Anal. Chem. 1999, 71, 550-556
Microfabricated Recessed Microdisk Electrodes: Characterization in Static and Convective Solutions Charles S. Henry and Ingrid Fritsch*
Department of Chemistry and Biochemistry, University of Arkansas, Fayetteville, Arkansas 72701
Construction and characterization of microfabricated recessed microdisk electrodes (RMDs) of 14- and 55-µm diameters and 4-µm depth are reported. For evaluation of electrode function, both faradaic current in Ru(NH3)63+/ KNO3 solution and charging current in KNO3 solution were measured with cyclic voltammetry. The experimental maximum current was measured and compared to calculated values, assuming radial and linear diffusion. A model for diffusion to a RMD best matches the behavior of the 14-µm RMD, which has a larger depth-to-diameter ratio than the 55-µm RMD. At fast scan rates (204 V s-1), where linear diffusion should dominate, there are large deviations from the linear diffusion model. Uncompensated resistance and overcorrection for background current contribute to this deviation. The dependence of capacitance on scan rate of the RMDs was found to be similar to that of a macroelectrode, indicating good adhesion between the insulator and the electrode. Chronoamperometry of Ru(NH3)63+ in KNO3 in both static and stirred solutions was performed using the RMDs and the current is compared to those from a 10-µm-diameter planar microdisk electrode (PMD). The signal-to-noise ratio of the 14-µm RMDs compared to the PMD is on average 4 times greater for stirred solutions. The 55-µm RMD exhibited no protection to convection of the stirred solution. Construction and characterization of microfabricated recessed microdisk electrodes (RMDs) are reported. These RMDs are an intermediate structure obtained during a more complex process that we developed for microcavities containing multiple electrodes.1,2 Electrochemical analysis of microfabricated RMDs is presented and the quality of the electrode seal with the insulator is discussed. The recessed nature of these RMDs is also shown to be valuable under convective conditions. Microelectrodes, in general, have been studied extensively and reviews have been published.3-6 Microelectrodes have been used * Corresponding author: (phone) 501-575-6499; (fax) 501-575-4049; (e-mail)
[email protected]. (1) Henry, C. S.; Fritsch, I. Submitted to Anal. Chem. (2) Henry, C. S.; Fritsch, I. Submitted to J. Electrochem. Soc. (3) Montenegro, M. I.; Queiros, M. A.; Daschbach, J. L. In Microelectrodes: Theory and Applications; Montenegro, M. I., Queiros, M.A., Daschbach, J. L., Eds.; NATO ASI Series E, 197; Kluwer Academic Publishers: Boston, 1991.
550 Analytical Chemistry, Vol. 71, No. 3, February 1, 1999
in complex media such as blood4 and urine.7 They can provide spatially resolved chemical information from surfaces8 and cell membranes.10-12 A host of biosensor applications has been reported.13-17 The most commonly studied microelectrode geometry is the disk because it is relatively simple to construct and can attain true steady-state current.5,18 Both planar microdisk electrodes (PMDs) and RMDs have been studied.17-24 A recessed microdisk resides at the bottom of a cavity whose walls are made of insulator material. Although the current measured at RMDs is typically less than that at PMDs of equal radius, it can be independent of convection outside of the cavity, depending on the cavity’s dimensions and the strength of the convective forces.18,19 RMDs were originally constructed from in-plane microdisk electrodes.17-19 Either chemical or electrochemical etching has (4) Fleishmann, M.; Pons, S.; Rolison, D. R.; Schmidt, P. Ultramicroelectrodes; Datatech Systems: Morganton, NC, 1987. (5) Oldham, K. B. In Microelectrodes: Theory and Applications; Montenegro, M. I., Queiros, M.A., Daschbach, J. L., Eds.; NATO ASI Series E, 197; Kluwer Academic Publishers: Boston, 1990; pp 35-50. (6) Wightman, R. M. Anal. Chem. 1981, 53, 1125A. (7) Buck, R. P.; Cosofret, V. V.; Erdosy, M.; Johnson, T. A.; Ash, R. B.; Neuman, M. R. Anal. Chem. 1995, 67, 1647-1653. (8) Bond, A. M.; Luscombe, D. L.; Davey, D. E.; Bixler, J. W. Anal. Chem. 1990, 62, 27-31. (9) Mirkin, M. V.; Shao, Y.; Fish, G.; Kokotov, S.; Palanker, D.; Lewis, A. Anal. Chem. 1997, 69, 1627-1634. (10) Antonenko, Y. N.; Bulychev, A. A. Biochim. Biophys. Acta 1991, 1070, 474480. (11) Matsue, T.; Shiku, H.; Yamada, H.; Uchida, I. J. Phys. Chem. 1994, 98, 11001-11003. (12) Pucacco, L. R.; Corona, S. R.; Carter, N. W. Anal. Biochem. 