Anal. Chem. 2004, 76, 3139-3144
Spectroelectrochemical Sensing Based on Multimode Selectivity Simultaneously Achievable in a Single Device. 17. Improvement in Detection Limits Using Ultrathin Perfluorosulfonated Ionomer Films in Conjunction with Continuous Sample Flow Sara E. Andria,† John N. Richardson,† Necati Kaval,‡ Imants Zudans,‡ Carl J. Seliskar,*,‡ and William R. Heineman*,‡
Departments of Chemistry, Shippensburg University, 1871 Old Main Drive, Shippensburg, Pennsylvania 17257, and University of Cincinnati, P.O. Box 210172, Cincinnati, Ohio 45221-0172
We report herein an attenuated total reflectance (ATR) absorbance-based spectroelectrochemical sensor for tris(2,2′-bipyridyl)ruthenium(II) ion [Ru(bpy)32+] that employs ultrathin (24-50 nm) Nafion films as the chargeselective layer. This film serves to sequester and preconcentrate the analyte at the optically transparent electrode surface such that it can be efficiently detected optically via electrochemical modulation. Our studies indicate that use of ultrathin films in tandem with continuous flow of sample solution through the cell compartment leads to a 100-500-fold enhancement in detection limit (10 nM) compared to earlier absorbance-based spectroelectrochemical sensors (∼1-5 µM); markedly shorter analysis times also result. We report the dependence of the measured absorbance on sample flow rate and Nafion film thickness, and also provide calibration curves that illustrate the linear range and detection limits of the sensor using a 24 nm film at a constant sample flow rate of 0.07 mL/min.
The characteristics of a chemical sensor that employs attenuated total reflectance (ATR) spectroelectrochemistry1 in conjunction with ultrathin (100 nm) should produce better results than thinner films, at least for absorbance studies. It was thought that thicker films could preconcentrate more analyte, thus producing a larger absorbance signal. In addition, thicker films serve to contain the evanescent wave such that it will not penetrate into the bulk solution. This is important because penetration of the evanescent wave into the bulk solution could compromise one mode of selectivity (i.e., film selectivity). However, some disadvantages of using thick films are also apparent. For example, the time required for the analyte to diffuse into the film far enough to be electro(18) Richardson, J. N.; Aguilar, Z.; Kaval, N.; Andria, S. E.; Shtoyko, T.; Seliskar, C. J.; Heineman, W. R. Electrochim. Acta 2003, 48, 4291-4299. (19) Zudans, I.; Seliskar, C. J.; Heineman, W. R. Unpublished results.
3140
Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
chemically modulated, and thus detected, is much longer than for thinner films due to slow diffusion in the film. Another issue that was considered was the type of optical detection that would produce the best results. It seemed obvious that a fluorescence-based sensor should be able to detect much lower concentrations of the analyte than the absorbance-based sensor. It would be ideal to use fluorescence for all detection purposes; however, few analytes fluoresce. With this in mind, we have developed an absorbance-based sensor to explore the issue of film thickness and its effect on detection limits, selectivity, and time of analysis. Furthermore, we have opted to modify the thinlayer flow cell used for the fluorescence studies for absorbance work, as reported previously.18 Therefore, by using Ru(bpy)32+ as the model analyte and a similar cell setup as used with the fluorescence-based sensor, comparisons may be readily drawn between the two sensors. We point out that Ru(bpy)32+ is a convenient probe for this and previous work because it readily absorbs and fluoresces in its reduced state, undergoes reversible electron transfer at the ITO electrode surface, and readily partitions into Nafion charge selective films. On the basis of our work with this model analyte, we hope to demonstrate the efficacy of both absorbance- and fluorescence-based spectroelectrochemical sensors to detection of other analytes of environmental interest that exhibit similar properties for sensing.
