A Signal Amplification Scheme for Ultrasensitive Amperometric

---

Anal. Chem. 1999, 71, 5101-5108

A Signal Amplification Scheme for Ultrasensitive Amperometric Detection in Flowing Streams Philip T. Radford, Marla French, and Stephen E. Creager*

Department of Chemistry, Clemson University, Clemson, South Carolina 29634

A new signal amplification scheme for ultrasensitive amperometric electrochemical detection of redox-active molecules in quiescent solution and in flowing streams is described. The method is based upon a continuous regeneration of electrochemically oxidized analytes by reaction with a sacrificial electron donor in solution. It utilizes a selective coating on the electrode that is chosen to have properties which allow for relatively facile electrooxidation of analyte but which also inhibit electrooxidation of the sacrificial electron donor. Ultrasensitive detection of hydroxymethylferrocene (HMFc) as a model analyte using ferrocyanide as the sacrificial electron donor is demonstrated at a dodecanethiol-coated gold electrode. Signal amplification factors of several hundred to several thousand are obtained in flow injection mode for analyte injections in a concentration range between 10-4 and 10-7 M where peaks can be discerned both with and without amplification. Even higher amplification factors are estimated for analyte concentrations below approximately 10-8 M, for which peaks without amplification are undetectable. Amperometric detection of 60 million injected HMFc analyte molecules (corresponding to either a 10µL injection at 10-11 M or a 1.0-mL injection at 10-13 M) is demonstrated using the new method in flow injection mode. Amperometric detection strategies for redox-active molecules can be inherently limited by the fact that each analyte molecule can accept or donate only a relatively small number of electrons. Because of this limitation and because of the noise inherent in measuring very small currents (e.g., currents of less than a few hundred femtoamperes are difficult to measure reliably), detection limits for amperometric electrochemical detection are often not as low as they might be. Most strategies which have been proposed for improving detection limits involve some form of analyte recycling, either physically (i.e., regeneration at a second nearby electrode)1-7 or chemically (i.e., regeneration via a coupled * Corresponding author: (fax) 864-656-6613; (e-mail) screage@ clemson.edu. (1) Fenn, R. J.; Siggia, S.; Curran, D. J. Anal. Chem. 1978, 50, 1067-1073. (2) Weber, S. G.; Purdy, W. C. Anal. Chem. 1982, 54, 1757-1764. (3) Goto, M.; Zou, G.; Ishii, D. J. Chromatogr. 1983, 268, 157-167. (4) Matsue, T.; Aoki, A.; Abe, T.; Uchida, I. Chem. Lett. 1989, 133-136. (5) Niwa, O.; Xu, Y.; Halsall, H. B.; Heineman, W. R. Anal. Chem. 1993, 65, 1559-1563. (6) Bard, A. J.; Fan, F. F. Acc. Chem. Res. 1996, 29, 572-578. (7) Brooks, S. A.; Kennedy, R. T. J. Electroanal. Chem. 1997, 436, 27-34. 10.1021/ac990477x CCC: $18.00 Published on Web 10/15/1999

