Imaging Fluorescent Nanoparticles To Probe Photoinduced

Optically transparent semiconductor electrodes are valuable tools that allow .... capturing long exposure (1.0 s) images of the illumination beam at h...
7 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/Langmuir

Imaging Fluorescent Nanoparticles To Probe Photoinduced Charging of a Semiconductor−Solution Interface Eric M. Peterson and Joel M. Harris* Department of Chemistry, University of Utah, 315 South 1400 East, Salt Lake City, Utah 84112-0850, United States ABSTRACT: Optically transparent semiconductors allow simultaneous control of interfacial electrical potential and spectroscopic observation of chemistry near the electrode surface. Care must be taken, however, to avoid unwanted photoexcitation-induced charging of the semiconductor electrode that could influence the results. In this work, we investigate the in situ surface charging by photoexcitation well below the band gap of an optically transparent semiconductor, indium−tin oxide (ITO) electrode. Using total-internalreflection fluorescence microscopy, the population of ∼100nm negatively charged carboxylate−polystyrene fluorescent nanoparticles at an ITO−aqueous solution interface could be monitored in situ. At positive applied potentials (∼0.7 V versus Ag/AgCl), nanoparticles accumulate reversibly in the electrical double-layer of the ITO surface, and the interfacial nanoparticle populations increase with 488-nm excitation intensity. The potential sensitivity of nanoparticle population exhibited no dependence on excitation intensity, varied from 0.1 to 10 W cm−2, while the onset potential for particle accumulation shifted by as much as 0.3 V. This shift in surface potential appears to be due to photoexcitation-induced charging of the ITO, even though the excitation radiation photon energy, ∼2.4 eV, is well below the primary band gap of ITO, >3.5 eV. A kinetic model was developed to determine the photon order of electron−hole generation relative to the electron−hole recombination. The photoexcitation process was found to be first-order in photon flux, suggesting one-photon excitation of an indirect band gap or defect sites, rather than two-photon excitation into the direct band gap. A control experiment was conducted with red-fluorescent carboxylate−polystyrene particles that were counted using 647-nm excitation, where the photon energy is below the indirect band gap or defect site energy and where the optical absorption of the film vanishes. Red illumination between 1 and 15 W cm−2 produced no detectable shifts in the onset accumulation potential, which is consistent with the negligible optical absorption of the ITO film at this longer wavelength.

■

INTRODUCTION Optically transparent semiconductor electrodes are valuable tools that allow simultaneous spectroscopic and electrochemical characterization of chemical processes at potential-controlled interfaces.1−3 Transparent electrodes have been used spectroscopically to study electrochemical kinetics2,4 and to detect electrochemically generated species.5 Transparent semiconductors have also been used to study photophysics in fluorescent molecules, including photoinduced electron transfer between semiconductors and single fluorophores6−8 and the effects of local electric field on fluorescence emission.9−11 When used in sensors, optically transparent semiconductors allow simultaneous spectroscopic and electrochemical detection of analytes, increasing measurement selectivity.12,13 In addition, control of electrical potential at the electrode surface allows for regenerable biosensors through electrostatic desorption14,15 and controlled immobilization of biomolecule probes in microarrays.16 Spectroelectrochemistry at semiconductor interfaces presents challenges, however, because the light used for spectroscopic detection may also excite electronic transitions in the semiconductor, leading to unexpected perturbations, such as © 2013 American Chemical Society

shifts in surface potential or conductivity. In this work, we report photoinduced shifts in the electrical potential of a transparent semiconductor electrode, indium tin oxide (ITO), from illumination with radiation having photon energy well below the primary band gap. Charged ∼100-nm particles are used as an in situ probe of electrical potential in the electrical double layer of an ITO−aqueous interface. Total-internalreflection fluorescence microscopy was employed to quantify the nanoparticle population in the ∼150-nm evanescent wave region at the interface by counting the individual fluorescent carboxylate-terminated polystyrene particles in the fluorescence images. The population of charged colloidal particles at the ITO interface increases with positive applied potential, and a Poisson−Boltzmann model describes the results based on two parameters: the potential sensitivity of the interfacial particle population and the onset potential for accumulation. By increasing the 488-nm illumination intensity at the interface, particle population versus applied potential curves shift to less Received: June 29, 2013 Revised: August 22, 2013 Published: August 23, 2013 11941

dx.doi.org/10.1021/la402468k | Langmuir 2013, 29, 11941−11949

Langmuir

Article

electrode.18 Metal clips hold the flow cell onto a stainless steel microscope-stage insert whose interior is covered with Parafilm to prevent shorts to ground. A Pine model AFCBP1 three-electrode potentiostat was used to control the potential of the ITO working electrode. TIRF Microscopy. Fluorescent nanoparticles at the ITO−solution interface were imaged with a Nikon TE-200 inverted microscope modified for through-the-objective TIRF illumination, previously described.19 For this work, two laser sources were used for fluorescence excitation; blue excitation was provided by the 488.0nm line from a Lexel model 95 argon ion laser coupled into a singlemode polarization-maintaining optical fiber (Thorlabs) with a focusing lens (Thorlabs PAF-X-5-A FiberPort). Red excitation was from the 647.1-nm line of a Coherent Innova 90C krypton ion laser coupled into a similar optical fiber and focusing lens (Thorlabs F240FC-B fiber collimator). Laser intensity emerging from the ends of either optical fiber ranged between 0.1 and 10.5 mW and was collected by a collimating plano-convex lens, directed through a band-pass filter (Semrock), and focused onto the back focal plane of the microscope objective by a second plano-convex lens. An oil-immersion microscope objective (60× apo TIRF 1.49 numerical aperture, Nikon) generated a collimated beam that illuminated the sample. Translating the optical fiber normal to the optical axis shifted the angle of the illumination beam relative to the ITO surface, generating total-internal reflection (TIR) at the ITO−solution interface. On the basis of calculations of the excitation intensity distribution versus incident angle for threephase dielectric thin film structures,20 the 280-nm thick ITO film (see below) on the glass coverslip does not significantly influence the angle at which an evanescent wave is generated at the ITO−aqueous solution interface;20 indeed, total-internal reflection at this interface was observed at comparable incident angles as with a bare glass coverslip. Images of fluorescence from the ITO−solution interface were acquired with an Andor iXon 897 electron multiplying charge coupled device (EMCCD) camera in a 53 × 53 μm (200 × 200 pixel) region using 15-ms integrations and 0.25-s time-lapse intervals. For images of the yellow-green nanospheres, electron-multiplying gain was reduced from 50× to 1× as the illumination intensity was increased, in order to avoid variations in the signal-to-noise ratio (SNR). Images of the darkred particles exhibited higher SNR than the 100 nm yellow-green particles, due to their ∼8 times greater volume and corresponding fluorophore content. Even at the lowest excitation power, 1 W cm−2, the electron multiplying gain could not be increased above 5× in images of dark-red particles without occasional bright particles saturating the CCD well depth. Because the electron multiplying gain can only be adjusted in integer units, there was not enough dynamic range in gain level to compensate for the range of excitation power density in the experiment. An alternative method of normalizing the particle count across excitation intensities was used, where the detection threshold was raised with increasing excitation intensity to achieve a constant particle count at small applied potentials (see below). All images were collected as 16-bit monochrome TIF image stacks using Andor Solis ver. 4.16. Calibrating Illumination Power Density. In order to report accurate illumination power densities in the imaging region, the diameter of the illumination beam at the sample must be calibrated. Illumination in this experiment was provided by a single transverse mode laser beam internally reflected at the ITO−aqueous interface. At the oblique illumination angles used in TIRF microscopy, the beam was elongated in the direction of reflection at the surface, resulting in an elliptical, two-dimensional Gaussian-shaped illumination region. The area of this illumination region was measured by capturing long exposure (1.0 s) images of the illumination beam at high power (10 mW) internally reflected at a glass−air interface. The illumination beam excited background fluorescence in the glass substrate, creating an elliptical spot in the middle of the image. A two-dimensional Gaussian function was fit to this background fluorescence profile using nonlinear least-squares optimization to provide parameters for the center coordinate and size of the beam in both directions. Using these fitted parameters, the total laser power in any region on the surface can

