J.; Force, R. K. Anal. Chem. 1990, 62, 678-680. (12) Cabalin, L. M.; Ruperez, A.; Laserna, J. J. Anal. Chim. Acta 1996, 318, 203-210. (13) Somsen, G. W.; Coulter, S. K.; Gooijer, C.; Velthorst, N. H.; Brinkman, U. A. Th. Anal. Chim. Acta 1997, 349, 189-197. (14) Seifar, R. M.; Dijkstra, R. J.; Brinkman, U. A. Th.; Gooijer, C. Anal. Commun. 1999, 36, 273-276. (15) Wegener, J. W.; Klamer, J. C.; Govers, H.; Brinkman, U. A. Th. Chromatographia 1987, 24, 865-875. 10.1021/ac000514g CCC: $19.00
© 2000 American Chemical Society Published on Web 10/21/2000
Figure 1. Anionic dyes used in this study.
reagents did not interfere with the SERRS effects. In fact, in the presence of ion pairs such as heptanesulfonate, the SERRS signals were intensified. In the present paper, the focus is on anionic dyes, analytes that are far more difficult to detect by means of SERRS than cationic dyes. It has been shown for such dyes that the observation of SERRS in aqueous solutions can be achieved by adding poly(L-lysine), a positively charged polymer that seems to act as a bridge between the negatively charged dye molecules and the negatively charged citrate layer of the silver sol particles.16 Of course, for the ion-pair LC separation of the anionic dyes, cationic ion pairs such as quaternary ammonium salts will have to be used. The main question to be answered in the present paper, therefore, is whether the demanding SERRS requirements and the LC separation conditions, i.e., gradient elution and the presence of quaternary ammonium salts, are compatible in an at-line LC-(TLC plate)-SERRS approach. As far as we are aware, TLC-SERRS of anionic dyes has not yet been reported in the literature. EXPERIMENTAL SECTION Materials. The anionic dyes, Food Yellow 3 (CI 15985), Acid Red 33 (CI 17200), Acid Orange 7 (CI 15510), Food Red 1 (CI 14700), Acid Violet 43 (CI 60730), and Acid Red 155 (CI 18130) (Figure 1) were kindly donated by. J. W. Wegener (Institute of Environmental Studies, IVM, Free University). Stock solutions (500 µg mL-1) were prepared in methanol-water (50:50 v/v). Prior to injection, sample solutions were prepared by mixing the stock solutions of the dyes and diluting them with the LC eluent. HPLC-grade methanol was purchased from J. T. Baker (Deventer, The Netherlands). Tetrabutylammonium bromide (TBABr) was obtained from Aldrich (Steinheim, Germany) and tetrabutylam(16) Munro, C. H.; Smith, W. E.; White, P. C. Analyst 1993, 118, 731-733.
monium nitrate (TBANO3) from Aldrich (Milwaukee, WI). Poly(L-lysine) hydrobromide, MW 7500, was obtained from Sigma (St. Louis, MO). Other chemicals were from stock and were of analytical-grade quality. Instrumentation. The chromatographic setup consisted of Gilson (Villiers-le-Bel, France) model 305 and 306 pumps and manometric modules and a Gilson dynamic mixer (Middelton, WI) solvent delivery pump, a six-port valve (Valco, Houston, TX) equipped with a 20-µL loop, a 100 mm × 3 mm i.d. LC separation column packed with 5-µm Hypersil ODS, and a HP series 1050 (Hewlett-Packard, Waldbrone, Germany) UV absorbance detector. The detection wavelength was 500 nm. The LC eluent flow rate was 0.7 mL min-1. At the outlet of the UV detector, the effluent was split in a T-piece flow splitter, 8% going to the spray-jet interface and the rest to waste. Two pieces of fused-silica capillary (100 µm i.d.) of different lengths were used to achieve the proper split ratio.14 A heated nitrogen flow (200 °C, 3 bar) was used to spray the effluent and evaporate the solvent. Precoated silica TLC plates with aluminum backing (Merck, Darmstadt, Germany) and Alugram Aloxn “aluminum oxide N” (Macherey-Nagel, Du¨ren, Germany) were used as deposition substrates. A modified Camag (Muttenz, Switzerland) Linomat III translation table was used to move the TLC plate at a speed of 2 mm min-1 during the deposition. The protrusion distance (distance between tip of the fused-silica tube and the interface) was 1 mm, and the capillaryto-substrate distance was 0.5 mm. The Raman setup consisted of a Spectra Physics (Mountain View, CA) series 2000 argon ion laser (excitation wavelengths 457.9 and 514.5 nm), a Spectra Physics model 336 beam expander, a Zeiss (Oberkochen, Germany) UEM microscope optically coupled to a SPEX (Metuchen, NJ) 1877 triplemate triple monochromator, a cooled EG&G Princeton Applied Research (PrinceAnalytical Chemistry, Vol. 72, No. 22, November 15, 2000