© 1999 American Chemical Society

chemical reaction).8-17 Recycling strategies in general are able to improve detection limits because they amplify the signal (thereby increasing sensitivity) but not the background. A characteristic feature of most chemical recycling strategies is that they require the presence of a reagent (or reagents) which can react with the oxidized or reduced form of analyte chemically, but for which direct reaction at the electrode is suppressed. It will always be the case that this reagent could react directly at the electrode at the potentials at which the analyte is oxidized or reduced. The recycling scheme will depend on this direct reaction being inhibited, usually because it is kinetically slow as a direct electrode reaction. In fact, it is often necessary to utilize some type of catalyst (e.g., a redox enzyme) to facilitate the recycling reaction. In such cases, the analyte is often either a cofactor for a redox enzyme or a redox mediator that delivers charge between an electrode and a redox enzyme. Amplification reactions that depend on enzyme catalysis can be very specific and can be applied in complex media; however, one must always consider the stability and reproducibility of the enzyme catalyst in such schemes. Also, turnover rates of redox-enzyme-catalyzed reactions are often not high, which can limit the amplification factors that are possible. We present here a relatively simple redox amplification scheme that can provide large signal enhancements for amperometric electrochemical detection of redox molecules in both quiescent solutions and flowing streams. The scheme utilizes simple redox molecules and requires no enzyme catalyst. As is always the case in such schemes, it does require that the medium include a quantity of a sacrificial electron donor that will serve to recycle the oxidized redox molecules via a solution-phase electronexchange reaction. A sample amplification scheme with hydroxymethylferrocene as analyte and ferrocyanide as the sacrificial (8) Doherty, A. P.; Stanley, M. A.; Leech, D.; Vos, J. G. Anal. Chim. Acta 1996, 319, 111-120. (9) Yao, T.; Suzuki, S.; Nakahara, T.; Nishino, H. Talanta 1998, 45, 917-923. (10) Cosnier, S.; Gondran, G. C.; Watelet, J. C.; DeGiovani, W.; Furriel, R. P. M.; Leone, F. A. Anal. Chem. 1998, 70, 3952-3956. (11) Moore, T. J.; Joseph, M. J.; Allen, B. W.; Coury, L. A. Anal. Chem. 1995, 67, 1896-1902. (12) Male, K. B.; Gartu, P. O.; Kamen, A. A.; Luong, J. H. T. Biotechnol. Bioeng. 1997, 55, 497-504. (13) Bauer, C. E.; Eremenko, A. V.; EhrentreichForster, E.; Bier, F. F.; Makower, A.; Halsall, H. B.; Heineman, W. R.; Scheller, F. W. Anal. Chem. 1996, 68, 2453-2458. (14) Blonder, R.; Katz, E.; Cohen, Y.; Itzhak, N.; Riklin, A.; Willner, I. Anal. Chem. 1998, 68, 3151-3157. (15) Katz, E.; Willner, I. J. Electroanal. Chem. 1996, 418, 67-72. (16) Lisdat, F.; Wollenberger, U. Anal. Lett. 1998, 31, 1275-1285. (17) Caruana, D. J.; Heller, A. J. Am. Chem. Soc. 1999, 121, 769-774.

Analytical Chemistry, Vol. 71, No. 22, November 15, 1999 5101

Figure 1. Illustration of the mechanism envisioned for redox amplification of hydroxymethylferrocene at a monolayer-coated electrode (right). The direct reaction of Fe(CN)64- is blocked by the monolayer (left).

electron donor is illustrated in Figure 1. The electrode reactions in this scheme are as follows:

HMFc f HMFc+ Fe(CN)64- + HMFc+ f Fe(CN)63- + HMFc

The reaction between hydroxymethylferricenium and ferrocyanide is approximately thermoneutral (the formal potentials are less than 10 mV apart) but should be relatively rapid since both molecules are characterized by relatively rapid electron selfexchange rate constants.18-21,22 Thus, this pair of reactants is wellsuited to an electrochemical amplification scheme for detecting ferrocene derivatives and ferrocene-tagged analytes. The scheme illustrated in Figure 1 requires that direct ferrocyanide oxidation at the electrode be inhibited but that direct hydroxymethylferrocene oxidation at the electrode occur rapidly. Coating the electrode with a self-assembled alkanethiolate monolayer provides this unusual reactivity pattern. It has been shown that such monolayers can serve as excellent barrier layers for preventing oxidation of highly charged and well-solvated metal complexes such as ferrocyanide,23-29 but that they are usually not good barriers for preventing oxidation of neutral and/or poorly solvated molecules such as most ferrocene derivatives.26,28,29 In the present case, the selectivity of a dodecanethiolate monolayer on gold is exploited to suppress the direct oxidation of ferrocyanide, thereby enabling the signal amplification scheme for hydroxymethylferrocene detection. EXPERIMENTAL SECTION Reagents. The buffer (pH 5) used in both quiescent solution and flowing stream experiments was composed of 0.10 M sodium (18) Yang, E. S.; Chan, M. S.; Wahl, A. C. J. Phys. Chem. 1980, 84, 3094-3099. (19) Yang, E. S.; Chan, M. S.; Wahl, A. C. J. Phys. Chem. 1975, 79, 2049-2052. (20) Nielson, R. M.; McManis, G. E.; Safford, S. K.; Weaver, M. J. J. Phys. Chem. 1989, 93, 2152-2157. (21) McManis, G. E.; Nielson, R. M.; Gocher, A.; Weaver, M. J. J. Am. Chem. Soc. 1989, 111, 5533-5541. (22) Shporer, M.; Ron, G.; Lowenstein, A.; Navon, G. Inorg. Chem. 1965, 4, 361364. (23) Chidsey, C. E. D.; Loiacono, D. N. Langmuir 1990, 6, 682-691. (24) Finklea, H. O.; Avery, S.; Lynch, M.; Furtsch, T. Langmuir 1987, 3, 409413. (25) Porter, M. D.; Bright, T. B.; Allara, D. L.; Chidsey, C. E. D. J. Am. Chem. Soc. 1987, 109, 3559-3568. (26) Groat, K. A.; Creager, S. E. Langmuir 1993, 9, 3668-3674. (27) Miller, C.; Cuendet, P.; Gratzel, M. J. Phys. Chem. 1991, 95, 877-886. (28) Creager, S. E.; Hockett, L. A.; Rowe, G. K. Langmuir 1992, 8, 854-861. (29) French, M.; Creager, S. E. Langmuir 1998, 14, 2129-2133.

