Flexible Gigahertz Transistors Derived from Solution-Based Single

Letter pubs.acs.org/NanoLett

Flexible Gigahertz Transistors Derived from Solution-Based SingleLayer Graphene Cédric Sire,† Florence Ardiaca,† Sylvie Lepilliet,‡ Jung-Woo T. Seo,§ Mark C. Hersam,§ Gilles Dambrine,‡ Henri Happy,‡ and Vincent Derycke*,† †

CEA Saclay, IRAMIS, Service de Physique de l’Etat Condensé (URA 2464), Laboratoire d’Electronique Moléculaire, F-91191 Gif sur Yvette, France ‡ Institut d’Electronique, de Microélectronique et de Nanotechnologie, UMR-CNRS 8520, BP 60069, Avenue Poincaré, F-59652 Villeneuve d’Ascq, France § Department of Materials Science and Engineering and Department of Chemistry, Northwestern University, Evanston, Illinois 60208-3108, United States ABSTRACT: Flexible electronics mostly relies on organic semiconductors but the limited carrier velocity in polymers and molecular films prevents their use at frequencies above a few megahertz. Conversely, the high potential of graphene for highfrequency electronics on rigid substrates was recently demonstrated. We conducted the first study of solution-based graphene transistors at gigahertz frequencies, and we show that solution-based single-layer graphene ideally combines the required properties to achieve high speed flexible electronics on plastic substrates. Our graphene flexible transistors have current gain cutoff frequencies of 2.2 GHz and power gain cutoff frequencies of 550 MHz. Radio frequency measurements directly performed on bent samples show remarkable mechanical stability of these devices and demonstrate the advantages of solution-based graphene field-effect transistors over other types of flexible transistors based on organic materials. KEYWORDS: Graphene, single-layer, solution-based, transistor, high-frequency, flexible electronics

T

he potential of graphene transistors1,2 for high-frequency electronics was recently demonstrated by several groups using exfoliated graphene,3−5 SiC-based graphene,6,7 and CVDbased graphene.8 The most recent studies reached deembedded current gain cutoff frequencies (f T) in the 100− 300 GHz range with room for improvement at both the material and device levels. Yet, field-effect transistors (FETs) based on III−V semiconductors have already reached f T above 640 GHz9 while silicon-based MOSFETs have reached 485 GHz.10 It is thus not yet fully established whether or not graphene will effectively compete with conventional crystalline semiconductors when approaching the terahertz range. In parallel, graphene is being explored for large scale electronics on flexible substrates via chemical vapor deposition (CVD) growth of graphene on metal foils associated with transfer methods.11 This progress is driven by the perspective of replacing ITO (indium tin oxide) as the material of choice for the transparent electrodes required in applications such as touch screens, flat panel displays or organic photovoltaic cells. However, the combination of these two properties, namely high speed and flexibility, remains an open challenge. In particular for the viable development of fast and flexible electronic applications in the areas of portable/wearable communicating devices with low power consumption, this combination should be achieved with a source of material adapted to low-cost manufacturing methods such as inkjet printing. © 2012 American Chemical Society

Printed electronics based on organic materials is a wellestablished field.12,13 Organic materials are particularly well suited for flexible circuits due to their mechanical resiliency.14 Yet, their low charge mobility limits their ultimate operating frequency. While several examples of organic devices and circuits operating in the kilohertz to megahertz range have been demonstrated on both rigid15,16 and flexible substrates,17−20 these approaches fall well short of the gigahertz range. Conversely, inorganic semiconducting materials such as III−V semiconductor nanowires and silicon thin films can reach the gigahertz range,21−24 but few studies have evaluated their performances upon severe bending and these inorganic materials are not ideally adapted for future printed technologies. Carbon-based nanomaterials potentially combine high-speed performance with the required mechanical properties. In particular, carbon nanotubes were used to develop highfrequency transistors on rigid substrates25−27 and to demonstrate flexible devices and circuits.28−32 They can be handled in the form of inks compatible with printed electronics.33 Recently, graphene transistors on flexible substrates were Received: September 23, 2011 Revised: January 11, 2012 Published: January 27, 2012 1184

