Organic Networks as Chemiresistor Coatings: The

Jul 22, 2008 - Gold Nanoparticle/Organic Networks as Chemiresistor Coatings: The Effect of ... The chemical sensitivity of the films was characterized...
0 downloads 0 Views 1MB Size
J. Phys. Chem. C 2008, 112, 12507–12514

12507

Gold Nanoparticle/Organic Networks as Chemiresistor Coatings: The Effect of Film Morphology on Vapor Sensitivity Yvonne Joseph,* Berit Guse,+ Tobias Vossmeyer,# and Akio Yasuda§ Materials Science Laboratory, Sony Deutschland GmbH, Hedelfinger Str. 61, D-70327 Stuttgart, Germany ReceiVed: February 15, 2008; ReVised Manuscript ReceiVed: April 18, 2008

Networked films comprising gold nanoparticles (4 nm core diameter) and dodecanedithiol were deposited via layer-by-layer self-assembly. The film thickness was controlled by the number of deposition cycles and ranged from submonolayer to multilayer coverage with up to 60 nm thickness. FE-SEM and XPS revealed island growth during the first four to five deposition cycles. At room temperature, films based on islands showed slightly nonlinear current-voltage curves, whereas thicker films gave Ohmic behavior. Between 100 and 300 K, the temperature dependence of conductance was consistent with an Arrhenius model for activated charge transport. The chemical sensitivity of the films was characterized by dosing them with vapors of toluene, 4-methyl-2-pentanone, 1-propanol, and water while monitoring their relative differential resistances. Thin films responded with a decrease in the resistance, whereas thicker films responded with a resistance increase. The results indicate that swelling along the film normal may be important for the underlying sensing mechanism of homogeneous multilayer films. Changes in permittivity and/or swelling-induced decrease of the interisland distance can be responsible for the observed decrease in resistance of the thinner films. Introduction Over the past few years, the use of gold nanoparticles for the development of novel biochemical and chemical sensors has attracted considerable attention.1–3 Such sensors are fabricated by depositing thin films of ligand-capped gold nanoparticles onto appropriate transducers, which are then used for a liquid or gas phase sensing operation. For gas phase applications, masssensitive sensors4–8 and, especially, chemiresistor devices5–19 are under investigation in several laboratories. In the sensor coating, the gold particles provide electrical conductivity to the material. Gold is the preferred choice of metal because of its chemical inertness. The capping ligands are usually selected from the class of organic thiols, which form a molecular monolayer around the metal core. This monolayer provides well-defined tunnel barriers between neighboring cores. In some cases the gold cores have been chemically linked through metal ion coordination,6,7 organic dithiols,5,8,16–19 or polyfunctional dendrimers,8,14,15 resulting in networked films with good mechanical stability. Wet chemical depositon of assemblies comprising nanoparticles stabilized by organic thiol ligands can be performed by spin-casting,20 spray-coating,9 deposition by Langmuir-Blodgett techniques,21 or by stamping.22 These assemblies can be later interlinked by subsequent ligand-linker exchange. Linked multilayer assemblies can also be prepared by direct ligand-linker exchange reactions followed by precipitation or, in a more controlled way, by a layer-bylayer self-assembly procedure.7,8,10,13–17,23–26 Despite the large efforts in research on the physical and chemical properties of the layer-by-layer assembled materials, investigations of the * Corresponding author. Phone: +49(0) 711-5858-836. Fax: +49(0) 7115858-484. E-mail: [email protected]. + Current address: Product Development, tesa AG, Quickbornstraβe 24, 20253 Hamburg. # Current address: Institut fu ¨ r Physikalische Chemie, Universita¨t Hamburg, Grindelallee 117, 20146 Hamburg. § Current address: Sony Bioinformatics Center, Tokyo Medical and Dental University, 1-5-45 Yushima Bunkyo-ku, Tokyo, 113-8510 Japan.

growth of the films and their final morphologies are sparse. The assembly of alkylthiol-, tetraoctylammonium bromide- or citratestabilized Au nanoparticles with alkyldithiols were investigated.10,26–28 However, in these studies, the film growth was mainly monitored by electrical measurements, and thus, the growth before network formation was only barely accessible. The conductivity of metal nanoparticle films has been discussed qualitatively in the context of an activated tunneling model.16,20,29 This gives the following expression for the conductivity of the film, σ:

( )

σ ∝ exp(-βδ) exp

-Ea kBT

(1)

where β is the tunneling decay constant (or electronic coupling coefficient), δ is the edge-to-edge separation of the metal cores, Ea is the activation energy, kB is the Boltzmann constant, and T is the absolute temperature. The first exponential term in relation 1 takes into account tunneling of charges between neighboring particles. The second exponential term includes the activation energy, Ea, for charge transport. Attempts have been made to explain Ea with the outer-sphere reorganization energy known from Marcus theory,29 with the classical Coulomb energy,20 or with the charging energy of the nanoparticles.30,31 Compared to conventional metal oxide-based sensors, an advantage of gold nanoparticle-based chemiresistors is seen in the possibility to operate these sensors at room temperature, or slightly above, which enables easy device integration and needs low power consumption. Additionally, these sensors have shown limits of detection in the parts-per-million16 or even parts-perbillion17,32 range, with remarkably fast and reversible response characteristics. In contrast to recent progress toward the applications of chemiresistors from ligand-capped gold nanoparticles and their integration into microanalytical systems on a chip,33 unfortunately, many of the underlying molecular mechanisms generating the electrical sensor signal remain only vaguely understood.

