Imparting Biomimetic Ion-Gating Recognition Properties to Electrodes

the degree of protonation−deprotonation of carboxylic groups at the interparticle ... Differences in structural networking, pH-tuning, and elect...
0 downloads 0 Views 323KB Size
Download PDF
Anal. Chem. 2000, 72, 2190-2199

Imparting Biomimetic Ion-Gating Recognition Properties to Electrodes with a Hydrogen-Bonding Structured Core-Shell Nanoparticle Network Wenxia Zheng, Mathew M. Maye, Frank L. Leibowitz, and Chuan-Jian Zhong*

Department of Chemistry, State University of New York at Binghamton, Binghamton, New York 13902

This paper presents findings of the creation of biomimetic ion-gating properties with core-shell nanoparticle network architectures. The architectures were formed by hydrogen-bonding linkages via an exchange-cross-linking-precipitation reaction pathway using gold nanoparticles capped with thiolate shell and alkylthiols terminated with carboxylic groups as model building blocks. Such network assemblies have open frameworks in which void space is in the form of a channel or chamber with the nanometer-sized cores defining its size, the geometric arrangement defining its shape, and the shell structures defining its chemical specificity. The formation of the network linkages via head-to-head hydrogen-bonded carboxylic terminals and the reversible pH-tuned structural properties between neutral and ionic states were characterized using infrared reflectance spectroscopic technique. The biomimetic ion-gating properties were demonstrated by measuring the pH-tuned network “open-close” responses to charged redox probes. Such redox responses were shown to depend on the degree of protonationdeprotonation of carboxylic groups at the interparticle linkages, core sizes of the nanoparticles, and charges of the redox probes. Differences in structural networking, pH-tuning, and electrochemical gating properties were identified between the network films derived from nanoparticles of two different core sizes (2 and 5 nm). The mechanistic correlation of these structural properties was discussed. These findings have added a new pathway to the current approaches to biomimetic molecular recognition via design of core-shell nanoparticle architectures at both nanocrystal and molecular scales. Biological membranes are built by protein or cell building blocks that form channels and receptors. Ion channels are membrane-spanning proteins that form a pathway for the flow of inorganic ions across cell membranes and for the conversion of external sensory signals to the electrical language of the nervous system. The channels can be closed and opened in response to chemical or biological stimulus so that the passage of ions or molecules can be regulated. The development of artificial mem* To whom correspondence should be addressed: (phone) 607-777-4605; (e-mail) [email protected].

2190 Analytical Chemistry, Vol. 72, No. 10, May 15, 2000

branes that mimic such ion-gating channel or receptor properties has attracted enormous interest.1-4 Many approaches are centered on the utilization of interfacial assemblies based on the selective binding of synthetic or biological receptors to analytes. Both intramolecular and intermolecular channels in such assemblies were exploited for molecular recognition applications.1-4 Despite the abundant demonstration, the ability to use well-defined building blocks to precisely manipulate the molecular recognition properties is largely limited. In this paper, we report a novel approach to the development of this ability, which explores the chemical reactivity at well-defined core-shell nanoparticles in terms of size, shape, and shell structural properties. Our goal is to impart biomimetic ion channel properties at electrode surfaces while gaining insights into the core-shell chemistry. Metallic or semiconductive nanoparticles with organic shell encapsulation 5-7 are interesting because the shell functionalities and the nanometer-sized cores offer potential technological applications including microelectronics, optic devices, magnetic materials, catalysis, and molecular recognition.8-10 Nanoconstruction using such materials as building blocks via molecular-linking strategies has recently been demonstrated, including, for example, stepwise layer-by-layer assembling10-13 and a DNA-based linking method.14 While there are a few examples demonstrating noncovalent linkages, most do not allow subsequent structural tuning because of the covalent nature of linkages. Noncovalent chemical (1) Fuhrhop, J.-H., Ed. Membrane and Molecular Assemblies: The Synkinetic Approach; Royal Society of Chemistry: Cambridge, U.K., 1994; p 149. (2) Mallouk, T. E.; Gavin, J. A. Acc. Chem. Res. 1998, 31, 209. (b) Crooks, R. M.; Ricco, A. J. Acc. Chem. Res. 1998, 31, 219. (3) Xiao K. P.; Buhlmann, P.; Umezawa, Y. Anal. Chem. 1999, 71, 1183. (b) Buhlmann, P.; Aoki, H.; Xiao, K. P.; Amemiya, S.; Tohda, K.; Umezawa, Y.; Electroanalysis 1998, 10, 1149. (4) Goldenberg, L. M.; Bryce, M. R.; Petty, M. C. J. Mater. Chem. 1999, 9, 1957. (5) (a) Hostetler, M. J.; Murray, R. W.; J. Curr. Opin. Colloid Interface Sci. 1997, 2, 42. (b) Chen, S. W.; Ingram, R. S.; Hostetler, M. J.; Murray, R. W.; Schaaff, T. G.; Khoury, J. T.; Alvarez, M. M.; Whetten, R. L. Science 1998, 280, 2098. (6) Kiely, C. J.; Fink, J.; Brust, M.; Bethell, D.; Schiffrin, D. J. Nature 1998, 396, 444. (7) Hu, K.; Brust, M.; Bard, A. J. Chem. Mater. 1998, 10, 1160. (8) Fendler, J. H., Ed. Nanoparticles and Nanostructured Films, Wiley-VCH: Weinheim, 1998. (9) Martin, C. R.; Mitchell, D. T. Anal. Chem. 1998, 70, 322A. (10) Shipway, A. N.; Lahav, M.; Blonder, R.; Willner, I. Chem. Mater. 1999, 11, 13. (11) Brust, M.; Bethell, D.; Kiely, C. J.; Schiffrin, D. J. Langmuir 1998, 14, 5425. (b) Brust, M.; Bethell, D.; Schiffrin, D. J.; Kiely, C. J. Adv. Mater. 1995, 7, 795. 10.1021/ac9912909 CCC: $19.00