dx.doi.org/10.1021/nl203316r | Nano Lett. 2012, 12, 1184−1188

Nano Letters

Letter

realized but their high-frequency performance was not evaluated.34,35 Here, we demonstrate that solution-based single-layer graphene ideally combines the required properties and presents important advantages over alternative graphene sources that are chemically grown or mechanically exfoliated. Several methods of producing stable graphene-based suspensions have been recently demonstrated.36 In particular, there have been extensive efforts to utilize graphene oxide as a solution-phase precursor for graphene in various applications.37 However, this approach requires subsequent chemical reduction treatments that preclude complete recovery of the superior electrical properties of pristine graphene. Alternatively, exfoliation of graphite in organic solvents38 or with polymeric surfactants in aqueous39 and nonaqueous40 solutions is also possible. While this procedure is effective at isolating few-layer pristine graphene, polydispersity in the thickness of graphene produced from these processes implies inferior performance compared to single-layer graphene in high-performance electronic applications. To overcome these issues, we employ solution-based, predominantly single-layer graphene flakes isolated via density gradient ultracentrifugation (DGU)41 to fabricate flexible transistors on organic substrates operating at gigahertz frequencies. The devices operate at low bias (VDS < 0.7 V), achieve current gain cutoff frequencies f T as high as 2.2 GHz before de-embedding (8.7 GHz after de-embedding), power gain cutoff frequency f MAX of 550 MHz, and a constant transconductance in the gigahertz range. In addition, we show that both the electron and hole conduction branches display high-speed performance, which is in contrast with previous reports where only one type of carrier was considered due to the either high n-type7,8 or p-type5 doping of the graphene used. Graphene suspensions were prepared by ultrasonication of graphite (natural graphite flakes, grade 3061, Asbury Graphite Mills) in an aqueous solution of 2% w/v sodium cholate. During DGU, the graphene flakes are separated by buoyant density, which implies separation by flake thickness.39 In particular, following DGU, predominantly single-layer graphene flakes are isolated with an average thickness of 1.1 nm and area of 16 000 nm2 as determined by atomic force microscopy.41 Following DGU, the density gradient medium was removed from the nearly monodisperse graphene dispersions via dialysis. Transistors in a top-gate geometry and adapted to radio frequency measurements (using a Ground-Signal-Ground (GSG) pad configuration) were fabricated on polyimide foils with a source-drain distance of ∼260 nm, a gate length of ∼170 nm, and a gate width of 40 μm (Figure 1). We used dielectrophoresis (DEP) as the graphene deposition method.42 DEP was performed at a deposition frequency of 1 MHz and a bias of 10 Vpp applied between the two ground electrodes of the GSG structure separated by 5 μm (before the realization of the final source drain electrodes). We note that DEP is not compatible with future printed electronics. It was used as a convenient way to evaluate the performance of the studied material. It could be replaced by other deposition techniques adapted to the ink form of the graphene flake solution used in this study. As gate dielectrics, we used 20 nm of yttrium oxide (εr ∼12−14) due to its known quality for graphene electronics.43 It was grown by four successive depositions of 3 nm thick yttrium films followed by exposure to pure oxygen for 10 min at room temperature. The DC transfer characteristics (ID(VGS)) (Figure 1d) of a typical as-prepared device displays the conventional V-shape

Figure 1. (a) Schematic cross-sectional and top views of the top-gate graphene transistor in GSG configuration. (b) AFM image of the graphene flakes deposited by DEP acquired before the deposition of the source, drain and, top-gate electrodes. (c) Graphene devices fabricated on a polyimide substrate. (d) Transfer characteristics ID(VGS) of an as-prepared (i.e., before annealing) device.