10.1021/jp8013546 CCC: $40.75  2008 American Chemical Society Published on Web 07/22/2008

12508 J. Phys. Chem. C, Vol. 112, No. 32, 2008 It is known that the chemical selectivity of the sensors can be well-controlled through functional groups and structural features of the organic ligands or linker decorating the nanoparticle cores.12,14,18 In addition, the amount of analyte sorbed into the film is easily accessible when quartz crystal microbalances are coated with the sensitive films and the change in resonant frequency during analyte exposure is measured. However, other effects of the sorption inherent to the electrical signal transduction in chemiresistors have been discussed only qualitatively. Relation 1 suggests that analyte sorption can have two counteracting effects. First, swelling may increase the resistance due to an increased interparticle tunnel distance, δ. Second, an increase in the permittivity of the organic matrix surrounding the metal cores decreases the resistance due to a decrease in the activation energy, Ea, and due to a reduction of the heights of the potential well barriers between the metal cores and thus decreases the tunneling decay constant β. It is well-known that the response of chemiresistors from ligand-capped Au nanoparticles can be either positive (increase in resistance) or negative (decrease in resistance). In qualitative agreement with relation 1, currently available data indicate that analytes with high dielectric constants, , tend to provoke a decrease in resistance (e.g., water,9,10 methanol,11 ethanol,12 propanol9,10). In contrast, analytes with low dielectric constant (e.g., toluene,5,9,10 n-hexane5) give an increase in the resistance. However, a decrease in the resistance was observed when dosing chemiresistors with toluene when using a rigid linker such as [4]staffane-3,3′′′-dithiol to interlink the nanoparticles. It was suggested that the rigid linker suppresses swelling of the material, and the change in permittivity becomes dominant.34 Mixed response signatures showing negative and positive signal contributions with different time constants have also been published.13 In addition, it has been shown that the direction of response can also depend on the analyte concentration. These results are, however, controversial. Crossovers from resistance increase to decrease with increasing analyte concentration10 and vice-versa12,35 have been reported. It was even reported in two different publications that the same material when dosed with the same amount of methanol responded with a decrease11 as well as with an increase12 in resistance. Nevertheless, in all cases, the results were discussed qualitatively by the counteracting effects of film swelling and permittivity increase according to relation 1. The controversial results outlined above clearly show that the experimental control on the response characteristics during sensor fabrication still relies heavily on practical experience and only partially on a fundamental understanding of the various parameters affecting the response signature. To address the questions regarding the growth, electrical properties, and response characteristics of chemiresistors, in this study, we investigate assemblies of dodecylamine-stabilized gold nanoparticles, networked through dodecanedithiol linker. In particular, we show that morphology and thickness of the gold nanoparticle network have a dramatic influence on the sensor response. Experimental Section Materials. Chemicals with reagent grade or of higher quality were purchased from Merck or Sigma-Aldrich and were used as received. The synthesis of the 1,12-dodecanedithiol (C12) is described elsewhere.16 Au nanoparticles with a dodecylamine stabilizing shell were prepared in a way similar to that described by Leff et al.36 and Brust et al.,37 resulting in particles with a core diameter of ∼4 nm with a standard deviation of ∼30%.

Joseph et al. Film Preparation. The chemiresistor films were prepared on the basis of the method described by Bethell et al.38 As substrates, glass or oxidized silicon wafer equipped with interdigitated gold electrode structures were used (50 finger pairs, 5 µm width and 100 nm height, 5 µm spacing, 900 µm overlap; Institut fu¨r Mikrotechnik Mainz GmbH, Mainz, Germany). The cleaning and silanization of the chemiresistor substrates is described elsewhere.16 The films were deposited via layer-by-layer self-assembly using Au nanoparticles and 1,12-dodecanedithiol (DT) as linker molecule. The concentration of the Au nanoparticles was measured and controlled by ensuring an absorbance of 0.4 at the maximum of the plasmon band (λmax ) 514 nm with a 2 mm path length), whereas the concentration of the linker solution was 25 µmol in 5 mL (5 mM) of toluene. The cleaned and silanized substrates were placed for 15 min alternatingly into the Au nanoparticle solution and into the DT solution. The exposure to both solutions equals 1 deposition cycle. This was repeated 2, 3, 4, 5, 8, 14, and 20 times to obtain different thicknesses of interlinked nanoparticle films. To avoid a precipitation of the solutions, the chemiresistors were rinsed in toluene baths between solutions. UV/Vis Spectroscopy. To observe the film growth during the layer-by-layer assembly, a Varian Cary 50 spectrophotometer was used. The absorbance at the maximum of the plasmon band of the Au nanoparticles was detected after each deposition cycle. Atomic Force Microscopy (AFM). The thicknesses of the samples were determined by using a Digital Instruments Dimension 3100 atomic force microscope with a Nanoscope IV controller. All images were taken in tapping mode. To calculate the thickness of the films, AFM images across scratches were taken. These scratches were made by hand by applying gentle pressure to hypodermic needles while moving across the surface of the films. Scanning Electron Microscopy (SEM). The film morphology was characterized using a Leo Gemini 1530 field emission scanning electron microscope. The samples were contacted with silver paint and conducting carbon pads to prevent charging in the SEM. To acquire the images, an Inlens detector was used to detect the electrons. The particle coverage was determined by making a binary of the image (performed by ImageJ software) and counting black and white pixels. X-ray Photoelectron Spectroscopy (XPS). The chemical composition of the films was investigated with a Kratos Axis Ultra system equipped with a DLD detector, a monochromated Al KR X-ray source, and a charge neutralizer. The pass energy of the analyzer was set to 20 eV, and the power of the X-ray source, to 120 W. From all spectra, a linear background was substracted, and the binding energy was calibrated to Au 4f ) 84.0 eV. The spectra were fitted using Voigt profiles (30% Lorentz, 70% Gauss) with the spin orbit splitting for Au 4f and S 2p set to 3.65 and 1.18 eV, respectively. The intensity ratio was kept at 4:3 (Au 4f7/2/Au 4f5/2) and 2:1 (S 2p3/2/S 2p1/2). The compositional atomic ratios of the film materials were determined from the integral intensities of the signals, which were corrected by Scofield factors for the instrument (C ) 0.278, S 2p ) 0.450, Au 4f ) 6.250). Charge Transport Measurements. For the observation of the current-voltage (IV) characteristics at variable temperatures (100-300 K), a home-built setup comprising a liquid nitrogen dewar, a computer-interfaced temperature controller (Lakeshore 330), and a source/monitor unit (HP4142B) was used. Sensor Measurements. The same instruments as described in ref 16 were used to perform in situ resistance measurements. Purified and dried air was used as carrier gas. As test vapors,

