Fabrication and Characterization of Sputtered-Carbon Microelectrode

Jun 1, 1996 - Small 2009 5 (10.1002/smll.v5:7), 776-788 ... Development of fast Fourier transformations with continuous cyclic voltammetry at an Au ...
2 downloads 0 Views 321KB Size
Anal. Chem. 1996, 68, 1858-1864

Fabrication and Characterization of Sputtered-Carbon Microelectrode Arrays G. Sreenivas,† Simon S. Ang,*,† Ingrid Fritsch,‡ William D. Brown,† Greg A. Gerhardt,§ and Donald J. Woodward⊥

Departments of Electrical Engineering and Chemistry and Biochemistry, University of Arkansas, Fayetteville, Arkansas 72701, Departments of Pharmacology and Psychiatry and Rocky Mountain Center for Sensor Technology, University of Colorado Health Sciences Center, Denver, Colorado 80262, and Department of Physiology and Pharmacology, Bowman Gray School of Medicine, Wake Forest University, Winston-Salem, North Carolina 27157

This paper describes a robust and reliable process for fabricating a novel sputter-deposited, thin-film carbon microelectrode array using standard integrated circuit technologies and silicon micromachining. Sputter-deposited carbon films were investigated as potential candidates for microelectrode materials. The surface properties and cross section of the microelectrode arrays were studied by atomic force microscopy and scanning electron microscopy, respectively. Electrical site impedance, crosstalk, and lifetime (dielectric integrity) of microelectrodes in the array were characterized. Electrochemical response of the microelectrodes to hexaammineruthenium(III) chloride and dopamine were investigated by fast-scan cyclic voltammetry and high-speed, computerbased chronoamperometry; results show that thin-film carbon microelectrodes are well-behaved electrochemically. The thin carbon films offer extremely good electrical, mechanical, and chemical properties and thus qualify as viable candidates for various electroanalytical applications, particularly acute neurophysiological studies. For over two decades, considerable research has involved development and characterization of new electrode materials with improved analytical performance for the detection of monoamines. In the early 1970s, Adams recognized that several neurotransmitters (small molecules which relay information between neurons) are easily oxidized and thus should be detectable with carbon electrodes.1 Very small voltammetric carbonaceous electrodes led to the development of a number of electrochemical techniques for the determination of these compounds, both in vitro and in vivo.2-5 Typically, microelectrodes in comparison with macroelectrodes offer higher sensitivity, smaller double-layer capacitance, and lower ohmic losses that result in a higher signal-tonoise (S/N) ratio.6 †

Department of Electrical Engineering, University of Arkansas. Department of Chemistry and Biochemistry, University of Arkansas. § University of Colorado Health Sciences Center. ⊥ Wake Forest University. (1) Adams, R. N. Anal. Chem. 1976, 48, 1126A-1138A. (2) Kissinger, P. T. Anal. Chem. 1977, 49, 447A-456A. (3) Ponchon, J. L.; Cespulgio, R.; Gonon, F.; Jouvet, M.; Pujol, J. F. Anal. Chem. 1979, 51, 1483-1486. (4) Ewing, A. G.; Wightman, R. M.; Dayton, M. A. Brain Res. 1982, 249, 361370. (5) Schenk, J. O.; Miller, E.; Rice, M. E.; Adams, R. N. Brain Res. 1983, 277, 1-8. ‡