5719
ton, NJ) intensified linear photodiode array detector, and an EG&G OMA III model 1460 optical multichannel analyzer. The entrance slit of the spectrograph was set at 100 µm. The full width, of the bands, at half-maximum (fwhm) was 14 cm-1. A 40 × (NA 0.35) Zeiss Epiplan-HD microscope objective was used to focus the laser beam and to collect the sample scatter signal. The incident power of the laser on the sample deposited on the TLC plate was 14 mW. The Raman scattering signals were collected using five accumulations of 10 s each for aqueous solutions and five accumulations of 2 s for TLC plates otherwise stated. The sample spot sizes on the TLC plates were ∼2 mm2. Reagents. A stock buffer solution was prepared by mixing equivalent amounts of ammonium acetate and acetic acid (pH 4.7). The pHs of the eluents were measured before the addition of organic solvent. For the gradient separation of the anionic dyes, two solutions were used; the first solution contained 25 mM acetate buffer (pH 4.7) and 25 mM TBANO3, while the second solution was pure methanol. All eluents except methanol were filtered over a 0.5-µm FH filter (Millipore, Etten-Leur, The Netherlands). The silver sol was prepared according to published procedures.17 All glassware was cleaned by aqua regia solution (HClHNO3, 3:1 v/v) followed by scrubbing with a soap solution and thorough rinsing with distilled water. A 10-mL portion of a 1% aqueous solution of trisodium citrate was rapidly added to a boiling aqueous solution of silver nitrate (500 mL; 180 mg L-1) in a flatbottom flask under vigorous stirring. The sol was allowed to boil for an additional 90 min. Subsequently, the solution was stored in a dark and cool place. A concentrated colloid was prepared by centrifugation (4000 rpm for 30 min) of the citrate-reduced silver colloid (10 mL) and subsequent removal of 97% of the supernatant. Finally, 5-µL portions of the concentrated colloid were used for deposited analytes on the TLC plate. The time span between addition of silver sol and recording the Raman spectra should not exceed a couple of minutes since the Raman signals tend to decrease as a function of time as will be outlined below. RESULTS AND DISCUSSION LC Separation of Anionic Dyes and Immobilization on TLC Plates. Figure 2 shows a gradient ion-pair reversed-phase LC separation of the test analytes on C18-bonded silica with a solution of a quaternary ammonium salt in aqueous methanol as eluent. The chromatogram was recorded using absorption detection at 500 nm. A simple spray-jet interface in combination with heated nitrogen gas (pressure, 2.5-5 bar) was used to immobilize the LC effluent on the TLC plate.13,14 The proper selection of the TLC material is of primary importance. Precoated silica and aluminum oxide TLC plates, both without a fluorescence indicator, were compared. The silica plates could easily accept LC effluents containing up to 60% water. Visual inspection revealed that neither the use of a gradient (i.e., a variation in eluent composition) nor the presence of a quaternary ammonium salt had an adverse effect on the deposition. However, when aluminum oxide plates were used, spreading of the analyte spots of the deposited chromatogram could not be prevented, not even at an elevated nitrogen gas temperature of ∼250 °C created by using an extra stream of hot air. The spots on aluminum oxide were ∼10-fold larger than those on silica, which will of course adversely affect detectability. (17) Lee, P. C.; Meisel, D. J. Phys. Chem. 1982, 86, 3391-3395.
