A Reproducible Surface-Enhanced Raman Spectroscopy Approach

Online SERS. Measurements in a Segmented. Microfluidic System. Katrin R. Strehle,† Dana Cialla,† Petra Ro1sch,† Thomas Henkel,‡ Michael Ko1hle...
0 downloads 12 Views 230KB Size

Anal. Chem. 2007, 79, 1542-1547

A Reproducible Surface-Enhanced Raman Spectroscopy Approach. Online SERS Measurements in a Segmented Microfluidic System Katrin R. Strehle,† Dana Cialla,† Petra Ro 1 sch,† Thomas Henkel,‡ Michael Ko 1 hler,§ and Ju 1 rgen Popp*,†,‡

Department for Physical Chemistry, Friedrich-Schiller-University Jena, Helmholtzweg 4, D-07743 Jena, Germany, Institute for Physical High Technology, Albert-Einstein-Strasse 9, D-07745 Jena, Germany, and Institut fu¨r Physik, Technische Universita¨t Ilmenau, Weimarer Strasse 32 (Faradaybau), D-98693 Ilmenau, Germany

The application of a liquid/liquid microsegmented flow for serial high-throughput microanalytical systems shows promising prospects for applications in clinical chemistry, pharmaceutical research, process diagnostics, and analytical chemistry. Microscopy and microspectral analytics offer powerful approaches for the analytical readout of droplet based assays. Within the generated segments, individuality and integrity are retained during the complete diagnostic process making the approach favored for analysis of individual microscaled objects like cells and microorganisms embedded in droplets. Here we report on the online application of surface-enhanced microRaman spectroscopy for the detection and quantization of analytes in a liquid/liquid segmented microfluidic system. Data acquisition was performed in microsegments down to a volume of 180 nl. With this approach, we overcome the well-known problem of adhesion of colloid/ analyte conjugates to the optical windows of detection cuvettes, which causes the so-called “memory effect”. The combination of the segmented microfluidic system with the highly sensitive SERS technique reaches in a reproducible quantification of analytes with the SERS technique. The application of narcotics, for example, is a difficult problem as the dose delivered should be kept as small as possible but has to be permanently kept over a special threshold. Therefore, the in-blood concentration of such drugs has to be controlled constantly. A microfluidic device where a minimal sample volume is needed seems to be an appropriate solution to monitor fluctuations of the in-blood concentrations of narcotics. For the detection of very low concentrations, surface-enhanced Raman spectroscopy (SERS) has been proven to be a powerful method.1-5 Compared to normal Raman spectroscopy, the signal intensity is significantly increased due to the interaction between * Corresponding author: (e-mail) [email protected]. † Friedrich-Schiller-University Jena. ‡ Institute for Physical High Technology. § Technische Universita¨t Ilmenau. (1) Petry, R.; Schmitt, M.; Popp, J. ChemPhysChem 2003, 4, 14-30. (2) Gessner, R.; Ro ¨sch, P.; Petry, R.; Schmitt, M.; Strehle, M. A.; Kiefer, W.; Popp, J. Analyst 2004, 129, 1193-1199.

