Anal. Chem. 2006, 78, 2859-2863
UV Excitation Thermal Lens Microscope for Sensitive and Nonlabeled Detection of Nonfluorescent Molecules Shinichiro Hiki,† Kazuma Mawatari,‡ Akihide Hibara,§ Manabu Tokeshi,†,‡ and Takehiko Kitamori*,†,‡,§
Institute of Microchemical Technology, KSP East207, 3-2-1 Sakado, Takatsu, Kawasaki, Kanagawa 213-0012, Japan, Microchemistry Group, Kanagawa Academy of Science and Technology (KAST), KSP East307, 3-2-1 Sakado, Takatsu, Kawasaki, Kanagawa 213-0012, Japan, Department of Applied Chemistry, Graduate School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo, Tokyo 113-8656, Japan
An ultrasensitive and nonlabeled detection method of nonfluorescent molecules on a microchip was developed by realizing a thermal lens microscope (TLM) with a 266nm UV pulsed laser as an excitation light source (UVTLM). Pulsed laser sources have advantages over continuous-wave laser sources in more compact size and better wavelength tuning, which are important for microchip-based analytical systems. Their disadvantage is difficulty in applying a lock-in amplifier due to the high (>104) duty ratio of pulse oscillation. To overcome this problem, we realized a quasi-continuous-wave excitation by modulating the pulse trains at ∼1 kHz and detecting the synchronous signal with a lock-in amplifier. The optimum pulse repetition frequency was obtained at 80 kHz, which was reasonable considering thermal equilibrium time. Furthermore, a permissible flow velocity in the range of 6.6-19.8 mm/s was found to avoid sensitivity decrease due to photochemical reactions and thermal energy dissipation. Under these conditions, we detected adenine aqueous solutions on a fused-silica microchip without labeling and obtained a sensitivity that was 350 times higher than that in a spectrophotometric method. The sensitivity was enough for detection on a microchip with an optical path length that was 2-3 orders shorter than that in conventional cuvettes. Finally, the UV-TLM method was applied to liquid chromatography detection. Fluorene and pyrene were separated in a microcolumn and detected in a capillary (50-µm inner diameter) with 150 times higher sensitivity than a spectrophotometric method. Our method provides highly sensitive and widely applicable detections for various analytical procedures and chemical syntheses on microchips. Miniaturization of analytical systems, called lab-on-a-chip or µ-TAS, has been viewed with great promise for chemical and * To whom all correspondence should be addressed: E-mail:
[email protected], Fax: +81-3-5841-6039. † Institute of Microchemical Technology. ‡ Kanagawa Academy of Science and Technology (KAST). § The University of Tokyo. 10.1021/ac051967u CCC: $33.50 Published on Web 03/10/2006
© 2006 American Chemical Society
biochemical analysis systems since the 1990s.1,2 The merits of miniaturization include reduced sample and reagent amounts, more effective reaction due to a large surface-to-volume ratio, smaller space requirements, and lower cost. These miniaturized systems require sophisticated microfabrication, microfluidics, and detection technologies; in particular, they put higher demands on the detection technology due to the small volume and short optical path length of the microchannels. Therefore, a sensitive detection method is desired in combination with wide applicability.3-9 Conventionally, laser-induced fluorescence (LIF) and spectrophotometric methods have been frequently utilized. The LIF method is known as one of the most sensitive optical detection methods and shows ultrahigh sensitivity at a single-molecule level under optimized conditions. However, its applicability is limited to fluorescent analytes. Derivatization reactions are not always available for analytes of interest, and they often interfere with quantification ability due to a matrix effect (change in the degree of derivatization).10 The spectrophotometric method is generally used due to its wide applicability. The problem is low sensitivity due to the 10-1000 times shorter optical path length in the microchannels than that in conventional cuvettes (mm∼cm scale). Thermal lens spectroscopy (TLS) is a promising method to overcome the low sensitivity of the spectrophotometric method.11,12 TLS measures thermal energy generated via nonradiative photo(1) Manz, A.; Graber, N.; Widmer, N. M. Sens. Actuators, B 1990, 1, 244248. (2) Harrison, D. J.; Manz, A.; Fan, Z.; Ludi, H.; Widmer, M. H. Anal. Chem. 1992, 64, 