Chemiresistor Coatings

J. Phys. Chem. C, Vol. 112, No. 32, 2008 12509

Figure 2. FE-SEM images of 4- and 5-cycle Au nanoparticle films. The substrate is clearly visible, indicating island growth at the beginning of film assembly.

Figure 1. (a) Absorbance and position of the Au plasmon band of the films versus the number of preparation cycles. (b) Conductance and conductivity at room temperature versus the number of preparation cycles.

TABLE 1: Thickness, Room Temperature Conductivity and Activation Energy for Films with Different Numbers of Preparation Cycles 4 cycles d

10.8 (nm)

σRT (S/cm) Ea (kJ/mol)

3.8 × 8.6

5 cycles 11.9

10-6

4.4 × 7.5

8 cycles 20.5

10-4

6.3 × 6.4

10-4

14 cycles

20 cycles

40.5

60.3

8.0 × 5.0

10-4

8.9 × 10-4 5.1

we used toluene (ε ) 2.38), 4-methyl-2-pentanone (4M2P, ε ) 13.1), 1-propanol (ε ) 20.8), and water (ε ) 80.1). These analytes were selected because they have similar vapor pressures. Therefore, differences in partitioning of the analyte in the sensor coating should depend mainly on their chemical properties; that is, polarity and structural features. The resistances of the chemiresistors were monitored with an applied bias of 100 mV. All experiments were carried out at room temperature. The sensor signal is expressed as the change in resistance, ∆R, divided by the baseline resistance, Rini; that is, the relative differential resistance response. Results The layer-by-layer self-assembly of the film was followed by UV/vis spectroscopy and resistivity measurements. A nearly linear increase of the absorbance after the first few cycles can be observed with an increasing number of deposition cycles, as shown in Figure 1a. The UV/vis spectra are given in Figure S1 of the Supporting Information. This confirms that within each preparation cycle, a similar amount of gold nanoparticles is deposited. Additionally, in Figure 1a, the corresponding position of the plasmon band λmax is indicated. During the first five cycles, the absorption maximum of the plasmon band shifts from

522 nm to ∼555 nm. For films with higher deposition cycles, the position stays constant. In Figure 1b, the conductance of the nanoparticle films measured after each completed deposition cycle is shown. As observed earlier,8,16 after the fourth deposition cycle, a clear increase in conductance is observed, which then further increases in an approximately linear fashion with the increasing number of deposition cycles. The film thicknesses, d, were determined by AFM, and the measured values are given in Table 1. Taking these values into account and also considering the geometry of the interdigitated electrode structure,39 room temperature conductivities, σRT, were calculated. They are included in Table 1 and Figure 1b and show that the particle network became conductive (σRT >10-7 S/cm) around the fourth deposition cycle. To characterize the structure of the nanoparticle films, field emission scanning electron microscopy (FE-SEM) was applied. Figure 2 shows FE-SEM images of films obtained after the fourth and fifth deposition cycles, respectively. Here, after the fourth cycle, the substrate is covered with islands, which form a partly fused network. The total coverage of the substrate area with islands is (32 ( 10)%. After the fifth cycle, the islands have fused into an almost continuous layer with coverage of (65 ( 10)%, and isolated islands of nanoparticles cannot be detected anymore. We emphasize that the exact structure of the island-dominated film is difficult to control by the number of deposition cycles and, thus, varies with preparation batch. To get an impression of the variations, SEM images of a different batch of films with slightly higher surface coverage are given in Figure S2 of the Supporting Information. To control the island structure in detail, patterning in the subnano regime would be necessary. The chemical composition of the films was investigated by X-ray photoelectron spectroscopy (XPS). Its dependence on the number of preparation cycles is shown in Figure 3. It was observed that the substrate signals (O 1s and Si 2p) decrease slightly up to the fourth cycle, then a strong decrease in these signals is visible. Between the 8th and 14th preparation cycles, the substrate signal vanishes. The signals of the film (Au 4f, C 1s, and S 2p) behave inversely. The same effect can be nicely observed in the region comprising Si 2s and S 2p signals shown in Figure 4. Here, the intensity of the S 2p peak increases with the number of preparation cycles, whereas the Si 2p peak decreases. The Si 2s only shows one species at 154.5 eV, which can be attributed to SiO2, according to the literature.40 The S 2p peak

12510 J. Phys. Chem. C, Vol. 112, No. 32, 2008

Figure 3. The dependence of film composition, determined by XPS, on the number of preparation cycles.