positive applied potentials, indicating a buildup of positive charge at the ITO−solution interface. A photochemical kinetic model, relating the rates of photoinduced charge separation and charge recombination, is consistent with a one-photon process, which is possibly the result of photoexcitation into an indirect band gap or photoexcitation of defect sites in the direct band gap of ITO. Potential scans with red-absorbing fluorescent particles imaged with 647 nm illumination far outside the ITO indirect band gap and defect energy gap show negligible shifts in potential with illumination intensity, confirming the photoinduced charge separation hypothesis.

■

EXPERIMENTAL SECTION

Chemicals and Materials. Carboxylate-modified fluorescent polystyrene nanospheres were purchased from Invitrogen (Carlsbad, CA) as 2% by weight suspensions in water. “Yellow-green” particles with 110-nm diameter exhibited fluorescence excitation and emission maxima of 505 and 515 nm, respectively, and “dark-red” 200-nm particles had excitation and emission maxima of 660 and 680 nm; dark-red nanospheres were unfortunately not available in the same 100-nm size of the yellow-green nanospheres. Both nanoparticle suspensions were characterized using a Nicomp 380ZLS for dynamic light scattering (DLS) and electrophoretic mobility measurements. The suspensions were diluted in 75 μM pH 8.0 phosphate buffer with 75 μM NaCl to concentrations of 0.01% by weight and 0.002% for 110- and 200-nm particles, respectively, for DLS and electrophoretic mobility measurements. Zeta-potentials were determined from electrophoretic mobilities using the Helmholtz−Smoluchowski equation for electrical double layers significantly smaller than the particle diameter;17 DLS measurements yielded particles diameters consistent with the supplier’s specifications. Buffers and solutions were prepared using water purified with a Barnstead NANOpure II system (Boston, MA) to a solution resistivity of approximately 18 MΩ·cm. Indium tin oxide (ITO) sputter-coated 22 × 22 mm no. 1 glass coverslips (sheet resistivity of 15−30 Ω/ square) were purchased from SPi Supplies (West Chester, PA). Buffers and supporting electrolyte solutions were prepared with ACS grade sodium phosphate monobasic monohydrate, sodium chloride, and sodium hydroxide from Mallinckrodt (Phillipsburg, NJ). Spectroscopy grade Omnisolv methanol from EMD chemicals (Darmstadt, Germany) was used for cleaning ITO coverslips. Conductive silver adhesive paste and Gold Seal glass 22 × 22 mm no. 1.5 coverslips were purchased from VWR (West Chester, PA). Nanosphere Suspensions. Prior to their use, the fluorescent particle suspensions were first sonicated for 15 min in a Fisher Scientific FS-28 ultrasonic bath to break up any aggregates. Solutions for imaging experiments were prepared by serial dilutions of the stock 2% suspensions into deionized water, followed by a final dilution into 75 μM pH 8.0 phosphate buffer and 75 μM sodium chloride (approximately 0.3 mM ionic strength). The particle concentration used for imaging was 5 pM for yellow-green nanospheres, and 2 pM for dark-red nanospheres. Buffers used to dilute bead suspensions were filtered with a 0.2-μm Whatman Puradisc polyether sulfone 25 mm syringe filter to remove particulate contamination. Preparation of ITO Substrates. ITO electrodes were prepared by first rinsing for 10 min each in 18 MΩ·cm water and methanol. The ITO-coated side of the coverslip was then cleaned for 25 min in a UV ozone cleaner (Jelight Co., model 342). Copper wire leads were attached to the edge of the ITO coated coverslip using silver conductive adhesive paste, which was then placed in a 120 °C oven overnight to allow the conductive paste to cure. The slides with attached wire leads were then UV ozone cleaned again for 25 min and assembled into an imaging electrochemical flow cell. This cell consists of a wired ITO slide separated from a top glass plate by a silicone gasket affixed to each surface with acrylic−polyester double-stick tape (3M, 9495MPF). The top plate has an inlet port and outlet port, which allow solution flow, and access by a platinum counter electrode and a Cypress Systems (ESA, Inc.) model EE009 Ag/AgCl reference 11942