with a Dirac point at low VGS typical of undoped graphene and a slightly sharper modulation for holes than electrons. The conductance per unit width is limited by (i) the fact that not the whole channel is covered with graphene flakes (Figure 1b), (ii) the potential presence of current pathways through multiple flakes having higher resistance, and (iii) the limited overlap between the source−drain electrodes and the graphene flakes. These three points will improve in the future with the increase of the average flake size. The transconductance is limited by the presence of nongated channel sections, which are important to limit parasitic capacitances at high frequency. However, the source−drain resistance level in the 100−200 Ω range allows the direct measurement of the 2-ports Sparameters in the 50 MHz to 20 GHz range using a Vectorial Network Analyzer. From the S-parameters, we obtained all the useful radio frequency (RF) metrics with high accuracy. In particular, from the measurement of the current gain (H21) at 510 MHz as a function of VGS (Figure 2a), we deduced the biasing conditions that yield the highest RF transconductance (gm = dID/dVGS) and current gain for holes (VGS ∼ −1 V) and electrons (VGS ∼ 1 V). We then acquired gm and H21 as a function of frequency at these VGS values. Figure 2b shows that gm is almost constant up to at least 5 GHz for electrons and holes and, as expected, very close to zero around the Dirac point. The as-measured cutoff frequencies for holes and electrons at VDS = −0.6 V are 1.3 GHz and 430 MHz, respectively. In addition, we see in Figure 2c that at the Dirac 1185

dx.doi.org/10.1021/nl203316r | Nano Lett. 2012, 12, 1184−1188

Nano Letters

Letter

Figure 3. Postannealing high-frequency performance: Evolution of the current gain (H21) and power gain (U) as a function of frequency measured at VGS = −0.6 V and VDS = −0.65 V (ID = −16.8 mA) for the same device after current annealing. H21(ext) is the as-measured (extrinsic) value, H21(int) is the post de-embedding (intrinsic) value. The cutoff frequencies are f T(ext) = 2.2 GHz, f T(int) = 8.7 GHz, and f MAX= 550 MHz. The dotted-line corresponds to a slope of −20 dB/ dec.

ensure strictly the same source−gate and drain−gate distance between the device under test and the “open structure”, which impacts the accuracy of the intrinsic f T extraction and leads to values from 6 to 15 GHz. From the de-embedding procedure, we also estimated the parasitic (Cpgs + Cpgs ∼ 40 fF) and intrinsic gate capacitances (Cg ∼ 11 fF). Using Cg and the RFtransconductance, we deduced a field-effect mobility of ∼102 ± 19 cm2 V−1 s−1 for holes. This value is low when compared with graphene not-handled in solution but very high in comparison with organic semiconductors. Indeed, most organic semiconductors have mobilities below 1 cm2 V−1 s−1 except for some crystalline organic materials that are not ideally suited for flexible electronics. Our device mobility is also comparable to the highest reported mobility for carbon nanotube inks (∼90 cm2 V−1 s−1).47 Generally speaking, f T is a figure of merit quite sensitive to parasitics, which makes comparisons of intrinsic f T from different studies questionable. Therefore, the power gains such as the Mason Gain (U) is a more significant figure of merit for analog RF applications, since its values remains constant with lossless impedance transforms. From the S-parameters, we also deduced the power gain U (Figure 3). From the latter, we extracted the power gain cutoff frequency f MAX (550 MHz) at which U is equal to 0 dB. This power gain, together with a high f T, makes graphene flexible transistors superior to their nanotube analogs29 and orders of magnitude faster than any other type of organic solution-processable transistors.13−20 To assess the mechanical stability of our devices, we also measured the RF performances of graphene transistors at different bending angles (Figure 4). For this purpose, we fabricated devices with the macroscopic contact pads of the GSG structure oriented so that the bending takes place along the channel direction. The devices were rigidly clamped on a round-shaped sample holder to allow improved stability of the RF probes during the measurements (see picture in Figure 4d). Figure 4c displays the cutoff frequency of such a transistor in five different configurations: flat (before bending), R = 71.5 mm, R = 25 mm, R = 12.5 mm and flat (after bending). From these characteristics, we extracted the maximum f T and reported the values in Figure 4d. The data show that up to R

Figure 2. Holes and electrons high frequency transport properties at VDS = −0.6 V. (a) Current gain (H21) measured at 510 MHz as a function of the gate bias. (b) Evolution of the transconductance as a function of frequency measured at VGS = −1 V (ID = −9.68 mA) for holes (red) and VGS = 1 V (ID = −8.39 mA) for electrons (blue) and at the Dirac point (black). (c) Evolution of the current gain (H21) as a function of frequency and associated cutoff frequencies f T for holes and electrons (same VGS as in b).