1858 Analytical Chemistry, Vol. 68, No. 11, June 1, 1996

The application of carbonaceous compounds as electrodes in the area of analytical measurements has been a matter of interest due to their good electrical and mechanical properties. Also, carbon electrodes exhibit lower background currents over a wider potential window than metal electrodes, particularly in the cathodic potential region.7 They also offer surface groups for chemical modification and controllable surface activity resulting from pretreatment.8 Furthermore, carbon tends to be a more biocompatible material than other commonly used electrode materials.9 Apart from its good electrical properties, the carbon electrode offers additional advantages of radiotransparency and high resistance to mechanical or chemical damage.10 These attractive properties of carbon enable its use in various other analytical applications. In recent years, a variety of carbon-based materials have been investigated as electrode materials. Arrays of microelectrodes have been fabricated using conventional techniques such as the suspension of carbon particles or fibers in an insulating matrix, the impregnation of porous carbon within an insulator, or the impregnation of the pores of a host membrane with conducting carbon particles.11,12 Furthermore, glassy carbon electrodes,6,12,13 reticulated vitreous carbon,14 and pyrolytic carbon7,15 arrays have also received considerable attention. Electrodes that are constructed by conventional techniques are subject to great variability in their physical, electrical, and electrochemical characteristics. Furthermore, these electrodes are difficult to reproducibly fabricate in multielectrode arrays; such arrays, when realized, can contain appreciable capacitive coupling between electrodes.16 Different types of carbon array electrodes have also been fabricated using photolithographic techniques,17,18 including mi(6) Horiuchi, T.; Niwa, O.; Morita, M.; Tabei, H. Anal. Chem. 1992, 64, 32063208. (7) Niwa, O.; Tabei, H. Anal. Chem. 1994, 66, 285-289. (8) McCreery, R. L. Electroanal. Chem. 1991, 17, 221-370. (9) Wightman, R. M.; May, J. L.; Michael, A. C. Anal. Chem. 1988, 60, 769A779A. (10) Spekhorst, H.; Sippensgroenewegen, A.; David, K. J.; Van Rijn, M. C.; Broekhuijsen, P. IEEE Trans. Biomed. Eng. 1988, 35, 402-406. (11) Wang, J.; Freiha, B. A. J. Chromatogr. 1984, 298, 79-87. (12) Weisshaar, D. E.; Tallman, D. E. Anal. Chem. 1983, 55, 1146-1151 (13) Poon, M.; McCreery, R. L. Anal. Chem. 1986, 58, 2745-2750. (14) Sleszynski, N.; Osteryoung, J.; Carter, M. Anal. Chem. 1984, 56, 130-135 (15) Rajo, A.; Rosenstratten, A.; Anjo, D. Anal. Chem. 1986, 58, 2988-2991. (16) Starr, A.; Wise, K. D.; Csongradi, J. IEEE Trans. Biomed. Eng. 1973, 20, 291-293. (17) Kittlesen, G. P.; White, H. S.; Wrighton, M. S. J. Am. Chem. Soc. 1985, 107, 7373-7380. S0003-2700(95)00881-X CCC: $12.00

© 1996 American Chemical Society

in an animal brian during in vivo measurements. Bonding pads, located on the opposite end of the substrate from the microelectrode array, allow for external electronic connections. In the present paper, electrode arrays have been characterized for their surface, electrical, and electrochemical characteristics in vitro.

Figure 1. Schematic diagram of a four-site carbon microprobe.