5102

Analytical Chemistry, Vol. 71, No. 22, November 15, 1999

perchlorate (Alfa Aesar), 0.10 M acetic acid (Fisher), and 0.18 M sodium acetate trihydrate (Alfa Aesar). Hydroxymethylferrocene (HMFc) was purchased from Strem. Sodium ferrocyanide decahydrate (Na4Fe(CN)6‚10H2O) was obtained from Fluka Chemika and mixed with the pH 5 buffer to make a 10-4 M solution of ferrocyanide in the buffer. Because of to the sensitivity of ferrocyanide in solution to reaction with dissolved oxygen, this buffer solution was prepared fresh daily. Mineral acids used to mix the dilute aqua regia (1:3:4 HNO3(concentrated):HCl(concentrated):H2O by volume) for etching electrodes were obtained from Fisher. 1-Dodecanethiol for monolayer preparation was purchased from Aldrich. Reagents were used as received from their respective manufacturers. All water for aqueous solutions was deionized using a Barnstead Nanopure system to a resistivity of 17 MΩ cm. Preparation of Stock Solutions and Standards. A 10 mM HMFc stock solution in ethanol was prepared by adding a measured amount of solid HMFc to ethanol in a 10-mL volumetric flask. One hundred microliters of the 10 mM HMFc stock solution was diluted with 900 µL of buffer using two Wheaton Socorex Micropipets (100 µL and 1000 µL) to make a 10-3 M HMFc solution. The solution was placed in a new 1.5-mL polypropylene microcentrifuge tube (Fisher) and inverted and shaken 30 times to facilitate mixing. The second dilution was performed by taking 100 µL of the 10-3 M HMFc solution and adding 900 µL of the buffer to make a 10-4 M HMFc solution. This process was repeated until all the standards, 10-3-10-13 M HMFc, were prepared. Two sets of HMFc standards were made, one using pure buffer (no ferrocyanide) and another using buffer containing 10-4 M ferrocyanide. Ferrocyanide was included in both the running buffer and the injected solution in FIA experiments to avoid small changes in the ferrocyanide concentration as the injected plug passes through the detector. Electrode Construction and SAM Preparation. Electrodes for cyclic voltammetry were constructed of 0.127-mm Au wire (Alfa, Premion grade, >99.999% pure) encased in epoxy (Epon 825, Shell) that was cross-linked with 1,4-diaminocyclohexane (Aldrich) and hardened for 3 h at 80 °C.26 The electrodes were then sanded and successively polished by hand using 25-, 5-, and 1-micrometer alumina with rinsing and a 1 min etch in dilute aqua regia between each polishing step.28 The 3-mm diameter gold electrode from the flowcell was mounted in a Minimet 1000 Grinder/Polisher and polished using the same process. This treatment produced an uncontaminated and stable gold foundation upon which the monolayers could be formed.26 After rinsing with water and 2-propanol, the electrodes were suspended in a 1.0 mM solution of alkanethiol in ethanol for 20-24 h to form the monolayer. Electrochemical Detection Apparatus and Procedure. Cyclic voltammetry was performed using a computer-based CHInstruments model 660 electrochemical workstation. A threeelectrode configuration with a platinum wire auxiliary electrode and a Ag/AgCl/saturated KCl reference electrode in pH 5 buffer electrolyte was used. The potential was swept over a range from +0.0 to +0.7 V at a scan rate of 0.1 V s-1. The flow injection apparatus consisted of an ISCO model 2350 HPLC pump fitted with a Rheodyne model 7125 injection valve connected to an ESA CouloChem II electrochemical detector with