dx.doi.org/10.1021/la402468k | Langmuir 2013, 29, 11941−11949

Langmuir

Article

be determined by numerically integrating the fitted 2D Gaussian function over the dimensions of the subregion of fluorescence acquisition and dividing it by the integral of the entire Gaussian function. The power density at the surface was thus determined from the fraction of the total beam area represented by the 53 × 53 μm (200 × 200 pixel) acquisition region at the center of the Gaussian beam profile and multiplying that fraction by the total laser power measured at the base of the microscope objective. Image Analysis for Counting Fluorescent Particles. Image analysis was performed using custom programs written in the Matlab (Mathworks) numerical analysis environment. Individual particles were located in TIRF images using methods previously described.18,19 Briefly, fluorescence spots were detected by locating 0.8 × 0.8 μm (3 × 3 pixel) regions with three or more pixels brighter than an intensity threshold, Ithold, set nstd background standard deviations, σbg, above the mean background intensity, μbg, so that Ithold = nstdσbg + μbg. The labeled particles were highly fluorescent and detected with high signalto-noise (S/N) ratios in all images (nstd > 6). Values for μbg and σbg were determined for each illumination intensity and camera acquisition setting by calculating the mean and standard deviation of the blank intensity. The intensity threshold was set between nstd = 6 and 20 for images of 110-nm nanospheres, depending on the SNR, and resulted in consistent populations across all illumination intensities for applied potentials less than 0.6 V, where a constant population in the evanescent wave is sampled. The 200-nm dark-red particles were significantly more intense than 110-nm nanospheres, and nstd was set between 10 and 100. After the potential steps, particles accumulated near the interface and reached an equilibrium population that did not change after approximately 60 s. At each applied potential, the particle populations were measured and averaged from 100 images, starting 60−80 s after the potential step. The number of nanoparticles counted in the images varied from frame-to-frame according to Poisson statistics, where the variance in particle number was proportional to the mean with a slope of 0.94 ± 0.24, indistinguishable from a theoretical slope of unity. Because of the large number of images (100) that were subsequently averaged, the resulting relative standard deviation of the mean particle count reported is small (∼1.4 at 200 molecules particles per frame), for data taken at a fixed potential and illumination intensity. Fluctuations in average particle counts between changes in applied potential are ∼3 times greater than the fundamental Poisson limit; however, the standard deviations continue to scale with the square root of the average particle count. Populations of counted fluorescent particles in images, measured as a function of applied potential and illumination intensity, were fit to a Poisson−Boltzmann model (see below), by optimizing the parameters in the model by nonlinear least squares. Uncertainties in reported parameters were estimated using the bootstrap method,21 generated by adding randomized residuals to a best fit curve, where randomized residuals were drawn from a Poisson distribution with a standard deviation of 2.6(NDET)1/2, matching the behavior of the experimental results. Standard deviations of each parameter were estimated by fitting a Gaussian distribution to a histogram of bootstrap-determined parameter values. The uncertainties in the reported parameters and error bars in the plots represent two standard deviations of the mean.

■

are on the same order of magnitude as the titrated carboxylate content. When positive potentials are applied to an ITO electrode in contact with a suspension of these particles, they accumulate at the interface, as shown in Figure 1A, which

Figure 1. Nanosphere population response to applied potential at an ITO surface. (A) Images of fluorescently labeled polystyrene particles at varying potentials; scale bar is 10 μm. (B) Plot of equilibrium populations after potential step (black squares), with fit to eq 3 (dashed line).

provides representative fluorescence images of carboxylate nanospheres near the ITO surface 60 s after the potential had been stepped to between 0 and 1.0 V versus an Ag/AgCl reference electrode. At applied potentials within this range, accumulated particles freely diffuse at the interface; adsorption of particles to the surface to produce motionless spots is rare, so electrodeposition of particles on the surface23 is avoided. Accumulation of the mobile particles at the solution−ITO interface at positive potentials is fully reversible, and populations return within 5 s to their original values when the potential is stepped back to 0 V. Fluorescent particles were identified and counted at each applied potential using an image analysis that requires a detected spot in a fluorescence image to exceed an intensity threshold in multiple adjacent pixels (see above). Because the fluorescent nanospheres were volume-labeled with thousands of fluorescent probe molecules, they could be detected at high S/ N ratio, 6−20 times the background noise depending on illumination intensity. Figure 1B shows a representative plot of equilibrium nanosphere populations versus applied potential. Populations were counted after interfacial particle populations reached equilibrium, typically 60−80 s after stepping the specified potential; potential steps were made sequentially in

RESULTS AND DISCUSSION

Charged Particle Response to Applied Potential. The fluorescent polystyrene nanospheres used in this work have a high density of carboxylate groups on their surface. In basic pH 8.0 phosphate buffer (well above the pKa of a carboxylate group), these particles are highly negatively charged, where the yellow-green nanospheres have ∼1.5 × 105 solution-accessible carboxylate groups per particle based on a titration analysis by the supplier. From measurements of the electrophoretic mobility of the nanoparticles (see Experimental Section), their zeta-potential, ζ = −52.1 ± 0.3 mV, indicates high surface charge density. Using an estimated Helmholtz layer capacitance22 of 15 μF cm−2, the total charges per particle (∼7 × 105) 11943

dx.doi.org/10.1021/la402468k | Langmuir 2013, 29, 11941−11949

Langmuir

Article

dependent behavior of the 110-nm nanosphere population near the ITO surfaces is clearly consistent with the model. The nanosphere population parameters, N0E and N0D, are both proportional to the bulk solution nanosphere concentration, so that the ratio N0D/N0E, the fraction of molecules in the evanescent wave region that responds to potential variation does not vary with nanosphere concentration as long as the Debye length remains constant. A significant fraction of detected nanospheres, N0D/N0E = 0.5 ± 0.1, respond to applied potential, a factor of 2 larger than was observed for 14.9 kbase-pair DNA plasmids,18 where N0D/N0E = 0.2 ± 0.02. This difference in response can be ascribed to several differences in the conditions for the two experiments: first, the nanoparticles were imaged with a higher numerical aperture objective lens than DNA, 1.49 NA versus 1.45 NA, which provides a higher total internal reflection angle, generating shorter evanescent field decay lengths and reducing N0E. In addition, the 100-nm nanoparticles are smaller than the ∼400-nm DNA plasmids, so the center of a nanoparticle must approach closer to the ITO surface than DNA in order for it to be detected in the evanescent field. The population of nanospheres exhibits a similar sensitivity to applied potential, β = 14 ± 2 V−1, versus β = 19 ± 1 V−1 for the DNA plasmid, which could be attributed to a smaller charge density of the nanoparticles compared to DNA, which is charged throughout its structure. Although the DNA and nanosphere data were collected with different bulk solution concentrations, the sensitivity to applied potential, β, characterizes the relative increase in NDET with applied potential and is independent of bulk solution concentration.18 Overall, populations of nanospheres near the polarized ITO surface behave qualitatively the same as populations of DNA plasmids and exhibit a comparable response to changes in applied potential. Nanosphere Response to Changes in 488-nm Illumination. During measurements of the nanosphere population response to applied potential, an anomalous behavior was noticed at high positive potentials when fluorescence excitation intensity, Iexc, was varied. At applied potentials above 0.7 V (versus Ag/AgCl), populations of nanoparticles near the interface increase with increasing excitation intensity. Figure 2A shows a series of representative images captured at equilibrium for 5 pM yellow-green nanospheres near an ITO surface poised at 0.82 V versus Ag/AgCl, illuminated at varying intensities of 488.0 nm light. At potentials below 0.6 V, illumination intensity had no measurable effect on the interfacial population, while at 0.82 V, higher illumination intensity greatly increased populations, as shown in the plot of nanosphere population versus illumination intensity in Figure 2B. These increasing populations are not simply a result of increasing the imaging depth due to increasing the total illumination intensity, because intensity thresholds were raised in proportion to the S/N to maintain constant baseline counts at low potentials. The lack of an intensity-dependent response in the interfacial population with 0.6-V applied potential confirms control over the detected particle population as the excitation intensity is varied. To investigate this intensity-dependent behavior in more detail, potential scan experiments were performed at varying excitation intensities. Plots of the yellow-green nanosphere populations versus applied potential at various 488.0-nm illumination intensities are shown in Figure 3. As these data show, increased illumination intensity shifts the threshold potential for accumulating particles to lower potentials, while