1986, 159, 43-49. (13) Akhtar, P.; Too, C. O.; Wallace, G. G. Anal. Chim. Acta 1997, 339, 211223. (14) Csoregi, E.; Gorton, L.; Marko-Varga, G.; Tudos, A. J.; Kok, W. T. Anal. Chem. 1994, 66, 3604-3610. (15) Padeste, C.; Kossek, S.; Lehmann, H. W.; Musil, C. R.; Gobrecht, J.; Tiefenauer, L. J. Electrochem. Soc. 1996, 143, 3890-3895. (16) McRipley, M. A.; Linsenmeier, R. A. J. Electroanal. Chem. 1996, 414, 235246. (17) Morita, K.; Shimizu, Y. Anal. Chem. 1989, 61, 159-162. (18) Bond, A. M.; Luscombe, D.; Oldham, K. B.; Zoski, C. G. J. Electroanal. Chem. 1988, 249, 1-14. (19) Tokuda, K.; Morita, K.; Shimizu, Y. Anal. Chem. 1989, 61, 1763-1768. (20) Morita, K.; Furuya, E. Anal. Chem. 1994, 66, 2197-2199. (21) Girault, H. H.; Ferrigno, R.; Brevet, P. F. Electrochim. Acta 1997, 42, 18951903. (22) Martin, C. R.; Brumlik, C. J.; Tokuda, K. Anal. Chem. 1992, 64, 12011203. (23) Osteryoung, J.; Aoki, K. J. Electroanal. Chem. 1981, 122, 19-35. (24) Heinze, J. J. Electroanal. Chem. 1981, 124, 73-86. 10.1021/ac980375r CCC: $18.00
© 1999 American Chemical Society Published on Web 12/22/1998
been used to etch the electrodes away from the surface plane of the insulator. The depth of the cavity and surface roughness of these electrodes are difficult to control. The early applications included chemical measurements in convective systems. Reproducible construction of RMDs using standard integrated circuit procedures7,25-28 (i.e., microfabrication) has been established more recently. The general principle behind this type of construction is the deposition and patterning of layers of materials. This is often accomplished through the use of photoactive polymers called photoresists. In general, the processing steps involved in microfabrication allow extensive control over dimensions and shape of one or multiple features that are vertical and lateral to the plane of the sample. Because of the batch-processing capabilities of microfabrication, it is well suited for transferring technology into commercial applications. The work reported here makes several new contributions to the current literature on microfabricated RMDs. First, the smallest of our microfabricated RMDs are smaller (14-µm diameter) than those reported by others (∼1 mm-200 µm), with greater depthto-diameter ratios (0.29 and 0.07, compared to 0.10 28 and 0.04 25), which should improve performance in convective systems. Scanning electron microscopy (SEM) was used to evaluate the general shape and quality of the cavities. Second, we present a detailed evaluation of the electrochemical responses of the microfabricated RMDs and compare them to models. The electrochemical response in a Ru(NH3)63+ solution is compared to models for linear and radial diffusion. Capacitance was determined from cyclic voltammetry in 0.5 M KNO3 and is compared to that for a macroelectrode. This latter comparison elucidates the quality of the seal between the insulator and the electrode. Third, evidence from chronoamperometry in static and stirred Ru(NH3)63+ solution shows the advantages of microfabricated RMDs vs PMDs in convective systems. EXPERIMENTAL SECTION Materials. All chemicals were reagent grade and used as received. Aqueous solutions were prepared with high-purity deionized water (Milli-Q, model RG). A gold coin (Credit Suisse, 99.99%) and a chromium-plated tungsten rod (R. D. Mathis) served as sources for thermal evaporation. Silicon wafers (5 in., (100)) were obtained from Silicon Quest International (Santa Clara, CA). Potassium nitrate, sulfuric acid, hydrochloric acid, nitric acid, and 30% hydrogen peroxide were purchased from Fisher Scientific. Hexaamineruthenium(III) chloride was obtained from Aldrich Chemical Co. Positive photoresist (AZ4330RS) and photoresist developer (AZ400K) were purchased from Hoechst-Celanese. Polyimide (Pyralin PI-2721) was purchased from DuPont. A gold 10-µm-diameter PMD (BioAnalytical Systems, BAS) was used for comparison to RMDs. Electrochemical Measurements. A BAS-100B potentiostat and PA-1 preamplifier with BAS-100W electrochemical software (Bioanalytical Systems, Lafayette, IN) were used to perform cyclic (25) Belmont, C.; Tercier, M.-L.; Buffle, J.; Fiaccabrino, G. C.; Koudelka-Hep, M. Anal. Chim. Acta 1996, 329, 203-214. (26) Madaras, M. B.; Popescu, I. C.; Ufer, S.; Buck, R. P. Anal. Chim. Acta 1996, 319, 335-345. (27) Kovacs, G. T. A.; Petersen, K.; Albin, M. Anal. Chem. 1996, 68, 407A412A. (28) Bratten, C. D. T.; Cobbold, P. H.; Cooper, J. M. Anal. Chem. 1997, 69, 253-258.
