Letter pubs.acs.org/NanoLett
Highly Porous Metal Oxide Networks of Interconnected Nanotubes by Atomic Layer Deposition Fengbin Li, Xueping Yao, Zhaogen Wang, Weihong Xing, Wanqin Jin, Jun Huang, and Yong Wang* State Key Laboratory of Materials-Oriented Chemical Engineering, College of Chemistry and Chemical Engineering, Nanjing University of Technology, Nanjing 210009, Jiangsu, People's Republic of China S Supporting Information *
ABSTRACT: Mesoporous metal oxide networks composed of interconnected nanotubes with ultrathin tube walls down to 3 nm and high porosity up to 90% were fabricated by atomic layer deposition (ALD) of alumina or titania onto templates of swelling-induced porous block copolymers. The nanotube networks possessed dual sets of interconnected pores separated by the tube wall whose thickness could be finely tuned by altering ALD cycles. Because of the excellent pore interconnectivity and high porosity, the alumina nanotube networks showed superior humidity-sensing performances. KEYWORDS: Atomic layer deposition, block copolymers, mesoporous materials, metal oxides
P
ore geometry and size, size distribution, interconnectivity, as well as surface chemistry of the pore wall predominate properties and performances of a porous material.1 In many applications, for example, separation, catalysis, sensing, and drug delivery, it is required that pores are interconnected as a continuous phase homogeneously distributed throughout the material matrix for a fast transport and an adequate exchange of substances.2,3 Such continuous, porous morphologies can be constructed in polymeric, inorganic, and metallic scaffolds using different approaches4,5 and metal oxide porous materials are of particular interest because of their biocompatible, dielectric, or semiconducting properties and are seeking promising applications in diverse fields from biomedicines to photovoltaics.6,7 Metal oxide materials with a continuous nanopore system are usually prepared by a template-aided sol−gel process followed by a template removal step that is difficult to obtain high porosity,8 or by sintering packed particles (gaps between particles define the pore system) that leads to an ill-defined pore morphology and a wide pore size distribution.9 It was shown that atomic layer deposition (ALD) of metal oxides on fibrous nanocelluloses10 is a simple and effective strategy to prepare metal oxides with finely adjustable, interconnected pore morphologies. However, because these fibrous templates were randomly stacked the derived porous oxides need to be improved in terms of pore size distribution and geometry regularity of pores. Therefore, it still remains a challenge to fabricate metal oxide materials with continuous pore morphologies as well as high porosity at no expense of structural uniformity and robustness. Here, we report the fabrication of metal oxide networks of interconnected hollow nanotubes with high porosity and ultrathin pore walls using swelling-induced mesoporous block © 2012 American Chemical Society
copolymers (BCPs) as templates and ALD as the replicating technique. Mesoporous materials with ordered pore systems can be generated by converting the minority domains in phaseseparated BCPs to void spaces.11−13 BCP-derived porous materials are extensively employed as templates for fabricating of a large variety of functional materials.14−17 However, most pore systems in these porous materials are discrete spherical, cylindrical, or lamellar ones, leading to collapsed replicas via the template synthesis process. The only continuous system, the gyroidal morphology, appears in a very narrow window in the phase diagram,18 and frequently requires a delicate and tedious annealing process to develop. A nice 3D interconnected nanopore system was created in an organo/organometallic BCP with an asymmetric cylindrical morphology by selectively etching away the organic components.19 Recently, it was demonstrated that the minority domains in amphiphilic BCPs could be swollen by selective solvents, yielding a continuous nanopore system after solvent evaporation.20 As displayed in Scheme 1, the initially isolated cylinders of the minority blocks in a BCP film are converted to a continuous void phase through the swelling process (panel a). The porous BCP template is then subjected to ALD of metal oxides, yielding a conformal thin layer of oxides that is tightly adhered to the BCP framework (panel b). The BCP component is burned off and converted to void spaces with likewise continuous morphology, whereas the coating oxide layer survives and maintains the integrity of the film (panel c). Received: July 31, 2012 Published: August 13, 2012 5033