5720 Analytical Chemistry, Vol. 72, No. 22, November 15, 2000
Figure 2. Separation of the six anionic dyes under gradient conditions on 5-µm Hypersil ODS, 100 × 3 mm i.d. column. Eluent: (first solution) 25 mM TBANO3, 25 mM acetate buffer pH 4.7; (second solution) methanol. Flow rate, 0.7 mL min-1; injection volume, 20 µL. Gradient indicated in the figure. Sample solution: mixture of anionic dyes, concentration of each 25 µg mL-1. Absorption detection at 500 nm.
Figure 3. UV-visible absorption spectra of the six anionic dyes. Concentration, 1 × 10-6 M in water: (a) Acid Violet 43, (b) Acid Red 155, (c) Acid Orange 7, (d) Food Yellow 3, (e) Food Red 1, and (f) Acid Red 33.
SERRS Experiments. SERRS of anionic dyes is complicated since electrostatic repulsion prevents the analytes from coming close to the surface of the negatively charged silver sol particles. It has been demonstrated that the citrate-reduced silver sols are negatively charged.18 The repulsion can be reduced by adding poly(L-lysine) to the mixture of silver sol and anionic dyes.16 In the present study, additional parameters were evaluated, i.e., the influence of the ion-pair reagents needed to perform the LC separation, the TLC plate material used for deposition, and also the laser excitation wavelength in SERRS. SERRS in Aqueous Solutions. Preliminary measurements for studying the SERRS effect were performed in aqueous (18) Munro, C. H.; Smith, W. E., Garner, M., Clarkson, J.; White, P. C. Langmuir 1995, 11, 3712-3720.
Figure 4. SERRS spectra of the six anionic dyes in aqueous solution. Spectra at 457.9 (lower trace) and 514.5 nm (upper trace) are presented. Concentration of dyes ∼10 µg mL-1.
solutions containing no cationic ion-pair reagent. The anionic dyes were added to a mixture of silver sol and poly(L-lysine) in a quartz cell and irradiated by the laser beam. An excitation wavelength of 514.5 nm, which is close to the maximum absorbance of most anionic dyes studied in this report (Figure 3), seemed to be most suitable. However, as Figure 4 shows, with excitation at 514.5 nm the SERRS spectra of the dyes Food Yellow 3 (CI 15985), Acid Orange 7 (CI 15510), and Food Red 1 (CI 14700) are partly or
almost completely obscured by an intense fluorescence background. By changing the excitation wavelength to 457.9 nm, this fluorescence interference can be largely eliminated. As regards the detailed spectroscopic features of the dyes in Figure 4, it should be realized thatsexcept for dye Acid Violet 43 (CI 60730), which has a molecular structure different from that of the other test compoundssin the present study we are dealing with R-hydroxylaryl azo dyes that can be present in two tautomeric Analytical Chemistry, Vol. 72, No. 22, November 15, 2000
5721
forms, i.e., the hydrazone and the azo form.19 According to the literature, under acidic conditions the hydrazone form dominates; this explains why the spectra in Figure 4 are similar to those reported in the literature for the azo dye, yellow 14, which was also studied in the hydrazone form.20 In the spectra of the monoazo dyes used in the present study, the strongest bands were observed in the region between 1100 and 1700 cm-1. The most intense bands can be assigned to vibrations with displacements of the bridging atoms (C-NHNdC) at about 1235, 1340, and 1495 cm-1 in the hydrazone structure.20 The bands at about 1555 and 1595 cm-1 correspond to C-C vibrations of the benzene and naphthalene groups, respectively. The less intense bands below 1100 cm-1 correspond to out-of-plane skeletal deformations and C-H bendings.19 Careful inspection reveals that for some compounds studied the vibration patterns observed under excitation at 457.9 and 514.5 nm are not fully identical. This is, for instance, obvious for the 400-500-cm-1 region recorded for dye Acid Orange 7 (CI 15510). From an identification point of view, this implies that SERRS spectra of unknowns and references should be recorded under the same laser wavelength excitation conditions. To explain the observed influence of the laser wavelength, one may speculate that upon excitation at 457.9 nm the azo tautomeric form gives a more significant contribution than at 514.5 nm, since the maximum absorptions of azo and hydrazone are at 430 and 503 nm, respectively. It should be noted that full molecular structure elucidation based on SERR spectra is not