1542 Analytical Chemistry, Vol. 79, No. 4, February 15, 2007

the surface plasmons of colloidal metal structures, analyte molecules, and incident laser light.6-8 In combination with the resonance Raman effect, which additionally enhances the signal, single-molecule detection can be achieved.5,9,10 Although this is a promising way, a quantitative assessment by means of SERS is difficult.11,12 The SERS signal intensity strongly varies depending on the size, shape, and aggregation behavior of the used colloid.13-15 In static SERS experiments, one often has to search for “hot spots”, which are positions of a drastically increased SERS signal compared to the rest of the probe volume, in an inhomogeneous solution. To overcome this problem, the implementation of flow cells is a promising way. At the beginning analyte, colloidal solution and aggregation agent were brought into a mixing chamber, where they were thoroughly mixed before being directed to a sample cell for detection.16-18 With this method, a relatively high amount of sample volume is necessary. Recently efforts have been made to reduce the required amount of the sample solution by the design (3) Gessner, R.; Ro ¨sch, P.; Kiefer, W.; Popp, J. Biopolymers 2002, 61, 327330. (4) Nie, S.; Emory, S. R. Science 1997, 275, 1102-1106. (5) Kneipp, K.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S. Springer Ser. Chem. Phys. 2001, 67, 144-160. (6) Campion, A.; Kambhampati, P. Chem. Soc. Rev. 1998, 27, 241-250. (7) Smith, W. E.; Rodger, C. In Handbook of Vibrational Spectroscopy; Chalmers, J. M., Griffiths, P. R., Eds.; Wiley & Sons: Chichester, 2002; Vol. 1, pp 775784. (8) Vo-Dinh, T.; Stokes, D. L. In Handbook of Vibrational Spectroscopy; Chalmers, J. M., Griffiths, P. R., Eds.; Wiley & Sons: Chichester, 2002; Vol. 2, pp 1302-1317. (9) Kneipp, K. Single Mol. 2001, 2, 291-292. (10) Kneipp, K.; Kneipp, H.; Itzkan, I.; Dasari, R. R.; Feld, M. S. Chem. Rev. 1999, 99, 2957-2975. (11) Jones, J. C.; McLaughlin, C.; Littlejohn, D.; Sadler, D. A.; Graham, D.; Smith, W. E. Anal. Chem. 1999, 71, 596-601. (12) Schneider, S.; Grau, H.; Halbig, P.; Nickel, U. Analyst 1993, 118, 689694. (13) Alvarez-Puebla, R. A.; Arceo, E.; Goulet, P. J. G.; Garrido, J. J.; Aroca, R. F. J. Phys. Chem. B 2005, 109, 3787-3792. (14) Emory, S. R.; Haskins, W. E.; Nie, S. J. Am. Chem. Soc. 1998, 120, 80098010. (15) Mock, J. J.; Barbic, M.; Smith, D. R.; Schultz, D. A.; Schultz, S. J. Chem. Phys. 2002, 116, 6755-6759. (16) Berthod, A.; Laserna, J. J.; Winefordner, J. D. Appl. Spectrosc. 1987, 41, 1137-1141. (17) Ni, F.; Sheng, R.; Cotton, T. M. Anal. Chem. 1990, 62, 1958-1963. (18) Sheng, R.; Ni, F.; Cotton, T. M. Anal. Chem. 1991, 63, 437-442. 10.1021/ac0615246 CCC: $37.00

© 2007 American Chemical Society Published on Web 01/20/2007

Figure 1. Molecular structure of crystal violet.

of chip-based mixing chambers, where an optical detection window is implemented.19-21 The new lab-on-a-chip microfluidics mainly aim at the detection of dye-labeled oligonuclotides.22 However, water pollutants like cyanide are also detected in such microsystems.23 McLaughlin et al. succeeded in the quantitative detection of mitoxantrone, a chemotherapeutic agent, in human serum by applying a microfluidic system.21 A problem with the miniaturization of the fluidic system is the adhesion of both, colloidal nanoparticles as well as analyte molecules, to the channel walls. For these problems, two possible solutions are feasible. The surface of glass-fabricated chips can be regenerated through rinsing with nitric acid or piranha solution, which dissolves the sedimented aggregates.21 A different approach comprises the fabrication of comparable low-cost poly(dimethylsiloxane) chips via soft lithography, which are disposable.20,23 In this contribution, we report on a highly reproducible SERS signal for a quantitative detection of crystal violet in a liquid/liquid microsegmented flow system.24-26 To the best of our knowledge, these are the first SERS experiments in a segmented microfluidic device, where sample volumes of ∼180 nL were investigated. The adhesion of nanoparticles to glass walls of the cell, which causes the so-called “memory effect”, is prevented, as the aqueous droplet containing the analyte solution is conducted in a stream of lipophilic tetradecane. Within such droplets, sample volumes down to a few nanoliters can be processed in a reliable way. Individuality and integrity of these volumes are retained during the complete diagnostic process, making the approach favored for analysis of individual microscaled objects such as cells and microorganisms embedded in droplets. EXPERIMENTAL SECTION Chemicals and Reagents. For testing the new cell concept, crystal violet (Sigma-Aldrich) was selected, as the signal of this aromatic dye (see Figure 1) can be strongly enhanced when there is a SERS-active metal substrate. A stock solution of 1 × 10-4 M was made to prepare the measured sample solutions by dilution. Gold Colloid. As a SERS-active metal substrate, a colloidal gold solution was used (λmax ) 550 nm), prepared according to a (19) Anderson, M. S. Anal. Chem. 2005, 77, 2907-2911. (20) Liu, G. L.; Lee, L. P. Appl. Phys. Lett. 2005, 87, 074101-074103. (21) McLaughlin, C.; MacMillan, D.; McCardle, C.; Smith, W. E. Anal. Chem. 2002, 74, 3160-3167. (22) Docherty, F. T.; Monaghan, P. B.; Keir, R.; Graham, D.; Smith, W. E.; Cooper, J. M. Chem. Commun. 2004, 118-119. (23) Yea, K.-h.; Lee, S.; Kyong, J. B.; Choo, J.; Lee, E. K.; Joo, S.-W.; Lee, S. Analyst 2005, 130, 1009-1011. (24) Atencia, J.; Beebe David, J. Nature 2005, 437, 648-655. (25) Zheng, B.; Tice, J. D.; Ismagilov, R. F. Anal. Chem. 2004, 76, 4977-4982. (26) Tice, J. D.; Song, H.; Lyon, A. D.; Ismagilov, R. F. Langmuir 2003, 19, 9127-9133. (27) Lee, P. C.; Meisel, D. J. Phys. Chem. 1982, 86, 3391-3395.