increasing order from 0 V. Populations in Figure 1B were measured for a 5 pM nanosphere suspension in 75 μM phosphate buffer with 75 μM sodium chloride. When nanosphere suspensions were initially added to the imaging flow cell, between 5 and 10 particles were found to adsorb to the surface in each imaging region. These immobile spots were excluded from the population counts because they were insensitive to applied potential. The population of nanospheres detected in the evanescent field volume, NDET, represents the total population of molecules in the evanescent field volume near the ITO surface. As Figure 1B shows, NDET increases exponentially with potential beyond a threshold applied potential of ∼0.7 V (versus Ag/AgCl). A simple Poisson−Boltzmann model18 based on averaging the ion accumulation in the double layer24,25 has successfully described the population of polyelectrolyte molecules (a DNA plasmid) attracted to the interface in response to changes in applied potential.18 In this model, the average population of charged polymer molecules, or in the present case charged nanoparticles, in the electrical double layer, ND, is modulated by the electrical potential, φ, according to a Poisson−Boltzmann relationship, where N0D is the bulk particle population, z is the charge of the particle in electron equivalents, e is the charge of an electron, kB is Boltzmann’s constant, and T is the temperature: ND = N 0 De−zeφ / kBT

(1)

The thickness of the electrical double layer, over which the applied potential influences the interfacial population, is described by the Debye length (κ−1), which is a function of the bulk ion concentration (C0), the electrical permittivity of a vacuum (ε0), and the relative permittivity (ε): ⎛ εε k T ⎞1/2 κ −1 = ⎜ 00 B2 2 ⎟ ⎝ 2C z e ⎠

(2) −1

For the buffer composition used in this experiment, κ is ∼18 nm, which is significantly smaller than the 100-nm evanescent wave that defines our detected population, NDET. When there is no net charge on the surface, NDET is equal to the particle population within the evanescent wave at zero charge, N0E. We simplify the argument of the exponential expression in eq 1 by defining a parameter, β = −ze/kBT, which describes the sensitivity of the potential response of a molecule or particle with a net electrical charge, z, averaged over the double-layer region. As can be seen in Figure 1B, a threshold potential must be reached before nanospheres begin accumulating at the interface. This threshold potential is designated as the outer Helmholtz plane (OHP) zero potential, E0OHP, the potential required to neutralize charged ITO surface groups and adsorbed anions on the solution side of the interface.18 Additional applied potential, E, is required to neutralize this excess surface charge, so that the net potential available to interact with ions and charged particles in solution is E − E0OHP. Combining these modifications to eq 1 yields a new expression for NDET as a function of applied potential: 0

NDET = (N 0 E − N 0 D) + N 0 De−β(E − E OHP)

(3)

The expression for NDET was fit to the particle accumulation data using a nonlinear least-squares search to determine values for N0E, N0D, β, and E0OHP, and the results are plotted as dashed lines in Figure 1B. On the basis of these results, the potential11944

dx.doi.org/10.1021/la402468k | Langmuir 2013, 29, 11941−11949

Langmuir

Article

Figure 2. Nanosphere interfacial population response to illumination intensity: (A) Images of nanospheres at 0.82-V applied potential versus Ag/AgCl at varying illumination intensity; scale bar is 10 μm. (B) Plot of equilibrium particle populations at 0.82 V (black squares) and 0.6 V (red circles) at varying 488.0-nm illumination intensity. Figure 4. Plot of parameters from eq 3 for illumination at 488.0 nm: (A) potential sensitivity, β; (B) fraction of particles under potential control, N0D/N0E; and (C) OHP zero potential, E0OHP.

in the OHP zero potential, E0OHP. As shown in Figure 4C, E0OHP shifts from 0.9 to 0.6 V as the illumination power density is varied from 0.14 and 10.4 W cm−2. This shift in the onset potential accounts for the changes in particle populations shown in Figure 2A, whereupon increasing the illumination from 0.14 to 10.4 W cm−2, the applied potential relative to the E0OHP has shifted from −0.08 to 0.22 V, producing a significant increase in the interfacial charged particle population. In our model for particle accumulation at the interface, E0OHP represents the potential needed to neutralize excess charge at the outer Helmholtz plane. The OHP is the distance of closest approach to the electrode surface for ions in solution, and it represents the thickness of an adsorbed layer of water and ions. This threshold potential is analogous to the potential of zero charge for metal electrodes22 and the flat band potential for semiconductor electrodes in contact with an electrolyte solution.26 The flat band potential is the applied potential required to neutralize all charge on the electrode side of the interface, while E0OHP is the potential required to neutralize charge on the solution side of the interface. The flat band potential for ITO is approximately −0.4 V versus Ag/AgCl in neutral pH solution,27 because of significant negative surface charge from anion adsorption28 and anionic deprotonated surface hydroxyl groups.29 The excess negative charge on the interface causes positive charge to build on the semiconductor side of the interface, requiring excess negative potential to neutralize this charge. Changes in interfacial chemical ion