dx.doi.org/10.1021/nl3028312 | Nano Lett. 2012, 12, 5033−5038
Nano Letters
Letter
and the pore sizes were correspondingly reduced until the surface pores were completely sealed at higher cycles, for example, 250 cycles (Figure 1 and Supporting Information Figure S2). Figure 1e represents the cross section of the 10cycle-deposited template, revealing that the branched skeleton rods have a solid interior. In addition, the deposited alumina layer did not peel off of the skeleton rods, confirming that the alumina layer was tightly adhered to the template surface. This strong adhesion should be contributed to the strong interaction between the ALD precursors and the pyridyl groups migrated on the template surface during the swelling process.22−24 The thickness of the deposited alumina layer was obtained by comparing the diameter of the skeleton rods of deposited membranes with that of the initial, undeposited one. For the 10-cycle sample, the thickness was determined to be ∼1.2 nm. However, as we will demonstrate later, there is also a subsurface growth of alumina through the sequential infiltration synthesis (SIS) mechanism.22−24 Transmission electron microscopy (TEM) reveals the total thickness of alumina deposited on the surface and grown underneath the surface, which is 3 nm for the 10-cycle sample. Therefore, the thickness of the alumina layer grown underneath the template surface can be estimated as ∼1.8 nm. The thickness of this layer is expected to remain nearly constant with increasing ALD cycles because the subsurface growth vanishes at higher ALD cycles. We can correlate the thickness of the alumina layer with the number of ALD cycles. As shown in Figure 1f, the thickness of the alumina layer rapidly increases during the first 20 cycles and continues
Scheme 1. Schematic Illustration for the Preparation of Metal Oxide Networks of Interconnected Nanotubes by ALD on Swelling-Induced Mesoporous BCP Templates
Tetrahydrofuran solutions of 3 wt % polystyrene-blockpoly(2-vinyl pyridine) (Mn (PS) = 50 000 g/mol; Mn (P2VP) = 16 500 g/mol) were spin-coated onto Si substrates and then swollen in ethanol at 60 °C for 15 h, yielding mesoporous membranes.20 As shown in Figure 1a and Supporting Information Figure S1, we can see pores with a relatively narrow size distribution mainly scattering within the range of 25−35 nm on the membrane surface. The skeleton of the porous membrane is composed of branched rods, which are interconnected forming an uninterrupted network. These rods have a uniform width of 45 nm. The mesoporous BCP membranes were subjected to alumina ALD with changing cycles. A conformal thin layer of alumina was uniformly deposited on the pore walls. As shown in Figure 1b−d, all of the deposited BCP membrane surfaces remained smooth and were free of particulates, implying that the alumina layer was actually grown via the ALD mode.21 With increasing numbers of ALD cycles, the skeleton rods became increasingly wider,
Figure 1. Top view SEM images of the initial BCP template (a) and the alumina-coated BCP templates subjected to 10 cycles (b), 40 cycles (c), and 80 cycles (d) of alumina ALD; cross-sectional view of the BCP template subjected to 10 alumina ALD cycles (e) and plots (f) correlating the thickness of the deposited alumina layers (the lower plot) and the porosity of calcined samples (the upper plot) with numbers of ALD cycles. Panels a−d have the same magnification and scale bars both in (d) and (e) correspond to 200 nm. 5034
dx.doi.org/10.1021/nl3028312 | Nano Lett. 2012, 12, 5033−5038
Nano Letters
Letter
Figure 2. Top view (a,c,e) and cross-sectional view (b,d,f) SEM images of calcined alumina-coated BCP templates subjected to 10 cycles (a,b), 40 cycles (c,d), and 80 cycles (e,f) of alumina ALD. Panels a, c, and e and panels b, d, and f have the same magnification, respectively, and scale bars both in panels e and f correspond to 200 nm.