© 2000 American Chemical Society Published on Web 03/24/2000

bonding is however crucial in chemical and biological interactions. For example, base pairing of nucleic acids and helix-coil transition of DNA depend on how such binding properties are involved in various biological organizations. The DNA-linked gold nanoparticle assembly via complimentary oligonucleotide-oligonucleotide interactions demonstrated by Mirkin and co-workers14 serves as an excellent example. Hydrogen bonding is one of the most common noncovalent interactions in chemical and biological systems which we view as an ideal model toward building functional nanomaterials at core-shell nanoparticles that mimic biological membrane properties. Shell functionalization and derivatization of metallic nanoparticles have recently been demonstrated, particularly by Murray and co-workers on exchangeplacement reactivities15 and by Schiffrin and co-workers on esterification reactivities.11,16 We have also recently exploited the core-shell chemistry for manipulating nanoparticle size and shape17 and for constructing network thin films,18 which has important implications to the novel approach reported herein. Chart 1 illustrates the basic concept using a hypothesized core-shell network architecture with ideal (111)-type packing. Depending on the packing, e.g., (111) and (100) types, the size and shape of the void frameworks of different geometry, e.g., 3-fold and 4-fold, are manipulated by the size of the cores and the geometric arrangements. For example, a (111)-type packing with an assumed interparticle shell length of 3.6 nm would ideally generate a net void area of 3.0 nm2 for 5-nm core size and of 1.3 nm2 for 2-nm core size. Importantly, such sites can be manipulated for molecular recognition in terms of size. The chemical specificity can also be manipulated by structure and functionality of the encapsulating or bridging shell components. As will be demonstrated in this work, void sites based on hydrogen bonding of carboxylic groups as illustrated by Chart 1 serves as an ideal model. Gold nanoparticles capped with carboxylic acid-terminated thiolates in solutions were previously reported.19 A rough estimate of the shell thiolates for a 5-nm core yields ∼350 carboxylic terminals18 as hydrogen-bonding sites toward strong supramolecule-like network architectures. Importantly, the tuning of functionality, charge, or binding in the core-shell architecture offers an intriguing pathway to impart molecular recognition properties to a desired substrate. (12) Musick, M. D.; Pena, D. J.; Botsko, S. L.; McEvoy, T. M.; Richardson, J. N.; Natan, M. J. Langmuir 1999, 15, 844. (b) Freeman, R. G.; Garbar, K. C.; Allison, K. J.; Bright, R. M.; Davis, J. A.; Guthrie, A. P.; Hommer, M. B.; Jackson, M. A.; Smith, P. C.; Walter, D. G.; Natan, M. J. Science 1995, 267, 17. (13) Demaille, C.; Brust, M.; Tsionsky, M.; Bard, A. J. Anal. Chem. 1997, 69, 2323. (14) Elghanian, R.; Storhoff, J. J.; Mucic, R. C.; Letsinger, R. L.; Mirkin, C. A. Science 1997, 277, 1078. (b) Mucic, R. C.: Storhoff, J. J.: Mirkin, C. A.: Letsinger, R. L. J. Am. Chem. Soc. 1998, 120, 12674. (15) Hostetler; M. J.; Green, S. J.; Stokes, J. J.; Murray, R. W. J. Am. Chem. Soc. 1996, 118, 4212. (b) Hostetler, M. J.; Templeton, A. C.; Murray, R. W. Langmuir 1999, 15, 3782. (16) Bethell, D.; Brust, M.; Schiffrin, D. J.; Kiely, C. J. J. Electroanal. Chem. 1996, 409, 137. (17) Maye, M. M.; Zheng, W. X.; Leibowitz, F. L.; Ly, N. K.; Zhong C. J. Langmuir 2000, 16, 490. (b) Zhong, C. J.; Zheng, W. X.; Leibowitz, F. L.; Eichelberger, H. H. Chem. Commun. 1999, 13, 1211. (18) Leibowitz, F. L.; Zheng, W. X.; Maye, M. M.; Zhong, C. J. Anal. Chem., 1999, 71, 5076. (19) Weisbecker, C. S.; Merritt, M. V.; Whitesides, G. M. Langmuir 1996, 12, 3763. (b) Johnson, S. R.; Evans, S. D.; Brydson, R. Langmuir 1998, 14, 6639.

Chart 1. Schematic Illustration of the Void Frameworks in (111)-Type Packing of Core-Shell Nanoparticle Networka

a The magnified view illustrates the chemical properties of the shell using carboxylic acid terminal as the model.

We have indeed discovered ion-gating properties using such core-shell nanoparticle architectures. The preparation of the architectures, following the abundantly demonstrated “placeexchange” chemistry of core-shell nanoclusters in solutions15 and our recent “exchanging-cross-linking-precipitation” chemistry for assembling thin films,18 exploits the shell hydrogen-bonding properties for forming a continuous network. The understanding of how the resulting size and shape of the network void sites depend on nanoparticle core size and shape, shell functionality and linkage, and structural tuning has broadly important implications for molecular recognition. We present electrochemical and infrared reflectance spectroscopic evidence to correlate the structure-recognition relationship. It is important to emphasize that the structural definition by both the core’s nanosize dimension and the shell’s molecular structure is unique, and such architectures are different from micrometer-sized materials or selfassembled monolayers on planar substrates.20 EXPERIMENTAL SECTION Chemicals. The main chemicals used included decanethiol (DT, 96%), octyldecanethiol (ODT, 98.5+%), 11-mercaptoundec(20) Liu, Y.; Zhao, M.; Bergreiter, D. E.; Crooks, R. M. J. Am. Chem. Soc. 1997, 119, 8720. (b) Chailapakul, O.; Crooks, R. M. Langmuir 1995, 11, 1329. (c) Bilewicz, R.; Sawaguchi, T.; Chamberlain, R. V.; Majda, M. Langmuir 1995, 11, 2256. (d) Steinberg, S.; Tor, Y.; Sabatani, E.; Rubinstein, I. J. Am. Chem., Soc. 1991, 113, 5176. (e) Schierbaum, K. D.; Weiss, T.; Thoden van Velzen, E. U.; Engbersen, J. F. J.; Reinhoudt, D. N.; Goepel, W. Science 1994, 265, 1413. (f) Flink, S.; Boukamp, B. A.; van den Berg, A.; van Veggel, F. C. J. M.; Reinhoudt, D. N. J. Am. Chem. Soc. 1998, 120, 4652. (g) Rojas, M. T.; Koniger, R.; Stoddart, J. F.; Kaifer, A. E. J. Am. Chem. Soc. 1995, 117, 336. (h) Li, M.; Wong, K. M.; Mann, S. Chem. Mater. 1999, 11, 23.