modified Lee and Meisel procedure.27 The colloid was aged for a few months. In the UV/visible absorption spectrum (see Supporting Information Figure S1), the formation of a shoulder around 650 nm can be observed. When aggregation occurs, the intensity of the SERS signal arises strongly, since the excitation laser wavelength (633 nm) is now in resonance with the surface plasmon frequency of the metal colloid.28,29 The addition of crystal violet did not cause any further aggregation. Flow Cell. For detection and generation of segmented sample streams, an all glass microfluidic device, dimensions 25 × 16 mm, was mounted on the microscope stage. The device includes facilities for segment generation and dosing operations. Microchannels are prepared by isotropic wet etching of the substrates (Borofloat33, Schott Jena) with hydrofluoric acid using a nickel chromium metallization as mask. Two identical half-channels were bonded face to face by the process of anodic bonding, using a 100-nm silicon layer as bond support layer. Injectors for dosing operations and segment generation are equipped with nozzles. The general design is based on results of parameter and geometry studies.30 The microchannels used in these experiments have a main channel height of 270 µm, a channel width of 570 µm, and cross section of 0.138 mm2. Nozzles of the injector structure are prepared with a cross section of 0.017 mm2. An image of the chip module and the microscope slide-sized chip mount is shown in Figure 2A, and a schematic depiction of the setup is presented in Figure 2B. A ceDoSys4-channel highprecision syringe pump system (Cetoni GmbH) was used for the automated, independent control of the required three flow channels. Standard 1/16-in. HPLC capillaries (PTFE) with an inner diameter of 0.5 mm and common HPLC fittings and ferrules were used for interconnection of the chip device at the syringe pump. In order to provide the required optimum wetting conditions for the separation medium, the internal surfaces of the microchannels were C18-functionalized by treatment with a solution of octadecyltrichlorosilane in anhydrous n-heptane at 50 °C for 3h. To remove redundant octadecyltrichlorsilane, the cell was extensively washed with n-heptane, 2-propanol, and water before being used. Instrumentation. SERS spectra were recorded with a conventional micro-Raman setup (LabRam Invers, Horiba Jobin-Yvon), consisting of a He/Ne laser (Coherent) operating at 633 nm and a spectrometer with a focal length of 800 mm, equipped with a 300 lines/mm grating. The incident laser power on the sample was ∼2 mW. The scattered light was detected by a CCD camera operating at 220 K. UV/visible absorption spectra were recorded with a Cary5000UV-vis-NIR spectrophotometer (Varian). To minimize the required sample volume, a fiber-optical microcell was used (TrayCell, Hellma). Spectra were processed with the LabSpec software provided by Horiba Jobin-Yvon. (28) Faulds, K.; Littleford, R. E.; Graham, D.; Dent, G.; Smith, W. E. Anal. Chem. 2004, 76, 592-598. (29) Garcia-Ramos, J. V.; Sanchez-Cortes, S. J. Mol. Struct. 1997, 405, 13-28. (30) Henkel, T.; Bermig, T.; Kielpinski, M.; Grodrian, A.; Metze, J.; Ko ¨hler, J. M. Chem. Eng. J. (Amsterdam, Netherlands) 2004, 101, 439-445.

Analytical Chemistry, Vol. 79, No. 4, February 15, 2007


Figure 2. (A) Used chip module on the microscope slide-sized chip mount with HPLC capillaries connected to the chip; (B) schematic diagram of the flow cell setup.