EXPERIMENTAL SECTION Chemical Reagents. The following reagents were used: tris(2,2′-bipyridyl)ruthenium(II) chloride hexahydrate ([Ru(bpy)3Cl2‚ 6H2O], Aldrich), potassium nitrate (Fisher), and Nafion (5% solution in lower aliphatic alcohols and water, Aldrich). All reagents were used without further purification. Ru(bpy)32+ solutions were prepared by dissolving the appropriate amounts in 0.1 M KNO3 solution (prepared with deionized water from a Barnstead water purification system). Indium tin oxide (ITO) coated glass slides (Corning 1737F and 7059, 11-50 Ω/sq, 130 nm thick film on 1.1 mm glass, Thin Film Devices, USA) with dimensions of 10 mm × 40 mm were used as OTEs. Preparation of Nafion Films on ITO Slides. The ITO slides were first scrubbed with soapy water (Alconox) and then thoroughly rinsed with deionized water. The cleaned slides were rinsed with methanol and dried in air prior to use. The Nafion films were prepared by using a Model 1-PM101DT-R485 spin coater from Headway Research, Inc. Films with thicknesses ranging from 24 to 416 nm were prepared using a Nafion solution, as summarized in Table 1. The Nafion solutions were diluted from a 5% stock solution with the appropriate ratio of a 2-propanol/ water mixture (Table 1). The slides were taped at the ends so that a 2 cm × 1 cm area was exposed for film coverage. A 100 µL portion of diluted Nafion was transferred onto the slide and spun at the appropriate rate for a suitable time (see Table 1). The spincoated slides were stored in air until ready for use. Dried films were rehydrated by soaking in aqueous electrolyte until a constant baseline was obtained prior to exposure to Ru(bpy)32+. Electrochemical Measurements. All cyclic voltammograms were acquired using a Bioanalytical Systems BAS 100-B electrochemical workstation. The three-electrode cell assembly included a Ag/AgCl reference electrode (BAS), a Pt mesh auxiliary
Table 1. Recipes for Preparation of Nafion Films of Varying Thickness film thickness (nm)a
vol % Nafionb
spin rate (rpm)c
24 50 70 134 198 416
1 2 2 2 5 5
6000 6000 3000 1000 4000 1000
a Film thickness values were measured using spectroscopic ellipsometry.21 b Commercial Nafion (5% in lower aliphatic hydrocarbons and water) was diluted with a 4:1 (v:v) 2-propanol/water mixture. c All films were spun for 30 s.
electrode, and a Nafion film-coated ITO glass slide as the working electrode. The uptake of the analyte into the film was observed over a 60 min period. All voltammograms were acquired using iR compensation. ATR Spectroelectrochemical Measurements. The thin-layer spectroelectrochemical cell used for all ATR absorbance experiments has been described previously;12,18 in the present configuration, the cell employs 3-4 internal reflections. The process through which the ATR measurements were obtained is as follows. Light from a tungsten-halogen lamp (Ocean Optics LS-1) was directed through a multimode optical fiber (Romack, 400 µm core step index, NA ) 0.22) to a collimating objective (Newport, 10×, NA ) 0.25). The collimated light was then passed in and out of the ATR cell by way of two Schott SF6 coupling prisms (Karl Lambrecht). The prisms were attached to the slide by a high refractive index mounting compound (Cargille Meltmount, n ) 1.702). The attenuated light was then focused by an objective lens (Newport, 20×, NA ) 0.40) into another optical fiber, which directed the light into a monochromator (SpectraPro 300i, Acton Research Corp., 0.3 m focal length) outfitted with a photon counting PMT. Data collection was achieved using Acton Research NCL monochromator controller electronics and Spectrasense software. A potentiostat (Bioanalytical Systems CV-27) was used to perform the electrochemical modulations. The solutions were introduced into the cell at various flow rates using a syringe pump (model 341 B, Sage Instruments). All data analysis and manipulation were carried out using commercial spreadsheet and graphics algorithms. RESULTS AND DISCUSSION We have recently reported a spectroelectrochemical sensor using fluorescence that yields limits of detection on the order of 10-12 M for the model analyte Ru(bpy)32+.12 The ultralow limits of detection result in part from preconcentration of the analyte in a very thin Nafion film coating the surface of the optically transparent electrode (OTE). The response time for sensing using this device was also rapid in comparison to previous designs because a thin-layer cell design was used in which analyte solution flowed continuously across the film, thus replenishing analyte at the film surface. The much thinner Nafion films used in the fluorescence study (12-50 nm thickness) appear to offer many advantages to practical sensing, especially when coupled with solution pumping in the cell.12 These improvements, in addition to the decreased response