Analytical Chemistry, Vol. 72, No. 10, May 15, 2000

2191

anoic acid (MUA, 97%), hydrogen tetracholoroaurate (HAuCl4, , 99%), tetraoctylammonium bromide (TOABr, 99%), and sodium borohydride (NaBH4, 99%). Other chemicals included toluene (99.8%), hexane (99.9%), ethanol (99.9%), potassium bromide (99+%), hexaamineruthenium(III) chloride ([Ru(NH3)]6Cl3, 98%), potassium ferricyanide (K3[Fe(CN)]6, 99+%), potassium chloride (99+%%), and potassium hydroxide (99.99%). All chemicals (from Aldrich) were used as received. Water was purified with a Millipore Milli-Q water system. Synthesis. Gold nanoparticles of 2-nm core size encapsulated with an alkanethiolate monolayer shell (Au2-nm) were synthesized by the standard two-phase method.21,22 Briefly, AuCl4- was first transferred from aqueous HAuCl4 solution (10 mM) to toluene solution by the phase-transfer reagent TOABr (36 mM). Thiols, e.g., DT, were added to the solution at a 2:1 mole ratio (DT/Au), and an excess (12×) of aqueous NaBH4 was slowly added for the reaction. The produced DT-encapsulated Au nanoparticles (∼1.9 nm) was subjected to solvent removal and multiple cleanings using ethanol. Nanoparticles of 5-nm core size (Au5-nm) were derived from the above Au2-nm by heating-induced evolution in size and shape, details of which were recently described.17 This processing route produced particles with cores narrowly sized at 5.2 nm. The final products were subsequently suspended in ethanol and centrifuged at least 4 times to ensure a complete removal of solvent and possible byproducts. Preparation. Although details of the preparation of the thinfilm formation are described elsewhere,24 the general procedure is briefly summarized as follows. Decanethiolate-encapsulated gold nanoparticles of 2- (Au2-nm) and 5-nm (Au5-nm) core sizes and COOH-terminated alkanethiols were used as the networking precursors. While we have studied COOH-terminated thiols of several chain lengths, this report focuses on the results from MUAbased systems. The thin films were prepared via an “exchangingcross-linking-precipitation” route,18 which involved an exchange of the MUA with the gold-bound thiolates followed by cross-linking and precipitation via hydrogen bonding at the terminals. The nanoparticles and MUA were mixed in a hexane or toluene solvent with a controlled ratio (typically, [MUA]/[Au5-nm] 50-500 with 0.1-0.5 µM Au5-nm and 0.05-0.2 mM MUA). The thickness of the precipitated thin films was controlled by immersion time. The films were thoroughly rinsed with the solvent and dried under argon. Three types of substrates, i.e., metals, glass, and glassy carbon, were used for the thin-film preparation. Glassy carbon (GC) disks, polished with 0.03-µm Al2O3 powders, were mainly used for electrochemical measurements. Precleaned glass slides were used for transmission UV-visible measurement. Gold substrates, (21) Brust, M.; Walker, M.; Bethell, D.; Schiffrin, D. J.; Whyman, R. J. Chem. Soc., Chem. Commun. 1994, 801. (b) Brust, M.; Fink, J.; Bethell, D.; Schiffrin, D. J.; Kiely, C. J. J. Chem. Soc., Chem. Commun. 1995, 1655. (22) Hoetetler, M. J.; Wingate, J. E.; Zhong, C. J.; Harris, J. E.; Vachet, R. W.; Clark, M. R.; Londono, J. D.; Green, S. J, Stokes, J. J.; Wignall, G. D.; Glish, G. L.; Porter, M. D.; Evans, N. D.; Murray, R. W. Langmuir 1998, 14, 17. (b) Hostetler, M. J. Zhong, C. J.; Yen, B. K. H.; Anderegg, J.; Gross, S. M.; Evans, N. D.; Porter, M. D.; Murray, R. W. J. Am. Chem. Soc. 1998, 120, 9396. (23) Smalley, J. F.; Chalfant, K.; Feldberg, S. W. J. Phys. Chem. B. 1999, 103, 1676. (24) Han, L.; Zheng, W. X.; Maye, M. M.; Leibowitz, F. L.; Zhong, C. J. Manuscript in preparation.