1926-1932. (3) Hisamoto, H.; Horiuchi, T.; Uchiyama, K.; Tokeshi, M.; Hibara, A.; Kitamori, T. Anal. Chem. 2001, 73, 1382-1386. (4) Mawatari, K.; Kitamori, T.; Sawada, T. Anal. Chem. 1998, 70, 5037-5041. (5) Tokeshi, M.; Minagawa, T.; Uchiyama, K.; Hibara, A.; Sato, K.; Hisamoto, H.; Kitamori, T. Anal. Chem. 2002, 74, 1565-1571. (6) Sato, K.; Yamanaka, M.; Takahashi, H.; Tokeshi, M.; Kimura, H.; Kitamori, T. Electrophoresis 2002, 23, 734-739. (7) Hibara, A.; Saito, T.; Kim, H.; Tokeshi, M.; Ooi, T.; Nakao, M.; Kitamori, T. Anal. Chem. 2002, 74, 6170-6176. (8) Kikutani, Y.; Hisamoto, H.; Tokeshi, M.; Kitamori, T. Lab Chip 2004, 4, 328-332. (9) Hibara, A.; Iwayama, S.; Matsuoka, S.; Ueno, M.; Kikutani, Y.; Tokeshi, M.; Kitamori, T. Anal. Chem. 2005, 77, 943-947. (10) Ragozina, N.; Putz, M.; Heissler, S.; Faubel, W.; Pyell, U. Anal. Chem. 2004, 76, 3804-3809. (11) Dovichi, N. J.; Harris, J. M. Anal. Chem. 1979, 51, 728-731. (12) Bialkowski, S. E. Photothermal Spectroscopy Methods for Chemical Analysis; Wiley: New York, 1996.
Analytical Chemistry, Vol. 78, No. 8, April 15, 2006 2859
thermal relaxation after light absorption. For the first time, we realized a thermal lens microscope (TLM) for sensitive detection in a microspace, and TLM was proved to detect 0.4 molecule in a 7-fL detection volume.13 TLM has wide applicability comparable to the spectrophotometric method and has been coupled with a microchip for applications to various analytical fields such as environmental analysis, medical diagnosis, and single-cell imaging.14 In addition, the TLM system has been miniaturized to desktop- and palmtop-sized systems for practical use.15,16 In TLS, a visible laser has been used as an excitation light source due to its easy operation, small size, and inexpensive optical component requirement. However, for further effective application of miniaturized TLS, a UV laser excitation is desired for detection of various chemicals and biomolecules without labeling. For example, Ragozina et al. reported a crossed-beam TLS with continuous-wave (CW) excitation (wavelength 257 nm) and applied it to capillary electrophoresis detection.10,17 They reported that the sensitivity was 30 times higher than that with a UV spectrophotometer. However, the crossed-beam geometry cannot be applied to measurements on microchips and needs laborious optical adjustments. In addition, a UV pulsed laser has several advantages over CW lasers, such as miniaturization and better wavelength tuning over a wide range, which promote its use. However, pulsed lasers have a very small pulse width (ns to fs) and very small duty ratio, which might be disadvantageous for synchronous detection by a lock-in amplifier. Furthermore, pulse-to-pulse reproducibility is ordinarily not so high (typically ∼5%) while shortterm stability control is a millisecond order. In this paper, we developed a novel TLM method with a pulsed UV laser (UV-TLM) for sensitive detection of nonfluorescent molecules without labeling. A pulsed laser (wavelength 266 nm) was used as an excitation laser beam. A quasi-continuous-wave (QCW) excitation mode18 was combined with UV-TLM to get a high signal-to-noise ratio with a lock-in amplifier, and the optimum pulse repetition frequency was determined. The pulsed laser is expected to induce photochemical reactions when the excitation beam is tightly focused on samples. Therefore, we investigated the dependence of sensitivity on flow velocity and laser power of excitation beam and determined the permissible range of flow velocity. Aqueous solutions of adenine were measured on a microchip by the UV-TLM method without labeling under the determined conditions. The sensitivity was compared with that of spectrophotometry, and a 2-order higher sensitivity was confirmed. Finally, we applied UV-TLM for HPLC detection and verified that the UV-TLM method had applicability to conventional separation systems. EXPERIMENTAL SECTION Principle of QCW mode. Figure 1 illustrates waveforms of the thermal lens signals with CW laser (left column) and pulsed (13) Tokeshi, M.; Uchida, M.; Hibara, A.; Sawada, T.; Kitamori, T. Anal. Chem. 2001, 73, 2112-2116. (14) Kitamori, T.; Tokeshi, M.; Hibara, A.; Sato, K. Anal. Chem. 2004, 76, 53A60A. (15) http://www1.odn.ne.jp/imt/index.html. (16) Tokeshi, M.; Yamaguchi, J.; Hattori, A.; Kitamori, T. Anal. Chem. 2005, 77, 626-630. (17) Ragozina, N.; Heissler, S.; Faubel, W.; Pyell, U. Anal. Chem. 2002, 74, 4480-4487. (18) Ito, K.; Makiko M.; Harata, A.; Kitamori, T. Chem. Phys. Lett. 1997, 275, 349-354.