Figure 2. Cyclic voltammograms of ferrocyanide and/or hydroxymethylferrocene in buffer solutions at gold electrodes. Voltammetric scans of the top series are both acquired in solutions containing 1 × 10-4 M Fe(CN)64-, using a bare gold electrode (left) and the dodecanethiolatecoated electrode (right). Scans on the bottom series are acquired in solutions that contain 1 × 10-6 M HMFc with the SAM coated electrode; HMFc alone (left) and HMFc and 1 × 10-4 M Fe(CN)64- (right).

an ESA model 5041 flow cell (flow channel dimensions were 10 mm × 3 mm × 25 µm). The pump delivered buffer at a preset flow rate of 1 mL min-1. The buffer was continuously degassed by bubbling with house nitrogen prior to pumping through the FIA system. Hydrodynamic voltammetry was performed in the flow cell by setting the applied potential to the desired value, allowing the current to stabilize, and injecting a series of 10-µL aliquots of a 1 × 10-6 M HMFc solution both with and without 1 × 10-4 M ferrocyanide present. The voltammetry was conducted at both a bare gold electrode (etched for 1 min in dilute aqua regia) and at a dodecanethiol-coated gold electrode. The potential was stepped in +0.1 V increments for subsequent injections. Current-time traces following injections were recorded using a Lineseis Ly1610011 chart recorder. Following each injection at progressively positive potentials, a repeat injection at +0.1 V vs reference (an internal palladium-hydrogen electrode) was made to establish whether the changes in response with increasingly positive applied potential were reversible. Current vs time traces for different HMFc concentrations were recorded for the series of injections of progressively more dilute HMFc solutions in pure buffer (ferrocyanide-free) and also in buffer solutions containing 1 × 10-4 M ferrocyanide. A short guard column (SiO2-packed) was included between the injection port and the detector to help damp the pressure pulse originating from

the injection event. Detection was performed at an applied potential of +0.4 V vs reference for injections containing between 10-4 and 10-13 M HMFc for both buffer series. RESULTS AND DISCUSSION The series of cyclic voltammograms in Figure 2 illustrates the viability of the proposed electrochemical signal amplification scheme. In the first panel (top left), ferrocyanide in a pH 5 buffer solution containing 1 × 10-4 M ferrocyanide is oxidized and then reduced during a voltammetric scan at a bare gold electrode. This reaction has a formal potential of approximately +0.236 V vs Ag/ AgCl/saturated KCl and is chemically reversible (anodic and cathodic peak currents are nearly equal). The oxidation yields a peak current (Ipeak) of 48 nA. The second panel (top right) corresponds to the same experiment but at an electrode that has been coated with a dodecanethiolate monolayer. There is no observable current for ferrocyanide oxidation near the formal potential, even though the current scale has been made more sensitive by an order of magnitude. This indicates that the ferrocyanide electrooxidation reaction has been effectively “blocked” by the monolayer. This effect, which has been noted previously,23-29 reflects the fact that the monolayer prevents close approach of ferrocyanide ions to the electrode, which, in turn, causes the standard electron-transfer rate constant for ferrocyanide to be greatly diminished relative to that at an uncoated electrode. The Analytical Chemistry, Vol. 71, No. 22, November 15, 1999