crodisks,19 microbands,20,21 and interdigitated array electrodes.7,22,23 Such photolithographically produced arrays of microelectrodes exhibit reduced capacitive coupling and can be fabricated reproducibly. They can be used to perform multiple recordings simultaneously at different, closely spaced, well-defined locations throughout the tissue volume in the brain, which is of great significance in neuroscience research.24 We have investigated characteristics of an electrode material, sputter-deposited thin-film carbon, which exhibits desirable properties for microelectrodes and which has not been previously explored. Niwa et al.7 reported that it is difficult to fabricate microarray electrodes from carbon films using lithography techniques, especially sputter- or vapor-deposited carbon films, because of poor adhesion to the substrate. Glass et al.18 used a niobium layer to enhance adhesion of electron-beam-evaporated carbon in multielement microelectrode arrays. Although some success with carbon films formed from chemical vapor deposition (CVD) has been made, CVD is an elaborate procedure requiring control of many more deposition parameters than for vacuumdeposited or sputtered carbon. In this paper, we report a robust, high-yield, reliable, and reproducible process for fabricating a sputter-deposited thin-film carbon microelectrode array using standard integrated circuit technologies and silicon micromachining. This technique enables fabrication of many extremely small, well-defined structures of different geometries that are capable of making multiple, simultaneous recordings at spatially distinct sites which are of interest in neurophysiological research. Our arrays show recording characteristics similar to those of single microelectrodes and microarrays previously studied. Figure 1 shows a schematic of the structure of one such four-site array microprobe.25 The sensing electrodes of 5580 µm2 areas (155 µm × 36 µm) reside on the shank of the microprobe and are spaced 200 µm from the centers of adjacent sites. The probe is 10.65 mm long and has a 1.98 mm long shank which tapers from 120 µm near the base to less than a few micrometers near the tip. Such small dimensions are critical for minimizing tissue damage (18) Reller, H.; Kirowa-Wisner, E.; Gileadi, E. J. Electroanal. Chem. 1984, 161, 247-268. (19) Cassidy, J.; Ghoroghchain, J.; Sarfarazi, F.; Smith, J. J.; Pons, S. Electrochim. Acta 1986, 31 (6), 629-636. (20) Bard, A. J.; Crayston, J. A.; Kittlesen, G. P.; Shea, T. V.; Wrighton, M. S. Anal. Chem. 1986, 58, 2321-2331. (21) Matsue, T.; Aoki, A.; Ando, E.; Uchida, I. Anal. Chem. 1990, 62, 407-409. (22) Aoki, K.; Morita, M.; Niwa, O.; Tabei, H. J. Electroanal. Chem. 1989, 267, 291-297. (23) Sanderson, D. G.; Anderson, L. B. Anal. Chem. 1985, 57, 2388-2393. (24) van Horne, C. G.; Bement, S.; Hoffer, B. J.; Gerhardt, G. A. Neurosci. Lett. 1990, 120, 249-252. (25) In this article, the array of electrodes is sometimes referred to as the microprobe, and each electrode in the array is referred to as a site.

EXPERIMENTAL SECTION Reagents. Ethylenediamine (99+%), catechol (99+%), pyrazine (99+%), Nafion (5% solution in water), ascorbic acid (AA), potassium nitrate (99+%), dopamine (DA), and hexaammineruthenium(III) chloride (99+%), all from Aldrich Chemical Co., were used as received unless otherwise specified. Also, dibasic sodium phosphate (Na2HPO4‚7H2O), monobasic sodium phosphate (NaH2PO4‚H2O), and sodium chloride (NaCl) obtained from Fisher Scientific were used as received. Deionized (DI) water was prepared using a three-filter purification system from Continental Water Systems (Modulab DI recirculator, service deionization polisher). Sulfuric acid, hydrogen peroxide, and isopropyl alcohol were CMOS electronic grade (all from J.T. Baker, Inc.). For photolithography, Hunt HPR 204 positive resist, HR 200 negative resist, PLSI and PLSI-T2 positive resist developer, WNRD negative resist developer, PF thinner, Microstrip 2001 (all from OCG Microelectronic Material, Inc.) and Universal photoresist stripper Nophenol 922 (EKC Technology, Inc.) were used. The silicon substrates used in this study were supplied by Virginia Semiconductor, Inc. A conductive carbon target (Kurt J. Lesker Co., Part No. EJTCXXX502A2) was used for sputter deposition. Epo-tek H20E-PFC (Epoxy Technology, Inc.) conductive silver epoxy, and ABLEBOND 967-3 and 933-1.5 (Ablestik Laboratories) nonconductive epoxies were also used. Electrode Fabrication. A typical process sequence is shown in Figure 2. The fabrication process began with a three-mil, n-type, 〈100〉 oriented, g10 Ω-cm, 2-in. diameter silicon wafer as the host substrate and involves three masking steps. A sufficiently thick layer (∼1 µm) of silicon nitride (Si3N4), deposited by plasmaenhanced chemical vapor deposition (PECVD), served as a lower dielectric layer between the electrodes and the host silicon wafer substrate and as an etch mask during final chemical etching of silicon. Such Si3N4 films were deposited by gas-phase dissociation of 100 cm3(STP) min-1 of 20% monosilane (SiH4) in argon, (99.99%), 80 cm3(STP) min-1 of ammonia (NH3, 99.9%), and 200 cm3(STP) min-1 of ultrahigh-purity grade nitrogen (N2) (Air Products, Ltd.) in a Texas Instruments Reinberg-type, capacitively coupled, parallel-plate cold wall reactor (Model A24C) at a radio frequency (rf) (13.56 MHz) power density of 28 mW/cm2 to achieve a deposition rate of 120 Å/min. The chamber pressure during deposition was maintained at 1 Torr. The mixture of gases was excited by an rf plasma in which high-energy electrons dissociate reactant gases to allow deposition of solid material on the substrate at moderate temperatures (300 °C). The outline of the microprobe was defined with a first-level photomask using a standard UV photolithographic process (Quintel mask aligner/ exposure system), which was optimized for our equipment. The Si3N4 was subsequently selectively removed from exposed areas by reactive ion etching (Plasma Therm Model 520 parallel-plate reactive ion etcher) using a tetrafluoromethane (99.9%, Air Products, Ltd.) plasma. An rf power density of 201 mW/cm2 was used at a chamber pressure of 300 mTorr to obtain an etch rate of 500 Å/min. The flow rates during deposition and etching processes were controlled by electronic mass flow controllers (Unit Instruments, Inc.). Analytical Chemistry, Vol. 68, No. 11, June 1, 1996

