Anal. Chem. 2003, 75, 6555-6559
Diffusion of Redox Probes in Hydrated Sol-Gel-Derived Glasses. Effect of Gel Structure Mandakini Kanungo and Maryanne M. Collinson*
Department of Chemistry, 111 Willard Hall, Kansas State University, Manhattan, Kansas 66506-3701
The diffusion coefficients of redox probes entrapped in a silica matrix prepared by the sol-gel process were measured using a combination of cyclic voltammetry and chronoamperometry at an ultramicroelectrode. In this study, the porosities of the gels were varied to assess the importance of constrained environments vs intermolecular interactions on the translational mobility of guests entrapped in this solid host matrix. The average pore diameter of the gels was varied from 40 to 400 Å by utilizing different catalysts (HCl, NH3, NaF) or different silicon precursors (tetramethoxysilane or Ludox colloidal silica). The diffusion coefficients of cobalt(II) tris(bipyridine), ferrocenemethyltrimethylammonium ion, and dicyanobis(phenanthroline)iron(II) and their rate of change as the gel dried were found to be nearly identical for gels prepared from TMOS and catalyzed with either HCl, NH3, or NaF. When trapped in gels prepared from Ludox, ferrocenemethanol and potassium ferricyanide diffused at rates identical to that measured in solution. In contrast, Dapp for ferrocenemethyl(trimethylammonium) dropped 1 order of magnitude over a 30-day drying period. These results attest to the importance of intermolecular interactions in governing diffusion in sol-gel-derived materials. Mesoporous silicate materials can easily be prepared under ambient conditions using the sol-gel process.1,2 In a typical procedure, tetramethoxysilane (TMOS) is mixed with water in a mutual solvent such as methanol in the presence of a catalyst (i.e., HCl, NaF, or NH3). After a period of time that typically ranges from seconds to days, the sol gels.1-2 The average pore size and pore size distribution of the gel depends on a number of factors, most notably the type of catalyst used.1-4 Gels prepared from acidcatalyzed sols consist of linear chains that become entangled upon drying and yield a less porous gel.1-5 Base-catalyzed sols are more particulate in nature and yield a material that has greater interstitial porosity.1-5 By using Ludox colloidal silica (22-nm diameter) as the starting precursor in place of TMOS, gels that have even higher interstitial porosity can be prepared. * Corresponding author. Phone: 785-532-1468. E-mail:
[email protected]. (1) Brinker, J.; Scherer, G. Sol-Gel Science; Academic Press: New York, 1989. (2) Brinker, C. J. J. Non-Cryst. Solids 1988, 100, 31-50. (3) Brinker, C. J.; Scherer, G. W. In Ultrastructure Processing of Ceramics, Glasses, and Composites; Hench, L. L., Ulrich, D. R., Eds.; John Wiley: New York, 1984; Chapter 5, pp 43-59. (4) Collinson, M. M. In Handbook of Advanced Electronic and Photonic Materials, Vol. 5: Chalcogenides Glasses and Sol-gel Materials; Nalwa, H. S., Ed.; Academic Press: New York, 2001. (5) Buckley, A. M.; Greenblatt, M. J. Chem. Educ. 1994, 599-602. 10.1021/ac034658h CCC: $25.00 Published on Web 10/16/2003
© 2003 American Chemical Society