2192

Analytical Chemistry, Vol. 72, No. 10, May 15, 2000

including gold thin films evaporated on Cr-primed glass slides and polycrystalline gold disks, were used for infrared reflectance spectroscopic characterization. The surfaces were precleaned by immersion in 1:3 H2O2 (30%)-H2SO4 (concentrated) solution and rinsing in deionized water and ethanol. (Caution: the H2O2-H2SO4 solution reacts violently with organic compounds and should be handled with extreme care). The gold surface was precoated with an octyldecanethiolate monolayer, which effectively enhances the film adhesion because of its strong hydrophobic interaction with the core-shell nanoparticles.24 Instrumentation. Infrared reflectance spectra (IRS) were acquired with a Nicolet 760 ESP FT-IR spectrometer that was purged with boil-off from liquid N2. The spectrometer was equipped with a liquid nitrogen-cooled HgCdTe detector and a variable-angle specular reflectance device. IRS measurements were performed in an external reflection mode using p-polarized light at an incident angle of 82° with respect to the surface normal. A gold slide coated with octadecanethiolate-d37 monolayer was used as the reference. The spectral deconvolution was based on the convergence of a Marquardt-Levenberg algorithm to values that gave the best least-squares fit to the spectral data. The peak position, absorbance, width, and area were determined by spectral deconvolution of the IRS data based on a Lorentzian-type band shape. Electrochemical measurements were performed using a computer-interfaced potentiostat (EG&G model 273A). A threeelectrode cell was employed, with a platinum coil as the auxiliary electrode and an Ag/AgCl (saturated KCl) electrode as the reference electrode; all potentials are given with respect to this reference. Disk electrodes included glassy carbon (geometric area, 0.28 cm2) and gold (geometric area, 0.55 cm2). All electrolytic solutions were deaerated with high-purity argon before the measurement. The pH was adjusted by adding 0.11 M NaOH into a 0.1 M KClO4 + 0.01 M H3PO4 solution (constant ionic strength 23). UV-visible spectra were acquired with a HP 8453 spectrophotometer. The nanoparticle samples were supported as thin films on glass slides for transmission measurements. Transmission electron microscopy (TEM) was performed on Hitachi H-7000 electron microscope (100 kV). The thin films were deposited on a carbon-coated copper grid sample holder. Conductivity measurements were performed using an interdigitated microelectrode array (IME, Microsensor Systems) as the substrate on which the thin film was deposited.18 Quartz crystal microbalance (QCM) measurements were performed on an EG&G instrument. AT-cut quartz crystals (9 MHz, 5.7 ng/cm‚Hz) with gold electrodes (geometric area, 0.63 cm2) were used. RESULTS AND DISCUSSION General Structural Properties of the Network Films. This section describes briefly the morphological, structural, and electronic properties for the MUA-linked network films derived from the 5- and 2-nm core sizes, denoted as MUA-Au5-nm and MUA-Au2-nm, respectively. A detailed assessment of the film growth is reported elsewhere.24 In general, the reaction pathway is similar to the one-step exchange-cross-linking-precipitation reactivity recently reported for dithiolate-Au nanoparticle films.18 The distinction is in the interparticle reactivity defined by hydrogen bonding at CO2H groups, as represented by

Figure 1. TEM images of the network thin films for MUA-Au5-nm (A) and MUA-Au2-nm (B). The films were prepared by short immersion times (∼3 h) to ensure a low coverage that can be imaged by TEM.

Aum-SC10 + HSC10CO2H f Aum-SC10CO2H f ... f [-Aum-SC10CO2H‚‚‚HO2CC10S-]nV in which the exchange between the bonded SC10 and the solution HSC10CO2H yields MUA encapsulation and a further hydrogenbonding cross-linking to precipitate a continuous network film. The films were uniform and stable and displayed colors ranging from blue to green depending on thickness. The films were insoluble in toluene or hexane, suggestive of the displacement of the original hydrophobic DT shell by the hydrophilic MUA shell. The conductivity of the MUA-Au5-nm film was found to be in the order of ∼10-4 S/cm, ∼3 orders of magnitude lower than a typical nonanedithiol(NDT)-linked Au5-nm film.18 The conductivity for MUA-Au5-nm film was even smaller due to the core size effect.18 The low conductivity is qualitatively consistent with the interparticle MUA head-to-head hydrogen-bonding distance being twice the NDT-defined distance. Such low levels of conductivity significantly reduce the effectiveness of electron hopping through the Au cores in the film. Figure 1 shows two representative TEM images for MUAAu5-nm (A) and MUA-Au2-nm (B) (submonolayer range). These samples were prepared by immersing the carbon film-coated TEM grid in the preparation solution for 2-3 h. In contrast to largely evenly spaced features observed for DT-Au nanoparticles evaporated on the grid,17,18 the images, especially for the MUA-Au5-nm film (A), display regions of collectively connected particles forming partially continuous network and domains including parallel lines, rings, and other type of patterns. As the film grows, these domains or patterns extend through the film. It is important to note that the morphologies at the submonolayer coverage are far from the idealized illustration (Chart 1), but provide a comparison of the two core sizes at the initial networking stage. AFM imaging of the detailed morphologies of thicker films on electrode surfaces is planned.

Figure 2. UV-visible spectra of MUA-Au5-nm (a) and MUAAu2-nm (b) films on glass substrates. The dashed lines are UV-visible spectra of solution samples of the Au5-nm (a′) and Au2-nm (b′) particles capped with DT.

The measurement of the surface plasmon (SP) resonance band of nanoparticles, a well-characterized optical property for thiolatecapped gold nanoparticles of different sizes,22,25 provides additional evidence for the individual and connection properties in the nanoparticle network films. Figure 2 presents UV-visible spectra for both MUA-Au5-nm (a) and MUA-Au2-nm (b) on glass slides. The solution sample spectra for the SP of Au5-nm (520 nm) and the weak and superimposed SP of Au2-nm (dashed lines) are included for comparison. The spectral difference is evident between the two films and largely resembles the solution (25) Alvarez, M. M.; Khoury, J. T.; Schaaff, T. G.; Shafigullin, M. N.; Vezmar, I.; Whetten, R. L. J. Phys. Chem. B 1997, 101, 3706.

Analytical Chemistry, Vol. 72, No. 10, May 15, 2000

2193

Table 1. Absorbance (A) of the SP Resonance Band at 520 nm and Mass Loading (∆m) for the Formation of the Network Thin Films at Different Immersion Times from a Hexane Solution of 0.1 µM DT-Capped Au5-nm and 0.05 mM MUA UV-vis measurement

QCM measurement

network films

time (min)

A

time (min)

∆m (µg/cm2)