Figure 4. XPS spectra of the S 2p and Si 2s regions of films with increasing number of preparation cycles. The vertical lines indicate different sulfur and silicon species (for details, see text). The pie diagrams give the fractions of the different sulfur species in the respective films.

can be fitted by two main species. According to previously published data,16,41–44 we assign the first species at 162.3 eV (S 2p3/2) to sulfur bound to gold. The second peak at 163.7 eV can be assigned to sulfur of free thiol groups, which are not bound to the particle surface. Around 164.5 eV, a minor third species can be observed, especially for the thicker films. We attribute this tentatively to sulfur damaged by radiation. However, it can also be due to slightly asymmetric peak shapes,

Joseph et al.

Figure 5. Arrhenius plots of the conductivity for the films. The number of preparation cycles is given.

which were not considered in the data evaluation. Signals from oxidized sulfur species, expected at higher binding energies, are not observed, even though the films are stored under ambient conditions for some weeks. The pie diagrams in Figure 4 display graphically the relative amounts of the three sulfur species in the film material. It can be clearly seen that the relative amounts of free thiols decrease strongly up to the fifth cycle and then stay approximately constant. The charge transport properties of the films were determined by measuring current (I)-voltage (V) curves in the range of (1 V and conductivities in dependence of the temperature. Samples with voltage-induced degradation have not been observed during the measurements. At room temperature, the four-cycle assembly showed slightly nonlinear current-voltage curves, whereas thicker films gave ohmic behavior. The I-V curves at different temperatures are given in Figure S3 of the Supporting Information. Between 100 and 300 K, the temperature dependence of the conductivity was consistent with the Arrhenius model for activated charge transport according to equation 120,29,45 (Figure 5). From the linear fits in Figure 5, the activation energies were determined (Table 1). With increasing cycle number, the activation energy decreased from 8.6 to 5.1 kJ/mol. The chemical sensitivities and selectivities of the films were investigated by monitoring their resistances during the exposure to toluene, 1-propanol, 4-methyl-2-pentanone (4M2P), and water vapors. In Figure 6, the response traces toward 5000 ppm of analyte are shown for the indicated number of preparation cycles (the corresponding SEM images are shown in Figure S2 of the Supporting Information). All films responded very quickly, showing t90 times below 300 ms, and the responses were completely reversible. The resulting response signatures have an almost ideal rectangular shape. Interestingly, the thin samples (cycles 3-4) responded with a decrease in resistance, whereas in contrast, thicker films responded with an increase in resistance. The response isotherms for a concentration range of 100-5000 ppm are given in Figure S4 of the Supporting Information. Because the amplitude of the sensor responses depends on the exact film structure, which is difficult to control, as already mentioned above, three different sensor batches with two or three sensors per selected cycle number were investigated to obtain statistical information. For direct comparison of the measurements, all sensors within one batch were normalized to the mean sensor response of the 14 cycles sensors vs 5000 ppm toluene. The results are given in Figure 7. Despite large

Chemiresistor Coatings

J. Phys. Chem. C, Vol. 112, No. 32, 2008 12511

Figure 6. Response traces of the films with the indicated number of preparation cycles toward 5000 ppm of toluene, 1-propanol, 4-methyl2-pentanone, and water.

Figure 8. Model for film growth: (insert) model for a nanoparticle interconnection, (a) island nuclei formation and island growth, (b) film at the percolation threshold, and (c) film growth after the percolation threshold.

Figure 7. Response amplitudes of three sensor batches normalized to the signal of the 14-cycle film toward toluene in dependence of the preparation cycle number. (The 8- and 20-cycle film was not measured toward 4-methyl-2-pentanone.)

error bars, it can be seen that thin film samples responded to the analytes with a decrease in resistance. Further, it is seen that in the beginning, with increasing number of deposition cycles, the response amplitudes increase but then reach saturation around the 14th deposition cycle when further increasing the film thickness. For nonpolar analytes, this increase appears for lower deposition cycle numbers than for polar analytes (in the following order: toluene ∼ 4-methyl-2-pentanone > 1-propanol . water). Discussion (a) Film Growth. All results presented above led to the film growth model schematically depicted in Figure 8. During the first preparation cycle, some Au nanoparticles adsorb at the surface of the substrate and act in the following cycles as nucleation centers for island formation (compare Figure 8a), and the assembly is characterized by mostly individual nanoparticles. This is indicated by the high ratio of free thiols as compared to gold-bound sulfur on their surface and absorption maximum of the plasmon band at quite low wavelengths. In the next preparation cycles, new nucleation centers form, and those from earlier deposition cycles grow to larger islands. This is confirmed by SEM, XPS, and UV/vis as presented above. As expected, the number of sulfurs bound to the Au nanoparticles surfaces increases due to interlinking, the substrate signals become attenuated because of higher substrate coverage and