One of the attractive features of sol-gel-derived materials is that they can be readily used as stable host matrixes for the entrapment of specific reagents such as proteins, enzymes, organic dyes, and redox probes.6-9 The dopant can just simply be added to the sol prior to its gelation. Among numerous applications, solgel-derived materials have been used in solid-state electrochemical devices,10 chemical sensors,6-9,11,12 and catalysts.13 Of utmost importance to these and other applications is to understand the diffusivity and stability of the reagents trapped inside the solgel matrix. Diffusion in sol-gel-derived solids can be significantly more complex than that observed in pure liquids because both the constrained pore structure and the extent of surface interactions can influence the rate at which reagents can move in and out of the matrix.14-16 In previous work, Raman spectroscopy, time-dependent methods, and ATR FT-IR have been used to study diffusion in xerogel monoliths and thin films.17-21 Electrochemical investigations, particularly cyclic voltammetry and chronoamperometry, have been shown to be a promising and convenient approach for the in situ characterization of diffusion in hydrated or partially hydrated gels.16, 22-26 Both macroscopic electrodes and ultramicroelectrodes (UMEs) have been employed. UMEs offer the (6) Avnir, D. Acc. Chem. Res. 1995, 28, 328. (7) Lev, O.; Tsionsky, M.; Rabinovich, L.; Glezer, V.; Sampath, S.; Pankratov, I.; Gun, J. Anal. Chem. 1995, 67, 22A. (8) Dave, B. C.; Dunn, B.; Valentine, J. S.; Zink, J. I. Anal. Chem. 1994, 66, 1120A-1127A. (9) Collinson, M. M. Trends Anal. Chem. 2002, 21, 30-38. (10) Dunn, B.; Farrington, G. C.; Katz, B. Solid State Ionics 1994, 70/71, 3-10. (11) Wolfbeis, O. S.; Reisfeld, R.; Oehme, I. Structure and Bonding; SpringerVerlag Berlin, 1996; pp 51-98. (12) Lin, J.; Brown, C. W. Trends Anal. Chem. 1997, 16, 200-211. (13) Blum, J.; Avnir, D.; Schumann, H. Chemtech 1999, 29, 32-38. (14) Badjic, J. D.; Kostic, N. M. J. Mater. Chem. 2001, 11, 408-418. (15) Badjic, J. D.; Kostic, N. M. J. Phys. Chem. B 2000, 104, 11081-11087. (16) Howells, A. R.; Zambrano, P. J.; Collinson, M. M. Anal. Chem. 2000, 72, 5265-5271. (17) Watson, J.; Zerda, T. W. Appl. Spectrosc. 1991, 45, 1360-1365. (18) Wallen, S. L.; Nikiel, L.; Yi, J.; Jonas, J. J. Phys. Chem. 1995, 99, 1542115427. (19) Sieminska, L.; Zerda, T. W. J. Phys. Chem. 1996, 100, 4591-4597. (20) Koone, N., Shao, Y.; Zerda, T. W. J. Phys. Chem. 1995, 99, 16976-16981. (21) Rivera, D.; Harris, J. M. Anal. Chem. 2001, 73, 423. (22) Y. Zhang, Ph.D. Dissertation, University of North Carolina, Chapel Hill, 1991. (b) Oliver, B. N.; Coury, L. A.; Egekeze, J. O.; Sosnoff, C. S.; Keller, C.; Umana, M. X. In Biosensor Technology. Fundamentals and Applications; Buck, R. P., Hatfield, W. E., Umana, M., Bowden, E. F., Eds.; Marcel Dekker: New York, 1990. (23) Audebert, P.; Griesmar, P.; Sanchez, C. J. Mater. Chem. 1991, 1, 699-700. (b) Audebert, P.; Griesmar, P.; Hapiot, P.; Sanchez, C. J. Mater. Chem. 1992, 2, 1293-1300. (24) Cattey, H.; Audebert, P.; Sanchez, C.; Hapiot, P. J. Phys. Chem. B 1998, 102, 1193-1202.