MUA-Au5-nm

5 57 260 309 1380

0.109 0.242 0.297 0.087 0.231

5 40 220

3.19 6.22 11.99

MUA-Au2-nm

counterparts, suggestive of the absence of direct core-core aggregation in the network. A small red shift was observed for the MUA-Au5-nm film with respect to the solution Au5-nm (575590 nm). The shift was also accompanied by a color change from red (solution) to blue (film). On the basis of similar features reported previously for DNA-linked14 and dithiolate-linked18 gold nanoparticles, we attribute the shift to an interparticle distance less than the core size. This change is insignificant for MUAAu2-nm film because the interparticle distance is comparable with the core size. Shell cohesive interactions, as known for monolayers on planar gold,26 may also contribute partially to the overall shift. As growth continued, the films displayed a gradual increase of absorbance in the 500-600-nm region with increasing immersion time, as shown in Table 1. The film thickness could be controlled by monitoring the change of SP band absorbance (A) with immersion time. The corresponding mass loading (∆m) was also determined by frequency change of the QCM device. Table 1 includes several values of A and ∆m vs the immersion time. Clearly, both A and ∆m increase with increasing immersion time for the film formation. As an estimate of the thickness, we consider a simple spherical model for the particles,27 by which the Au5-nm counts to ∼3850 Au atoms and ∼370 thiolates and the Au2-nm counts to ∼250 Au atoms and ∼60 thiolates. As an ideal model, e.g., (111)-type packing, a one-layer coverage of the particles with a head-to-head hydrogen-bounded MUA interparticle distance (∼3.6 nm) would generate a total mass of 2.18 (Au5-nm) or 0.37 µg/cm2 (Au2-nm). On the basis of the measured frequency changes for a number of samples, the mass loading data yielded, in average, an equivalent number of layers ranging from 1 to 7 for films prepared from lower precursor concentrations and 20 to 30 for those from higher concentrations. Electrochemical Characterization of pH-Tunable IonGating Properties. The assessment of ion-gating properties is largely based on the transformation of shell COOH groups (linked and free) between carboxylic acid and carboxylate forms. The ion passages through the network, regulated by both shell electrostatic and particle size effects, are determined by pH dependence of voltammetric responses to ionic redox probes of two different signs, i.e., [Fe(CN)]63-/4- and Ru(NH3)63+/2+. Since the data for the two couples at a bare electrode are not notably affected by (26) Bain, C. D.; Troughton, E. B.; Tao, Y. T.; Evall, J.; Whitesides. G. M.; Nuzzo, R. G. J. Am. Chem. Soc. 1989, 111, 321. (27) The number of Au atoms, NAu, and of thiolates, NRS, are calculated by NAu ) (4/3)πR3/Vg and NRS ) 4πR2/ARS, in which R, Vg, and ARS are core radius, core density, and thiolate density, respectively.

2194 Analytical Chemistry, Vol. 72, No. 10, May 15, 2000

Figure 3. Cyclic voltammetric curves for MUA-Au2-nm film coated on a GC electrode in 1.0 mM Ru(NH3)63+/0.5 M KCl solutions at pH 2.7 (a) and 10.5 (b), and for a bare GC (c). Scan rate, 50 mV/s.

pH,20a the detection of any difference in redox responses must be associated with the barrier and gating properties of the film. Figure 3 shows a set of cyclic voltammetric curves at a GC electrode coated with a MUA-Au2-nm film for the [Ru(NH3)]63+ probe. To demonstrate the contrast, the data are shown at two pH extremes, at which the COOH groups are in protonated (a) or deprotonated (b) forms. The curve at the bare GC electrode (c) is included for comparison. At low pH, the redox currents (a) are basically shut off by the film (“close” state), with only small capacitive currents being detectable. In contrast, the response at high pH (b) is large (“open” state) and comparable with that at the bare GC. The pH-tuned redox responses are reversible, as evidenced by data of up to three pH cycles. Note also that there is a subtle difference of the redox wave symmetry between curves b and c, which shows the peak current of the reduction wave is noticeably larger than the oxidation current. Interestingly, the magnitude of the difference increases with increasing film thickness, which reflects to a certain degree effects of redox probe preconcentration (or incorporation) and current rectification. The effect of preconcentration relates the current enhancement to the incorporation of analytes into the film as often reported for chemically modified electrodes based on electrostatic interactions; the effect of rectification relates the fact that the cathodic current is larger than the anodic to a current-rectifying behavior. The electrostatic effect is clearly operative for the ion-gating responses. When pH is high, the network carries negative charges (CO2-) and admits the positively charged redox species via an attractive force. When pH is low, the neutral network does not admit the probe at all, suggesting that the channels in the neutral network are too small for the probe to penetrate. In other words, the channels created by the hydrogen-bonded nanoparticle network are in the size range of the redox probes. To confirm the electrostatic effect, Figure 4 presents a set of voltammetric curves for the negatively charged probe [Fe(CN)]63at the MUA-Au2-nm network film electrode at two pHs (a and b).

Figure 5. Cyclic voltammetric curves for MUA-Au5-nm film coated on a GC electrode in 1.0 mM Ru(NH3)63+/0.5 M KCl solutions at pH 2.7 (a) and 10.8 (b). Scan rate, 50 mV/s.

Figure 4. Cyclic voltammetric curves for MUA-Au2-nm film coated on GC electrode in 1.0 mM Fe(CN)63-/0.5 M KCl solutions at pH 2.7 (a) and 10.8 (b), and for a bare GC (c). Scan rate, 50 mV/s.

The curve at the bare GC electrode (c) is included for comparison. It is evident that the redox responses at both pHs are effectively suppressed, which is in sharp contrast to the result observed for the cationic redox probe. The effective blockage at high pH is indeed consistent with the repulsive effect between the probe and the negative charges residing at the rim of the channels. We also note a small difference in the faradaic leakage currents at these two pHs. The current at low pH was slightly larger than at high pH, varying from sample to sample. While a further investigation is needed, it is possible that an additional hydrogen-bonding interaction between protonated [Fe(CN)]63-/4- species at low pH, e.g., Hn[Fe(CN)]6-(3-n), and the film’s CO2H groups partially promotes the penetration of the probe through the network to the underlying substrate. We have further examined the particle core size effect by using the MUA-Au5-nm film. Figure 5 shows a set of voltammetric responses for Ru(NH3)63+ at two pHs. In comparison with the MUA-Au2-nm film, the MUA-Au5-nm film exhibits largely similar ion-gating behavior. The cationic probe is admitted at high pH but blocked at low pH. Two subtle differences in terms of the redox current are noticed. At high pH, the redox current (b) is in general smaller than that at the bare GC. Thicker films tend to attenuate the current more than thinner films. This is in contrast to the enhanced current response observed for the MUA-Au2-nm film upon increasing film thickness. On the other hand, the leakage currents at the MUA-Au5-nm film at lower pH (a) are larger than that observed at the MUA-Au2-nm film, indicating that the redox currents are not completely shut off by the protonated network. This observation is probably due to defects associated with larger void sizes in the linked Au5-nm network. The MUA-Au5-nm film is also examined for its barrier properties toward the [Fe(CN)]63- probe. Again, the general characteristic of the redox responses at different pHs is very similar to those observed for the same redox probe at the MUA-