The calcined replicas gained another set of pores (the interior void spaces of the tubes) in addition to the existing pore system in the template. The size of the new pore system was determined by the diameter of the skeleton rods in the template that was fixed for samples with changing ALD cycles, whereas the dimension of the inherent pores in the template could be finely tuned by altering ALD cycles. Importantly, both pore systems were continuous and they were separated from each other by the tube wall. Although the tube interiors were not directly exposed to the atmosphere, they were still accessible by small molecules like water because of the permeable nature of the ultrathin tube walls. The permeability of the ultrathin tube walls was confirmed by the effective release of the degrading components, for example, H2O and CO2, of the BCP template during calcination10,16 and by the evidence that ALD precursors could pass through the alumina layer grown via SIS in the initial stage for further growth. Because of the dual pore systems and the ultrathin tube wall, the alumina nanotube networks were expected to have very high porosity. We could roughly estimate the porosity (Φ) of an alumina nanotube network with a thickness of tube wall (d), using the following equation
increasing thereafter but at a lower growth rate with a mean value of ∼1.6 Å/cycle, which is close to that of ZnO ALD on a P2VP suface.25 We calcined the alumina-coated membranes in air at 540 °C to remove the BCP templates. Figure 2 shows the surface and cross-sectional morphology of the calcined samples. Strikingly, for the membrane subjected to 10 ALD cycles that had an alumina layer as thin as 3 nm, the entire framework of the membrane was faithfully preserved and its surface morphology appeared very similar to that before calcination (Figure 1b vs Figure 2a). However, the cross-sectional SEM image showed that the embedded BCP template was completely removed. The initial solid skeleton rods were transformed to a network of hollow tubes (Figure 2b) whose size and arrangement were similar to that of the skeleton rods in the membrane before calcination (Figure 1e vs Figure 2b). The precise replication of the fine structural features in the template and the wellpreserved integrity of the template skeleton should be attributed to uniform and conformal deposition of alumina on the template and the strong adhesion between the coating layer and the template surface. For the other calcined samples that were subjected to various ALD cycles, their surface morphologies unexceptionally resembled that of their corresponding counterparts before calcination, whereas the crosssectional SEM images revealed that all of them were composed of interconnected hollow tubes (Figure 2). For the samples with varying ALD cycles, their constituent tubes had an almost constant inner diameter with an average value of ∼35 nm. In contrast, the thickness of the tube wall increased with the increase in ALD cycles. The thickness of the alumina networks reduced to ∼350−550 nm (Figure 2b,d,f) from initial thickness of ∼400−600 nm (Figure 1e) of the BCP templates.
2 ⎛ d⎞ Φ = 1 − Φ0 + ⎜1 − ⎟ Φ0 ⎝ r⎠
(1)
where Φ0 is the porosity of the original porous BCP template and r is the radius of the pore in the original BCP template, assuming that the pore is ideally cylindrical with a uniform pore diameter (r = ∼16 nm because the template has an average pore diameter of ∼32 nm as mentioned above). The BCP template used in this work had a porosity of approximately 5035
dx.doi.org/10.1021/nl3028312 | Nano Lett. 2012, 12, 5033−5038
Nano Letters
Letter
30%.20 Because the thickness of tube wall was equal to that of the deposited alumina layer that was already obtained (the lower plot, Figure 1f), we were able to calculate the porosity of calcined samples subjected to different ALD cycles. The 10cycle sample had a vey high porosity of ∼90% because of its ultrathin tube wall of 3 nm. With rising ALD cycles, the porosity decreased as a consequence of the thickening tubewalls (the upper plot, Figure 1h). However, there is a lower limit for Φ which is Φ0 (70%) because even the pores in the BCP template are completely filled by alumina, burning off the template will lead to a porosity of at least 70%. We tried to experimentally determine the porosity of the alumina networks using the nitrogen adsorption method but failed because this method is practically inapplicable to slightly weighted thin-film samples like the alumina networks here. Alternatively, we used thermogravimetric (TG) analysis to evaluate the high porosity of the alumina networks. TG analysis on the sample subjected to 10 ALD cycles showed that only ∼6.5% of the sample mass remained after combustion in oxygen up to 1200 °C (Supporting Information Figure S3). Therefore, the deposited alumina only accounted for a weight ratio of ∼6.5% in the sample. Considering that the density of alumina was nearly four times larger than the polymer, the volume fraction of the alumina in the composite should be even smaller, clearly indicating a porosity higher than 90% of the alumina network. Surprisingly, we found that even the nanotube film with the thinnest tube wall survived a harsh ultrasonication challenge (10 min at 100 W). No cracks or changes in morphology could be found on the nanotube film under SEM after ultrasonication. This strong mechanical robustness of the alumina nanotube films is believed to be related to the self-supporting nature of the interconnected network structures.26 Because porous metal oxide films with a micrometer-scale thickness are desired for many applications, we also attempted to prepare thicker alumina nanotube networks. In both the swelling and the ALD process diffusion of the swelling agent or the ALD precursors into thick media might not be sufficient, resulting in a distribution in the pore size of the BCP template and in the thickness of the deposited alumina layer, respectively. However, we successfully suppressed the diffusion effect by adopting a higher swelling temperature (70 °C) and a longer exposure time (20 s) for precursors and produced uniform alumina nanotube networks with a total thickness up to several tens of micrometers. Figure 3a exhibits the large-field view of the cross section of a nanotube network with a total thickness of 5 μm. Similar to the morphology of thin alumina