Monolithic Capillary Electrophoresis Device with Integrated

A CMOS-Microfluidic Chemiluminescence Contact Imaging Microsystem. Ritu Raj Singh , Lian Leng , Axel Guenther , Roman Genov. IEEE Journal of Solid-Sta...
0 downloads 0 Views 241KB Size
Anal. Chem. 2001, 73, 1622-1626

Monolithic Capillary Electrophoresis Device with Integrated Fluorescence Detector J. R. Webster,*,† M. A. Burns,‡ D. T. Burke,§ and C. H. Mastrangelo†

Center for Integrated Microsystems, Department of Electrical Engineering and Computer Science, Department of Chemical Engineering, and Department of Human Genetics, University of Michigan, Ann Arbor, Michigan 48109-2122

A monolithic capillary electrophoresis system with integrated on-chip fluorescence detector has been microfabricated on a silicon substrate. Photodiodes in the silicon substrate measure fluorescence emitted from eluting molecules. The device incorporates an on-chip thin-film interference filter that prevents excitation light from inhibiting the fluorescence detection. A transparent AZO conducting ground plane is also used to prevent the high electric fields used for the separation from interfering with the photodiode response. Separations of DNA restriction fragments have been performed in these devices with femtogram detection limits using SYBR Green I intercalating dye. Capillary electrophoresis has benefited greatly from the microfabrication of devices. Fast, highly efficient separations have been routinely performed in these devices.1-3 Besides the benefits of miniaturization, microfabrication techniques can also produce complex devices with a high degree of functionality. The potential for complex DNA microprocessors by microfabrication has been demonstrated.4 The majority of capillary electrophoresis microdevices have been fabricated using a bonded, glass substrate technology and utilize off-chip photomultipliers for laser-induced fluorescence detection. The advantages of miniaturization are reduced, however, when these systems rely on equipment several orders of magnitude larger. The use of on-chip detection techniques can make portable DNA diagnostic instruments practical. One way to accomplish this is by using nonoptical detection techniques. While on-chip electrochemical DNA detection schemes are currently under development,5 fluorescence detection methods are very attractive because many of the current biochemistry protocols already incorporate fluorescent labels. Reactions such † Center for Integrated Microsystems, Department of Electrical Engineering and Computer Science. ‡ Department of Chemical Engineering. § Department of Human Genetics. (1) Woolley, A. T.; Mathies, R. A. Proc. Natl. Acad. Sci. U.S.A. 1994, 91, 11348-11352. (2) Woolley, A. T.; Mathies, R. A. Anal. Chem. 1995, 67, 3676-3680. (3) Bousse, L.; Dubrow, B.; Ulfelder, K. High-performance DNA separations in microchip electrophoresis systems. In Proceedings of the µTAS ‘98 Workshop, Banff, Canada, 1998; pp 271-275. (4) Burns, M. A.; Johnson, B. N.; Brahmasandra, S. N.; Handique, K.; Webster, J. R.; Krishnan, M.; Sammarco, T. S.; Man, P. F.; Jones, D.; Heldsinger, D.; Mastrangelo, C. H.; Burke, D. T. Science 1998, 282, 484-487. (5) Woolley, A. T.; Lao, K.; Glazer, A. N.; Mathies, R. A. Anal. Chem. 1998, 70, 684-688.