Figure 6. Scan rate dependence of the voltammetric curves for MUA-Au2-nm (A) and MUA-Au5-nm (B) films on a GC electrode in a 1.0 mM Ru(NH3)63+/0.5 M KCl solution at pH 10.4. Scan rate (υ, from bottom to top), 20, 50, 100, and 200 mV/s. The insets represent plots of the cathodic (ipc, closed circles) and anodic (ipa, open circles) peak currents vs υ.

Au2-nm film. The redox currents are basically blocked at both high and low pHs. The relative magnitude of the leakage currents at both pHs are however slightly larger than those observed for the MUA-Au2-nm film. This difference may reflect in part a difference in film thickness and in part a difference of the actual sizes of void sites and defects in these two films. The above data reveal that the ion-gating capability of the MUA-Au2-nm film is more effective than the MUA-Au5-nm in terms of the enhancement of the response current at the “open” state and the shutting-off current at the “close” state. An examination of the scan rate (υ) dependence of the voltammetric currents at the “open” state (Figure 6) further supports the assessment. While the MUA-Au2-nm film displays an enhanced asymmetric waves, the MUA-Au5-nm film exhibits noticeably attenuated symmetric waves. Plots of the ip vs υ1/2 show a linear relationship, indicative of a diffusion-controlled process for the redox probe through the film (Note that the linearity for the Au2-nm is slightly Analytical Chemistry, Vol. 72, No. 10, May 15, 2000

2195

poorer than that for the Au5-nm.). The ip-υ1/2 slopes, i.e., 13.9 (ipc) and 6.0 (ipa), are clearly larger for the MUA-Au2-nm than those (2.3 (ipc) and 1.8 (ipa)) for the MUA-Au5-nm (The slopes are 4.0 (ipc) and 3.9 (ipa) for a bare GC electrode.). For the MUAAu5-nm film, the redox current decreases with increasing film thickness, with the ip-υ1/2 slope being similar to that for the bare GC electrode. In contrast, the redox currents for the MUA-Au2-nm film increase with increasing film thickness, with the relative ratio of ipa to ipc showing a slight dependence on thickness. For example, a thicker film (SP band A∼520 nm ) 0.27) shows ip-υ1/2 slopes of 13.9 (ipc) and 6.0 (ipa), whereas a thinner film (SP band A∼520 nm ) 0.09) displays slopes of 4.1 (ipc) and 2.0 (ipa). The data are again supportive of the preconcentration and rectification assessments for the redox responses in the MUA-Au2-nm film. While a quantitative delineation of the diffusion coefficients and concentrations of the probes in the network is part of our ongoing work, we qualitatively conclude that the “open” film exhibits membrane-type ion-exchanging properties. Similar behavior was known for redox species that were electrostatically incorporated into polyelectrolyte films.28 Interestingly, while thinner films of both core sizes exhibited similar redox potential close to the E° of [Ru(NH3)]63+ (-0.16 V), thicker films of the MUA-Au2-nm showed a negative shift of the potential (∼100 mV) and a reduced peak separation (∼40 mV) with increasing υ. These features are indicative of electron self-exchanging transport within the surface network28 and again supportive of the redox probe incorporation into the film. The mechanistic structure-reactivity relation will be further studied. We note that the peak-shaped, rather than sigmoidal-like, voltammetric waves for the above redox responses at the “open” state are evident. This type of response can be interpreted by a radial diffusion-controlled process at microelectrode arrays with very close spacing. On the basis of the theoretical model for partially blocked electrode surfaces by Amator et al.,29 we could simulate the observed voltammetric waves with a reasonable satisfaction.30 In the model simulation for an ensemble of nanoelectrodes, we treated the void space as being spherical and separation distance as being defined by the core-shell size. The result yielded a comparable peak-shaped wave for the thin-film network at the “open” state. Three major factors determining whether the film admits, partially admits, or blocks the redox probes can be summarized as follows: (1) the degree of protonation-deprotonation at the interparticle linkages, (2) the sizes of the particle core, and (3) the charges and sizes of the redox probe. The combination of these factors allows the manipulation of gating properties at both nanocrystal and molecular levels. Although the CO2H/CO2switching properties have been used for ion gating in systems such as self-assembled monolayers, dendrimers, and polymers, 1-2,20 the core-shell nanoparticle combination in terms of shell linkage and core size is to our knowledge the first example of imparting biomimetic gating properties to electrodes. We further (28) Doblhofer, K.; Armstrong, R. D. Electrochim. Acta 1988, 33, 453. (b) Murray, R. M. In Electroanalytical Chemistry; Bard, A. J., Ed.; Marcel Dekker, Inc.: New York, 1984; Vol. 13, p 191. (29) Amatore, C.; Saveant, J. M.; Tessier, D. J. Electroanal. Chem. 1983, 147, 39. (b) Finklea, H. O.; Snider, D. A.; Fedyk, J.; Sabatani, E.; Gafni, Y.; Rubinstein, I. Langmuir 1993, 9, 3660. (30) Zheng, W. X.; Maye, M. M.; Zhong, C. J., unpublished work.