1859

Figure 2. Typical process sequence for the thin-film carbon microprobe fabrication. (Layers are not to scale.)

Because conventional etching procedures on sputter-deposited carbon films have not been successfully developed, a lift-off process was employed. Hence, the sensing electrodes, conducting lines, and bonding pads were defined into photoresist by photolithography using a second-level mask. This was followed by a 30 s descum (i.e., O2 plasma etch) to ensure complete removal of residual photoresist in the exposed regions of the pattern. A 2500 Å thick carbon film was sputter deposited onto the patterned substrate in a modified Perkin Elmer Model 2400 dc magnetron sputtering system. The system was pumped down to a base pressure of 3 µTorr, and argon was introduced to maintain a pressure of 20 mTorr during deposition. A deposition rate of 21 Å/min was achieved, based on the step-height measurement using a surface profilometer (Sloan Dektak 3030). The unwanted carbon overlying the photoresist was then removed by dissolving the photoresist in a lift-off process in boiling Nophenol 922 at 90 ( 5 °C. A top layer of Si3N4 was deposited (∼1.5 µm) and photolithographically defined using the level-three mask and etched to provide openings for sensing electrodes and bonding pads. Finally, the carbon microprobes were separated from the host silicon substrate by an orientation-dependant etchant, ethylenediamine pyrocatechol in water (EPW).26,27 Pyrazine (6 gm/L of ethylenediamine) was added to the mixture to eliminate silicon dioxide residue contamination on the electrodes during the silicon etching process. The final thickness of the silicon microprobe was about 40 µm. The fabrication processes were controlled to (26) Finne, R. M.; Klein, D. L. J. Electrochem. Soc. 1967, 114 (9), 965-970. (27) Reisman, A.; Berkenblit, M.; Chan. S. A.; Kaufman, F. B.; Green, D. C. J. Electrochem. Soc. 1979, 126 (8), 1406-1415.