the plasmon band shift to higher wavelengths due to stronger interactions between particles. Around preparation cycle four, a relatively high surface coverage is reached, as observed by SEM. The free thiol (S-H) to gold-bound sulfur (S-Au) ratio and the position of the plasmon band are near their final values. As depicted in Figure 8b, the islands are now closer to each other and grow together. As a result, a 1-dimensional percolation pathway, allowing charge transport, is formed, and the material becomes conductive. The continued deposition of nanoparticles after cycle four (compare Figure 8c) grows the islands completely together into a continuous film, and a 3-dimensional growth of the material is established. The UV/vis plasmon band position as well as the chemical composition remains unchanged. The substrate signals measured by XPS become significantly attenuated until the film thickness exceeds the information depth of the method and the substrate cannot be detected anymore. With increasing film thickness, the number of possible percolation pathways increases, resulting in a higher conductivity. (b) Charge Transport. According to the literature,28,46 two metal nanoparticles separated from each other can be electrically treated as a capacitor in parallel with a resistor. This is schematically depicted in the inset in Figure 8. The organic matrix separates the two metal particles in a manner similar to a dielectric medium separating a plate capacitor. However, if the distance between the nanoparticles becomes small enough, charge can tunnel from one particle to the other, resulting in a resistive behavior of the junction. Thus, it is important how close the nanoparticles are in an assembly to determine whether the resistive or capacitive behavior becomes dominant. The grown nucleation centers, given in Figure 8a, are separated by large distances. In an equivalent electrical circuit diagram, the structure is represented by very low capacitance capacitors in series. Thus, the individual nuclei are still too far

12512 J. Phys. Chem. C, Vol. 112, No. 32, 2008

Figure 9. Directions of film swelling of (a) a multilayer film and (b) a “film” with islandlike structures.

separated from each other to form a conductive material and show no current when applying a bias. The island-dominated structure near the percolation threshold in Figure 8b gives rise to a changed equivalent electrical circuit diagram. Within an island, a resistive behavior becomes dominant due to the short distances between the nanoparticles, and the number of islands represents the number of resistors in the diagram. The junctions between different islands can still be described as capacitors, but due to the smaller island-toisland distance, they show higher capacitance. If the distance between the islands becomes small enough to allow charge tunneling between them, conductivity in the film is enabled. In the equivalent electrical circuit diagram, the capacitors are now in parallel with a resistor. It has to be emphasized that these junctions between the islands are bottleneck junctions in the 1-dimensional percolation pathway, meaning that they dominate the charge transport. Because not all junctions within the device of 1 × 1 mm were imaged with SEM, a quantitative evaluation of the data according to equation 1 is not possible. However, the origin of the nonlinearity in the observed I-V curves can be understood as follows: The bottleneck junctions are singleelectron-charging barriers in the percolation pathway. The energy of the barrier is inversely proportional to the capacitance and gives rise to Coulomb blockade behavior of the junction. Due to variations in film morphology and fluctuations in local fields induced by trapped charges, the energy barriers are not energetically equal and not distributed symmetrically between the contact electrodes. Thus, the “originally” stepped I-V curves become averaged and thus S-shaped. The same averaging effect can be observed for single electron transistors operated in series with floating gate voltages. In a 3-dimensional networked film, as shown in Figure 8c, the film exhibits junctions only where particles are interlinked with neighboring particles, and the resistive properties dominate the charge transport. Thus, the resulting I-V curves are linear, demonstrating simple ohmic properties. (c) Sensing Properties. Results from several research groups, including our own work, have shown that gold nanoparticle/ alkylenedithiol films usually respond with an increase in resistance when dosed with solvent vapors. This observation is explained by the assumption that swelling of the films increases the tunneling distances between neighboring particles. As schematically depicted in Figure 9a, the closed homogeneous films used in these investigations may swell only along the film normal, and thus, only the percolation pathways with contributions parallel to the film normal are influenced by the swelling. This means a resistance increase due to swelling to the surface normal may appear only in 3-dimensional networks and can be observed only when the resistive character of the film dominates the charge transport. Interestingly, at around cycle number 14 (∼40 nm), the responses are saturating, indicating that either the swelling reaches a limit or that the swelling does not affect the conductive percolation pathways any further. However,