5103

slight rise in current at potentials more positive than approximately +0.6 V probably reflects the onset of long-range electron tunneling as a mechanism for oxidizing ferrocyanide through the monolayer. This is an inherently slow process, as indicated by the fact that such a large positive overpotential is required to drive the oxidation. The lower left panel of Figure 2 shows a voltammogram corresponding to oxidation of a small concentration (1 × 10-6 M) of hydroxymethylferrocene (HMFc) at a dodecanethiolate-coated gold electrode in the pH 5 buffer. A small oxidative peak (Ipeak ) 0.10 nA) is detected, though the current scale had to be made more sensitive by another order of magnitude to detect the peak. The oxidative peak potential of approximately +0.39 V is shifted positive by approximately 130 mV relative to the reversible oxidative peak potential expected at a bare electrode. (The formal potential for hydroxymethylferrocene was independently determined to be +0.228 V vs Ag/AgCl/saturated KCl under these conditions.) The positive shift indicates that HMFc oxidation is slower at the monolayer-coated electrode than at a bare electrode. Both the positive shift in peak potential and the lack of a reductive peak on the return scan are typical of HMFc voltammetry at alkanethiolate-coated electrodes26,28 and apparently reflect the fact that HMFc+ reduction is even less rapid than HMFc oxidation at the coated electrode. Even so, HMFc oxidation is not blocked nearly as effectively as is ferrocyanide oxidation at the coated electrode. This fact, which has been noted previously,26 is thought to reflect the fact that HMFc is poorly solvated in water and therefore interacts much more strongly with the SAM surface than does ferrocyanide. The last panel (lower right) in Figure 2 illustrates the behavior at the SAM-coated electrode when both ferrocyanide (1 × 10-4 M) and HMFc (1 × 10-6 M) are present in the buffer. The peak current of 27 nA arising from the coupled HMFc/ferrocyanide oxidation is much larger (by 270 times) than that observed for HMFc alone, so much so that the current scale had to be changed back to the setting appropriate for the first panel of the figure to see the entire peak. The increase in current undoubtedly reflects the fact that the oxidized HMFc molecules have been recycled by the solution-phase electron-transfer reaction between oxidized HMFc and ferrocyanide. This recycling corresponds formally to a signal amplification factor of 270 for HMFc detection. Note that the peak current for the mediated reaction is relatively close (within a factor of 2) to that which was observed for the direct oxidation of ferrocyanide on bare gold (top left). This probably reflects the fact that the current in both cases ultimately becomes limited at least in part by concentration polarization of ferrocyanide. In general, the amplified current at any potential could be limited by any of the steps in the mediated reaction including the initial oxidation of HMFc at the electrode, the cross-reaction between HMFc+ and ferrocyanide, and mass transfer of HMFc and/or ferrocyanide to the electrode. Electrochemical detection is often employed in liquid chromatography, capillary electrophoresis, flow injection analysis, and other applications in which detection takes place in a liquid flow cell.30 Thus, it is of interest to examine the behavior of the HMFc/ ferrocyanide/alkanethiol-coated-gold system in an electrochemical flow cell. The amplification reaction can be quantitatively evaluated (30) Buchberger, W. Fresenius’ J. Anal. Chem. 1996, 354, 797-802.