Figure 1. Diagram of a cross-section view (not to scale) of a single cavity.
voltammetry (CV) and chronoamperometry (CA). The electrochemical cell contained a Pt flag auxiliary electrode and Ag/AgCl (saturated KCl) reference electrode. For CV experiments, a solution of 5.0 mM Ru(NH3)63+ and 0.5 M KNO3 was purged with Ar (zero grade, AirGas, Radnor, PA) to minimize the oxygen content.. The experiment was repeated 256 times, and the results were averaged. Digital simulations were performed using Digisim (Version 2.1, BioAnalytical Systems) Stirring studies involved CA and were performed in an electrochemical cell seated on a Corning PC-320 stir plate and containing a 1/2-in. magnetic stir bar (Fisher Scientific). The solution was not purged prior to CA. The rotation rate was determined by counting the rotations of the stir bar over a given time period. Uncompensated resistance measurements were made by applying a dc potential of 0.1 V in a 0.5 M KNO3 solution. Potential steps, (0.25 V, were made around the dc potential, and the current was sampled at 54 and 72 µs after the step was applied. The experiment was repeated 256 times, and the results were averaged. Au macroelectrodes were (approximately 1 cm × 2 cm) made by depositing 15 Å of Cr followed by 2000 Å of Au on a separate oxidized Si wafer. Electrodes were diced to size by hand. These macroelectrodes (not the on-chip microfabricated macroelectrodes) were used in the capacitance studies. The preparation of the Si wafer and formation of a passivating oxide layer is the same as described below for the microfabrication. Construction of Recessed Microdisk Electrodes. The fabrication of RMDs was accomplished by forming a hole through a Au and polyimide layer, exposing an underlying Au disk. Figure 1 shows a cross-sectional diagram of an RMD. The top layer of Au, while not used in the electrochemical measurements, is essential in the fabrication process so that cavities with welldefined, vertical walls can be produced. Cavity devices that accommodate two individually addressable electrodes are described elsewhere.1,2 Here, we focus on the use of the same design for a more simplified electrode configuration. The fabrication of RMDs consists of four steps. The process is shown as a cross-section schematic in Figure 2. A 2-µm SiO2 film was grown on a Si wafer by thermal oxidation. The wafer was spin-coated with positive photoresist and exposed to UV light (400 W, 300 nm) through a photolithographic mask (HTA Photomask). The photoresist was developed, leaving the pattern of a series of parallel lines, which eventually become the contact leads and microdisk electrodes. A 15-Å Cr film, which serves as Analytical Chemistry, Vol. 71, No. 3, February 1, 1999
551
Figure 2. Fabrication schematic, cross-section view.