point H21(dB) is always negative, which is in agreement with the very low transconductance at this point. It is known that driving significant current in a graphene device can improve its conductivity.44,45 We used this Joule annealing technique by driving high currents in our devices (this step could be replaced by a global vacuum annealing at 350 °C compatible with the polyimide substrate). We consistently observed an increase in performance (higher ID and gm) accompanied by a p-type doping effect bringing the Dirac point toward more positive VGS values. After such annealing, the same device as in Figure 2 reached an f T of 2.2 GHz (Figure 3) for holes at VGS = −0.6 V and VDS = −0.65 V. It is important to note that 2.2 GHz is the as-measured value. It thus compares very well with the 2.4 GHz value reported in ref 5 before de-embedding for devices built from exfoliated graphene deposited on silicon. A de-embedding procedure consists of extracting from the measured data the intrinsic performances that the device could ideally reached if the parasitic contributions from the measurement geometry could be minimized. While de-embedding can provide insight into the ultimate performance of a high-frequency device, it is very sensitive to assumptions and process variability and thus needs to be carefully considered. In the present work, we prepared socalled open structures similar to the device structures but without graphene in the active channel, and used them to evaluate the parasitic contributions. By using a similar procedure as in ref 46 (namely the subtraction of the Yparameters of the open structure from the Y-parameters of the tested device), we obtained an estimate of the intrinsic cutoff frequency: f T ∼ 8.7 GHz. We tested several open structures and noted the sensitivity of the de-embedding procedure to the open structure. Indeed, on organic substrates it is difficult to 1186