Figure 3. Equilibrium nanoparticle populations versus applied potential (vs Ag/AgCl) for 0.3 W cm−2 (black squares), 1.4 W cm−2 (red circles), 3.6 W cm−2 (blue triangles), and 10.4 W cm−2 (green diamonds) illumination at 488.0 nm, with fits to eq 3.

having little impact on the sensitivity to applied potential as reflected in the rate of exponential rise. To quantify the changes in potential response, the data in Figure 3 were fit with eq 3 to provide parameters describing the potential sensitivity and the potential threshold relative to illumination intensity. As shown in Figure 4A, the Boltzmann exponential factor, β, exhibits no dependence on Iexc. Above the threshold potential for accumulation, the populations of nanospheres all show the same exponential dependence on applied potential. In addition, illumination intensity does not affect the fraction of molecules in the evanescent field under potential control, N0D/N0E, as shown in Figure 4B. All of the observed changes in nanoparticle potential response with illumination intensity arise from shifts 11945

dx.doi.org/10.1021/la402468k | Langmuir 2013, 29, 11941−11949

Langmuir

Article

composition, such as pH, can change the density of surface charge groups and shift the flat band potential.27 On the solution side of the interface, this excess negative charge must be neutralized by the positive applied potential, E0OHP, to reverse the potential on the solution side and electrostatically attract anions. Therefore, a shift in E0OHP with increasing Iexc suggests a photoinduced change in ITO surface charging and a corresponding shift in the ITO surface potential. Optical trapping is another possible mechanism for particle accumulation at high illumination intensities. High laser intensity gradients can cause high refractive index particles to migrate due to optical scattering forces.30 Focused laser beams have been used to optically trap polystyrene particles at the center of a laser focus.31 Because of the short decay length of the evanescent wave, the intensity gradient normal to the interface in an evanescent wave can exceed the gradient in a tightly focused laser beam,32 so evanescent fields might be expected to trap polystyrene particles at an interface. In the present work, however, optical trapping could not be a significant contribution to particle accumulation. The power densities used in this fluorescence imaging generate intensity gradients in the evanescent wave that are 103 times smaller than are required for optical trapping of 200-nm polystyrene particles in water and nearly 104 times smaller than required to optically trap 100-nm particles.30 Photoinduced Charge Separation in ITO. We hypothesize that shifts in E0OHP with higher illumination intensity are a result of photoinduced charging of the ITO surface. Excitation of the direct band gap of ITO, 3.5−4.2 eV33,34 (340−280 nm), with ultraviolet light has been used to drive electrochemical reactions35 and to spatially control accumulation of charged particles at the ITO surface.36 However, in the present work, the sample is illuminated with light that is nearly 1 eV lower in photon energy than the ITO band gap, so that the observed surface charging is unlikely due to one-photon excitation into the direct band gap. ITO could be undergoing two-photon excitation into the direct band gap or weak one-photon excitation of a possible indirect band gap37−41 or of lattice defect sites.41,42 Either process would generate excited electrons in the conduction band and vacancies in the valence band of the lattice or “holes.” Two-photon excitation is a nonlinear optical process where an electronic transition is accessed by the combined energies of two photons that results in a secondorder rate of photoexcitation with respect to photon flux.43,44 Indirect band gap photoexcitation would correspond to weakly allowed single-photon excitation of electrons into a conduction band with a different momentum wave vector than the valence band,45 while lattice defects can lead to accessible exciton states at energies well below the direct band gap;41,42 both processes would follow first-order kinetics with respect to photon flux. To determine whether one-photon or two-photon excitation governs photoexcitation in our samples, we have developed a kinetic model to evaluate the relative rate order of the photoexcitation and electron−hole recombination process, as outlined in Figure 5A. In this model, ITO lattice sites are excited by a number, n, of photons (hν), to form electron (e−) and hole (h+) pairs or excitons: n(hν) + ITO → e− + h+

Figure 5. (A) Diagram of the photoexcitation model; see the text for details. (B) Plot of the natural log of the charge density (μC cm−2) versus the natural log of the illumination intensity (W cm−2), with weighted linear least-squares fit (dashed line).

d[h+] = k′1[ITO][hν]n = k1[hν]n dt

(5)

Due to the negative charge on the ITO surface, holes migrate to the space-charge layer at the electrode surface, neutralizing negative surface charge and shifting E0OHP to more positive potentials. At the interface, excitons can be destroyed via electron−hole recombination, which is typically a nonradiative process that occurs at the numerous lattice defect sites in amorphous semiconductors like ITO:45 e− + h+ → [ITO]

(6)

Because electrons and holes are generated in pairs by the same process, the concentration of electrons is equivalent to the concentration of holes. The resulting rate expression for electron hole recombination with rate constant k2 is d[h+] = −k 2[h+][e−] = −k 2[h+]2 dt

(7)

At steady-state, the concentration of holes is constant, and rates of hole generation and exciton recombination are equivalent: d[h+] = 0 = k1[hν]n − k 2[h+]2 dt k1[hν]n = k 2[h+]2

so that (8)

The excess hole concentration, [h+], during photoexcitation is proportional to the excess charge density, σ, near the interface, and the photon flux is proportional to the excitation intensity, Iexc. By taking the natural logarithm of eq 8, we obtain a linear expression that reveals the order of the exciton generation process in terms of photon flux and surface charge density:

(4)

The expression for the rate of hole generation in the absence of charge carrier saturation of the ITO is given by 11946

dx.doi.org/10.1021/la402468k | Langmuir 2013, 29, 11941−11949

Langmuir

Article

n 1 ln(Iexc) + [ln(k1) − ln(k 2)] (9) 2 2 The experimentally measured shift in surface potential was determined by subtracting E0OHP at each illumination intensity from its value at very low light levels, which averaged E0OHP ≈ 0.9 V. Surface charge density was estimated by multiplying the shift in potential by an estimated Helmholtz layer capacitance22 of 15 μF cm−2. We can estimate the order of this process in photons, n, from the slope of a plot of ln(σ) versus ln(Iexc), shown in Figure 5B, where the slope of a line fit from linear least-squares, n = 1.2 ± 0.4, indicates that the photoexcitation process is first-order in photon flux. This result is reasonable, because two-photon excitation processes have very low absorption cross sections, typically requiring 106 greater illumination power densities than those used in this work.46 If a one-photon process is indeed responsible for 488-nm photoexcitation of surface charge in ITO films, there should be evidence in the optical absorption spectrum of the ITO film (see Figure 6). As can be seen in these results, absorption by