an adhesion layer, and 1000-Å Au layer were deposited on the photoresist by thermal evaporation (Edwards 306 Auto). The wafer was sonicated for 15 min in acetone, which dissolves the photoresist, causing liftoff of the metal on top and leaving behind parallel lines of Au/Cr. After drying for 30 min at 125 °C, the wafer was spin-coated with polyimide (4 µm thick). The polyimide was polymerized by exposure to a blanket of UV light and then cured at 150 °C for 30 min and 250 °C for 30 min to cross-link the polymer. Cr (15 Å) and Au (1000 Å) were deposited on top of the polyimide by thermal evaporation. The wafer was spin-coated with positive photoresist. The photoresist was patterned by UV exposure through a second photolithographic mask (HTA Photomask). The top layer of Au and Cr were etched with 50% aqua regia (1 HNO3/3 HCl). The remaining photoresist was stripped with acetone and the wafer was dried for 30 min at 125 °C. This step leaves a layer of Au/Cr covering the entire electrode substrate with the following areas left exposed: (a) one end of one microfabricated lead to serve as an on-chip microelectrode for experiments not reported here and (b) the opposite ends of all of the underlying Au/Cr lines to serve as contact pads. The wafer was spin-coated with photoresist and patterned using a third photolithographic mask (Photronics). This step leaves a circular opening through the photoresist over each line in the first gold layer. The topmost layer of Au/Cr was etched using radio frequency (rf) sputtering (5 min, 50 sccm Ar, 30 mT, 500 V). Then, the polyimide was etched with reactive ion etching (RIE) (13 min, 40 sccm O2, 10 sccm SF6, 300 mT, 300 W). Before use, the electrodes were cleaned by sonicating in acetone for 30 s. Figure 3 shows a top down view of a completed device. There are five electrodes, one macroelectrode and four cavities with possible diameters of 50, 10, 5, and 2 µm. Only the 50- and 10-µm cavities could be formed with this set of microfabrication conditions. Electrical contact was made through wires attached to the contact pads. SEM was performed with a Hitachi S-2300 scanning 552 Analytical Chemistry, Vol. 71, No. 3, February 1, 1999
Figure 3. Top down schematic of a single microfabricated device containing 5 electrodes. The cavities (50, 10, 5, and 2 µm in diameter) are too small to be shown to scale in this figure. (The macroelectrodes on the same devices as the cavities are for future experiments. They are not the same kind of macroelectrodes that were used for the capacitance studies described herein. See the Experimental Section for more details.)
electron microscope (20-kV accelerating voltage). A profilometer (Dektak 3030) was used to measure the polyimide thickness. RESULTS AND DISCUSSION Physical Characterization. RMDs were characterized by SEM to determine shape and dimensions. Figure 4 shows top down electron micrographs of RMDs of 14- and 55-µm diameter. The circle defines the edge of the disk at the bottom of the cavity. The larger cavity appears to have a smooth, circular opening at 1000× magnification. The opening of the 14-µm-diameter cavity appears less regular but it is at higher magnification (5000× magnification). The 55-µm RMD also shows slightly irregular edges upon higher magnification. The bright halo that appears in Figure 4a is probably due to charging effects. The measurement of this cavity was made to the inner limit of the halo and is consistent with measurements of the cavity opening by optical microscopy, which does not cause charging. The dark-gray halo that is around the lighter center disk in Figure 4b is likely due to good depth perception of the walls in the SEM. Thus, the cavity diameter was measured to the outer limit of the dark halo. When an optical microscope, which has poorer depth of field, is focused to the opening of the cavity, it shows a diameter that has a similar dimension. When focusing through the cavity, optical microscopy does not reveal any protruding polyimide lip between the top layer of Au and the RMD, which would provide an alternative explanation for the dark halo in the SEM. The average diameter of the small RMDs is 14 ( 0.28 µm (n ) 3 electrodes). The average diameter of the large RMDs is 55.2 µm ( 0.0 (n ) 3 electrodes). The small diameter of the cavities prevented the use of atomic force microscopy (AFM) or profilometry to measure the depth
Figure 5. Comparison of CV responses in 5 mM Ru(NH3)63+ and 0.5 M KNO3 of a (a, b) 14-µm RMD and (c, d) 55-µm RMD. The potential was cycled at 0.1 (a, c) and 50 V s-1 (b, d).
Figure 4. Scanning electron micrographs of (a) 55- and (b) 14-µm recessed disk microelectrodes that are 4-µm deep. The 14-µm RMD micrograph shows some irregularities along the walls of the opening.