Analytical Chemistry, Vol. 75, No. 23, December 1, 2003 6555
advantage of minimizing the cracking of the gels upon drying. In addition, their reduced double layer capacitance enables high sweep rates to be used, ensuring the integrity of the gel-electrode interface.26 UMEs also provide a relatively simple means for calculating the apparent diffusion coefficient, Dapp, without prior knowledge of the solution concentration, which is important because the concentration of the entrapped reagents changes as the gel dries due to solvent evaporation and subsequent gel shrinkage.16 The present work is a continuation of previous work in our laboratory in which we entrapped various redox probes (selected for their size and charge) in silica monoliths and studied their diffusion.16,26 Specific interactions between the redox probes and pore walls were deemed to be important. In the present work, the structure of the sol-gel network has been varied by utilizing different catalysts (e.g., HCl, NH3, and NaF) and different precursors (e.g., Ludox colloidal silica) in an attempt to more clearly understand what role porosity plays on the diffusion coefficient of reagents trapped within constrained environments. EXPERIMENTAL SECTION Reagents. Tetramethyl orthosilicate (TMOS 99%), ferrocenemethanol (Fc-CH2OH), and Ludox AS-40 colloidal silica were purchased from Aldrich. The Ludox colloidal silica consist of uniform discrete particles of silica of 22-nm diameter in an alkaline medium (pH 9). Dicyanobis(1,10-phenanthroline)iron(II) (Fe(CN)2(phen)20/+) was obtained from GFS chemicals, Inc. Hydrochloric acid, potassium ferricyanide (Fe(CN)63-/4-), potassium chloride, ammonium hydroxide,, and methanol were purchased from Fisher Scientific. The chemicals were used as received without any further purification. Cobalt(II) tris(bipyridine) (Co(bpy)32+/3+), cobalt(II)phenanthroline (Co(phen)32+/3+), and ferrocenylmethyl trimethylammonium hexaflurophosphate (FcN+PF6-) was synthesized as previously described.16 Water was purified to type I using a Labconco Water Pro PS four-cartridge system. Procedures. The microelectrode assembly consists of a glassencased Pt microelectrode (r ) 13.3 µm) and a Ag/AgCl electrode secured in a nylon cap with epoxy. A Parafilm plug blocked a ∼3mm-diameter hole in the cap.26 The preparation and polishing of Pt UMEs have been described previously.26 The HCl-catalyzed silica gels were prepared by combining TMOS with water, methanol, and 0.01 M HCl in the mole ratio of 1:10:3:2 × 10-4 Si/H2O/MeOH/HCl. The NaF-catalyzed gels were prepared using a Si/H2O/MeOH/NaF mole ratio of 1:10:3:2 × 10-4. The sols were stirred for 20-30 min, poured into polystyrene vials, capped with the microelectrode assembly, and secured in a Faraday cage. For the NH3-catalyzed materials, the sol was made as described above using HCl as the catalyst. After 20-30 min, 0.6 mL of 0.01 M NH3 was added to the ∼20 mL of sol. The sol was quickly poured into the polystyrene vial containing the electrodes and secured in the Faraday cage. The Ludox sols were prepared by mixing 15 mL of Ludox with 1.5 mL of MeOH (referred to as low-methanol gels) or 15 mL of MeOH (referred to as high-methanol gels). HCl was added to decrease the pH to ∼5.0-5.5 as measured with pH paper. (25) Cox, J. A.; Wolkiewicz, A. M.; Kulesza, P. J. J. Solid State Electrochem. 1998, 2, 247-252. (26) Collinson, M. M.; Zambrano, P. J.; Wang, H.; Taussig, J. S. Langmuir 1999, 15, 662-668.
6556 Analytical Chemistry, Vol. 75, No. 23, December 1, 2003
Figure 1. Relative change in the mass of gels prepared from Ludox colloidal silica (high-methanol preparation (O); low-methanol preparation (b)) during aging and drying.
The concentration of the redox species and supporting electrolyte in the gels were 3-5 mM and 0.15 M KCl, respectively. The gelation time for the HCl-, NaF-, and NH3-catalyzed gels was approximately 12 h, 45 min, and 10 min, respectively. The Ludox sols gelled within 30 min. The Parafilm plug was removed from the cap after the gels were 3-days old, and the gels were allowed to dry under high-humidity condition (65-70%) for 3-8 weeks. The pore size of the gels was determined after 3 weeks by measuring the N2 adsorption-desorption isotherms. The 3-weekaged gels were first converted to “ambigels” to minimize the collapse of the wet gel network during drying.27 This was done by first washing the gels three times with acetone with one wash per day to remove any residual water, alcohol, or unreacted precursors. After 3 days, the acetone-washed gels were solventexchanged with a hexane, a low surface tension solvent, for three additional washings of 1 day each. The hexanes-exchanged gels were then dried at room temperature for 1 day. Instrumentation. Cyclic voltammetry and chronoamperometry were performed with a BAS 100-W potentiostat equipped with a model PA-1 preamplifier. The data were collected at regular intervals as the samples were aged and dried. N2 adsorption and desorption isotherms was measured on a Quantachrome Autosorb 1MP analyzer with an equilibrium time of 2 min. All the samples were outgassed at 120 °C for 24 h. The surface area of the gel was obtained using multipoint BET. The average pore diameter of the gels were determined from the desorption branch using the BJH model. RESULTS All the gels were aged for 3 days inside a darkened Faraday cage and then were allowed to dry in a high-humidity environment (65-70% humidity). During this time, the gel contracts and pulls away from the sides of the polystyrene vials. The change in the mass of the gels can be recorded as the gel dried to obtain an idea about how much solvent is lost. For gels prepared from TMOS sols, ∼25% loss in mass was observed after a 30-day period.16 Figure 1 shows the relative change in the mass of ferricyanide-encapsulated Ludox gels during drying for the lowmethanol- and high-methanol-content gels. For the low-methanol gels, the drying was very slow and the mass loss was ∼7% after 30 days whereas for the high-methanol content the mass loss of the gels was 22-25%, which was comparable to that of TMOS gels. (27) Harreld, J. H.; Dong, W.; Dunn, B. Mater. Res. Bull. 1998, 33, 561-567.