nanotube films, the thick film was also composed of interconnected, branched nanotubes uniformly distributed throughout the entire thickness of the film. Insets in Figure 3a show the detailed view of the nanotube network at positions near the upper side, in the middle, and near the bottom, respectively. Though these examined positions varied remarkably in the distance to the bottom side, there was no obvious difference in the morphology, diameter, and wall thickness for nanotubes located at different positions. The micrometer-thick alumina film could be partially detached from the Si substrate using a sharp scalpel, allowing us to perform further characterizations on the unsupported alumina network. Figure 3b displays the TEM image of the 10-cycle alumina network, demonstrating once again its interconnected morphology composed of branched nanotubes with thin but uniform-inthickness tube walls. As shown in Figure 3c, magnified TEM imaging indicates that the wall thickness of the nanotube
Figure 3. The cross-sectional SEM image (a), TEM images with different magnifications (b,c) of a 10-cycle alumina nanotube network with a total thickness of ∼5 μm, and the photograph (d) of a thick 10cycle alumina nanotube network held by a pair of tweezers, demonstrating its mechanical robustness. Insets in (a) from top to bottom are magnified SEM images of positions near the upper surface, in the middle, and near the bottom of the cross section, respectively, showing the uniformity of the constituent nanotubes along the thickness of the alumina network. The scale bar in panel a corresponds to 1 μm and scale bars in panels b and c and insets in panel a correspond to 100 nm.
prepared from 10 ALD cycles is 3 nm, which is much larger than the value determined by SEM imaging (∼1.2 nm). This discrepancy suggests an infiltration mechanism of the alumina growth in the BCP in the initial stage because of the strong interactions of P2VP chains with both TMA and water.22−24 Moreover, the inner diameter of the nanotubes prepared with 10 cycles is measured to be 35 nm which agrees well with the SEM determination. Selected area electron diffraction (SAED) analysis (Supporting Information Figure S4) reveals that the alumina film calcinated at 540 °C is amorphous while it can be transformed to be crystalline by calcinating at elevated temperatures, for example, 1000 °C. We note that the 10cycle sample maintains its structure at a calcination temperature of 540 °C but collapses at 1000 °C (Supporting Information Figure S5). In contrast, samples with higher ALD cycles, for instance, 20 cycles, withstand calcination at 1000 °C and their fine structure is well-preserved, showing no discernible difference to the one calcinated at 540 °C (Supporting Information Figure S6). Interestingly, ALD on unsupported thicker BCP templates followed by calcination leads to stable, free-standing alumina nanotube films. As demonstrated in Figure 3d, the free-standing alumina film subjected to 10 ALD cycles is transparent and mechanical robust as it can be easily held and moved using a pair of tweezers without breaking it. Similarly, we are able to fabricate nanotube networks of other metal oxides, for example, titanium oxide by ALD onto swelling-induced porous BCP templates. The obtained titanium oxide replica possesses a similar morphology with dual 5036
dx.doi.org/10.1021/nl3028312 | Nano Lett. 2012, 12, 5033−5038
Nano Letters
Letter
values. The lower current for the alumina nanotube networks prepared at higher ALD cycles should be attributed to the lower porosity and thicker tube wall which slowed down the diffusion of water molecules. The increase in the effective surface area of the nanotubes with higher ALD cycles only played a minor role in affecting the sensing properties. Note that the interconnectivity of the tube wall framework also contributed significantly to the sensing properties since it assured the formation of a correspondingly continuous adsorbed water layer other than disconnected one leading to conduction failure. Figure 4b presents the change of current with time of the alumina nanotube network prepared at 10 ALD cycles when RH was switched between 30% and 5%. The film had a response time of 12 s when RH increased from 5% to 30% and a recovery time of 5 s when RH decreased from 30% to 5%. The fast recovery time from high RH to low RH suggested a quick and complete desorption of water molecules from the surface of alumina tubes. Compared to many other nanostructure-based sensors,29,30 the alumina nanotube networks showed typically 10times faster response time and recovery time, which should be attributed to the highly porous and interconnected nanotube network structures. Moreover, almost identical profiles of the current change with time during the three cycles indicated the excellent reproducibility and robustness of the alumina film sensors. Furthermore, we stored the 10-cycle alumina nanotube sensor in ambient conditions for 3 months and rechecked its sensing performances. We found it worked as well as the fresh one, indicating its excellent long-term stability and reversibility. In stark contrast, we observed none or very weak humiditysensing properties of four different control samples (inset in Figure 4a): (i) the pristine, uncoated porous BCP templates, (ii) the BCP membrane subjected to 10 cycles of alumina deposition before removal of the template, and the smooth alumina film that was ALD-deposited directly onto the substrate with a thickness of ∼200 nm (iii) before and (iv) after calcination at 540 °C. Therefore, we can conclude that the interconnected pore system with high porosity is essential for the superior humidity sensing properties of the alumina nanotube networks. In conclusion, we prepared highly porous networks of interconnected metal oxide nanotubes by atomic layer deposition of metal oxides (alumina and titania) onto swelling-induced mesoporous block copolymer templates. The alumina nanotube networks were featured as dual sets of continuous pore systems with high porosity (∼70−90%) and ultrathin tube walls down to 3 nm. The alumina nanotube networks exhibited significantly improved humidity-sensing performances. We believe such nanotube networks will also find important applications in the fields of catalysis, separation, photovoltaics, and so forth due to their continuous pore system and high porosity.