Joseph et al. direct observation of swelling would require an ambient pressure electron microscope with excellent resolution. As explained above, the chemical sensitivity of the sensor coating is determined by the properties of the used organic linker. In the case of the dodecanedithiol used in this investigation, the highest amounts of sorbed analyte are expected for nonpolar analytes, whereas polar analytes should give only minor sorption. The 14 cycles film in Figure 6 shows the expected behavior, that toluene gives the highest response, followed by 4M2P, 1-propanol, and water. As described before, sometimes a decrease in resistance was reported when analytes with relatively high permittivity; that is, 1-propanol (ε ) 20.8), ethanol (ε ) 25.3), methanol (ε ) 33.0), or water (ε ) 80.1) were used or when the material is prevented from swelling by using a rigid linker. This general finding is consistent with the assumption that a decrease in resistance is caused by an increase in the average permittivity of the nanoparticles’ environment. The results presented in this article reveal that the appearance of a decrease in the resistance is not exclusively controlled by the permittivity of the analyte, but that thickness and structure of the sensitive layer play an important role. This becomes obvious from the results obtained with films at the percolation threshold (cycle 3 in Figure 6), which gave negative responses to all analytes. Two explanations for this behavior are possible. For the first explanation, it has to be assumed that the particles are pinned to the surface by dispersion forces within the island and by interaction with the aminosilane. This would imply that the material is not able to swell. Within or close enough to the bottleneck junctions, which are not completely screened by the linker/ligand analyte, molecules can still adsorb. Thus, an increase in permittivity alone, induced by sorption of anlyte at the bottleneck junctions, can decrease the resistance.34 Such an increase in permittivity mainly decreases the activation energy and thus facilitates the charge transport. However, even when the material is dosed with toluene, a resistance decrease was observed. This is remarkable because the permittivity of toluene (ε ) 2.38) is only slightly higher than that of the linker molecule (ε ) 2.01 for dodecanedithiol).16 For the second explanation, it was assumed that the islandlike structure may allow swelling, not only along the surface normal as in thicker films, but also in all directions except toward the substrate, as depicted in Figure 9b. Now swelling would result in a decrease in the island-to-island distance (bottleneck junctions) in the assembly. As described above, these junctions dominate the charge transport, and reducing the distance between the islands leads to an enhancement of the conductivity (or decrease in the resistance) of the material. Both effects, change in permittivity and swelling-induced decrease in the distance of the bottleneck junctions, have to be taken into account to explain the result that for thin films, 4M2P gives the highest (negative) response, followed by 1-propanol, toluene and water. It has to be noted that already, with a slightly increased layer thickness, the previously reported trend of resistance decrease to analytes with higher permittivity is reproduced. This can be easily seen by comparing the responses of the fifth cycle film in Figure 6. The response of the nonpolar analyte toluene is already highly positive, whereas the response of 4M2P is only slightly positive. The more polar analytes 1-propanol and water give nearly no response. Conclusions In this study, we investigated the growth, charge transport and sensor response characteristics of chemiresistors based on gold nanoparticles, which were networked through dode-

Chemiresistor Coatings canedithiol linker. The material assembly starts with the adhesion of nuclei on the substrate. The nuclei grow to islands, and the islands coalesce to form a homogeneous closed film. Further assembly leads to a 3-dimensional film. In dependence of the film morphology different chemical sensing behaviors can be observed: • For individual-nanoparticle-dominated morphology, the structure is nonconductive and the chemiresistor device is not measurable. • For island-dominated morphology, charge transport is possible after a 1-dimensional percolation pathway has formed. This percolation pathway still contains island-to-island contacts with larger distances, which are bottleneck junctions for charge transport due to their Coulomb blockade behavior. Changes in permittivity and swelling-induced decrease of the distance of the bottleneck junctions can be responsible for the observed decrease in resistance. • For homogeneous three-dimensional films, many percolation pathways are possible, and the material shows resistive behavior. It is assumed that when the device is dosed with analytes, swelling of the material along the surface normal occurs, and pathways with contribution in this direction are affected. Due to an increase in the interparticle distances along these percolation pathways, a resistance increase can be observed. Finally, it has to be emphasized that controlling the film structure during self-assembly is important for reproducible electrical and sensor results. Acknowledgment. The authors thank M. Rosenberger, L.M. Seegert, P.-P. Henkel, and K. Wiegers for technical support, as well as Dr. A. Simonis, Dr. N. Krasteva, and Dr. J. Wessels for fruitful discussions and Dr. W. Ford for corrections of the manuscript. Supporting Information Available: Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Franke, M. E.; Koplin, T. J.; Simon, U. Metal and Metal Oxide Nanoparticles in Chemiresistors: Does the Nanoscale Matter? Small 2006, 2, 36–50. (2) Shipway, A. N.; Katz, E.; Willner, I. Nanoparticle Arrays on Surfaces for Electronic, Optical, and Sensor Applications. ChemPhysChem 2000, 1, 18–52. (3) Daniel, M.-C.; Astruc, D. Gold Nanoparticles: Assembly, Supramolecular Chemistry, Quantum-Size-Related Properties, and Applications toward Biology, Catalysis, and Nanotechnology. Chem. ReV. 2004, 104, 293–346. (4) Grate, J. W.; Nelson, D. A.; Skaggs, R. Sorptive Behavior of Monolayer-Protected Gold Nanoparticle Films: Implications for Chemical Vapor Sensing. Anal. Chem. 2003, 75, 1868–1879. (5) Han, L.; Daniel, D. R.; Maye, M. M.; Zhong, C. J. Core-Shell Nanostructured Nanoparticle Films as Chemically Sensitive Interfaces. Anal. Chem. 2001, 73, 4441–4449. (6) Zamborini, F. P.; Leopold, M. C.; Hicks, J. F.; Kulesza, P. L.; Malik, M. A.; Murray, R. W. Electron Hopping Conductivity and Vapor Sensing Properties of Flexible Network Polymer Films of Metal Nanoparticles. J. Am. Chem. Soc. 2002, 124, 8958–8964. (7) Leopold, M. C.; Donkers, R. L.; Georganopoulou, D.; Fisher, M.; Zamborini, F. P.; Murray, R. W. Growth, conductivity, and vapor response properties of metal ion-carboxylate linked nanoparticle films. Faraday Discuss. 2004, 125, 63–76. (8) Joseph, Y.; Krasteva, N.; Besnard, I.; Guse, B.; Rosenberger, M.; Wild, U.; Knop-Gericke, A.; Schloegl, R.; Krustev, R.; Yasuda, A.; Vossmeyer, T. Gold-nanoparticle/organic linker films: self-assembly, electronic and structural characterisation, composition and vapour sensitivity. Faraday Discuss. 2004, 125, 77–97. (9) Wohltjen, H.; Snow, A. W. Colloidal Metal-Insulator-Metal Ensemble Chemiresistor Sensor. Anal. Chem. 1998, 70, 2856–2859. (10) Foos, E. E.; Snow, A. W.; Twigg, M. E.; Ancona, M. G. ThiolTerminated Di-, Tri-, and Tetraethylene Oxide Functionalized Gold Nano-