RESULTS AND DISCUSSION Measuring Principle. A schematic representation of the liquid/liquid microsegmented flow device is given in Figure 2B. Via connector a, the channel of the microchip is filled with lipophilic tetradecane. The flow velocity of the oil was set to 0.1 µL/s. The crystal violet solution is inserted dropwise via connector b to this constant stream of tetradecane with a reduced flow rate of 0.02 µL/s. Afterward, the crystal violet-containing droplet is combined with the SERS-active substrate, which is introduced via connector c with a flow rate of 0.02 µL/s. In this case, the volume of the combined droplets is ∼180 nL each, which can be calculated from the given dimensions of the microchannel and a droplet length of 1.5 mm.30 The size of the microdroplets may be controlled by the flow rates of the fluids, resulting in droplets with a volume between 60 and 180 nL.30 For regular droplet formation, the flow rates were kept constant. One has also to pay attention that the tubes are free of air bubbles as the presence of air bubbles inhibits the regular formation of the analyte-containing droplets, which is necessary for a quantitative detection. Raman spectra were taken consecutively with the incident laser beam perpendicular to the glass surface and with an acquisition time of 1 s for each spectrum. The signal detection was placed at the end of the microchannel, to increase the contact time between colloid and analyte. Preliminary studies showed an inhomogeneous signal distribution within the channel of the cell, so the focus of the laser beam was kept in the same position in x, y, and z directions. Therefore, fluctuations in the signal intensity caused by an irregular distribution of the colloidal particles within the droplets could be eliminated. Figure 3 illustrates the applied measuring principle. Spectrum A was taken in the oil phase showing a Raman spectrum of tetradecane. The two prominent bands at wavenumber positions around 1303 and 1442 cm-1 result from -(CH2)n- in phase twisting modes and antisymmetrical -CH3 bending modes, respectively.31 Spectrum B displays the SERS spectrum of crystal violet taken in an aqueous droplet. The strong band at 1621 cm-1 is caused by an in-plane C-C stretching vibration of the ring (νip(ring-CC)). The broadened peak at 1381 cm-1 might be the result of an overlap of an in-plane C+-C stretching vibration (νip(C+-C)) and an in-plane stretching vibration between nitrogen atoms and the phenyl rings (νip(N-Ph)). Peaks at wavenumber positions at 918 and 807 cm-1, respectively, evoke from out-of-plane deformation 1544

Analytical Chemistry, Vol. 79, No. 4, February 15, 2007

vibrations of the phenyl rings (δoop(ring-C-H)) in comparison to out-of-plane deformation vibrations of the Ph-C+-Ph bond at 443 and 425 cm-1 (δoop(Ph-C+-Ph)). The marked band at 1177 cm-1, used to calculate the integrated Raman intensities, results from an in-plane deformation mode of the ring (δip(ring-C-H)).32,33 The left panel of Figure 3 shows the integrated Raman intensity between 1130 and 1217 cm-1 plotted against the measuring time. One can clearly distinguish between a crystal violet-containing segment and the tetradecane, which shows no bands in the plotted wavenumber region. A regularly alternating pattern can be seen, proofing a constant droplet formation within the microchip. With a flow rate of 0.02 µL/s at a measuring time of ∼900 s for each series of measurements and 20 µL of dead volume, 40 µL of colloidal and sample solution is necessary for a quantitative result. Reproducibility. Figure 4 shows the reproducibility of the SERS signal for a time period of 800 s. Within this period ∼80 crystal violet-containing segments are generated and measured in the cell, interrupted by the tetradecane. The intensity of the integrated Raman signal, which is plotted against the measuring time, fluctuates only in a small range (4.9%), resulting from the unsteady formation of aggregates. For each concentration, a series of measurements over 15 min was performed. In order to quantitatively determine the concentration of the investigated solution, an average value and the standard deviation of the peak height of the integrated Raman intensity are calculated. Memory Effect. Preliminary studies showed that when the colloid/analyte mixture is pumped through the microchannel without any separation medium, the so-called memory effect can be observed. In this case, a SERS spectrum of crystal violet could be detected even when the flow of the analyte solution was stopped, caused by the photodeposition of analyte/colloid conjugates to the channel walls. This could be proved, when investigating the glass walls of the used microchips with dark field illumination. In Figure 5, the integrated Raman intensity is plotted against the measuring time when the flow of crystal violet is stopped during the measuring period leading to an abrupt stop of the regular oscillations after 80 s. When recording a Raman spectrum (31) Lin-Vien, D.; Colthup, N. B.; Fateley, W. G.; Grasselli, J. G. The Handbook of Infrared and Raman Characteristic Frequencies of Organic Molecules; Academic Press: San Diego, 1991. (32) Gicquel, J.; Carles, M.; Bodot, H. J. Phys. Chem. 1979, 83, 699-706. (33) Liang, E. J.; Ye, X. L.; Kiefer, W. J. Phys. Chem. A 1997, 101, 7330-7335.