5104

Analytical Chemistry, Vol. 71, No. 22, November 15, 1999

in such a cell in terms of an improvement in detection limit for injected analytes. The flow cell used in this work differs from the static cell used to acquire the data in Figure 2 in two important ways. First, the buffer solution in the flow cell is flowing past the planar gold disk in a laminar fashion such that mass transfer to the electrode is governed by coupled convective and diffusive transport. This will affect the concentration-distance profiles (and therefore the current) of both analyte and sacrificial donor in the diffusion and reaction layers near the electrode surface. Second, this flow cell incorporates an internal palladium hydride reference electrode, for which the reference potential is shifted by a fixed but uncertain amount relative to the Ag/AgCl/saturated KCl electrode used to acquire the data in Figure 2. Figure 3 presents three hydrodynamic voltammograms acquired in flow injection mode for a series of 10-µL injections of a 1 × 10-6 M HMFc solution without (top, middle) and with (bottom) the addition of 1 × 10-4 M ferrocyanide in the flowing buffer stream. The voltammograms were recorded by systematically varying the potential (Edc) applied to the working electrode of the flow cell for each injection. These experiments were performed to identify the plateau region in which the peak current reaches a limiting value with respect to applied potential; to observe the differences in half-wave potential for HMFc oxidation at bare gold, SAM-coated gold, and SAM-coated gold with amplification; and to explore the potential limits beyond which the monolayer stability is compromised. The top voltammogram in Figure 3 exhibits a half-wave potential of approximately -0.04 V and a limiting current plateau at potentials more positive than approximately +0.2 V. (Note that from this half-wave potential and the HMFc formal potential of +0.228 V vs Ag/AgCl/saturated KCl determined previously, we can estimate the potential of the internal reference electrode in the flow cell to be approximately +0.27 V vs Ag/AgCl/saturated KCl.) The middle voltammogram is similar to the top one except that it was acquired at a dodecanethiol-coated electrode. It exhibits a half-wave potential of approximately +0.12 V, which is shifted positive by 160 mV relative to that at the bare electrode. As discussed above, this reflects the fact that HMFc oxidation is slower at the coated electrode than at the bare electrode, such that a more positive applied potential is required to drive the reaction at the mass-transfer-limited rate. The bottom voltammogram in Figure 3 was acquired at the same electrode used to acquire the middle voltammogram, except that 10-4 M ferrocyanide was present in the injected solution and the running buffer. A further positive shift in half-wave potential (to +0.32 V) and a large increase in the limiting current are evident. Both effects are indicative of the amplification phenomenon discussed above. The increase in current reflects the analyte recycling brought about by reaction of oxidized HMFc with ferrocyanide, and the positive shift in half-wave potential reflects the need for an even more positive overpotential to continuously reoxidize the recycled HMFc at the mass-transfer limited rate. The stability of the monolayer at the more positive applied potentials was assessed by following each injection in the middle voltammogram of Figure 3 with an injection at an applied potential of +0.1 V, for which the current should be relatively sensitive to the presence of defects in the monolayer. It was found that the

Figure 3. Hydrodynamic voltammograms for HMFc detection in the flow cell. Ipeak vs EDC, for a 10-µL injection of 1 × 10-6 M HMFc for each potential using a bare Au electrode (top), dodecanethiolate-coated electrode (middle), and a dodecanethiolate monolayer with the addition of 1 × 10-4 M Fe(CN)64- in both the buffer and in the injection (bottom). The flowrate for the buffer is 1 mL/min.

current for this follow-up injection remained constant for applied potentials to +0.7 V, but that at potentials more positive than +0.70 V the current for the follow-up injection was higher than it was

prior to application of the extreme positive potential. This indicates that the monolayer barrier properties have been compromised by application of the positive potential. Thus, applied potentials Analytical Chemistry, Vol. 71, No. 22, November 15, 1999

5105

Figure 4. Representative current-time traces for injection of HMFccontaining buffer solutions of variable HMFc concentration in buffer. Top series is for ferrocyanide-free buffer, and middle and bottom series are for buffer containing 1 × 10-4 M ferrocyanide. Bottom series represents injections for ferrocyanide-containing buffer of solutions containing 1 × 10-9 M HMFc (left), 1 × 10-11 M HMFc (center), and no HMFc (right). HMFc concentrations for each injected solution (10µL injection volume) are listed on the panels.

were kept between +0.4 and +0.7 V in the dilution experiments described below. We now consider quantitatively the role of the amplification reaction in enhancing detection limits for HMFc. Figure 4 presents several representative current vs time traces acquired for injection of a series of HMFc-containing buffer solutions for which the HMFc concentration is systematically varied. (Note that a short SiO2-packed guard was included in the flow stream for these experiments between the injector and the detector. This was done to help suppress the background current pulsations associated with the reciprocating HPLC pump and the injection event.) The top row corresponds to detection in the absence of ferrocyanide and the middle and bottom rows to detection in the presence of 1 × 10-4 M ferrocyanide. Inspection of the traces quickly reveals that the peak currents for HMFc detection with ferrocyanide present are always larger than those for HMFc alone. For example, the peak current for a 1 × 10-6 M HMFc injection without ferrocyanide is approximately 0.18 nA, whereas that for a similar HMFc injection with 1 × 10-4 M ferrocyanide present is approximately 31 nA, an increase of over 170 times. This corresponds formally to an amplification factor of 170 for HMFc detection. Note that, for the particular traces shown in Figure 4, a steadystate background current of approximately 70 nA is passing at the detector electrode. This current is presumably associated with direct oxidation of ferrocyanide at the electrode, either by longrange electron transfer across the monolayer or at monolayer defect sites. (The fact that background current is seen at this 5106 Analytical Chemistry, Vol. 71, No. 22, November 15, 1999