A one-photon excitation process is, therefore, consistent with the weak but detectable optical absorption of the film at 488 nm. This absorption may arise from indirect band gap excitation, and several researchers have reported indium oxide indirect band gaps between 450 and 500 nm,37,38,40 close to our 488-nm excitation wavelength. However, recent density functional theory calculations predict that there is no indirect band gap near 500 nm and that light absorption in the visible region arises from photoexcitation of surface states and defects caused by the high level of n-type doping in ITO.41 Whether the photoexcitation is into an indirect band gap or due to defect sites, both processes correspond to one-photon excitations that are linear in photon flux, consistent with the weak but detectable optical absorption of the film at 488 nm. Photoinduced ITO Charging with 647-nm Illumination. If the apparent charging of the ITO−solution interface is a result of photoexcitation into an indirect band gap or into defect sites with energies of ∼2.5 eV (488 nm), we would expect illumination with photon energy far below the energy to have a negligible effect on surface charging. For the present ITO film, the contribution to transmittance loss from optical absorption at wavelengths longer than 630 nm is negligible (Figure 6, inset), and it has been reported that photoexcitation of tin-doped indium oxide films decreases significantly with lower photon energy.47 To test this hypothesis, the potentialcontrolled nanosphere accumulation experiment was repeated with 647-nm excitation. In this experiment, populations of 200nm dark-red fluorescent carboxylated-polystyrene nanospheres were counted near an ITO surface at potentials between 0 and 1.6 V versus an Ag/AgCl reference electrode. The dark-red nanospheres had high charge density similar to that of the yellow-green nanospheres, 1 × 106 carboxylate groups per nanosphere based on a titration analysis by the supplier and 2 × 106 excess negative charges per particle based on zeta-potential measurements. Supporting electrolyte and buffer concentration were the same used to image yellow-green nanospheres, and the particle concentration in solution was 2 pM. Potential scans were performed at illumination power densities between 1 and 15 W cm−2, and the results are plotted in Figure 7A. Equation 3 was fit to each curve to determine parameters describing the potential sensitivity, β, and E0OHP, which are plotted in parts B and C of Figure 7, respectively. The lower particle concentration used for the dark-red nanospheres compared to the yellow-green nanospheres should not influence β, and E0OHP in eq 3, since these parameters govern relative changes in interfacial population (see above). As Figure 7 shows, there is little sensitivity in either β or E0OHP to illumination intensity at photon energy far below the indirect band gap or defect energy. Because the power densities of the long-wavelength excitation were even greater than those used for 488-nm experiments, the lack of an intensity-dependent response provides further evidence that there is no optical trapping by the evanescent field and that the shifts in E0OHP with illumination at higher photon energies are indeed due to a photoexcitation of charge carriers in the ITO film.

ln(σ ) ∝

Figure 6. UV/vis absorption spectrum of ITO substrate. Inset: Transmittance of ITO substrate with measured spectrum (black), model spectrum from glass and ITO optical constants for a film thickness of 280 nm (blue), and model spectrum without absorption term (red).

the ITO film at ultraviolet wavelengths less than 350 nm is strong, consistent with the expected direct band gap of ITO.33,34 At longer wavelengths, some modulated loss is observed, but this could also arise due to reflections from the high refractive index ITO film. To account for this modulation, the optical constants for the ITO film material, provided by the supplier between 400 and 700 nm, were used in a three-phase thin-film interference model20 to fit the transmittance data. Wavelength-dependent complex refractive indices of the ITO and glass substrate were used to predict the transmittance through the substrate, including reflection off the back of the glass slide, by varying a single parameter, the ITO film thickness. The optimal fit indicates an ITO film thickness of 280 nm and agrees quantitatively with the transmittance data without amplitude scaling (see Figure 6, inset) over the range from 400 to 550 nm, which includes the 488-nm excitation wavelength. To test the contribution of optical absorption to the measured thin film transmittance results, the three-phase model was modified to predict a transmittance spectrum without absorption by setting the imaginary refractive index to zero. The results show that the actual film has a measurable absorption loss at wavelengths as long as 600 nm, which becomes undetectable at wavelengths longer than ∼630 nm (see Figure 6, inset).

■

SUMMARY In this work, we investigated the charging of ITO surfaces by photoexcitation of electron−hole pairs at photon energies well below the direct band gap of the material. We demonstrated the use of charged fluorescent polystyrene nanospheres as probes of surface charge density by counting the particle populations attracted to an ITO electrode with applied potential. The 11947

dx.doi.org/10.1021/la402468k | Langmuir 2013, 29, 11941−11949

Langmuir

Article

much as 0.3 V when using power densities between 6 and 10 W cm−2 (see Figure 4C). For this reason, low excitation intensities and fluorescence excitation wavelengths well above 600 nm are recommended to avoid photoexcitation of charge carriers in ITO and shifting its surface potential. The detection of potential shifts from photoexcited charge carriers by counting of charged colloidal particles at the interface is a sensitive method to investigate these photophysics in small illuminated areas; electrochemical (potentiometric) detection of these shifts might be possible if one had sufficient laser power (6−10 W) to illuminate a macroscopic ITO (1 cm2) electrode surface.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS This research was supported in part with funds from the U.S. Department of Energy under Grant DE-FG03-93ER14333. REFERENCES