directly. The tip of the profilometer was too large for the cavity. AFM showed a smooth surface around the cavity and a fairly smooth rim. However, because of the dimensions and shape of the cavity, the base of the AFM tip could not clear the rim of the cavity before the tip traversed to the opposite side. Thus, obtaining cavity depth or an image of the bottom was not possible under the conditions we used. Instead, to obtain an approximate measure of the depth, the thickness of the polyimide layer was measured using a profilometer where the contact pad had been etched clear. The thickness of the polyimide was consistently 4 µm. Faradaic Response. Cyclic voltammetry was used to characterize the electrochemical response of the RMDs. Ru(NH3)63+ was chosen as a probe because of its well-established electrochemical properties.29 One advantage of using disk microelectrodes is that true steady-state currents can be attained.5,6,18 Steadystate current is a result of constant flux to the electrode surface. For static systems it is obtained when the mass transport is dominated by radial diffusion. In a steady-state CV, at very slow scan rates where the diffusion layer is large relative to the size of the electrode, the current of the reverse scan should retrace that of the forward scan in a sigmoidal shape. Pseudo steady state occurs when the time scale is short enough and the diffusion layer is thin enough so that transport involves both linear and radial diffusion. Cyclic voltammograms in this scan rate regime are (29) Saveant, J. M. In Microelectrodes: Theory and Applications; Montenegro, M. I., Queiros, M.A., Daschbach, J. L., Eds.; NATO ASI Series E, 197; Kluwer Academic Publishers: Boston, 1991; pp 307-340.
expected to be sigmoidal, with a separation between the forward and reverse currents.5 Figure 5 shows CV responses from the 14-µm (a, b) and 55µm (c, d) RMDs. All four were obtained in 5.0 mM Ru(NH3)63+ and 0.5 M KNO3 at either 0.1 (a, c) or 50 V s-1 (b, d). At 0.1 V s-1, neither of the microelectrodes exhibits true steady-state behavior. The deviation occurs for two reasons. As electrode size increases, the contribution of linear diffusion to the total flux for a given time scale increases. This is the case in comparing CV responses from the 14- and 55-µm RMDs. Second, the walls of the cavity may prevent radial diffusion from occurring as long as the diffusion layer is within the cavity. At 50 V s-1, CVs for both sizes of RMDs show peaked behavior. The peak splitting is greater than the ideal 60 mV. This is probably due to uncompensated resistance in the system. The microelectrodes were further investigated to understand the effects of the cavity. The thickness of the diffusion layer is inversely proportional to the square root of the scan rate and can be approximated by eq 1,30 where x is the diffusion layer thickness,
x ) (2Dt)1/2
(1)
D is the diffusion coefficient for Ru(NH3)63+ (7.8 × 10-6 cm2 s-1),36 and we have defined t as the time spent on the reducing side of E°′. The scan rate at which the diffusion layer thickness is equal to the depth of the cavity (4 µm) is calculated to be 58.5 V s-1. Tokuda et al.19 analyzed RMDs using computer simulations. Three current regions are defined. At slow scan rates, the current should be independent of scan rate (i.e., steady state). The (30) Bard, A. J.; Faulkner, L. R. Electrochemical Methods: Fundamentals and Applications; John Wiley and Sons: New York, 1980. (31) Wang, J.; Park, D. S.; Pamidi, P. V. A.; White, C.; Renschler, C. L. J. Electroanal. Chem. 1997, 425, 179-182. (32) Aoki, K.; Akimoto, K.; Tokuda, K.; Matsuda, H.; Osteryoung, J. J. Electroanal. Chem. 1984, 171, 219-230. (33) Bowyer, W. J.; Odell, D. M. Anal. Chem. 1990, 62, 1619-1623. (34) Nagale, M. P.; Fritsch, I. Anal. Chem. 1998, 70, 2908-2913. (35) WehMeyer, K. R.; Wightman, R. M. J. Electroanal. Chem. 1985, 196, 417. (36) Licht, S.; Cammarata, V.; Wrighton, M. S. Science 1989, 243, 1176.
Analytical Chemistry, Vol. 71, No. 3, February 1, 1999
553
equation used to calculate this current was first established by Bond et al.18 and is shown below.
iss ) (4πnFC*Dr2)/(4L + πr)
(2)
where n is the moles of electrons per mole of analyte involved in the reaction, F is the Faraday constant (96 485 C/mol of electrons), C* is the concentration of Ru(NH3)63+, L is the depth of the cavity, and r is the radius of the disk. A decrease in current relative to PMDs of the same radius is expected. For PMDs (where L ) 0), the steady-state current is23
iss ) 4nFC*Dr
(3)
At faster scan rates, there will be a transition region where neither steady-state nor linear diffusion models alone completely apply. At fast scan rates relative to the depth of the cavity and area of the electrode, the current should follow the model for linear diffusion and be proportional to the square root of scan rate, ν1/2 (at 25 °C)30
ip ) (2.69 × 105)n3/2AD1/2ν1/2C* (πr2).