2860
Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
Figure 1. Principle of conventional CW laser excitation and QCW pulsed laser excitation for thermal lens signal generation.
laser (right column) excitations. In conventional TLM, the CW laser beam is modulated at ∼1 kHz. The rise and fall of the thermal lens signal presented in Figure 1 correspond to on and off of the excitation irradiation, respectively. The rise and fall times are related to a characteristic thermal time constant Tc expressed as12
Tc ) w2/4D
(1)
where w is the excitation beam radius in the sample volume determined by numerical aperture (NA) of an objective lens and D is solution thermal diffusivity. The time to achieve equilibrium is roughly estimated by 25Tc.12 When we assume a NA of 0.4, which corresponds to w of 0.4 µm, and D of 1.4 × 10-7 m2/s, 25Tc is calculated to be 7 µs. When the time interval between the pulses is set shorter than this value, a similar time course of the thermal lens response to that of CW laser excitation is expected to be observed (Figure 1). When the time course is detected by a lock-in amplifier, a synchronous signal intensity with the modulation frequency (∼1 kHz) can be obtained. Thus, we can use the pulsed laser beam with the QCW mode. UV-TLM System. The basic principle of TLM was explained previously.19 Here, we briefly describe the experimental setup of the present UV-TLM system (Figure 2). The UV-TLM system was constructed by modification of a desk-top TLM system (Institute of Microchemical Technology, ITLM-10). A UV laser (Cyber Laser Inc., SPICA266-HR; wavelength 266 nm, pulse repetition frequency 1-80 kHz, pulse width 18-25 ns) was used as an excitation beam. The power in the QCW mode was set at 4.3 mW under an objective lens unless otherwise mentioned. The excitation beam was modulated at ∼1 kHz with a light chopper (NF Electronic Instruments, 5584A). Two lenses were used for expanding the diameter of the excitation beam by three times magnification and adjusting the focus position under an objective lens by setting the distance between the lenses. Then, the excitation beam was introduced into the microscope (Nikon, EclipseE600). A laser diode (Coherent Japan Inc., VLM2-3RL; wavelength 670 nm) fixed on the microscope was used as a probe beam. These two beams were aligned coaxially by a dichroic mirror and introduced into an objective lens (Nikon Engineering Corp.; UV objective lens, 50 times magnification, NA 0.4). The focal length for the excitation beam was made shorter by ∼20 µm than that for the probe beam because of sensitivity enhancement.13 The objective lens focused the laser beams into a sample solution where the thermal lens (19) Uchiyama, K.; Hibara, A.; Kimura, H.; Sawada, T.; Kitamori, T. Jpn. J. Appl. Phys. 2000, 39, 5316-5322.
Figure 2. Illustration of UV-TLM system, microchip and HPLC system.
effect was induced, and the probe beam was converged by the effect. The two transmitted beams were collimated by a condenser lens. Just the probe beam was reflected by another dichroic mirror and introduced into a photodiode through a cut filter for the excitation beam and a pinhole. All optical components were designed to minimize absorption of UV light resulting large background signals. The amplitude and phase signals from a lockin amplifier were recorded as the thermal lens signal by a computer through GPIB interface and analyzed by LabVIEW program (National Instruments Japan Corp.). A fused-silica microchip (Institute of Microchemical Technology; microchannel size 100 µm wide × 45 µm deep) was utilized for basic evaluation of the performance. For HPLC application, a reversed-phase microcolumn (Micro-Tech Scientific Inc., ZM-5-C18SBX; i.d. 0.5 mm) with an injector (Valco Instruments Co. Inc., C14W.06; 60nL) was coupled to a detection capillary (Polymicro Technologies, LLC., WWP050375; i.d. 50 µm). The separated samples were detected in the capillary. For comparison of sensitivity, a spectrophotometric detector (Jasco Corp., CE971-UV) was used. Chemicals. Sunset yellow (SY) (266 nm ) 12 200 M-1cm-1, 532 nm ) 4700 M-1 cm-1) and adenine (266 nm ) 13 200 M-1 cm-1) from Wako Pure Chemical Industries were used for basic evaluation of the performances. They were prepared by stepwise dilution of stock solution. Fluorene (266 nm ) 16 330 M-1 cm-1, fluorescent quantum yield φ ) 0.7) and pyrene (266 nm ) 13 870 M-1 cm-1, φ ) 0.6) were used for HPLC applications because they are used as standard samples for investigating separation performance. The dissolved oxygen was removed by ultrasonic bath prior to use. RESULTS AND DISCUSSION QCW Mode. To verify the time course of the QCW mode thermal lens signal, a waveform was recorded with an oscilloscope as shown in Figure 3a, where an aqueous solution of 10-2 M SY was driven at 6.6 mm/s in the microchannel. The repetition frequency of pulse radiation was set at 80 kHz, and this gave the maximum signal intensity in ∼500 µs as CW excitation. To analyze the signal generation process, the frequency component was extracted by fast Fourier transform analysis as shown in Figure
Figure 3. Measurement of a dye solution by UV-TLM: (a) waveform and (b) fast Fourier transform analysis of the waveform.