times mentioned previously, include improved reproducibility from film to film and film regeneration by flushing with ethyl alcohol between analyses. Choice of Chemical System. In order to successfully demonstrate the function of the absorbance sensor, Ru(bpy)32+ was chosen as a model analyte. This species easily met the criteria required for a successful spectroelectrochemical probe. Specifically, this analyte undergoes facile electron transfer at the OTE surface and provides different spectral characteristics corresponding to its different oxidation states such that one state has a significantly higher absorbance than the other at the analytical wavelength (vide infra). In addition, the analyte exhibited stability in both of its oxidation states. Ru(bpy)32+ was also chosen because it was used in the previous fluorescence study.12 With the information gained from the experiments described in this paper, we are able to compare the performance characteristics of both types of sensors. Characterization of Ru(bpy)32+. Before characterization of the sensor was accomplished, it was necessary to evaluate the electrochemical properties of the Ru(bpy)32+/3+ couple. Panel A of Figure 1 illustrates the cyclic voltammetry associated with the uptake of 0.01 mM Ru(bpy)32+ into a 24 nm Nafion film on an ITO OTE acquired at a potential scan rate of 25 mV/s. We note also that the solution was continuously pumped across the film at a rate of 0.07 mL/min during the experiment. Apparent is the initial increase in both anodic and cathodic peak currents with each new potential cycle; this behavior is consistent with gradual uptake of analyte into the Nafion film. By the eighth cycle, uptake has reached equilibrium, and the film is considered to be loaded. The time required to load the film is relatively short (ca. 5.3 min). For the sake of comparison, uptake of 0.1 mM Fe(bpy)32+ into a thicker Nafion film (200 nm) without pumping as measured using cyclic voltammetry required about 80 min to achieve equilibrium.11 Panel B of Figure 1 illustrates the optical modulation monitored at 450 nm that results from cycling the potential of a 0.01 mM solution of Ru(bpy)32+ loaded into a 24 nm film between 0.8 and 1.3 V (vs Ag/AgCl ref) at a potential scan rate of 20 mV/s. Here, one can easily see that the complex ion absorbs in the reduced (2+) state, but remains nonabsorbing in the oxidized state (3+). Modulation between the two oxidation states therefore aids in the monitoring of the absorbance signal by introducing a mode of selectivity (i.e., selected potential window) to the sensor design. At any time, the magnitude of the optical signal depends on the penetration depth of the evanescent wave into the film. As a rule of thumb, the penetration depth is on the order of one wavelength of the probing radiation, but is dependent on many variables such as the angle of incidence and relative refractive indices of all layers (ITO, film) present on the glass substrate. Given the dimensions of the 24 nm films, it is likely that the evanescent wave extends beyond the film and into solution in this case. However, this has in practice not been shown to be detrimental to Ru(bpy)32+ detection because (1) the concentration of analyte loaded into the film is far in excess of that in the bulk of solution and (2) the evanescent wave decays exponentially as it extends from the substrate ITO surface. Therefore, optical signals stemming from species outside the film appear to be negligible compared to those from species in the film. Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
3141
Figure 2. Trend of the rate of uptake of 1 × 10-6 M Ru(bpy)32+ (in 0.1 M KNO3) into a 24 nM Nafion film on an ITO OTE with increasing flow rate. Rate of uptake is expressed as the slope of the initial linear region of a plot of absorbance vs time for each flow rate.