straightforward: only those Raman active bands that are amenable to both surface enhancement and resonance effects are sufficiently intense to be recorded. Nonetheless, the spectra are sufficiently detailed to enable analyte identification provided that reference spectra have been recorded under identical conditions. Finally, one should note that the presence of a cationic ion pair such as TBANO3 (in the absence of poly(L-lysine)) also induces SERRS effects. However, these were much weaker than observed in the presence of poly(L-lysine). SERRS on TLC Plates. To optimize the deposition and SERRS identification of anionic dyes, preliminary experiments were performed by manual deposition of the analytes, silver sol, poly(L-lysine), and other additives on aluminum oxide and silica TLC plates using micropipets. Aluminum Oxide. On aluminum oxide, good SERRS effects were observed for all anionic dyes of this study. However, slightly better SERRS signals could be observed when using the sequence silver sol, poly(L-lysine), and dye. Presumably, due to strong interaction between the silver particles and the hydrophilic aluminum oxide,21 the silver particles are less spread out, which leads to regions with a high density of these particles. Another point to be noticed is the gradual disappearance of the signal as a function of time after addition of silver sol. This has been shown in Figure 5 traces a-c. In line with the observations of Pe´rez et al., who used filter paper as a substrate,22 it was found that wetting of the TLC material by simply adding a drop of water is quite (19) Munro, C. H.; Smith, W. E.; White, P. C. Analyst 1995, 120, 993-1003. (20) Munro, C. H.; Smith, W. E.; Armstrong, D. R.; White, P. C. J. Phys. Chem. 1995, 99, 879-885. (21) Soper, S. A.; Ratzlaff, K. L.; Kuwana, T. Anal. Chem. 1990, 62, 1438-1444. (22) Pe´rez, R.; Ruperez, A.; Laserna, J. J. Anal. Chim. Acta 1998, 376, 255263.
5722 Analytical Chemistry, Vol. 72, No. 22, November 15, 2000
Figure 5. SERRS spectra of manually deposited Food Yellow 3 on aluminum oxide recorded after (a) 0, (b) 10, and (c) 40 min and (d) following addition of water after 40 min. Laser excitation at 514.5 nm.
useful. It leads to a partial restoration of signals as shown in Figure 5d. Similar results were obtained for other dyes. This observation illustrates why it is advantageous to carry out SERRS on a TLC plate rather than in aqueous solutions. In the latter case, SERS or SERRS intensities disappear in a very short time because of the tendency of the colloids to precipitate.22 Silica. Contrary to the results obtained for aluminum oxide, initially no SERRS effects were observed on a silica plate. This was rather disappointing, since silica is the substrate of choice from the deposition point of view. Fortunately, the problem could be solved satisfactorily. The application of a drop of 1 M HNO3 on an analyte spot prior to (but not after) the application of the silver sol immediately provided intense SERRS signals for all anionic dyes of this study, even in the absence of poly(L-lysine). The latter aspectsno need to use poly(L-lysine)sis of distinct importance because it opens the possibility to develop a very simple and robust at-line LC-SERRS technique to identify anionic dyes. Although we do not have a full explanation for the role of HNO3, one may speculate that at low pH the negative charge of the silica particles and, consequently, the repulsion experienced by the silver particles is decreased. Furthermore, it has been shown that HNO3 can induce aggregation of the silver colloids.20 LC-TLC-SERRS. From the results presented in previous sections, it is obvious that silica is the substrate of choice in atline LC-(TLC)-SERRS. It can accommodate LC effluents with high water content, and high-quality SERRS spectra of the anionic dye are recorded if HNO3 is applied prior to the silver sol. Figure 6 shows the spectra for the six anionic dyes, using 457.9- and 514.5-nm excitation. The spectra show a strong similarity with those of Figure 4. The similarity of the spectra in Figure 4 (in
Figure 6. At-line LC-(TLC)-SERRS spectra of the six anionic dyes. The spectra were recorded at 457.9 (lower trace) and 514.5 nm (upper trace). Dye concentration, 25 µg mL-1 injected on LC column; ∼40 ng deposited on TLC plate. The SERRS signals were collected using five accumulation of 1 s each.