(1) Kuwana, T.; Darlington, R. K.; Leedy, D. W. Electrochemical Studies Using Conducting Glass Indicator Electrodes. Anal. Chem. 1964, 36 (10), 2023−2025. (2) Kuwana, T.; Heineman, W. R. Study of Electrogenerated Reactants Using Optically Transparent Electrodes. Acc. Chem. Res. 1976, 9 (7), 241−248. (3) Heineman, W. R.; Jensen, W. B. Spectroelectrochemistry Using Transparent Electrodes. In Electrochemistry, Past and Present; American Chemical Society: Washington, DC, 1989; Vol. 390, pp 442−457. (4) Bowden, E. F.; Hawkridge, F. M.; Blount, H. N. The Heterogeneous Electron Transfer Properties of Cytochrome c. In Electrochemical and Spectrochemical Studies of Biological Redox Components; American Chemical Society: Washington, DC, 1982; Vol. 201, pp 159−171. (5) Pinyayev, T. S.; Seliskar, C. J.; Heineman, W. R. Fluorescence Spectroelectrochemical Sensor for 1-Hydroxypyrene. Anal. Chem. 2010, 82 (23), 9743−9748. (6) Lu, H. P.; Xie, X. S. Single-Molecule Kinetics of Interfacial Electron Transfer. J. Phys. Chem. B 1997, 101 (15), 2753−2757. (7) Palacios, R. E.; Chang, W.-S.; Grey, J. K.; Chang, Y.-L.; Miller, W. L.; Lu, C.-Y.; Henkelman, G.; Zepeda, D.; Ferraris, J.; Barbara, P. F. Detailed Single-Molecule Spectroelectrochemical Studies of the Oxidation of Conjugated Polymers. J. Phys. Chem. B 2005, 113 (44), 14619−14628. (8) Palacios, R. E.; Fan, F.-R. F.; Bard, A. J.; Barbara, P. F. SingleMolecule Spectroelectrochemistry (SMS-EC). J. Am. Chem. Soc. 2006, 128 (28), 9028−9029. (9) Ara, A. M.; Iimori, T.; Nakabayashi, T.; Maeda, H.; Mizuno, K.; Ohta, N. Electric Field Effects on Absorption and Fluorescence Spectra of Trimethylsilyl- and Trimethylsilylethynyl-Substituted Compounds of Pyrene in a PMMA Film. J. Phys. Chem. B 2007, 111 (36), 10687−10696. (10) Smith, T. M.; Kim, J.; Peteanu, L. A.; Wildeman, J. Electric Field Effects on Internal Conversion: An Alternative Mechanism for FieldInduced Fluorescence Quenching of MEH-PPV and Its Oligomers in the Low Concentration Limit. J. Phys. Chem. C 2007, 111 (27), 10119−10129. (11) Nakabayashi, T.; Hino, K.; Ohta, Y.; Ito, S.; Nakano, H.; Ohta, N. Electric-Field-Induced Changes in Absorption and Fluorescence of the Green Fluorescent Protein Chromophore in a PMMA Film. J. Phys. Chem. B 2011, 115 (26), 8622−8626. (12) Shi, Y.; Slaterbeck, A. F.; Seliskar, C. J.; Heineman, W. R. Spectroelectrochemical Sensing Based on Multimode Selectivity

Figure 7. Interfacial population of 200-nm particles excited with 647nm light at varying power density: (A) Population versus applied potential at 3.6 W cm−2 (black squares), 7.1 W cm−2 (red circles), and 10.7 W cm−2 (blue triangles) with fitted functions from eq 3. (B) Potential sensitivity, β, versus laser power density and (C) OHP zero potential, E0OHP, from the fit to eq 3.

population response to potential was fit to a simple Poisson− Boltzmann model, which provides parameters describing the potential sensitivity and the onset potential for particle accumulation. Increasing excitation intensity shifted the threshold potential for particle accumulation to more positive potentials, while the potential sensitivity of particle accumulation was unaffected. We attributed the shift in onset potential to photoinduced charge separation in ITO, which caused positively charged holes to accumulate at the ITO−solution interface. From a steady-state kinetics analysis of the change in surface charge density versus incident photon flux, the surface charging was found to be consistent with one-photon photoexcitation of the indirect band gap or defects sites in ITO. The photoexcitation hypothesis was confirmed by a control experiment which showed no photoinduced charging of the interface by illumination at photon energies well below the indirect band gap or defect energy where evidence of optical absorption is not detectable. The illumination intensities used this work were typical for fluorescence imaging measurements, and excitation wavelengths below 530 nm are commonly used to image fluorophores such as FITC,48 Alexafluor 488,49 cy3,50 and green fluorescent protein11,15 near ITO surfaces. Our results suggest that moderate illumination intensities even at wavelengths 1 eV below the direct band gap can influence the surface charge density of ITO, shifting surface potentials by as 11948

dx.doi.org/10.1021/la402468k | Langmuir 2013, 29, 11941−11949

Langmuir

Article

Properties of Indium Tin Oxide Thin Films. J. Appl. Phys. 2000, 88 (10), 6021−6025. (35) Tansil, N. C.; Xie, H.; Gao, Z. Photoelectrochemical Behavior of Oxalate at an Indium Tin Oxide Electrode. J. Phys. Chem. B 2004, 108 (43), 16850−16854. (36) Hayward, R. C.; Saville, D. A.; Aksay, I. A. Electrophoretic Assembly of Colloidal Crystals with Optically Tunable Micropatterns. Nature 2000, 404 (6773), 56−59. (37) Weiher, R. Optical Properties of Indium Oxide. J. Appl. Phys. 1966, 37 (1), 299. (38) McCann, J. F.; Bockris, J. O. M. Photoelectrochemical Properties of n-Type In2O3. J. Electrochem. Soc. 1981, 128 (8), 1719−1723. (39) Ray, S.; Banerjee, R.; Basu, N.; Batabyal, A. K.; Barua, A. K. Properties of Tin Doped Indium Oxide Thin Films Prepared by Magnetron Sputtering. J. Appl. Phys. 1983, 54 (6), 3497−3501. (40) van den Meerakker, J. E. A. M.; Meulenkamp, E. A.; Scholten, M. (Photo)electrochemical Characterization of Tin-Doped Indium Oxide. J. Appl. Phys. 1993, 74 (5), 3282−3288. (41) Erhart, P.; Klein, A.; Egdell, R. G.; Albe, K. Band Structure of Indium Oxide: Indirect versus Direct Band Gap. Phys. Rev. B 2007, 75 (15), 153205. (42) Gonzalez, G. B.; Cohen, J. B.; Hwang, J.-H.; Mason, T. O.; Hodges, J. P.; Jorgensen, J. D. Neutron Diffraction Study on the Defect Structure of Indium−Tin-Oxide. J. Appl. Phys. 2001, 89 (5), 2550−2555. (43) Kaiser, W.; Garrett, C. G. B. Two-Photon Excitation in CaF2: Eu2+. Phys. Rev. Lett. 1961, 7 (6), 229−231. (44) Bechtel, J. H.; Smith, W. L. Two-photon absorption in semiconductors with picosecond laser pulses. Phys. Rev. B 1976, 13 (8), 3515−3522. (45) Pleskov, Y. V.; Gurevich, Y. Y. Semiconductor Photoelectrochemistry; Consultants Bureau: New York, 1986; p 422. (46) McClain, W. M. Two-photon molecular spectroscopy. Acc. Chem. Res. 1974, 7 (5), 129−135. (47) Hamberg, I. Evaporated Sn-Doped In2O3 Films: Basic Optical Properties and Applications to Energy-Efficient Windows. J. Appl. Phys. 1986, 60 (11), R123. (48) Hwang, H.; Park, J.-K. Measurement of Molecular Diffusion Based on Optoelectrofluidic Fluorescence Microscopy. Anal. Chem. 2009, 81 (21), 9163−9167. (49) Li, L.; Tian, X.; Zou, G.; Shi, Z.; Zhang, X.; Jin, W. Quantitative Counting of Single Fluorescent Molecules by Combined Electrochemical Adsorption Accumulation and Total Internal Reflection Fluorescence Microscopy. Anal. Chem. 2008, 80 (11), 3999−4006. (50) Schlapak, R.; Armitage, D.; Saucedo-Zeni, N.; Hohage, M.; Howorka, S. Dense Passivating Poly(ethylene glycol) Films on Indium Tin Oxide Substrates. Langmuir 2007, 23 (20), 10244−10253.