electrode even though the same background current was apparently suppressed at the electrode in Figure 2 indicates that the monolayer on the electrode in the flow cell is not as effective as the monolayer on the electrode in Figure 2 at suppressing direct ferrocyanide oxidation.) Fortunately, even though background current in the flow cell is relatively large, it was also found to vary quite slowly with time. This makes it relatively easy to compensate for by simply adjusting the zero level on the detector, which in turn facilitates detection of small peaks associated with analyte injection. Peak detection in the absence of amplification was difficult for injected solutions in which the HMFc concentration was below approximately 1 × 10-8 M. In contrast, peaks could be readily detected for very much more dilute HMFc injections when ferrocyanide was present in the buffer stream to bring about amplification. This is illustrated in the bottom three traces of Figure 4, corresponding to injection of solutions containing 1 × 10-9 M HMFc, 1 × 10-11 M HMFc, and no HMFc. The peak is quite easily detected for the 1 × 10-9 M solution (which corresponds to injection of approximately 10 fmol or 6 billion HMFc molecules) and is smaller but still easily discernible from background for the 1 × 10-11 M solution (which corresponds to injection of approximately 0.1 fmol or 60 million HMFc molecules injected). There is clearly no peak present for injection of buffer containing no HMFc; also, no peak was observed for injection of a solution that had been progressively diluted until it contained only 1 × 10-15 M HMFc (data not shown). These latter two results indicate that the peaks obtained at low HMFc concentration are due to HMFc in the injected solution and not just to a baseline transient associated with the injection. It is impossible to directly calculate an amplification factor for the low-concentration HMFc injections since no peak could be observed in the absence of amplification. Even so, a lower limit of the amplification factor may be obtained by integration of the detected peak and comparison of the charge under the peak with the charge expected for detection of all of the injected HMFc with 100% coulometric efficiency. For example, the area under the peak for the 1 × 10-11 M injection corresponds to approximately 3 × 10-8 C of charge. The charge required to fully oxidize 60 million HMFc molecules is 1 × 10-11 C; therefore, the signal amplification factor must be at least 3000 for the injected molecules to enable passage of the observed amount of charge. In fact, the real amplification factor per HMFc molecule is probably greater than this, since most of the HMFc molecules probably pass through the detector without ever being oxidized, regardless of whether amplification takes place. Work in progress will aim to provide more quantitative estimates of the amplification factors at low concentrations and to provide a clear mechanistic picture of why the amplification factors appear to be so large at very low concentrations. Figure 5 presents a series of current vs time traces similar to those in Figure 4, except that the injection volume has been increased from 10 µL to 1 mL. As expected, the traces are much broader for all injections. Amplification of the HMFc signal still occurs; for example, a 1 × 10-7 M HMFc injection in pure buffer yields a peak current of less than 80 pA, whereas the current for the same injection with ferrocyanide in the buffer is over 420 nA, which corresponds to an amplification factor of more than 5000.

Figure 6. Peak current vs HMFc concentration for a series of 10µL injections (solid) and 1-mL injections (open) of HMFc in the absence (circles) and presence (triangles) of 1 × 10-4 M Fe(CN)64in the buffer.

Figure 5. Same as Figure 4 but with a larger, 1-mL, injection volume and alterations in HMFc concentration as noted on the panels.