Figure 2. Cyclic voltammetric curves for cobalt(II) tris(bipyridine) encapsulated in a sol-gel prepared using NaF as the catalyst just prior to gelation (D0) and 15 days (D15) after gelation. Scan rate, 2 mV/s.
The voltammetry of the redox species immobilized into different gels (i.e., HCl-catalyzed gels, NH3-catalyzed gels, NaFcatalyzed gels, and Ludox gels) was recorded as a function of time as the sols gelled, aged, and dried. Most of this work employed large redox probes as these have been shown to exhibit the largest change in Dapp when entrapped in the gel. The voltammetry was recorded both at high sweep rates and at low sweep rates as previously described.16,26 At high sweep rates, the peak current, ip, is proportional to (scan rate)1/2, indicative of planar diffusion, and at low sweep rates the steady-state current, iss, was found to be independent of the scan rate, indicative of radial diffusion and steady-state voltammetry.28 The experiments were stopped when a distortion of the characteristic sigmoidal or peak-shaped waveforms was observed, indicating that the gel was pulling away from the electrode surface.26 The gels were run typically for 3-5 weeks for TMOS gels and as long as 7-8 weeks for the low-methanol Ludox gels. Figure 2 shows representative CVs for Co(bpy)32+ encapsulated in a gel catalyzed by NaF at different drying times. At this sweep rate (2 mV/s), the steady-state limiting current is proportional to the product of Dapp and concentration (C). As the gel dries, solvent evaporates and the matrix shrinks, the concentration of dopant increases, and the surface area-to-volume ratio of the gel increases. If D decreases faster than C increases, the steady-state limiting current, iss, will decrease as the gel dries. With the exception of the negatively charged redox probes (i.e., Fe(CN)63-) entrapped in the TMOS gels, this is what is observed. The exact magnitude of the drop is highly dependent on the size and charge of the entrapped probe. In this case, the redox potential shifts negative by ∼70 mV, consistent with the stabilization of the oxidized form of the redox probe (Co(bpy)33+). The best way to evaluate diffusion in these gels is to calculate Dapp. This can easily be done with chronoamperometric data collected at an ultramicroelectrode. The normalized CA response of the ultramicroelectrode is given by the following equation,29
id(t)/iss ) 0.7854 + 0.4431(Dappt/r2)-1/2 + 0.2146 exp(-0.3911(Dappt/r2)-1/2) (1)
where Dapp is the diffusion coefficient of the redox probes, t is (28) Wightman, R. M.; Wipf, D. O. In Electroanalytical Chemistry; Bard, A. J., Ed.; Marcel Dekker: New York, 1989; Vol 15. (29) Shoup, D.; Szabo, A. J. Electroanal. Chem. 1982, 140, 237-245.
Figure 3. Normalized chronoamperometric response for cobalt(II) tris(bipyridine) encapsulated in a sol-gel prepared using NaF as the catalyst 2 h (a) and 24 days (b) after gelation. Solid lines: nonlinear regression fit to the experimental data (points), R 2 ) 0.98.