(4)
where A is the area of the electrode A scan rate study was performed to compare our RMDs with these models. The maximum current (imax) was measured from the CV responses. If no peak is present, then imax was measured as the plateau current. If a peak is present, then imax was measured to the peak. The faradaic current was corrected for charging current by subtracting the extrapolated charging current at the beginning of the CV response. In Figures 6 and 7, imax for the microelectrodes is compared to iss and ip from eqs 2 and 4 as a function of ν1/2. The scan rate (58.5 V s-1) is marked in the figures, corresponding to conditions when the diffusion layer is expected to equal the depth (eq 1). The top half (a) of each figure shows scan rates from 0.01 to 327 V s-1. The bottom half (b) is an expanded view of scan rates from 0.01 to 10 V s-1. For the 55-µm RMD (L ) 4 µm), steady-state current persists up to a scan rate of 0.1 V s-1. The steady-state current for the 55-µm RMD is 53.50 ( 0.48 nA (n ) 4 electrodes), that predicted by eq 2 is 34.9 nA, and that predicted by eq 3 is 41.4 nA. The 55-µm RMDs follow theory for PMDs better than for RMDs. If the diameter of the cavity is actually larger than determined by SEM, then a diameter of 71, not 55 µm, would be needed (using eq 3) to account for the experimental current of 53.5 nA, an error in diameter measurement of 29%, which is unlikely. Another possible explanation is that, at this scan rate, there is a noticeable but small component of linear diffusion (as is evident in the hysteresis in Figure 5c) which adds to the current. At faster scan rates, the current at the 55-µm RMD increases with scan rate. At 204 V s-1, where the diffusion layer is thin and the electrodes should follow theory for linear diffusion (eq 4), the maximum current is 784.6 ( 41.2 nA (n ) 4). The current predicted by eq 4 is 1274 nA. This deviation between measured and predicted current is discussed below. The transition away from steady-state current occurs at the same scan rate where the theoretical curves cross for ip and iss vs ν1/2. The transition region 554
Analytical Chemistry, Vol. 71, No. 3, February 1, 1999
Figure 6. Comparison of maximum current taken from CV in 5 mM Ru(NH3)63+ and 0.5 M KNO3 at a 55-µm RMD to radial and linear diffusion models. (b) is an expanded view of the region between 0.01 and 10 V s-1.
Figure 7. Comparison of maximum current taken from CV in 5 mM Ru(NH3)63+ and 0.5 M KNO3 at a 14-µm RMD to radial and linear diffusion models; (b) is an expanded view of the region between 0.01 and 10 V s-1.
where imax does not follow either eq 2 or eq 4 is similar to that predicted by Tokuda et al.19 For the 14-µm RMD, steady-state current persists up to a scan rate of 1 V s-1. The maximum current measured at the 14-µm
RMDs (5.39 ( 0.96 nA, n ) 5) fits the RMD model of eq 2 (6.10 nA) better than the PMD model of eq 3 (10.5 nA). For the 14-µm RMD, the curves for ip and iss vs n1/2 cross at 1 V s-1. Above this scan rate, the current increases with increasing scan rate. At 204 V s-1, the current measured at the 14-µm RMDs is 35.37 ( 9.51 nA (n ) 6). The current predicted by eq 4 is 82.58 nA. Again, there is a transition region between steady-state and linear models for this size of RMD, as expected. At fast scan rates, the current at both the 55- and 14-µm RMDs should be predicted by eq 4. Uncompensated resistance, Ru, was evaluated as a possible cause for the discrepancy between the experimental and theoretical current. The Ru was found to be 96.6 kΩ for the 55-µm RMD and 209 kΩ for the 14-µm RMD, which are higher than that at the 10-µm PMD (49 kΩ). The effect of Ru on current has been studied previously for RMDs17 and for recessed microelectrode ensembles,31 however, the resistance in those cases (15-30 kΩ) was not as high. The effects of uncompensated resistance on peak current was modeled by computer simulations. Simulations were based on the linear diffusion model, which should be appropriate at high scan rates. The electrode geometry in the simulations was a planar electrode of area equal to the area of the electrodes experimentally investigated. Although this is not an exact representation of the true geometry, it can serve as an approximation at fast scan rates. The peak current at two scan rates, 50 and 204 V s-1, was measured from simulations at both Ru ) 0 Ω and either Ru ) 96.6 or 209 kΩ, depending on electrode diameter. A simple