1860

Analytical Chemistry, Vol. 68, No. 11, June 1, 1996

obtain a typical yield (with microscopic inspection) in the 7080% range, to yield about 30 probes on a typical run using a 2-in. silicon wafer. Microprobes were mounted onto the tip of a custommade printed circuit board (PCB) carrier, and conducting wires were bonded with silver epoxy. The other end of the PCB contained dual-in-line (DIP) pins for connecting to instrumentation. Some of the arrays only had two or three functional sites. Topography and Cross-Sectional Studies. A Digital Instruments Nanoscope III Dimension 3000 atomic force/scanning probe microscope (AFM/SPM) was used in the contact mode to study surface topography of vacuum-deposited carbon and silicon nitride films. This tool provides qualitative information about surface features, particulate contamination, and film roughness (under Section Analysis mode). Electrical Characteristics. A custom-made impedance tester was used to measure the total impedance of electrodes. The instrument designed was based on that described by Clark et al.28 A sine wave of 1 V peak-to-peak, centered at 0 V (ground), was applied across a resistor (470 kΩ) in series with the electrochemical cell. The electrochemical cell consisted of the microelectrode probe and counter electrode in electrolyte. The rms potential drop across the electrochemical cell (between the two electrodes) was measured. A dummy cell with a variable resistor, R, and capacitor, C, in series was switched into the circuit in place of the electrochemical cell. The value of R was varied until the rms potential of the dummy cell equaled the value of that measured for the electrochemical cell. This R value is considered to represent the total impedance, Z, of the electrochemical cell. The approximation Z ≈ R can be made because Z ) {R2 + (2πfC)-2}1/2, where (2πfC)-1 is the reactance of the capacitor, and C was chosen so that (2πfC)-1 , R. The lifetime (dielectric integrity) of the microelectrode array was measured for an extended period of time (over 180 h) in a physiological saline solution by periodically monitoring the changes in the site impedance. Crosstalk, a source of signal distortion between conducting paths, was also investigated. Shunt capacitance (Csh) and coupling capacitance (Cc) were measured in air using a HP 4280 CV meter. Statistical analyses report errors as one standard deviation. Electrochemical Experiments. Cyclic voltammetry (CV) was performed using a computer-interfaced Bioanalytical Systems, Model 100B potentiostat with BAS 100W software. Chronoamperometry (CA) was performed on a high-speed computer-based in vivo electrochemistry recording system (IVEC-10; Medical Systems Corp.). The reference was an aqueous Ag/AgCl (3 M NaCl) electrode (Bioanalytical systems, Model RE-5). Platinum gauze was used as an auxiliary electrode in CV experiments. Cyclic voltammetry was performed with 10 mM hexaammineruthenium(III) chloride in 1 M potassium nitrate (KNO3). A typical procedure consisted of running CV measurements in a mixture of 20 mL of KNO3 and hexaammineruthenium(III) chloride solution purged with argon for ∼20-30 min to remove the oxygen in the solution. Scan rates were varied from 0.025 to 100 V/s, and the triangular waves were performed between +0.10 and -0.45 V vs Ag/AgCl. Hexaammineruthenium(III) chloride was used because it is a well-behaved, reversible redox couple used to characterize electrodes. Also, extensive literature has documented properties of this molecule. (28) Clark, S. T.; Paul, D. W.; Sundgren, H.; Lundstrom, I. Microchem. J. 1992, 46, 225-233.

Figure 3. A 3 µm scan of a typical carbon microelectrode surface performed at 5 Hz and a resolution of 256 data points per line using AFM.

A solution of 0.5 µM dopamine in the presence of 250 µM ascorbic acid in 0.1 M phosphate-buffered saline solution, pH ) 7.2-7.4, was used to obtain calibration curves and evaluate selectivities, sensitivities or detection limit, and ratio of reduction to oxidation current of the carbon electrodes. To a 40 mL phosphate-buffered saline solution (very close to extracellular body fluid in composition) was added a 500 µL solution of ascorbic acid (20 µM) to achieve a final concentration of 250 µM. Increments of 10 µL of dopamine (0.5 µM) were then added to the stock solution to raise the dopamine concentration by 0.5 µM intervals. CA measurements were performed in this solution in the form of square-wave pulses at 5 Hz from 0.0 to +0.55 V vs Ag/AgCl. Measurements were obtained at varying gain settings. RESULTS AND DISCUSSION Figure 3 shows the surface topography of a carbon electrode obtained by AFM. An area of 3 µm × 3 µm was scanned at 5 Hz, and 256 data points per line were taken. The figure shows that the film is uniform and free from protrusions, pinholes, and delaminations, which indicates good planar coverage. The typical root-mean-square (rms) roughness of the film ranged from 34 to 84 Å and averaged 57 ( 16 Å (n ) 8 electrodes). However, the rms roughness of the bonding pads was found to be less by ∼20 Å than that of the sensing sites. To determine electrode site impedance, probes were immersed beyond the uppermost electrode site in a 0.9% saline aqueous solution. Site impedance depends on the material, area, surface roughness, electrolyte, signal frequency, and current density.29 Figure 4 shows a typical plot of site impedance of one electrode