Simultaneously Achievable in a Single Device. 1. Demonstration of Concept with Ferricyanide. Anal. Chem. 1997, 69 (18), 3679−3686. (13) Andria, S. E.; Seliskar, C. J.; Heineman, W. R. Simultaneous Detection of Two Analytes Using a Spectroelectrochemical Sensor. Anal. Chem. 2010, 82 (5), 1720−1726. (14) Asanov, A. N.; Wilson, W. W.; Oldham, P. B. Regenerable Biosensor Platform: A Total Internal Reflection Fluorescence Cell with Electrochemical Control. Anal. Chem. 1998, 70 (6), 1156−1163. (15) Luo, W.; Westcott, N. P.; Pulsipher, A.; Yousaf, M. N. Renewable and Optically Transparent Electroactive Indium Tin Oxide Surfaces for Chemoselective Ligand Immobilization and Biospecific Cell Adhesion. Langmuir 2008, 24 (22), 13096−13101. (16) Tang, C. S.; Dusseiller, M.; Makohliso, S.; Heuschkel, M.; Sharma, S.; Keller, B.; Vörös, J. Dynamic, Electronically Switchable Surfaces for Membrane Protein Microarrays. Anal. Chem. 2005, 78 (3), 711−717. (17) Hiemenz, P. C. Principles of Colloid and Surface Chemistry, 2nd ed.; Marcel Dekker, Inc.: New York, 1986; pp 737−790. (18) Peterson, E. M.; Harris, J. M. Single-Molecule Fluorescence Imaging of DNA at a Potential-Controlled Interface. Langmuir 2013, 29 (26), 8292−8301. (19) Peterson, E. M.; Harris, J. M. Quantitative Detection of Single Molecules in Fluorescence Microscopy Images. Anal. Chem. 2009, 82 (1), 189−196. (20) Hansen, W. Electric Fields Produced by the Propagation of Plane Coherent Electromagnetic Radiation in a Stratified Medium. J. Opt. Soc. Am. 1968, 58 (3), 380−390. (21) Efron, B. Bootstrap Methods: Another Look at the Jackknife. Ann. Stat. 1979, 7 (1), 1−26. (22) Grahame, D. C. The Electrical Double Layer and the Theory of Electrocapillarity. Chem. Rev. 1947, 41 (3), 441−501. (23) Böhmer, M. Situ Observation of 2-Dimensional Clustering during Electrophoretic Deposition. Langmuir 1996, 12 (24), 5747− 5750. (24) Gouy, L. G. Sur la Constitution de la Charge Électrique à la Surface d’un Électrolyte. J. Phys. Theor. Appl. 1910, 9 (1), 457−467. (25) Chapman, D. L. A Contribution to the Theory of Electrocapillarity. Phil. Mag. 1913, 25 (148), 475−481. (26) Laidler, K. J.; King, M. C. Development of Transition-State Theory. J. Phys. Chem. 1983, 87 (15), 2657−2664. (27) Bolts, J. M.; Wrighton, M. S. Correlation of Photocurrent− Voltage Curves with Flat-Band Potential for Stable Photoelectrodes for the Photoelectrolysis of Water. J. Phys. Chem. 1976, 80 (24), 2641− 2645. (28) Swint, A. L.; Bohn, P. W. Effect of the Interfacial Chemical Environment on In-Plane Electronic Conduction of Indium Tin Oxide: Role of Surface Charge, Dipole Magnitude, and Carrier Injection. Langmuir 2004, 20 (10), 4076−4084. (29) Donley, C.; Dunphy, D.; Paine, D.; Carter, C.; Nebesny, K.; Lee, P.; Alloway, D.; Armstrong, N. R. Characterization of Indium−Tin Oxide Interfaces Using X-ray Photoelectron Spectroscopy and Redox Processes of a Chemisorbed Probe Molecule: Effect of Surface Pretreatment Conditions. Langmuir 2001, 18 (2), 450−457. (30) Ashkin, A.; Dziedzic, J. M.; Bjorkholm, J. E.; Chu, S. Observation of a Single-Beam Gradient Force Optical Trap for Dielectric Particles. Opt. Lett. 1986, 11 (5), 288−290. (31) Cherney, D. P.; Bridges, T. E.; Harris, J. M. Optical Trapping of Unilamellar Phospholipid Vesicles: Investigation of the Effect of Optical Forces on the Lipid Membrane Shape by Confocal-Raman Microscopy. Anal. Chem. 2004, 76 (17), 4920−4928. (32) Walz, J. Y.; Prieve, D. C. Prediction and Measurement of the Optical Trapping Forces on a Microscopic Dielectric Sphere. Langmuir 1992, 8 (12), 3073−3082. (33) Mason, M. G.; Hung, L. S.; Tang, C. W.; Lee, S. T.; Wong, K. W.; Wang, M. Characterization of Treated Indium−Tin-Oxide Surfaces Used in Electroluminescent Devices. J. Appl. Phys. 1999, 86 (3), 1688−1692. (34) Kim, H.; Horwitz, J. S.; Kushto, G.; Pique, A.; Kafafi, Z. H.; Gilmore, C. M.; Chrisey, D. B. Effect of Film Thickness on the 11949

dx.doi.org/10.1021/la402468k | Langmuir 2013, 29, 11941−11949