J. Phys. Chem. C, Vol. 112, No. 32, 2008 12513 particles: A Water-Soluble, Charge-Neutral Cluster. Chem. Mater. 2002, 14, 2401–2408. (11) Evans, S. D.; Johnson, S. R.; Cheng, Y. L.; Shen, T. Vapour Sensing Using Hybrid Organic-Inorganic Nanostructured Materials. J. Mater. Chem. 2000, 10, 183–188. (12) Zhang, H. L.; Evans, S. D.; Henderson, J. I.; Miles, R. E.; Shen, T. Vapor Sensing Using Surface Functionalized Gold Nanoparticles. Nanotechnology 2002, 13, 439–444. (13) Vossmeyer, T.; Guse, B.; Besnard, I.; Bauer, R.; Mu¨llen, K. H.; Yasuda, A. Au-Nanoparticle/Polyphenylene Dendrimer Composite Films: Preparation and Vapor-Sensing Properties. AdV. Mater. 2002, 14, 238–242. (14) Krasteva, N.; Besnard, I.; Guse, B.; Bauer, R. E.; Mu¨llen, K.; Yasuda, A.; Vossmeyer, T. Self-assembled Gold Nanoparticle/Dendrimer Composite Films for Vapor Sensing Applications. Nano Lett. 2002, 2, 551– 555. (15) Krasteva, N.; Guse, B.; Besnard, I.; Yasuda, A.; Vossmeyer, T. Gold Nanoparticle/PPI-Dendrimer Based Chemiresistors: Vapor-Sensing Properties as a Function of the Dendrimer Size. Sens. Actuators, B 2003, 92, 137–143. (16) Joseph, Y.; Besnard, I.; Rosenberger, M.; Guse, B.; Nothofer, H. G.; Wessels, J.; Wild, U.; Knop-Gericke, A.; Su, D.; Yasuda, A.; Vossmeyer, T. Self-Assembled Gold Nanoparticle/Alkanedithiol Films: Preparation, Electron Microscopy, XPS-Analysis, Charge Transport, and Vapor-Sensing Properties. J. Phys. Chem. B 2003, 107, 7406–7413. (17) Joseph, Y.; Guse, B.; Yasuda, A.; Vossmeyer, T. Chemiresistor coatings from Pt- and Au-nanoparticle/nonanedithiol films: sensitivity to gases and solvent vapors. Sens. Actuators, B 2004, 98, 188–195. (18) Vossmeyer, T.; Joseph, Y.; Besnard, I.; Harnack, O.; Krasteva, N.; Guse, B.; Nothofer, H.-G.; Yasuda, A. Gold-nanoparticle/dithiol films as chemical sensors and first steps towards their integration on chip. SPIE Proc. 2004, 5513, 202–212. (19) Iban˜ez, F. J.; Gowrishetty, U.; Crain, M. M.; Walsh, K. M.; Zamborini, F. P. Chemiresistive Vapor Sensing with Microscale Films of Gold Monolayer Protected Clusters. Anal. Chem. 2006, 78, 753–761. (20) Andres, R. P.; Bielefeld, J. D.; Janes, D. B.; Kolagunta, V. R.; Kubiak, C. P.; Mahoney, W. J.; Osifchin, R. G. Self-Assembly of a TwoDimensional Superlattice of Molecularly Linked Metal Clusters. Science 1996, 273, 1–7. (21) Chen, S. Langmuir-Blodgett Fabrcation of Two-Dimensional Robust Cross-Linked Nanoparticle Assemblies. Langmuir 2001, 17, 2878–2884. (22) Liao, J.; Bernard, L.; Langer, M.; Scho¨nenberger, C.; Calame, M. Reversible Formation of Molecular Junctions in 2D Nanoparticle Arrays. AdV. Mater. 2006, 18, 2444–2447. (23) Harnack, O.; Raible, I.; Yasuda, A.; Vossmeyer, T. Lithographic patterning of nanoparticle films self-assembled from organic solutions by using a water-soluble mask. Appl. Phys. Lett. 2005, 86, 034108-1–0341083. (24) Zamborini, F. P.; Hicks, J. F.; Murray, R. W. Quantized Double Layer Charging of Nanoparticle Films Assembled Using Carboxylate/(Cu2+ or Zn2+)/ Carboxylate Bridges. J. Am. Chem. Soc. 2000, 122, 4514–4515. (25) Wessels, J. M.; Nothofer, H. G.; Ford, W. E.; von Wrochem, F.; Scholz, F.; Vossmeyer, T.; Schroedter, A.; Weller, H.; Yasuda, A. Optical and electrical properties of three-dimensional interlinked gold nanoparticle assemblies. J. Am. Chem. Soc. 2004, 126, 3349–3356. (26) Snow, A. W.; Ancona, M. G.; Kruppa, W.; Jernigan, G. G.; Foos, E. E.; Park, D. Self-assembly of gold nanoclusters on micro- and nanoelectronic substrates. J. Mater. Chem. 2002, 12, 1222–1230. (27) Trudeau, P. E.; Kwan, E.; Orozco, A.; Dhirani, A.-A. Competitive transport and percolation in disordered arrays of molecularly linked Au nanoparticles. J. Chem. Phys. 2002, 117, 3978–3981. (28) Suganuma, Y.; Dhirani, A. A. Gating of Enhanced ElectronCharging Thresholds in Self-Assembled Nanoparticle Films. J. Phys. Chem. B 2005, 109, 15391–15396. (29) Wuelfing, W. P.; Green, S. J.; Pietron, J. J.; Cliffel, D. E.; Murray, R. W. Electronic Conductivity of Solid-State, Mixed-Valent, MonolayerProtected Au Clusters. J. Am. Chem. Soc. 2000, 122, 11465–11472. (30) Nanoparticles - From Theory to Application, Wiley-VCH Verlag GmbH & Co KGaA: Weinheim, 2005. (31) Zabet-Khosousi, A.; Trudeau, P. E.; Suganuma, Y.; Dhirani, A. A. Metal to Insulator Transition in Films of Molecularly Linked Gold Nanoparticles. Phys. ReV. Lett. 2006, 96, 156403-1156403-4. (32) Briglin, S. M.; Gao, T.; Lewis, N. S. Detection of Organic Mercaptan Vapors Using Thin Films of Alkylamine-Passivated Gold Nanocrystals. Langmuir 2004, 20, 299–305. (33) Steinecker, W. H.; Rowe, M.; Matzger, A.; Zellers, E. T. Chemiresistor Array with Nanocluster Interfaces as a Micro-GC Detector. IEEE Transducers 2003, 3E44P, 1343–1346. (34) Joseph, Y.; Peic, A.; Chen, X.; Michl, J.; Vossmeyer, T.; Yasuda, A. Vapor Sensitivity of Networked Gold Nanoparticle Chemiresistors: Importance of Flexibility and Resistivity of the Interlinkage. J. Phys. Chem. C 2007, 111, 12855–12859.