dx.doi.org/10.1021/nl203316r | Nano Lett. 2012, 12, 1184−1188

Nano Letters

Letter

Figure 4. (a,b) Transfer and output characteristics of the device used for the RF measurements upon bending. (c) As-measured (extrinsic) current gain cutoff frequency as a function of gate bias for the sample in the flat configuration (before and after bending) and for three different bending angles. (d) Evolution of the maximum extrinsic f T extracted from figure (c) and picture of the device under test at a bending angle of 25 mm. (5) Liao, L.; Lin, Y.-C.; Bao, M.; Cheng, R.; Bai, J.; Liu, Y.; Qu, Y.; Wang, K. L.; Huang, Y.; Duan, X. Nature 2010, 467 (7313), 305−308. (6) Moon, J. S.; Curtis, D.; Hu, M.; Wong, D.; McGuire, C.; Campbell, P. M.; Jernigan, G.; Tedesco, J. L.; VanMil, B.; Myers-Ward, R.; Eddy, C. Jr.; Gaskill, D. K IEEE Electron Device Lett. 2009, 30 (6), 650−652. (7) Lin, Y. M.; Dimitrakopoulos, C.; Jenkins, K. A.; Farmer, D. B.; Chiu, H. Y.; Grill, A.; Avouris, P. Science 2010, 327 (5966), 662−662. (8) Wu, Y.; Lin, Y.-m.; Bol, A. A.; Jenkins, K. A.; Xia, F.; Farmer, D. B.; Zhu, Y.; Avouris, P. Nature 2011, 472 (7341), 74−78. (9) Kim, D. H.; del Alamo, J. A. IEEE Electron Device Lett. 2010, 31 (8), 806−808. (10) Lee, S.; Jagannathan, B.; Narasimha, S.; Chou, A.; Zamdmer, N.; Johnson, J.; Williams, R.; Wagner, L.; Kim, J.; Plouchart, J. O.; Pekarik, J.; Springer, S.; Freeman, G. Record RF performance of 45 nm SOI CMOS technology. IEDM Tech. Dig. 2007, 255−258. (11) Bae, S.; Kim, H.; Lee, Y.; Xu, X.; Park, J.-S.; Zheng, Y.; Balakrishnan, J.; Lei, T.; Kim, H. R.; Song, Y. I.; Kim, Y.-J.; Kim, K. S.; Ozyilmaz, B.; Ahn, J.-H.; Hong, B. H.; Iijima, S. Nat. Nanotechnol. 2010, 5 (8), 574−578. (12) Sirringhaus, H.; Kawase, T.; Friend, R. H.; Shimoda, T.; Inbasekaran, M.; Wu, W.; Woo, E. P. Science 2000, 290 (5499), 2123− 2126. (13) Noh, Y.-Y.; Zhao, N.; Caironi, M.; Sirringhaus, H. Nat. Nanotechnol. 2007, 2 (12), 784−789. (14) Sekitani, T.; Zschieschang, U.; Klauk, H.; Someya, T. Nat. Mater. 2010, 9 (12), 1015−1022. (15) Baude, P. F.; Ender, D. A.; Haase, M. A.; Kelley, T. W.; Muyres, D. V.; Theiss, S. D. Appl. Phys. Lett. 2003, 82 (22), 3964−3966. (16) Wagner, V.; Woebkenberg, P.; Hoppe, A.; Seekamp, J. Appl. Phys. Lett. 2006, 89 (24), 243515. (17) Cantatore, E.; Geuns, T. C. T.; Gelinck, G. H.; van Veenendaal, E.; Gruijthuijsen, A. F. A.; Schrijnemakers, L.; Drews, S.; de Leeuw, D. M. IEEE J. Solid-State. Circuits 2007, 42 (1), 84−92. (18) Marien, H.; Steyaert, M. S. J.; van Veenendaal, E.; Heremans, P. IEEE J. Solid-State. Circuits 2011, 46 (1), 276−284. (19) Myny, K.; Steudel, S.; Vicca, P.; Genoe, J.; Heremans, P. Appl. Phys. Lett. 2008, 93 (9), 093305. (20) Rotzoll, R.; Mohapatra, S.; Olariu, V.; Wenz, R.; Grigas, M.; Dimmler, K.; Shchekin, O.; Dodabalapur, A. Appl. Phys. Lett. 2006, 88 (12), 123502. (21) Ahn, J.-h.; Kim, H.-s.; Lee, K. J.; Zhu, Z.; Menard, E.; Nuzzo, R. G.; Rogers, J. A. IEEE Electron Device Lett. 2006, 27 (6), 460−462. (22) Sun, Y.; Menard, E.; Rogers, J. A.; Kim, H.-S.; Kim, S.; Chen, G.; Adesida, I.; Dettmer, R.; Cortez, R.; Tewksbury, A. Appl. Phys. Lett. 2006, 88 (18), 183509−183509.

= 25 mm the curvature has essentially no impact on f T. At higher bending angle, we observed a slight shift of the gate bias at which f T is maximized (∼+0.1 V) and a ∼6.6% decrease of f T. This moderate loss of performance is irreversible. We attribute it principally to poorly clamped graphene flakes that become disconnected from the source or drain electrodes upon bending and to an increase in the resistance of current pathways involving multiple flakes. We anticipate that increasing the average size of graphene flakes will be an effective future strategy to further increase both the RF performance and mechanical stability. In conclusion, we have developed fast and flexible FETs from solution-based graphene that show low resistance, extrinsic and intrinsic current gain cutoff frequency of 2.2 and 8.7 GHz, respectively, and maximum frequency of oscillation of 550 MHz. Importantly, these performances are achieved at very low bias (VGS = −0.6 V and VDS = −0.65 V) and both electrons and holes have excellent performance up to high frequencies. The combination of electrical performance, mechanical stability, and compatibility with large area/low cost fabrication methods places solution-based graphene at the top of the short list of materials that are suitable for flexible RF electronics.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS M.C.H. acknowledges support from the National Science Foundation (DMR-1006391 and DMR-1121262) and the Nanoelectronics Research Initiative. We thank S. Fregonese, S. Saada, and T. David for expert assistance.