continuous pore systems (Supporting Information Figure S7), indicating the generality of this strategy. As a demonstration of applications of metal oxides with well-defined, interconnected porosity, we explored the humidity-sensing properties of the produced alumina nanotube networks. Alumina itself is electrical nonconductive. However, it has a strong capability to adsorb water from the ambience. Absorbed water condenses on the surface of alumina materials and protons will be conducted on the formed water layer. Higher adsorption of water leads to increase in proton conduction and decrease in impedance, which means an increase in current flow under a constant voltage.27 We applied a voltage of 2 V to alumina nanotube networks with a total thickness of ∼500 nm and measured electric currents across the alumina films exposed to different relative humidity (RH). As shown in Figure 4a, there
Figure 4. (a) The electric current (I)−RH curves of alumina nanotube networks prepared at 10 cycles (■), 20 cycles (●), 40 cycles (▲), 60 cycles (▼) of alumina ALD and of four different control samples: (i) the pristine, uncoated porous BCP templates, (ii) the BCP membrane subjected to 10 cycles of alumina deposition before removal of the template, and the smooth alumina film that was ALD-deposited directly onto the substrate with a thickness of ∼200 nm (iii) before and (iv) after calcination at 540 °C. Inset in panel a shows the magnified I−RH curves of the four control samples. (b) The current− time curve of the alumina nanotube network prepared at 10 ALD cycles exposing to RH switching between 5 and 30% from which the response time and the recovery time were determined.
■
ASSOCIATED CONTENT
S Supporting Information *
was a remarkable current increase up to 4 orders of magnitude with RH rising from 5 to 90% for all the tested alumina nanotube networks. The alumina nanotube network prepared at 10 ALD cycles showed significantly higher current than others in the RH range of 30−90% because it had a larger porosity, facilitating faster diffusion and consequently enhanced adsorption of more moisture onto the alumina tube surface.28 Furthermore, it was observed that alumina nanotube networks prepared at higher ALD cycles gave lower current at same RH
Experimental details, TG, SAED, and additional SEM images are included. This material is available free of charge via the Internet at http://pubs.acs.org.
■
AUTHOR INFORMATION
Corresponding Author
*Fax: 0086-25-8317 2292. E-mail:
[email protected]. 5037
dx.doi.org/10.1021/nl3028312 | Nano Lett. 2012, 12, 5033−5038
Nano Letters
Letter
Notes
(29) Kuang, Q.; Lao, C.; Wang, Z. L.; Xie, Z.; Zheng, L. J. Am. Chem. Soc. 2007, 129, 6070−6071. (30) Yao, L.; Zheng, M.; Li, H.; Ma, L.; Shen, W. Nanotechnology 2009, 20, 395501−395507.
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS This work is supported by the National Basic Research Program of China (2011CB612302), the Natural Science Foundation of China (21004033, 21176120, and 21125629), the Jiangsu Natural Science Fund for Distinguished Young Scholars (BK2012039), the Fok Ying Dong Education Foundation (131046), the Open Research Fund Program of the State Key Laboratory of Polymer Physics and Chemistry (Changchun Institute of Applied Chemistry), and the Project of Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD).