aqueous phase) and in Figure 6 (after deposition) underlines the analyte identification power of the method at hand. Apparently, the deposition process and the interaction of the analytes with the silica matrix do not disturb the spectral shapes. The use of nitric acid instead of poly(L-lysine) does hardly cause changes in vibrational structures; the only exception is dye Food Red 1 (CI 14700) for which significant differences are recorded. For example, upon addition of poly(L-lysine), a strong band is observed at 1358 cm-1, which is absent if nitric acid is applied. It is of course interesting to estimate the detection limits obtained with the LC-(TLC)-SERRS technique. Ideally, these should not differ too much from the detection limits normally
achieved in UV absorbance detection, so that both quantitative and qualitative (structural) information can be obtained. The spectra of Figure 6 were recorded using an injected dye concentration of 25 µg mL-1. With an injection volume of 20 µL and a split ratio of 8%, the total amount of each dye on the TLC plate therefore was ∼40 ng, so that the identification limits were in the range 10-20 ng. It should be emphasized that these amounts were dispersed on the silica plate in spots of ∼2 mm2. Only a minor fraction of such a spot was irradiated by the laser beam, which had a cross section of ∼10 µm2 when the 40 × microscope objective was used. In fact, the spectra recorded represent no more than a few picograms of dye and it will be clear that redesigning Analytical Chemistry, Vol. 72, No. 22, November 15, 2000
5723
the setup to make better use of this excellent detect ability will have a high priority in future studies. CONCLUSION At-line LC-(TLC)-SERRS was found to be a useful technique for the identification of both anionic (this paper) and cationic dyes.14 For the anionic dyes, published ion-pair LC separations can be adopted without any modification, also if gradients with high water contents are applied. The spray-jet assembly used as an interface can handle flow rates of up to ∼60 µL min-1. SERRS spectra can be recorded without any problem. For both types of dye, silica plates are the substrate of choice. It should be noted that, in separation as dealt with in the present paper, the use of SERRS has some distinct advantages over Fourier transform infrared (FT-IR) spectroscopy. Higher (variable) eluent water contents and flow rates can be applied since silica TLC is much more amenable to water than ZnSe, the substrate material used in at-line LC-FT-IR. Furthermore, in contrast to FT-IR, ion-pair reagents can be used without the involvement of a postcolumn liquid-liquid extraction device.23 The ion-pair reagent necessary for the LC separation of the dyes does not disturb the analyte deposition on the plate or the recording of the SERRS spectra. (23) Somsen, G. W.; Gooijer, C.; Brinkman, U. A. Th. J. Chromatogr., A 1999, 856, 213-242.
5724
Analytical Chemistry, Vol. 72, No. 22, November 15, 2000
Only one additional action is required, i.e., the addition of a drop of strong acid (1 M HNO3) prior to the application of silver sol. In future studies, the improvement of sensitivity will have a high priority. It is expected that the on-line coupling of solid-phase extraction (SPE) techniques and LC to achieve SPE-LC-(TLC)SERRS should enable the introduction of milliliters of sample, rather than the 20-µL volumes presently applied, so that the detection limits in the nanogram per milliliter instead of microgram per milliliter range can be reached. As a spinoff, the present paper also gives relevant information regarding the development of TLC-SERRS procedures, i.e., on how to interface TLC and Raman spectroscopy. To all probability, silica will be a better substrate than aluminum oxide here also, because of the more compact spots and consequently improved analyte performance. ACKNOWLEDGMENT We thank Drs. J. W. Wegener (Institute of Environmental Studies, IVM, Free University, Amsterdam) for supplying the anionic dyes and providing useful background information. Received for review May 3, 2000. Accepted August 22, 2000. AC000514G