Figure 1. Panel A: cyclic voltammograms illustrating uptake of 1 × 10-5 M Ru(bpy)32+ in 0.1 M KNO3 into a 24 nm Nafion film coated onto an ITO OTE. The potential sweep rate was 25 mV/s. Panel B: voltammetrically induced optical modulation of 1 × 10-5 M Ru(bpy)32+ loaded into a 24 nm film taken at 450 nm and at a potential sweep rate of 20 mV/s between 0.7 and 1.3 V (vs Ag/AgCl ref); this illustrates the differences in the optical properties (i.e., molar absorptivity) between the oxidized and reduced forms of Ru(bipy)32+/3+.
Flow Rate and Sensor Response. Earlier absorbance-based ATR spectroelectrochemical sensors we have reported have employed a large-volume cell into which analyte solution was introduced.2 The solution was not stirred or circulated in any way once in the cell, so mass transport to the film was limited by diffusion. In efforts to miniaturize the ATR cell design, we have moved toward small thin-layer flow cells that require constant circulation of solution such that depletion of analyte does not occur.12,18 In the attempt to characterize the response time of the sensor (as defined by the time taken for analyte partitioning into the film to reach equilibrium), we have investigated its dependence on flow rate. The effect of solution flow rate on the sensor’s uptake time is demonstrated in Figure 2, which is a plot of slope of analyte uptake (AU/s) into a 24 nm Nafion film vs flow rate (mL/min). The flow rates investigated ranged from 0.01 mL/min to 0.5 mL/ min. It is obvious from Figure 2 that there is a direct correlation between rate of uptake and flow rate. As the flow rate was increased, the slope of the uptake also increased. However, the rate of increase, while quite pronounced at flow rates less than 0.07 mL/min, becomes essentially constant at a smaller slope at larger flow rates. We model this behavior on the basis of the 3142 Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
biphasic nature of the bulk solution/film interface. The uptake rate generally increases with solution flow rate because increased convection serves to decrease the thickness of the Nernst diffusion layer in the bulk solution (as in a rotated disk experiment), thus increasing the rate of transport of the analyte to the film surface. Once in the film, however, the optical response will depend upon the diffusion coefficient of the analyte in the film, as well as the film thickness (vide supra), thus eventually becoming independent of solution flow rate. With this model in mind, circulation of analyte solution in conjunction with use of ultrathin films should be instrumental in decreasing analysis times considerably, thus allowing lower analyte concentrations to be detected much more rapidly than before. Film Thickness and Sensor Response. To study the effects of film thickness on sensor response, we investigated the uptake of Ru(bpy)32+ into Nafion films of varying thickness at a constant flow rate of 0.07 mL/min. In this investigation, films ranging in thickness from 24 to 416 nm were tested; Figure 3 compares two of those films: a 24 nm film and a 70 nm film. The 24 nm film exhibits a much faster rate of uptake than the 70 nm film, as shown in panel A. After uptake equilibrium had been reached, the potential was cycled between 0.7 and 1.3 V to electrochemically modulate the analyte. Panel B of Figure 3 shows the absorbance signals that were recorded. The 24 nm film offered a much larger signal with the modulations returning to baseline giving a corresponding change in absorbance (∆A) of 0.21 AU. In contrast, the modulated optical signal for the 70 nm film was much smaller, giving a ∆A value of 0.085 AU, though we note that the absolute absorbance is larger in the case of the 70 nm film. The magnitude of the electrochemically modulated absorbance signal depends strongly on two factors: (1) the penetration depth of the evanescent wave into the film (as discussed earlier) and (2) the rate of diffusion of the analyte within the film. While theory predicts that for a purely optical sensor a larger signal should be obtained using a thicker film (i.e., film thickness > penetration depth of the evanescent wave) since it can accommodate more analyte at equilibrium, thinner films in effect may actually be more efficient for most spectroelectrochemical sensing applications. First, in order to electrochemically modulate the analyte in the film, it must diffuse to the OTE surface to be oxidized or reduced. Even if a thick film is fully loaded with analyte, optical modulation of the bulk of the analyte may not be complete because slow
Figure 4. Time-based electrochemically induced optical modulation illustrating the uptake of 5 × 10-8 M Ru(bpy)32+ into a 24 nm Nafion film spin-coated on an ITO OTE. The solution flow rate was 0.07 mL/ min, and the potential was cycled between 0.7 and 1.3 V (vs Ag/ AgCl ref) at 20 mV/s. Initially, 0.1 M KNO3 was pumped through the cell in order to establish an absorbance baseline; the Ru(bpy)32+/0.1 M KNO3 solution was introduced 250 s into the run.