12514 J. Phys. Chem. C, Vol. 112, No. 32, 2008 (35) Pang, P.; Guo, X.-L.; Wu, S.-H.; Cai, Q. Humidity effect on the dithiol-linked gold nanoparticles interfaced chemiresistor sensor for VOCs analysis. Sens. Actuators, B 2006, 114, 799–803. (36) Leff, D. V.; Brandt, L.; Heath, J. R. Synthesis and Characterization of Hydrophobic, Organically Soluble Gold Nanocrystals Functionalized with Primary Amines. Langmuir 1996, 12, 4723–4730. (37) Brust, M.; Walker, D.; Bethell, D.; Schiffrin, D. J.; Whyman, R. Synthesis of Thiol-Derivatised Gold Nanoparticles in a Two-Phase LiquidLiquid System. Chem. Commun. 1994, 7, 801–802. (38) Bethell, D.; Brust, M.; Schiffrin, D. J.; Kiely, C. From monolayers to nanostructured materials: an organic chemist’s view of self-assembly. J. Electroanal. Chem. 1996, 409, 137–143. (39) Dimensions of the electrode structure: 50 electrode finger pairs (n ) 50), 10 µm gap between the electrodes (g ) 10 µm), 1800 µm overlap of the electrode fingers (l ) 1800 µm). Conductivities were calculated according to σRT ) g/[(2-1)ldR], where d is the film thickness and R is the resistance. (40) Seyama, H.; Soma, M. Bonding-State Characterization of the Constituent Elements of Silicate Minerals by X-ray Photoelectron Spectroscopy. J. Chem. Soc., Faraday Trans.1 1985, 81, 485–495.

Joseph et al. (41) Bourg, M. C.; Badia, A.; Lennox, R. B. Gold-Sulfur Bonding in 2D and 3D Self-Assembled Monolayers: XPS Characterization. J. Phys. Chem. B 2000, 104, 6562–6567. (42) Maye, M. M.; Luo, J.; Lin, Y.; Engelhard, M. H.; Hepel, M.; Zhong, C. J. X-ray Photoelectron Spectroscopic Study of the Activation of Molecularly-Linked Gold Nanoparticle Catalysts. Langmuir 2003, 19, 125– 131. (43) Castner, D. G. X-ray Photoelectron Spectroscopy Sulfur 2p Study of Organic Thiol and Disulfide Binding Interactions with Gold Surfaces. Langmuir 1996, 12, 5083–5086. (44) Cavalleri, O.; Oliveri, O.; Dacca, A.; Parodi, R.; Rolandi, R. XPS measurements on L-cysteine and 1-oxtadecanethiol self-assembled films: a comparative study. Appl. Surf. Sci. 2001, 175 (176), 357– 362. (45) Wuelfing, W. P.; Murray, R. W. Electron Hopping through Films of Arenethiolate Monolayer-Protected Gold Clusters. J. Phys. Chem. B 2002, 106, 3139–3145. (46) Torma, V.; Schmid, G.; Simon, U. Structure-Property Relations in Au55 Cluster Layers Studied by Temperature-Dependent Impedance Measurements. ChemPhysChem 2001, 5, 321–325.

JP8013546