Figure 4. Site impedance at different frequencies for one of the sites of a carbon microprobe.

at different frequencies of physiological interest, 500 Hz to 5 kHz. The site impedance ranged from 1 to 14 MΩ. At 1 kHz, the site impedance was ∼8 MΩ, which is in the appropriate range for a good neuronal recording.30 In applications involving neuronal recordings, electrodes implanted in the brain for extended periods of time must be able to withstand long-term exposure to electrolytes in the extracellular (29) Prohaska, O. J.; Olcaytug, F.; Pfunder, P.; Dragaun, H. IEEE Trans. Biomed. Eng. 1986, 33 (2), 223-229. (30) Bement, S. L.; Wise, K. D.; Anderson, D. J.; Najafi, K.; Drake, D. L. IEEE Trans. Biomed. Eng. 1986, 33, 230-241.

Analytical Chemistry, Vol. 68, No. 11, June 1, 1996

1861

Figure 5. Site impedance versus soak time of a carbon microelectrode.

fluid and should be capable of delivering large amounts of charge without dissolution into the extracellular fluid or delamination from the conductor surface.31,32 Thus, as shown in Figure 5, tests of microprobe lifetime for continuous immersion in physiological saline solution were conducted over an extended period of time (180 h) on one of the recording sites. The recording site impedance, which is directly proportional to the measured voltage, remained essentially unchanged over the test period, assuring that the layers of dielectric are adequate for long-term encapsulation of the probe. There are also no indications of degradation of adhesion between conductor and insulator layers. A major issue of concern in scaling down the size of features of the microprobes is the increase in electrical crosstalk between adjacent electrodes. In a multielectrode environment, width of and spacing between interconnect lines determine shunt capacitance and coupling capacitance, respectively. Hence, as features of recording electrodes are reduced in size, electrical crosstalk due to capacitive coupling will increase. Figure 6a shows the cross section of a carbon probe sandwiched between two layers of Si3N4 dielectric material.33 The shunt capacitance was found to range between 39 and 47 nF/cm2, averaging 45 ( 4 nF/cm2 (n ) 8). The coupling capacitance ranged from 63 to 68 nF/cm2 and averaged 65 ( 2 nF/cm2 (n ) 8). These values computed for the array in air would be somewhat larger in saline, where the conducting saline is separated from the conducting leads by the Si3N4 dielectric insulator. Other investigators have measured a comparative capacitance of 16 µF/cm2 at similar gold multielectrode array surfaces.34 An approximate equivalent circuit for the microprobe used for PSPICE (Microsim Corp.) simulation is shown in Figure 6b. This circuit is similar to one previously studied,33 where Vin is the potential generated in extracellular ionic flow, Rs is the equivalent resistance of neural tissue, Re and Ce are the electrical double-layer resistance and capacitance, respectively, Rm is the resistance of the carbon lead, Cc is the coupling capacitance to the neighboring channel, Csh is the shunt capacitance, and Ra and Ca are the input resistance and capacitance, (31) Najafi, K.; Wise, K. D. J. Solid-State Circuits 1986, 21 (6), 1035-1044. (32) Anderson, D. J.; Najafi, K.; Tanghe, S. J.; Evans, D. A.; Levy, K. L.; Hetke, J. F.; Xue, X.; Zappia, J. J.; Wise, K. D. IEEE Trans. Biomed. Eng. 1989, 36 (7), 693-703. (33) Najafi, K.; Ji, J.; Wise, K. D. IEEE Trans. Biomed. Eng. 1990, 37 (1), 1-11. (34) Pochay, P.; Wise, K. D.; Allard, L. F.; Rutledge. IEEE Trans. Biomed. Eng. 1979, 26 (4), 199-206.

1862 Analytical Chemistry, Vol. 68, No. 11, June 1, 1996

Figure 6. (a) Cross section of a carbon microprobe insulated between bottom and top dielectric layers (Si3N4) used for crosstalk simulation. (b) Equivalent circuit model used in PSPICE to evaluate the electrical crosstalk.

respectively, of the first-stage amplifier. The crosstalk was