2196

Analytical Chemistry, Vol. 72, No. 10, May 15, 2000

note that, in addition to the probe penetration at the interparticle sites as described above, a direct electron exchange between the solution redox probes and the nanocrystal core may be a possible pathway that is partially responsible for the observed redox responses. This assessment is based to our recent findings18 that NDT-linked 5-nm cores are partially accessible to redox probes probably via packing defects at the crystal corners or edges. Although a precise differentiation of contributions between the interparticle sites and the packing defects remain to be investigated, we currently believe that the former mechanism is predominant for three major reasons. First, the electron exchange between the core and the probe requires a sufficient electronic conductivity of the film (e.g., electron hopping through Au cores as in the NDT-Au5-nm film18). The conductivity of a MUA-linked film is however ∼3 orders of magnitude lower than the NDTlinked films, which, especially for the MUA-Au2-nm film, is unlikely to be effective for electron hopping. Second, the fact that ion gating at the MUA-Au2-nm film is more effective than at the MUA-Au5-nm film is suggestive of an insignificant contribution of defects, because, on the basis of the studies of NDT-Au5-nm and -Au2-nm films,18 more defects are present on the larger-sized particles. Also, defect sites larger than several particle sizes due to packing may display similar redox responses at “open” states, but cannot display blocking behavior at the “close” state. Finally, no significant increase of the voltammetric peak separation was observed for both probes at both films. The decrease in conductivity usually increases the peak separation. As an additional confirmation of the gating properties, we have also looked at redox probes of catechol (neutral) and trimethylaminomethylferrocenyl (cationic) at these films. Both showed blocking effect at low pH, and the latter showed gating responses similar to the Ru(NH3)63+ species. IRS Characterization of the Network and the pH-Tuned Structural Properties. An infrared reflectance spectroscopic (IRS) technique18,31 is used to characterize the detailed hydrogenbonding structures and the pH-tuned network structures of MUAAu5-nm and MUA-Au2-nm films. As we have recently demonstrated,31 the carboxylic acid groups provide a diagnostic handle for probing the shell structure and reactivity. Figure 7 presents a set of IRS spectra in the low- (A) and high- (B) energy regions for the MUA-Au2-nm (a) and MUA-Au5-nm (b) films. The spectrum for a MUA monolayer (c) is included for comparison. The assignments of bands diagnostic of -CO2H, -CO2-, and C-H stretching vibrations are listed in Table 2. We begin by analyzing the bands diagnostic of the COOH (ν(COOH)) bands in the low-energy region (Figure 7A). The spectral components under the ν(COOH) vibration envelope in the 1740-1650-cm-1 region differ significantly between these films, whereas those in the 1500-1400-cm-1 region (mainly C-H bending modes) show little variations. Note that both spectra have an overall absorbance of ∼50 times larger than the monolayer, presumably due to a combination of the nanoparticle multilayer (10-20 equivalent) and surface enhancement32 effects. Each spectrum exhibits an envelope of multiple band components, as represented by the dashed curves based on spectral deconvolution (31) Zheng, W. X.; Maye, M. M.; Leibowitz, F. L.; Zhong, C. J. Analyst 2000, 125, 17. (32) Brown, C. W.; Yue, L.; Seelenbinder, J. A.; Pivarnik, P.; Rand, A.; Letcher, S. V.; Gregory, O. J.; Platek, M. J. Anal. Chem. 1998, 70, 2991.

Chart 2. Cis and Trans Configurations of the Head-to-Head Hydrogen-Bonding Linkage Structure in the Network (only a Fraction of the Network Is Drawn)a

a In cis configuration, the C -C bond is on the same side of R β the CdO group; the CR-Cβ is on the opposite side of the CdO group for trans configuration.

Figure 7. IRS spectra in low- (A) and high- (B) energy regions for a MUA-Au5-nm film (a), a MUA-Au2-nm film (b), and a MUA monolayer (c) on Au/glass substrates. The dashed lines represent spectral deconvolution based on Lorentzian-type peak profiles. Each of the νCdO(CO2H) spectral components is numerically labeled as 1-4. Table 2. Mode Assignments for the Diagnostic IRS Bands (cm-1) in the High- and Low-Energy Regions for MUA-Au5-nm Film, MUA-Au2-nm Film, and MUA Monolayer mode assignment

MUA-Au5-nm

νa(CH3) νs(CH3) νa(CH2) νs(CH2) νCdO(CO2H)a ν1 ν2 ν3 ν4 νs(CO2-) νs(CO2-)

∼2956b ∼2872b 2920 2850 1739 1710 1668 1578 ∼1435

MUA-Au2-nm

MUA

2920 2850

2920 2849

1738 1709 1693 1665 1569 ∼1432

1741 1717 1695 1556 1431

a See Figure 7 and the text for the assignments of the ν to ν 1 4 components of the ν(COOH) bands. b Weak shoulder peaks.

(numbered 1-4). The two network films (a and b) show remarkable distinctions from the MUA monolayer (c) in terms of the band position and relative weight. For the MUA monolayer, the three ν(CdO) bands were previously assigned33 to -COOH groups of free or non-hydrogen-bonded mode (1741 cm-1), sideby-side dimeric hydrogen-bonded mode (1718 cm-1), and polymeric hydrogen-bonded (∼1690 cm-1) modes, respectively. In

comparison, the three bands identified for the MUA-Au5-nm network film are displayed at 1740 (1), 1710 (2), and ∼1670 cm-1 (4). While the free acid band (1) remains basically unchanged, the shifts in the hydrogen-bonded modes (2, 4) are remarkable. We attribute these bands to the formation of head-to-head hydrogen bonding of the COOH groups in the nanostructure. In fact, the wavenumber of band 2 is identical to the band position observed for the head-to-head hydrogen-bonding dimer with cis configuration in condensed phases of alkanoic acids.34 Such dimer structures are depicted in Chart 2. The trans configuration is basically absent in the MUA-Au5-nm film. The broad band 4 is mostly due to polymeric hydrogen bonding. Its difference from the MUA monolayer in band position reflects the difference of the hydrogen-bonding properties between head-to-head and sideby-side configurations. The head-to-head binding modes are impossible in the 2D MUA monolayers. Moreover, while the 2D monolayer exhibits ∼40% free acid component, the MUA-Au5-nm film shows only a level of 6, demonstrating the transformation from -CO2H to -CO2- in the network. In pH 6-10, the detection of both -CO2H and -CO2- species is indicative of partial protonation-deprotonation. The pH-tuning is reversible, as evidenced by the reappearance of the ν(CdO) bands upon returning to low pH (g), with a ∼10% absorbance decrease probably due to a structural reorganization. How these hydrogen-bonded and non-hydrogen-bonded COOH components respond to pH-tuning is an important issue that is 2198 Analytical Chemistry, Vol. 72, No. 10, May 15, 2000