1622 Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

as Sanger sequencing and the polymerase chain reaction (PCR) have been adapted to fluorescent labeling methods.6 In fact, quantitative PCR amplification reactions are being performed with high efficiency using fluorescent labels.7 Therefore, chip-based assays can easily be incorporated into existing protocols without changing the biochemistry. Furthermore, fluorescence techniques are chemically decoupled from the analysis steps. Therefore, fluorescence detection can be used in a variety of applications on an analysis chip. In this device, we use a combination of integrated silicon photodiodes with a thin-film optical filter to detect fluorescence signals. Silicon photodiode detectors are widely used in commercial sequencing machines as an alternative to photomultiplier tubes, and they can be microfabricated and integrated easily with other components. Silicon photodiodes have been used for many years in the measurement of electromagnetic radiation from the near-UV range to the near-infrared.8-11 Devices have been reported with noise equivalent powers (NEP) in the femtowatt range.12 In this device, the fluorescence excitation is provided using collinear llumination provided by an inexpensive blue light-emitting diode (LED).13 DEVICE STRUCTURE Figure 1 shows a cross-sectional schematic of the device. The device basically consists of a photodiode built on a silicon substrate, an optical interference filter, and a thin-film plastic (parylene14) electrophoresis channel above. The successful integration of microfluidic, optical, and electronic devices requires several layers of isolation. Chemical isolation is needed to prevent the ions in the separation solution from reaching the ion-sensitive silicon detectors below. Chemical isolation is provided by the thick (5 µm) lower wall of the electrophoresis channel. Similarly, the (6) Smith, L. M.; Sanders, J. Z.; Kaiser, R. J.; Hughes, P.; Dodd, C.; Connell, C. R.; Heiner, C.; Kent, S. B. H.; Hood, L. E. Nature 1986, 321, 674-679. (7) Woudenberg, T. M.; Stevens, J. Quantitative PCR by Real Time Detection. In Proceedings of ultrasensitive biochemical diagnostics; SPIE: Bellingham, WA, 1996; pp 306-315. (8) Chamberlain, S. G. J. Appl. Phys. 1979, 50, 7228-7231. (9) Chamberlain, S. G.; Roulston, D. J.; Desai, S. P. IEEE Trans. Electron Devices 1978, 25, 241-246. (10) Hamstra, R. H.; Wendland, P. Appl. Opt. 1972, 11, 1539-1547. (11) Muench, W. V.; Gessert, C.; Koeniger, M. E. IEEE Trans. Electron Devices 1976, 23, 1203-1207. (12) Fisher, R. Appl. Opt. 1968, 7, 1079-1083. (13) Bruno, A. E.; Maystre, F.; Krattiget, B.; Nussbaum, P.; Gassmann, E. Trends Anal. Chem. 1994, 13, 190-198. (14) Webster, J. R.; Burke, D. T.; Burns, M. A.; Mastrangelo, C. H. An inexpensive plastic technology for microfabricated capillary electrophoresis chips. In Proceedings of the µTAS ‘98 Workshop, Banff, Canada, 1998; pp 249-252. 10.1021/ac0004512 CCC: $20.00

© 2001 American Chemical Society Published on Web 02/23/2001

Figure 1. Schematic of electrophoresis device cross section.

Figure 2. Optical micrograph of integrated electrophoresis device.

photodiodes must be dielectrically isolated from the high voltages needed by the separation stage. This is provided by the 8-µmthick combination of parylene wall and optical filter. Field isolation is also needed as the high voltages can invert the silicon surface near the photodiode causing very high noise levels. Field isolation is achieved using a ground plane; however, the ground plane must be transparent as it must permit the transmission of the optical fluorescent signal. The transparent ground plane is implemented using a thin layer of aluminum zinc oxide (AZO) placed between the top of the optical filter and the bottom channel wall. Figure 2 shows an optical micrograph of the microfabricated electrophoresis device with on-chip fluorescence detection. The channel is 200 µm wide, and the total channel length is ∼1.3 cm. EXPERIMENTAL SECTION Microfabrication. The fabrication process for the electrophoresis device consists of 13 lithography steps. Figure 3 shows a simplified process flow. The starting wafer is a 1000-2000 Ω‚ cm boron doped, p-type, 〈100〉 float zone silicon substrate. The high resistivity substrate is used to increase the depletion region and therefore the efficiency of the photodiode p-n junction. The wafers are first implanted with a field implant of 50-keV boron with a dose of 1 × 1012 cm-2. The field implant prevents inversion of the substrate below the metal lines near the surface. Using a photoresist mask, the silicon substrate is etched with a wet

isotropic silicon etchant for 10 min. The photodiode is constructed in the substrate by using two successive ion implantation steps. For each a photoresist mask is used. The n-type implantation is done with 50-keV phosphorus at a dose of 5 × 1014 cm-2. The p-type implantation for the substrate contact is done with 30-keV boron at a dose of 1 × 1015 cm-2. The photoresist masks from the implantations were stripped using PRS2000 at 110 °C followed by a H2SO4/H2O2 1:1 cleaning. A passivation thermal oxide was next grown on the wafer under wet, trichloroethane (TCA) conditions. The 0.2-µm-thick silicon dioxide layer was grown at 900 °C. Contacts to the substrate and diodes are made using a wet buffered hydrofluoric acid (BHF) etchant. A chromium/gold metallization was next deposited and patterned to make contact with the photodiodes. A short, 10-s BHF dip was used to clean the contact areas. The wafer is then immediately placed into a vacuum chamber and 0.05 µm of chromium followed by 0.2 µm of gold was evaporated on the substrate. The Cr/Au metallization was then patterned using a photoresist mask and wet metal etchants. Next, a thinfilm interference filter was deposited on top of the wafer. The filter layers were designed and deposited by ZC&R Coatings for Optics. The filter consists of alternating layers of SiO2 and TiO2 at nearly quarter-wavelength thicknesses. The filter consists of ∼20 layers and is ∼3 µm thick. On top of the interference filter, a 2.5-µmthick layer of parylene-C is deposited. Adhesion of the parylene layer to the filter is assisted by a silanation procedure. The immediate deposition of the parylene layer serves as a passivation layer for the filter to maintain the integrity of the optical properties. After the parylene-C layer, the ground plane is deposited. First a 0.18-µm-thick layer of SiO2 is sputtered on top of the paryleneC. Next a 0.15-µm-thick layer of AZO is sputtered on top of the SiO2. The AZO layer has a greater than 90% transmission with a sheet resistance of ∼50 Ω/0. The SiO2 is used as an adhesion layer for the AZO. The AZO is then patterned using photoresist and a wet etchant consisting of 1:4:500 HNO3/HCl/H2O.15 The etch rate of the AZO is extremely fast. Contact to the AZO layer is made by depositing a 0.02 µm/0.08 µm Cr/Au layer by liftoff. A second SiO2 layer is sputtered on top of the AZO. This layer will provide excellent adhesion to parylene-C and provide additional passivation to the AZO layer underneath. Another (15) Lan, J.; Kanicki, J.; Catalano, A.; Keane, J.; Boer, W. D.; Gu, T. J. Electron. Mater. 1996, 25, 1806-1817.

Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

1623

Figure 4. On-chip fluorescence detection system.

Figure 3. Simplified process flow for the electrophoresis device.

parylene-C layer is deposited on top of the SiO2. This layer is 4.8 µm thick and forms the bottom layer of the electrophoresis channel. Next the contact holes must be etched through the two parylene-C layers to access the photodiode metalization. This etch step is performed in two lithographies using the same mask. The first parylene-C layer is etched in an O2 plasma RIE using a 20µm-thick photoresist mask. The SiO2 underneath the parylene is etched using 10% BHF prior to stripping the photoresist. The resist 1624 Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

is then stripped in acetone, and the parylene is etched a second time again using a 20-µm-thick photoresist mask. This time, the filter layers beneath the parylene are etched using a 10% BHF solution for 8 min prior to stripping the photoresist. The undercut of the etching was found to be quite large (20 µm); however, fine features were not necessary since only metal pad openings were made. The parylene is etched a third time to open contact to the AZO metallization. Again the oxide is etched in the same way. Next 0.1-µm-thick gold electrophoresis electrodes are patterned on the parylene by liftoff. The sacrificial photoresist is then patterned using a 20-µm-thick layer of AZ9260 (Clariant). The top of the channel is then formed by depositing a 5.3-µm-thick parylene-C layer. The adhesion of this layer is assisted by a short oxygen plasma. The final parylene layer is etched in a O2 plasma RIE using a thick photoresist. Channel reservoir openings as well as pad openings to all metal layers are made in this step. Photodefinable silicone rubber rings previously demonstrated on plastic electrophoresis devices can be patterned on this device at this point using the same procedure outlined earlier.14 In the device tested here, this step was omitted out of convenience and silicone rubber rings were applied by hand at the time of device testing. The wafer was diced using a diamond tip saw. The final photoresist mask was left in place to protect devices during dicing. The devices were then released in acetone for 20 h followed by an isopropyl alcohol (IPA) rinse and dried with N2. Separation Conditions. Separations were performed in these devices using a sieving matrix consisting of 0.5% (w/v) hydroxyethylcellulose (HEC) (Polysciences, Inc., Warrington, PA; MW 90 000-105 000) and 0.1× Tris/boric acid/EDTA buffer (Sigma Chemical Co., St. Louis, MO). Approximately 3 µL of the buffer solution is loaded in one reservoir and allowed to fill the entire channel by capillary action. The remaining reservoirs are filled with ∼3 µL of buffer solution. The channel was preelectrophoresed at 300 V/cm for 10 min. This was found to concentrate the HEC in the separation column to more than 0.5% resulting in a higher resolving power. A 0.2 µg/µL sample of DNA was labeled with SYBR Green I (Molecular Probes, Eugene, OR) intercalating dye at an intercalation ratio of 1:5 dye/DNA bp. Approximately 2 µL of the DNA sample was loaded into the injection reservoir. The samples were cross-injected using a pinched injection scheme.16 The electric fields used during separation were 110 V/cm. Fluorescence Detection. A schematic of the on-chip fluorescence detection system is shown in Figure 4. A Stanford Research SR830 lock-in amplifier was used to measure the photodiode current. A GaN blue LED was used as an excitation source with a 450 ( 27 nm band-pass filter (Omega Optical). The LED was driven with an in-house built amplifier, 5-V power supply, and the (16) Jacobson, S. C.; Hergenroder, R.; Koutny, L. B.; Warmack, R. J.; Ramsey, J. M. Anal. Chem. 1994, 66, 1107-1113.