3b. In addition to a peak at ∼1 kHz (the chopper modulation frequency), a clear peak was detected at 80 kHz corresponding to the pulse repetition frequency. The 80-kHz peak was observed just for the rise signals and was not observed for the fall signals. Thus, the basic principle of the QCW mode for UV-TLM was confirmed. Next, we investigated dependence of the signal intensity and noises on the repetition frequency in the range of 1-80 kHz with a lock-in amplifier. The results are shown in Figure 4. The output power of the excitation beam in the QCW mode was kept at 0.8 mW in the microchannel. The long-term noises (Figure 4b) were evaluated by measuring values of standard deviation in the signal intensity during 5-min measurement, and the short-term noises were obtained from the maximum fluctuation of the signal intensity. In Figure 4a, the sensitivity increased with the repetition frequency and leveled off at a higher frequency than 40 kHz. The Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
2861
Figure 5. Measurement of a dye solution by UV-TLM without flow.
Figure 4. Pulse repetition frequency dependence of (a) signal intensity, (b) signal-to-standard deviation (long-term noise), and (c) S/N (short-term noise).
value is the same order with the simple theoretical calculations in the former section (principle of QCW mode) considering small confocal length (∼2 µm) compared with the depth of the microchannel (45 µm) and expansion of beam diameter along the optical axis. The signal-to-noise ratio also increased in long-term (Figure 4b) and short-term (Figure 4c) noise evaluations. Therefore, 80-kHz repetition frequency was selected as an optimized condition from the viewpoints of S/N and the output power. Flow Condition. A tightly focused UV pulsed laser has a high power density in time and space, which induces nonlinear phenomena such as multiphoton absorption. The sample and solvent can be photochemically decomposed by the UV pulsed laser irradiation. Figure 5 shows a time course of the lock-in amplifier output signal of 1 × 10-4 M SY solution without flow. As expected, the signal immediately began to decrease after turning on the excitation beam. Some photochemical reactions could be induced by the tightly focused pulsed UV laser beam, and their occurrence decreased the thermal lens signal intensity. To optimize the sensitivity, we investigated flow condition for velocities of 0-34 mm/s. The result is shown in Figure 6. The thermal lens signal decreased with increasing flow velocity due to thermal energy dissipation by flow.20,21 On the other hand, the photochemical decomposition also reduced the signal but the (20) Burgi, D. S.; Nolan, T. G.; Risfelt, J. A.; Dovichi, N. J. Opt. Eng. 1984, 23, 756-758. (21) Burgi, D. S.; Dovichi, N. J. Appl. Opt. 1987, 26, 4665-4669.
2862 Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
Figure 6. Dependence of signal intensity on flow velocity: (a) 10-4 M SY solution and (b) pure water (blank).
decrease began at a lower flow velocity. Therefore, convolution of the photochemical reactions and thermal energy dissipation determined the sensitivity of the thermal lens signal as shown in Figure 6a. As a reference for the decomposition process, the thermal lens signal of pure water was measured (Figure 6b). The thermal lens signal of pure water simply decreased with the flow velocity by thermal energy dissipation because no photochemical reaction was induced in pure water. As a result, the permissible flow velocity was determined from the range of 6.6-19.8 mm/s, which gave a sensitivity decrease of 5% from the optimum sensitivity at 13.3 mm/s, and 6.6 mm/s was selected as the flow condition in this range. To investigate the signal generation mechanism, the thermal lens signal dependences on the excitation laser power were measured for the flow velocity of 6.6 mm/s. Figure 7a shows the thermal lens signal of 1 × 10-4 M SY solution as a function of the excitation laser power. The signal linearly depended on the laser power because the effect of photodecomposition was small in this flow condition. Figure 7b shows the thermal lens signal of pure water as a function of the excitation laser power. In contrast to signals for SY solution, the signal depended on the square of the laser power, and therefore, the thermal lens signal of water was
Figure 8. Detection on HPLC by UV-TLM: (a) fluorene (F) 1.0 × 10-3 M, pyrene (P) 8.0 × 10-4 M; (b) F 6.0 × 10-4 M, P 5.0 × 10-4 M; (c) F 3.0 × 10-4 M, P 3.0 × 10-4 M; (d) F 7.5 × 10-5 M, P 1.0 × 10-4 M; (e) F 5.0 × 10-5 M, P 5.0 × 10-5 M; and (f) pure water. Table 2. Comparison of Performance for HPLC Detection spectrophotometer Figure 7. Dependence of signal intensity on power of the excitation beam: (a) 10-4 M SY solution and (b) pure water (blank).