ratio was obtained by dividing the current influenced by uncompensated resistance by the ideal current at Ru ) 0 Ω. A similar ratio was obtained from the experimental data by dividing the experimentally measured current by the current calculated from eq 4 for a given scan rate. A value of 1 implies no effect of Ru on current. For the 55-µm RMD, the experimental ratio at 204 V s-1 is 0.616. At 50 V s-1, the ratio is 0.784. The values from simulations are 0.809 and 0.885, respectively. For the 14-µm RMD, the experimental ratio at 204 V s-1 is 0.428 and at 50 V s-1 is 0.511. The values from simulations are 0.965 and 0.982, respectively. Consequently, the deviation of experimental current from the linear model at fast scan rates can be explained only partially by uncompensated resistance. Another possible explanation for the deviation is that the product of the total resistance of the electrochemical system (between auxiliary and working electrodes) and the current might be greater than the compliance voltage of the potentiostat ((12 V in our case). This is unlikely, because the theoretical current at the RMDs is too low to achieve this condition. Also, the 55-µm RMD, which has a larger current, shows a smaller deviation from theory than the 14-µm RMD. Another more plausible factor that might result in the low apparent faradaic current at fast scan rates is the subtraction of charging current. At high scan rates, charging current (which scales linearly with ν) can become large relative to the faradaic current (which is a constant for radial diffusion or scales with ν1/2 for linear diffusion). If the charging current, which is extrapolated from the beginning of the CV response, is overestimated even by a small amount, then the background-subtracted faradaic current would be significantly lower. This outcome would be more apparent at the 14-µm RMD at high scan rates, which
Figure 8. Quality of fabrication and seal as determined by the dependence of the area-normalized capacitance on scan rate. The charging current was measured from cyclic voltammograms in 0.5 M KNO3. Comparison is made to that from a Au macroelectrode.
has a larger radial diffusion component, and therefore weaker dependence of faradaic current on scan rate, than the 55-µm RMD. An alternative method for background subtraction that might minimize this problem is to subtract the CV response in pure electrolyte from the one in solution containing the redox species.32 Charging Current. Capacitance studies were used to evaluate the quality of the construction of the microelectrodes. The charging current was measured from CV responses in 0.5 M KNO3 electrolyte. The following equation was used to calculate the areanormalized capacitance.
C ) ic/νA
(5)
where ic is the charging current. The current was measured at 0.3 V vs Ag/AgCl (saturated KCl). Figure 8 shows a log-log plot of the capacitance as a function of scan rate for a Au macroelectrode, 14-µm RMD, and 55-µm RMD. Representation of capacitance data in this form has been used previously to determine the quality of microelectrode fabrication.33-35 This is because the dependence of capacitance on scan rate is influenced by the presence of submicroscopic cracks or defects in the insulating material around the electrode. It has been suggested that a small dependence of measured capacitance on scan rate, which was previously observed for conventionally sized electrodes, may also be due to these phenomena, but that the small circumference-toarea ratio makes the effect of defects in the seal less apparent.35 The results reported here in Figure 8 for the macroelectrode show a slope that is also shallow, like that reported earlier. The slopes for the RMDs are similar to that of the macroelectrode, indicating that the seal between the polyimide insulator and electrodes is good, and no cracking has occurred. The seal appears superior to that reported previously with edge band electrodes in which silicon nitride and epoxy were used as insulating layers.34 For a given scan rate, the absolute magnitude of the areanormalized capacitance of the 14- and 55-µm RMDs are within only a factor of 10 of each other and of the macroelectrode. These Analytical Chemistry, Vol. 71, No. 3, February 1, 1999
555
Figure 9. Comparison of chronoamperometric responses of a PMD and RMD in 5 mM Ru(NH3)63+ and 0.5 M KNO3. The solution is (a) static, (b) stirred at 70 rpm, and (c) stirred at 150 rpm.