Figure 5. Relative quantum efficiency of photodiode with on-chip filter and transparent ground plane.

TTL output of the lock-in amplifier. The lock-in amplfier output was recorded using an RS232 port, and a in-house-written data collection program for an IBM-compatible PC. Data were recorded at a rate of 64 Hz. Instrumentation. A stereomicroscope (Olympus SZX12) equipped with a fluorescence illuminator (100-W mercury lamp illuminator) was used to observe the separations. The filter cube consisted of 470-nm band-pass excitation filter, 495-nm long-pass dichroic mirror, and a 500-nm long-pass emission filter. Separations were recorded to SVHS videotape using a Dage 68 SIT camera. Optical characterization was performed using a 150-W halogen lamp (Micro lite, Three Rivers, MA), a Monospec 18 monochromator (Thermo Vision, CO), an optical power meter (Ophir Optronics, Israel), and a picoammeter (Keithley). RESULTS AND DISCUSSION Figure 5 shows the relative efficiency of the photodiode combined with the on-chip interference filter. The filter blocks the excitation light up to ∼495 nm. The photodiode itself is most efficient in the red spectrum due to the finite junction depth (∼0.65 µm) of the diode. The absolute efficiency could not be measured since diffusion of carriers dominates the photo response. The maximum efficiency shown is calculated to be between 90 and 100%. The ripples in the passband are due to the on-chip filter. The photodiode response to fluorescently labeled DNA in a microchannel was tested using various concentrations of DNA labeled with YOYO-1 (Molecular Probes) at a ratio of 1:5 dye/ DNA bp. For DNA concentrations between 6.25 and 200 ng/µL, the photodiode current varied linearly between 3 and 6 nA. Separations of the HaeIII digestion of ΦX174 RF DNA are shown in Figure 6. A video image of four fragments is shown in Figure 6A. The separation distance from injection to detection is 0.9 cm. The electropherogram generated by the on-chip detection of fragments is shown in Figure 6B corresponding to roughly 30 000 effective theoretical plates. The baseline has been extracted

Figure 6. Separation of the HaeIII digest of ΦX174 RF DNA in the integrated CE device: (A) video image of separated fragments; (B) electropherogram using on-chip fluorescence detection.

using a simple curve fit. No other data processing has been done. The 194 234-bp fragments are no longer baseline separated due to the finite width of the detection region. The effective detector width is in fact much larger than the photodiode itself due to diffusion of carriers and the finite distance of the DNA from the diode. The bottom of the electrophoresis channel is 11 µm from the silicon surface. Since the blue LED excitation is a broad area excitation source, and the emitted fluorescence is isotropic, the DNA fluorescence can be seen nearly 200 µm from the detector. A simple geometric model17 predicts an effective detection width of 175 µm. This therefore contributes 2600 µm2 to the band variance and significantly alters the resolution of the separation. The use of small area excitation sources would lower the detector width to a more acceptable size, as well as the use of small detector apertures. The limits of detection have been estimated from the photocurrent noise and electrophoresis peak heights. The standard deviation of the photocurrent noise was ∼0.7 pA. The limit of detection for the 72-bp fragment is 75 fg of DNA (S/N ) 2). This limit is slighly higher than that reported with electrochemical detection schemes,5 but it can be substantially lower when a brighter LED or laser excitation source is used. A perpendicular laser excitation would provide the lowest detection (17) Webster, J. R. Monolithic Structures for Integrated Capillary Electrophoresis Systems. Thesis, University of Michigan, Ann Arbor, 1999.

Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

1625

limits. Such a scheme could be realized via an integrated waveguide. CONCLUSIONS With the integration of on-chip detection for microfabricated electrophoresis devices, practical hand-held diagnostic instruments may be possible. In this paper, we discussed the implementation and testing of a microfabricated integrated electrophoretic separation system with on-chip fluorescence detection (filters, detectors, and field shields). The fluorescence was excited using a blue LED. Separations carried out using SYBR-green-labeled dsDNA and HEC as the sieving matrix indicate an effective resolution of ∼30 000 theoretical plates on a 0.9-cm-long microchannel with a minimum detectable signal of 75 fg of sample. Detection limit calculations indicate that lower limits can be achieved with the

1626

Analytical Chemistry, Vol. 73, No. 7, April 1, 2001

use of narrowband, brighter laser excitation. The use of compact, inexpensive solid-state lasers with these systems could make such devices attractive with sensitivities rivaling those of more traditional analysis equipment. ACKNOWLEDGMENT This work was supported by the National Institute of Health under Grant NIH-R01-H601044. We thank George Finney of Olympus America for the device photograph.

Received for review April 19, 2000. Accepted December 17, 2000. AC0004512