LQD for fluorene (M) LQD for pyrene (M)
10-5
8.2 × 7.8 × 10-5
UV-TLM 2.8 × 10-7 5.2 × 10-7
Table 1. Comparison of Detection Limits for Adenine Solutions
LOD concentration (M) absorbance (abs) LQD concentration (M) absorbance (abs)
spectrophotometer
UV-TLM
5.0 × 10-6 3.3 × 10-4
1.4 × 10-8 8.3 × 10-7
6.8 × 10-6 4.5 × 10-4
1.5 × 10-8 8.9 × 10-7
generated by two-photon absorption and the following photothermal process. Pure water has a strong absorption at a wavelength below 139.3 nm (>8.9 eV),22 and the present UV pulse having a wavelength of 266 nm (4.7 eV) could induce the two-photon absorption. Therefore, the background signal from an aqueous solution was found to be generated mainly by two-photon process. Evaluation of Performance. We evaluated the UV-TLM method performance under the investigated conditions using aqueous solutions of adenine in the range of 2.5 × 10-8-1.5 × 10-7 M. For comparison of sensitivity, adenine solutions were measured in the range of 1.0 × 10-5-1.0 × 10-4 M with a spectrophotometer. For UV-TLM detection, the signal intensity linearly increased with a correlation coefficient of 0.992. The limit of quantitative determination (LQD) was calculated from the value of twice the standard deviation (2σ) in the calibration curve. The lower limit of detection (LOD) was calculated from the conditions of S/N ) 2 in the time-dependent signal of 2.5 × 10-8 M. For comparison, the LOD and LQD with a UV spectrophotometer were evaluated utilizing a fused-silica capillary (50-µm i.d.). These values are summarized in Table 1. As a result, 350 and 450 times higher sensitivity in concentration was demonstrated. Therefore, high sensitivity of UV-TLM for microchip analysis was verified. Application to HPLC. We applied UV-TLM to a HPLC system detection to demonstrate the applicability of UV-TLM for conventional separation systems. Fluorene and pyrene were used as as (22) Crowell, R. A.; Bartels, D. M. J. Phys. Chem. 1996, 100, 17940-17949.
standard samples for performance evaluations. The repetition frequency and the flow velocity were kept as 80 kHz and 6.6 mm/ s, respectively. The concentrations were changed in the range of 3.0 × 10-7-6.0 × 10-6 M for fluorene and 2.4 × 10-7-4.9 × 10-6 M for pyrene. The chromatograms are shown in Figure 8. Two clear peaks were observed, and each peak was assigned by injection of a single-component sample. The peak heights and areas showed linear dependence on the concentrations. The LQDs were determined by the peak area calculations. The results are summarized in Table 2. As a result, 290 and 150 times higher sensitivity was confirmed than those of spectrophotometric detection for fluorine and pyrene, respectively. In conclusion, UV-TLM method was developed for sensitive detection of nonfluorescent samples on microchips. The QCW mode was applied in order to utilize the pulsed UV laser excitation for lock-in amplifier detection. Although background signal generation by the two-photon absorption of water could limit the detection limit, the sensitivity was more than 2 orders of magnitude higher compared with a spectrophotometer, and the applicability for microchip systems was verified including detection for conventional separation systems. Development of on-chip HPLC is now in progress for which we are integrating an injection valve and separation column on the microchip. Our UV-TLM method is expected to be a powerful analytical tool for various applications on microchips. ACKNOWLEDGMENT We thank Mr. Tomohiko Kawakami and Mr. Masaya Kakuta of Institute of Microchemical Technology for valuable discussions. This research was partially supported by the New Technology Development Foundation. Received for review November 4, 2005. Accepted February 7, 2006. AC051967U Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
2863