the time, and r is the radius of the microelectrode. Dapp can be determined by fitting the normalized CA data to the above equation. Figure 3 shows a sample chronoamperomgram for gelentrapped Co(bpy)32+ that has been normalized by iss obtained from the cyclic voltammogram recorded at 2 mV/s. The inset in Figure 3 shows the actual chronoamperometric response of Co(bpy)32+ obtained at 2 h (Dapp ) 1.4 × 10-6 cm2/s) and 24 days (Dapp ) 0.1 × 10-6 cm2/s) after gelation. Both the experimental points and the regression curve (R2 ) 0.98) are shown. An R2 value less than 0.96 usually indicates that the electrode might be starting to separate from the gel as noted in prior work,16 and subsequently, these data were not used in the plots below. The diffusion coefficients for the gel-entrapped redox species initially were around (1-2) × 10-6 cm2/s and dropped 1 order of magnitude after ∼5-30 days. Figure 4 shows a plot of the relative change in the value of Dapp for different redox probes vs drying time for the NH3-catalyzed TMOS gels. The relative Dapp values of HCl-catalyzed gels are shown for comparison. Nearly identical plots were obtained for the same redox probes in the NaF-catalyzed gels (figure not shown). Figure 5 shows the variations of Dapp versus drying time for FcCH2OH, Fe(CN)63-, and FcN+ trapped in the low-methanol Ludox gels. The diffusion coefficients were also obtained for Ludox gels with higher methanol content whose drying time more closely resembled the TMOS gels in terms of solvent loss. However, the current becomes suddenly low after aging/drying for 4-5 days. This was attributed to the separation of the colloidal silica particles from the electrode surface since the gel dries more rapidly when the methanol content is high. This was problematic for the Ludox gels more so than for the TMOS-based gels and is likely due to the fact that Ludox sols/gels consist of large colloidal particles. The surface area, pore volume, pore size, and pore size distribution was determined for the different gels after a 3-week period. The gels were first converted into ambigels in attempt to preserve as close as possible the original structure of the gel.27 Figure 6 shows the N2 adsorption-desorption isotherms for the TMOS gels. The isotherms are of type IV characteristic of mesoporous materials. The inset in Figure 6 shows the pore size distribution for the HCl-, NH3-, and NaF-catalyzed gels, respectively. There is a small overlap in the pore size distribution at small-pore radii for the HCl- and NH3-catalyzed gels but essentially Analytical Chemistry, Vol. 75, No. 23, December 1, 2003
6557
Figure 4. Relative variations in the apparent diffusion coefficients (D) of Co(bpy)32+ (A), FcN+ (B), and Fe(CN)2(phen)2 (C) with gel drying time for gels prepared from either HCl (b) or NH3 (]) as the catalyst. Error bars reflect standard deviations from individual measurements from three to six gels.
Figure 5. Relative variations in the apparent diffusion coefficients (D) of Fe(CN)63- (A), FcCH2OH (B), and FcN+ (C) with gel drying time for gels prepared with Ludox colloidal silica. Error bars reflect standard deviations from individual measurements from three to six gels.
none for the HCl-catalyzed gels versus the NaF-catalyzed gels. The isotherms obtained for the Ludox gels were similar in shape only shifted to larger pressures compared to the HCl-, NaF-, and NH3-catalyzed gels. There was no overlap in the pore size distribution plots for the Ludox gels versus the HCl-catalyzed gels. The surface area, pore volume, and average pore radius of TMOS and Ludox gels are shown in Table 1. DISCUSSION Diffusion in sol-gel-derived solids can be complex because it will depend on the porosity of the gel and the extent of interactions between an entrapped dopant and the walls of the matrix.14-21 A question that needs to be addressed is what is more important in governing diffusion in these materials: porosity or intermolecular interactions. In this work, we aim to address this issue by selectively varying the porosity of the gel and the type of reagent trapped within the network. The average pore size and pore size distribution of sol-gelderived solids can be varied to a large extent by changing the nature and concentration of the catalyst used in the polymerization process. Under acid-catalyzed conditions, the polymerization proceeds in a manner analogous to organic polymerization, whereby linear chains are formed.1-3 When a base or nucleophile (i.e., F-) is used, more particulate sols are formed. These sols give rise to materials that have significantly different structures.1-3 6558 Analytical Chemistry, Vol. 75, No. 23, December 1, 2003
Figure 6. N2 adsorption-desorption isotherms for 3-week-old ambigels prepared from HCl (b), NH3 (9), and NaF (2). Solid points, adsorption branch. Open points, desorption branch. Inset, pore size distributions.