Figure 3. Panel A: uptake of 1 × 10-5 M Ru(bpy)32+ into 24 and 70 nm Nafion films on ITO OTEs taken at a constant solution flow rate of 0.07 mL/min. Panel B: time-based modulation of Ru(bpy)32+ in the 24 and 70 nm films after uptake taken by scanning the potential from 0.7 to 1.3 V (vs Ag/AgCl ref) at a potential sweep rate of 20 mV/s. The supporting electrolyte was 0.1 M KNO3.
diffusion in the film limits the mass transport of analyte to the electrode surface over the time frame of the voltammetric sweep such that only the fraction nearest the electrode surface is actually detected. Furthermore, optical sensing is most efficient in the region just adjacent to the electrode surface due to the exponential decay of the evanescent wave into the bulk of the film as noted earlier. This reasoning explains why in this case the thicker 70 nm film actually gave a smaller modulated optical signal than the 24 nm film, even though the absolute absorbance was larger in the case of the thicker film. We note also that ultrathin films, specifically those of 24 and 50 nm thickness, yielded much more reproducible film-to-film uptake profiles and ∆A values once equilibration had been reached than thicker films. In some instances the thick films provided larger signals than the thin films, but more often than not, the magnitudes of the modulated signals were quite limited. Much of this lack of reproducibility can be attributed to variations in film structure or porosity that appear to plague the thicker films due to inconsistencies in curing time or temperature after they are spin-coated onto the OTE surface. Detection Limits and Calibration Curves. In addition to characterizing the response time of the sensor, we also explored its linear calibration range and limits of detection. To do this, the modulated uptake of Ru(bpy)32+ into a 24 nm Nafion film was obtained as a function of concentration. We were able to detect an absorbance signal for the lowest concentration tested, 10 nM, by monitoring uptake for 5000 s. The absorbance signal for the
50 nM Ru(bpy)32+ sample is illustrated in Figure 4. Initially, the film was bathed in circulating 0.1 M KNO3 electrolyte solution in order to establish a baseline. After about 250 s, the 50 nM Ru(bpy)32+/0.1 M KNO3 solution was introduced at a constant flow rate of 0.07 mL/min; by ca. 650 s, the baseline begins to shift upward. The shifting baseline indicates the initial detectable partitioning of analyte into the film. After approximately 900 s, one begins to note the presence of modulations building in above the noise. These modulations continue to grow with time as the film continues to load; we note that by 5000 s, when the experiment was terminated, the film was still not at uptake equilibrium. The data in Figure 4 emphasize the fact that the combined use of ultrathin films and analyte solution circulation allow for detection of analyte concentrations using ATR absorbance spectroelectrochemical techniques far lower (i.e., 2 orders of magnitude) than reported previously. Such results are possible because of the combined advantages of rapid mass transport to the film surface, minimal film thickness to minimize diffusion time to the electrode surface, and intensity of the evanescent wave within the first few nanometers from the OTE surface. Figure 5 illustrates the calibration curves compiled from this study using a 24 nm Nafion film. Both curves are plots of ∆A vs concentration. In panel A, ∆A is the difference in the peak-tovalley absorbance caused by electrochemical cycling between 0.8 and 1.3 V at a potential scan rate of 20 mV/s taken near 1500 s. In panel B, ∆A is a measure of the unmodulated absorbance at 1500 s and the initial baseline absorbance. Ultimately, these calibration curves illustrate the linear relationship between ∆A and concentration up to 1 µM Ru(bpy)32+ in addition to the reproducibility of the runs as evidenced by the error bars shown. Each error bar represents 1 standard deviation on either side of the mean of three trials using a single film. In fact, all of the data in each calibration curve were acquired using a single film; the film was thoroughly rinsed by circulating ethyl alcohol through the cell after each uptake until the original baseline was obtained in 0.1 KNO3. Each individual concentration was run in random order such that no memory effects would be present. These calibration curves provided limits of detection of 50 and 10 nM for panels A and B of Figure 5, respectively. Though the Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
3143
Figure 5. Calibration curves for uptake of different concentrations of Ru(bpy)32+ into a 24 nm Nafion film. Absorbances were taken at 1500 s. The change in absorbance (∆A) measures the difference between the peak and the valley of the modulation (panel A) and the peak and the initial baseline (panel B). For panel A, the potential was cycled between 0.7 and 1.3 V (vs Ag/AgCl ref) at 20 mV/s; no modulation was employed for acquisition of data in panel B. The solution flow rate was held constant at 0.07 mL/min in both cases.