Figure 10. pH dependence of the IRS spectra in the low-energy region for the MUA-Au2-nm film emersed from solutions of different pH values: starting from pH 1.8 (a), 3.2 (b), 4.7 (c), 5.9 (d), 7.0 (e), 8.0 (f), 9.1 (g), and 10.1 (h) to 12.0 (i), and back to 1.8 (j).

examined by determining the peak area of each band. Figure 9 plots the peak areas against pH. Note that only the asymmetric νa(CO2-) band is plotted because the symmetric is overlapped with several bands (e.g., ν(CH2, ∼1415 cm-1), ν(C-O) + δ(OH) (1430-1460 cm-1)). While band 2 decreases gradually with increasing pH, bands 1 and 4 do not show a decrease until pH 5-6. This result suggests that pH-tuning is most effective for the head-to-head hydrogen-bonded linkage. Furthermore, the region between the disappearance of ν(CO2H) bands and the onset of

components and the CO2- component are around pH 6-7, and the cross point between the total COOHs and the CO2- is at pH ∼7.8. The overall pKa range is thus estimated to be within pH 5-9, which is about one pH unit lower than that for the MUAAu5-nm film. This difference demonstrates again that shell linkage properties in the networks are strongly dependent on the particle core sizes. The IRS data clearly demonstrate that the interparticle linking functionality can be reversibly tuned by pH between protonated and deprotonated states with little degradation of the network architecture. There are major differences between MUA-Au2-nm and MUA-Au5-nm films in the hydrogen-bonding fine structures and the associated pKa range, manifestations of the structural manipulation by the particle core sizes.

Figure 11. Peak areas for each of the νCdO(CO2H) bands, 1-4, and the νa(CO2-) band, as well as the total (1 + 2 + 3 + 4), as a function of the pH for the MUA-Au2-nm film. The peak areas were obtained by spectral deconvolution of IRS data (Figure 10).

the νa(CO2-) band reflects an approximate interfacial pKa range. The cross points between these CO2H components and CO2- are around pH 7-8, whereas that for the total CO2H is at pH ∼8.2. A precise determination of the pKa will need to consider other factors including interfacial charge distribution and solution concentrations.23,35-38 The IRS spectra for the MUA-Au2-nm film exhibit similar pH dependence (Figure 10). As pH increases (from a to i), the ν(Cd O) bands of 1, 2, and 4 display a gradual decrease in intensity and disappear at pH ∼10. The ν(CO2-) bands, i.e., 1578 (νa) and ∼1435 cm-1(νs), emerge at pH >6. Distinctively, the transhydrogen-bonded COOH, 1694 cm-1(3), displays only a small dependence on pH and remains even at pH 12 at which all other COOH components are completely converted to CO2-. The pH dependencies for each COOH component are shown in Figure 11. The cis configuration is strongly dependent on pH, whereas the trans mode is not. The cross points between the COOH (35) (a) White, H. S.; Peterson, J. D.; Cui, Q.; Stevenson, K. J. J. Phys. Chem. B 1998, 102, 2930. (b) Smith, C. P.; White, H. S. Langmuir 1993, 9, 0743. (36) Bryant, M. A.; Crooks, R. M. Langmuir 1993, 9, 385. (37) Hu, K.; Bard, A. J. Langmuir 1997, 13, 5114. (38) Lee, T. R.; Carey, R I.; Biebuyck, H. A.; Whitesides, G. M. Langmuir 1994, 10, 741. (39) Yaghi, O. M. In Access in Nanoporous Materials; Pinnavaia, T. J., Thorpe, M. F., Eds.; Plenum Press, New York, 1995; p 111. (b) Russell, V. A.; Evans, C. C.; Li, W.; Ward, M. D. Science 1997, 276, 575. (c) Johnson, S. A.; Ollivier, P. J.; Mallouk, T. E. Science 1999, 283, 963.

CONCLUSION In conclusion, we have demonstrated for the first time that the core-shell structure and reactivity of thiolate-encapsulated gold nanoparticles can be utilized to construct network architectures that impart biomimetic ion-gating properties. The nanoconstruction is via noncovalent head-to-head hydrogen-bonding linkages at the carboxylic shells. The network can be effectively tuned by pH between a neutral “close” and an ionic “open” or “close” state to exhibit electrochemical ion-gating properties. While still far from the real biological ion gate in cell membranes, the demonstrated properties are one important step toward the biomimetic function. The IRS characterizations have unraveled that the network structures are dependent on both the shell linkages of the noncovalent carboxylic linkages and the nanoparticle core sizes. The dependence is reflected by detection of fine structures of cis and trans carboxylic hydrogen-bonding linkages that can be manipulated by core sizes. The reversible network tuning serves as a model system of chemical manipulation at interparticle linkages and offers a new strategy for designing molecular recognition elements. Biomimetic applications of such designs are in fact subjects of our continuous work in view of the emerging core-shell chemistry5-6,15-18 and technologies based on nanoporous architectures.39 ACKNOWLEDGMENT We thank Mr. H. Eichelberger for TEM measurement, Dr. O. Sadik for using her QCM instrument, and Dr. M. Porter for providing initial Au/glass slides and octadecanethiolate-d37 reference slides. Received for review November 10, 1999. Accepted February 12, 2000. AC9912909

Analytical Chemistry, Vol. 72, No. 10, May 15, 2000

2199