Figure 3. (Left panel) Integrated Raman intensity in the wavenumber region between 1130 and 1217 cm-1 plotted against the measuring time. (Right panels) Spectra A and B were taken at the marked positions, showing bands resulting from tetradecane and a SERS spectrum of crystal violet, respectively.

Figure 4. Integrated Raman intensity in the wavenumber region between 1130 and 1217 cm-1. The intensity is nearly constant over a measuring period of 800 s, showing the high reproducibility that can be reached with the segmented flow setup.

in a colloid containing only water droplet (position A), neither a SERS signal nor a tetradecane spectrum can be seen (see spectrum A depicted in Figure 5). This behavior proves the fact that due to the oil film, which surrounds the water droplet, no

adhesion to the channel walls of either the gold nanoparticles or the analyte molecules occurs. Applying this concept, the problem of the memory effect when using microfluidic channels can be solved. Analytical Chemistry, Vol. 79, No. 4, February 15, 2007


Figure 5. Stopping the flow of crystal violet during the measuring period. The peaks representing the integrated Raman intensity abruptly disappears (see left panel). In the water segments (position A), which now consist of pure colloid without any crystal violet, neither a SERS spectrum nor a tetradecane signal can be detected.

Figure 7. Mean values of the integrated Raman intensity plotted against the concentration of the crystal violet solution showing a linear dependence.

Figure 6. Representative Raman spectra of different crystal violet concentrations ranging from 1 × 10-5 (A) to 1 × 10-6 M (I).

Quantitative Analysis. For a quantitative analysis, sample solutions with different concentrations of crystal violet ranging from 1 × 10-5 to 1 × 10-6 M were measured. The crystal violet solution is diluted in the droplet so that the correct concentrations of crystal violet with the measured droplets range from 5 × 10-7 to 5 × 10-6 M. Figure 6 depicts one representative SERS spectrum for each concentration. The SERS spectra are arranged according to a decreasing concentration of crystal violet from top to bottom. The intensity of the Raman peak at a wavenumber position of 1177 cm-1 was used for the quantification. One can see that with an increasing concentration of crystal violet the fluorescence background arises as well. Therefore, an automatically generated baseline correction was performed for each spectrum. The barriers for the baseline were set at 1130 and 1217 cm-1, a linear baseline generated, and afterward subtracted. The integrated Raman intensity in this wavenumber region is then plotted against the measuring time, resulting in a regular oscillation of the signal with the time (see Figure 4). The peak heights of these oscillations 1546 Analytical Chemistry, Vol. 79, No. 4, February 15, 2007

were used for quantification. As the integrated Raman intensity fluctuates slightly for different droplets (see Figure 4), which is the result of an irregular formation of aggregates, the mean value was calculated of all droplets measured during a time period of ∼900 s. Plotting the peak height against the concentration of the respective crystal violet concentration, a linear dependence should arise. Figure 7 shows the calculated mean values of the integrated Raman intensity as a function of the crystal violet concentration. The regression coefficient of the linear regression has a value of R2 ) 0.9823. Standard deviations are calculated on the basis of the integrated Raman intensity and are also given in Figure 7. CONCLUSION AND OUTLOOK In the presented article, we showed the successful realization of highly reproducible SERS experiments in a liquid/liquid microsegmented flow system on the model compound crystal violet. This setup can now be applied to more relevant systems such as the detection of drugs, water pollutants, or food additives where an online detection is highly appreciated. The limit of detection can be reduced when the integration time is increased from 1 s to higher acquisition times. For the realization of longer integration times, a trigger system has to be implemented so that only the analyte-containing droplets are readout. Another important topic of the presented contribution is the fact that separated sample volumes down to 60 nL can be

investigated; this seems to be a promising technique for an analysis of microscaled objects like cells and microorganisms embedded in aqueous droplets. In a next step, we are planning a further miniaturization of the whole apparatus followed by an automation, to make the presented technique interesting for potential users in analytical and clinical chemistry, pharmaceutical research, and process diagnostics. ACKNOWLEDGMENT Bernd Ku¨stner (Institute for Physical Chemistry, University of Wu¨rzburg, Germany) is thanked for his friendly help concern-

ing the preparation of the gold colloid. Financial support of the BMBF/VDI (16V1989, SERIZELL) is gratefully acknowledged. 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 August 16, 2006. Accepted December 6, 2006. AC0615246

Analytical Chemistry, Vol. 79, No. 4, February 15, 2007