In this work, we catalyzed the reactions with HCl, NH3, and NaF to produce gels with widely different pore sizes. In the case of the NH3-catalyzed materials, hydrolysis was initially started using HCl as the catalyst. NH3 was then added to predominately catalyze the condensation of the sol. As can be seen in Table 1, the average pore diameter of the NH3 gel is ∼1.5 times larger than the HCl-catalyzed gel, 38 Å. For the NaF-catalyzed gel,
Table 1. Results from N2 Adsorption-Desorption Isotherms gels
surface area, m2/g
pore vol, mL/g
av pore radius, Å
TMOS-HCl TMOS-NH3 TMOS-NaF Ludox-low MeOH
700-800 800-1000 800 50-100
0.6 1.2 1.6 1.2
19 28 48 200
greater than a 2-fold increase in pore size relative to HCl-catalyzed gels was observed with essentially no pores below ∼20 Å. These relative differences are consistent with the nature of the sol: more particulate sols will give rise to gels that have greater interstitial porosity. When the gel is prepared from a commercial sol consisting of large silica particles (22 nm), it has the largest porositys∼10 times higher than the HCl-catalyzed gel. Figures 4 shows a comparison of the rate of change of Dapp for Co(phen)32+, FcN+, and Fe(CN)2(phen)2 as gels prepared from sols catalyzed with HCl and NH3 were dried. Nearly identical results were obtained for the NaF-catalyzed gels. As expected, Dapp decreases as the gels dried due to a decrease in the porosity of the gel, an increase in the extent of intramolecular interactions due to the increase in surface area-to-volume ratio, or both. Large positively charged reagents such as Co(bpy)32+/3+ show a faster drop in Dapp compared to (Fe(CN)2(phen)2)0/+ and FcN+/0. The walls of the silicate matrix will be negative under these conditions (pI silica ∼2)30 so it is expected that cationic probes will strongly interact with the matrix. The approximate 2-fold increase in the porosity of gels measured when the gels were 3 weeks old does not seem to matter too much. The rate in change of Dapp for all three redox probes appears to be similar within error for the NH3 and NaF gels as compared to the HCl-catalyzed gel. This strongly suggests that the porosity of these hydrated gels does not govern how fast a molecule diffuses, but rather it is the extent of interactions between the walls of the matrix and the dopant. To further confirm this, gels were prepared from Ludox colloidal silica. As noted in Table 1, the average pore size is 10 (30) Iler, R. K. The Chemistry of Silica; Wiley: New York, 1979.
times higher than the HCl-catalyzed gels. For gel-encapsulated FcCH2OH and Fe(CN)63-, no change in Dapp was observed over a 30-40-day period. Because the matrix consists of negatively charged silica colloids, it is expected that these reagents will not interact with it and will reside in the center of the pore and diffuse at rates almost identical to that in solution. On the other hand, FcN+, which is similar in size to both FcCH2OH and Fe(CN)63-, shows an order of magnitude drop in Dapp over a 30-day period. Since the pores are at least 40 times larger than the size of the dopant, it is unlikely that this effect can be attributed to constrained geometries within the gel. Rather it can be ascribed to intermolecular interactions between the host and the guest. As the gels dry, the surface area-to-volume ratio increases and the extent of intermolecular interactions subsequently increases. Hence Dapp drops. CONCLUSIONS Diffusion in porous solids can be significantly more complex than that in solution. In the case of sol-gel-derived materials, intermolecular interactions (H-bonding, electrostatic, van der Waals) coupled with constrained environments can significantly impede diffusion. In this work, the porosity of the gels was varied more than 1 order of magnitude to evaluate its role on the diffusion of reagents trapped in these materials. These results indicate that intermolecular interactions between the guest and the host are the most important. This is particularly apparent in the gels prepared from Ludox, where the average pore size is significantly larger than the reagent (FcN+) yet the apparent diffusion coefficient is ∼1 order of magnitude smaller than it is in solution. The translational mobility of gel-entrapped reagents is very important in the development of solid-state electrochemical devices. This work suggests that the best way to improve diffusion in these materials is to control the interactions between the dopant and the pore walls. Future work will proceed in this direction. ACKNOWLEDGMENT We gratefully acknowledge support of this work by the Office of Naval Research. Received for review June 17, 2003. Accepted September 8, 2003. AC034658H
Analytical Chemistry, Vol. 75, No. 23, December 1, 2003
6559