unmodulated data provided a lower limit of detection, we note that one mode of selectivity, namely, potential, was sacrificed. Possible reasons for the difference in observed detection limit are 2-fold: (1) modulation of the ITO itself has been shown to interfere with the ∆A measurement at low concentration,4 and (2) the time scale of the electrochemical modulation may still be too short to exhaustively electrolyze all of the analyte in the film. The obtained detection limits, however, are even more remarkable when one realizes that Ru(bpy)32+ has a relatively small molar absorptivity (14600 L/mol-cm at 452 nm)12 when compared to other analytes we have investigated in Nafion films. CONCLUSIONS We have demonstrated in the work reported here that limits of detection for Ru(bpy)32+ on the order of 10-50 nM can be (20) Slaterbeck, A. F.; Ridgway, T. H.; Seliskar, C. J.; Heineman, W. R. Anal. Chem. 1999, 71, 1196-1203.
3144 Analytical Chemistry, Vol. 76, No. 11, June 1, 2004
achieved using an absorbance-based ATR spectroelectrochemical sensor employing ultrathin Nafion charge selective films in tandem with constant flow of analyte to the film. The magnitude of the analytical signal was also found to be highly dependent upon sample flow rate into the spectroelectrochemical cell and film thickness. Another advantage afforded by this improved sampling approach is considerably shorter analysis times compared to those reported in previous studies, especially at higher concentrations (>100 nM). While the limits of detection obtained in this work were considerably higher than those obtained using our recent fluorescence-based sensor, we emphasize that we have demonstrated a 100-500-fold improvement over earlier absorbance-based sensor designs. We also note that the absorbance sensor complements the fluorescence sensor since many potential analytes of interest do not exhibit measurable fluorescence. We expect that several improvements could be made to the sensor that would serve to further lower its limits of detection. For example, the present design uses a multiple internal reflectance medium that exhibits only 3-4 reflections, whereas previous designs employed a larger medium that possessed twice as many reflections. Toward this end, use of a medium employing a higher density of reflections (such as a conductive waveguide) would go a long way toward improving the limits of detection. Furthermore, since the uptake time required to achieve a measurable absorbance has been decreased with pumping and thinner films, signal averaging20 could become a viable means of lowering detection limits. We are presently working toward determination of Fe2+ in environmental water samples as a means of testing the sensor using real samples. In this case, we fall back on previous work in which the ligand 2,2′-bipyridine is preloaded into the Nafion film such that the colored complex ion Fe(bpy)32+ is formed in situ in the film. Studies such as these will also allow us to further investigate the selectivity of the sensor against the myriad ions in such samples that could potentially act as interferences. ACKNOWLEDGMENT The authors gratefully acknowledge support from the Environmental Management Science Program of the U.S. Department of Energy, Office of Environmental Management under Grant DEFG07-99ER62311-70010.
Received for review December 11, 2003. Accepted March 25, 2004. AC035467H