The HMFc concentration for which a peak could be reliably detected was extended to 1 × 10-13 M for the larger injection volume, although, because of the increased injection volume, the absolute amount detected was still approximately 0.1 fmol or 60 million HMFc molecules injected. Note that for this series, injection of HMFc-free buffer (lower right panel) yields a current vs time trace which appears to contain a small ripple. This feature could be the result of a pressure transient associated with the injection pulse, the presence of an air bubble caught in the larger 1-mL injection loop, a small difference in the ferrocyanide concentration in the running buffer and the injected solution, or possibly even a trace amount of HMFc still retained along the side walls of the injection loop, syringe, or needle. Further work will be required to determine the cause of this effect. Even so, the magnitude of the ripple is much smaller than the peak observed for the 1 × 10-13 M HMFc injection, which indicates that the peak detected for the HMFc injection is significant. Figure 6 presents a set of log-log plots of peak current vs HMFc concentration for a series of 10-µL (solid) and 1-mL (open) injections in the absence (circles) and presence (triangles) of ferrocyanide in the buffer. Some unusual features are evident in these plots. For example, the slope of both plots for the HMFconly injections is approximately one, which indicates a direct proportionality between peak current and HMFc concentration. This is the expected result for electrochemical detection without amplification. The behavior was very different for the case of the amplified peaks; the slopes of both log-log plots with amplification are less than one in all regions, indicating a shallow and nonlinear dependence of the amplified current on HMFc concentration. Indeed, these log-log plots are not really even linear, but rather exhibit a higher slope (but still less than one) at high HMFc concentrations than at low HMFc concentrations. The most

curious result is that there appears to be an approach to a plateau region at low HMFc concentration, such that the peak current is becoming nearly independent of HMFc concentration. The cause of this unusual effect is not yet completely understood and is still under investigation. One possibility is that it may reflect a strong tendency of HMFc to adsorb onto the hydrophobic surface of the dodecanthiolate-coated electrode. The electrochemical response would be dominated by the adsorbed HMFc molecules since they could be rapidly reoxidized by the electrode. Also, adsorbed molecules would have relatively long residence times in the detector cell since fluid flow through the detector should be laminar and the volume elements near the electrode surface would move the slowest. These two factors could combine to make the amplification factor for such molecules particularly high. The observation that peak current is nearly independent of HMFc concentration in the plateau region could reflect the fact that in this concentration regime the dissolved HMFc molecules contribute little to the response, and the monolayer surface is saturated with adsorbed HMFc. We note that other factors may also play a role in causing this unusual behavior and that further work will be required to determine if this way of thinking about the detector response at low analyte concentrations is correct. Such work is in progress. Finally, we note that the experimental apparatus used to acquire these data is not yet optimal, and several improvements could be made. In particular, the flow pulsations caused by the reciprocating HPLC pump could be eliminated by use of a syringe pump, and elimination of the pulse-damping guard column from the flow stream would minimize dilution of the injected analyte before it reaches the detector. The use of a postcolumn mixing chamber to add the ferrocyanide into the flow stream after the injection but prior to detection would mitigate concerns about dilution of the ferrocyanide caused by the injection. This would enable analysis of HMFc-containing samples that do not already contain ferrocyanide. It seems likely that further improvements in performance will be possible with this second-generation apparatus. SUMMARY A new electrochemical amplification scheme for detecting very small amounts of redox-active molecules has been described. The Analytical Chemistry, Vol. 71, No. 22, November 15, 1999

5107

reaction involves “recycling” of oxidized analyte molecules via a solution-phase electron-exchange reaction with a sacrificial electron donor. The scheme relies heavily on the action of a selective monolayer coating on the electrode which suppresses the direct oxidation of the sacrificial donor but permits the facile oxidation of analyte molecules. The scheme is demonstrated for detection of hydroxymethylferrocene at a dodecanethiolate-coated gold electrode with ferrocyanide as the sacrificial electron donor.

5108

Analytical Chemistry, Vol. 71, No. 22, November 15, 1999

ACKNOWLEDGMENT Financial support of this work by the National Science Foundation (CHE 9616370) is gratefully acknowledged. Also, a generous donation of electrochemical detector hardware by ESA Inc., Chelmsford, MA, is acknowledged. Received for review May 4, 1999. Accepted July 30, 1999. AC990477X