■
REFERENCES
(1) Tokarev, I.; Minko, S. Adv. Mater. 2010, 22, 3446−3462. (2) McKeown, N. B.; Budd, P. M. Chem. Soc. Rev. 2006, 35, 675− 683. (3) Jones, B. H.; Lodge, T. P. ACS Nano 2011, 5, 8914−8927. (4) Erlebacher, J.; Aziz, M. J.; Karma, A.; Dimitrov, N.; Sieradzki, K. Nature 2001, 410, 450−453. (5) Enke, D.; Janowski, F.; Schwieger, W. Microporous Mesoporous Mater. 2003, 60, 19−30. (6) Vallet-Regi, M.; Balas, F.; Arcos, D. Angew. Chem., Int. Ed. 2007, 46, 7548−7558. (7) Crossland, E. J. W.; Kamperman, M.; Nedelcu, M.; Ducati, C.; Wiesner, U.; Smilgies, D. M.; Toombes, G. E. S.; Hillmyer, M. A.; Ludwigs, S.; Steiner, U.; Snaith, H. J. Nano Lett. 2009, 9, 2807−2812. (8) McGrath, K. M.; Dabbs, D. M.; Yao, N.; Aksay, I. A.; Gruner, S. M. Science 1997, 277, 552−556. (9) Olevsky, E. Mater. Sci. Eng. R. 1998, 23, 41−100. (10) Korhonen, J. T.; Hiekkataipale, P.; Malm, J.; Karppinen, M.; Ikkala, O.; Ras, R. H. A. ACS Nano 2011, 5, 1967−1974. (11) Kim, S. W.; Han, T. H.; Kim, J.; Gwon, H.; Moon, H. S.; Kang, S. W.; Kim, S. O.; Kang, K. ACS Nano 2009, 3, 1085−1090. (12) Hillmyer, M. A. Adv. Polym. Sci. 2005, 190, 137−181. (13) Olson, D. A.; Chen, L.; Hillmyer, M. A. Chem. Mater. 2008, 20, 869−890. (14) Wang, Y.; Li, F. Adv. Mater. 2011, 23, 2134−2148. (15) Thurn-Albrecht, T.; Schotter, J.; Kastle, C. A.; Emley, N.; Shibauchi, T.; Krusin-Elbaum, L.; Guarini, K.; Black, C. T.; Tuominen, M. T.; Russell, T. P. Science 2000, 290, 2126−2129. (16) Ras, R. H. A.; Kemell, M.; Wit, J.; de; Ritala, M.; Brinke, G.; ten; Leskelä, M.; Ikkala, O. Adv. Mater. 2007, 19, 102−106. (17) Son, J. G.; Chang, J. B.; Berggren, K. K.; Ross, C. A. Nano Lett. 2011, 11, 5079−5084. (18) Tseng, W. H.; Chen, C. K.; Chiang, Y. W.; Ho, R. M.; Akasaka, S.; Hasegawa, H. J. Am. Chem. Soc. 2009, 131, 1356−1357. (19) Ramanathan, M.; Strzalka, J.; Wang, J.; Darling, S. B. Polymer 2010, 51, 4663−4666. (20) Wang, Y.; He, C.; Xing, W.; Li, F.; Tong, L.; Chen, Z.; Liao, X.; Steinhart, M. Adv. Mater. 2010, 22, 2068−2072. (21) George, S. M. Chem. Rev. 2010, 110, 111−131. (22) Peng, Q.; Tseng, Y. C.; Darling, S. B.; Elam, J. W. Adv. Mater. 2010, 22, 5129−5133. (23) Peng, Q.; Tseng, Y. C.; Darling, S. B.; Elam, J. W. ACS Nano 2011, 5, 4600−4606. (24) Tseng, Y. C.; Peng, Q.; Ocola, L. E.; Elam, J. WM.; Darling, S. B. J. Phys. Chem. C 2011, 115, 17725−17729. (25) Wang, Y.; Qin, Y.; Berger, A.; Yau, E.; He, C. C.; Zhang, L. B.; Goesele, U.; Knez, M.; Steinhart, M. Adv. Mater. 2009, 21, 2763− 2766. (26) Yang, S. Y.; Park, J.; Yoon, J.; Ree, M.; Jang, S. K.; Kim, J. K. Adv. Funct. Mater. 2008, 18, 1371−1377. (27) Chen, Z.; Lu, C. Sens. Lett. 2005, 3, 274−295. (28) Wang, W.; Virkar, A. V. Sens. Actuators B. 2004, 98, 282−290. 5038
dx.doi.org/10.1021/nl3028312 | Nano Lett. 2012, 12, 5033−5038