differences are possibly caused by varying surface conditions of the RMDs from microfabrication, cleaning, and electrochemical use. Convection Studies. A simple set of experiments were conducted to demonstrate the utility of these microfabricated RMDs in convective systems. CA was carried out in a solution of 5 mM Ru(NH3)63+ and 0.5 M KNO3. The responses at both sizes of recessed disks were compared to that at a 10-µm PMD. Figure 9 is an overlay of the chronoamperograms at the 14µm RMD and the 10-µm PMD, tested in both static solutions and solutions stirred at different rates; they are representative of all replicates. In the static solution, Figure 9a, the current for the PMD is 7.30 nA, while the current for the RMD is 7.04 nA. The steady-state current measured in the static solution can be used to determine the area of each electrode. Using eq 3, the effective diameter of the PMD is 9.70 µm. Using eq 2 for the RMD, the effective diameter is 15.5 µm within a reasonable error to the 14 µm determined by SEM. Figure 9b shows the response in a solution that was stirred at 70 rpm. The current measured with the PMD increased to 8.80 ( 0.68 nA (n ) 2) and the RMD was 6.65 ( 0.63 nA (n ) 5), within error of the current in static solution. Figure 9c shows the responses in a solution stirred at 150 rpm. The current magnitude increases to 13.27 ( 0.37 nA (n ) 2) for the 10-µm PMD, whereas that for the 14-µm RMD (7.184 ( 0.23 nA (n ) 5)) appears to be unaffected by the increased stirring. The results at the 14-µm RMDs indicate their utility in applications where analyte concentrations can be determined without recalibrating the electrode between variations in solution convection. The average current for these electrodes was determined by measuring the average current in the steady-state region for each electrode and then averaging them together for all electrodes of a given geometry. The standard deviations reported above were calculated based on these average current measurements. The performance of the RMD as an electrochemical detector in convective solutions is best evaluated by determining the signal556 Analytical Chemistry, Vol. 71, No. 3, February 1, 1999
to-noise ratio (SNR). The SNR for a single electrode was calculated by dividing the steady-state current averaged over the last 4 s of the reduction step by the standard deviation of the steady-state current over the same period of time for the same electrode. This represents one SNR data point and this procedure was repeated for five RMDs and two PMDs to obtain an average SNR in the static solution and in both stirred solutions. For the 10-µm RMD, the SNR at 70 rpm is 37.6 ( 10.4 and at 150 rpm is 9.89 ( 1.41. The SNR for the 14-µm RMD at 70 rpm is 116 ( 24.7 and at 150 rpm is 46.0 ( 12.9. The improvement in SNR from PMD to RMD at high stirring rates is significant even though the depth of the RMD electrode is only 4 µm and the diffusion layer extends well beyond the opening. A similar set of experiments was carried out for the 55-µm RMD. Unlike the 14-µm RMD, the noise increased with stir rate just as if it were a PMD. This result is not surprising, although the depth of the cavity is the same. Because of the electrode’s larger area, the center is less protected from convection than for the 14-µm RMD. Others have reported the use of RMDs in convective solutions. Bond et al.18 reported noise-free CA with cavities of ∼90-µm depth and 25-µm diameter. Tokuda et al.19 reported a dependence of noise on cavity depth, with the noise disappearing around a cavity depth of 50 µm for electrode arrays with individual electrode diameters of 7 µm. The depth-to-diameter value of both of these systems is larger (3.6 and 7.1, respectively) than that of the microfabricated 14-µm RMDs (0.29) generated in this work. Despite this difference, relatively noise-free CA was obtained under our conditions. The noise that is present, especially at the 50-µm RMD, would be further diminished with a deeper cavity. CONCLUSIONS We have reported construction and characterization of RMDs using microfabrication technology. The seal between the polyimide and Au layers is excellent as determined by capacitance studies. At high scan rates, the experimental faradaic current deviates from theory. The electrodes suffer only partially from the iRu drop due to high uncompensated resistance which has a greater effect on current at the 55-µm than at the 14-µm RMDs, as verified by digital simulations. Subtraction of an overestimated background current from the total current appears to be the major contributor to the deviation. At slow scan rates, the current predicted from a steady-state model best fits the experimental current at these electrodes. Furthermore, the 14-µm RMDs may serve as excellent candidates for electrochemical detectors in convective systems because the decrease in SNR is minimal with increasing convection as compared to planar microdisk electrodes. ACKNOWLEDGMENT This research was financially supported by the National Science Foundation under Grants CHE-9308946, EHR-9180762, and CHE9624114 (CAREER) and by the University of Arkansas. The work of Ben Bowen on initial convection studies, supported by an NSFfunded Research Experience for Undergraduates program, is appreciated. Special thanks go to Dr. Tim Lenihan, Dr. John Shultz, and Mr. Errol Porter. Dr. Len Schaper is acknowledged for use of the engineering facilities at the High Density Electronics Center at the University of Arkansas. We are also grateful to BioAnalytical Systems for donating the Digisim software. Received for review April 2, 1998. Accepted October 21, 1998. AC980375R