■

REFERENCES

(1) Avouris, P. Nano Lett. 2010, 10 (11), 4285−4294. (2) Schwierz, F. Nat. Nanotechnol. 2010, 5 (7), 487−496. (3) Meric, I.; Baklitskaya, N.; Kim, P.; Shepard, K. L. IEDM Tech. Dig. 2008, DOI: 10.1109/IEDM.2008.4796738. (4) Lin, Y. M.; Jenkins, K. A.; Valdes-Garcia, A.; Small, J. P.; Farmer, D. B.; Avouris, P. Nano Lett. 2009, 9 (1), 422−426. 1187

dx.doi.org/10.1021/nl203316r | Nano Lett. 2012, 12, 1184−1188

Nano Letters

Letter

(23) Yuan, H.-C.; Ma, Z. Appl. Phys. Lett. 2006, 89 (21), 212105− 212105. (24) Takahashi, T.; Takei, K.; Adabi, E.; Fan, Z.; Niknejad, A. M.; Javey, A. ACS Nano 2010, 4 (10), 5855−5860. (25) Kocabas, C.; Dunham, S.; Cao, Q.; Cimino, K.; Ho, X.; Kim, H.S.; Dawson, D.; Payne, J.; Stuenkel, M.; Zhang, H.; Banks, T.; Feng, M.; Rotkin, S. V.; Rogers, J. a. Nano Lett. 2009, 9 (5), 1937−1943. (26) Nougaret, L.; Happy, H.; Dambrine, G.; Derycke, V.; Bourgoin, J. P.; Green, A. A.; Hersam, M. C. Appl. Phys. Lett. 2009, 94 (24), 243505. (27) Rutherglen, C.; Jain, D.; Burke, P. Nat. Nanotechnol. 2009, 4 (12), 811−819. (28) Bradley, K.; Gabriel, J.-C. P.; Grüner, G. Nano Lett. 2003, 3, 1353−1355. (29) Chimot, N.; Derycke, V.; Goffman, M. F.; Bourgoin, J. P.; Happy, H.; Dambrine, G. Appl. Phys. Lett. 2007, 91, 15. (30) Cao, Q.; Kim, H.-s.; Pimparkar, N.; Kulkarni, J. P.; Wang, C.; Shim, M.; Roy, K.; Alam, M.; Rogers, J. Nature 2008, 454 (7203), 495−500. (31) Ha, M.; Xia, Y.; Green, A. A.; Zhang, W.; Renn, M. J.; Kim, C. H.; Hersam, M. C.; Frisbie, C. D. ACS Nano 2010, 4 (8), 4388−4395. (32) Sun, D.-M.; Timmermans, M. Y.; Tian, Y.; Nasibulin, A. G.; Kauppinen, E. I.; Kishimoto, S.; Mizutani, T.; Ohno, Y. Nat. Nanotechnol. 2011, 6 (3), 156−161. (33) Rouhi, N.; Jain, D.; Burke, P. J. ACS Nano 2011, 5 (11), 8471. (34) Kim, B. J.; Jang, H.; Lee, S.-K.; Hong, B. H.; Ahn, J.-H.; Cho, J. H. Nano Lett. 2010, 10 (9), 3464−3466. (35) Lee, S-K; Kim, B. J.; Jang, H.; Yoon, S. C.; Lee, C.; Hong, B. H.; Rogers, J. A.; Cho, J. H.; Ahn, J.-H. Nano Lett. 2011, 11 (11), 4642− 4646. (36) Green, A. A.; Hersam, M. C. J. Phys. Chem. Lett. 2010, 1 (2), 544−549. (37) Park, S.; Ruoff, R. S. Nat. Nanotechnol. 2009, 4 (4), 217−224. (38) Hernandez, Y.; Nicolosi, V.; Lotya, M.; Blighe, F. M.; Sun, Z.; De, S.; McGovern, I. T.; Holland, B.; Byrne, M.; Gun’ko, Y. K.; Boland, J. J.; Niraj, P.; Duesberg, G.; Krishnamurthy, S.; Goodhue, R.; Hutchison, J.; Scardaci, V.; Ferrari, A. C.; Coleman, J. N. Nat. Nanotechnol. 2008, 3 (9), 563−568. (39) Seo, J.-W. T.; Green, A. A.; Antaris, A. L.; Hersam, M. C. J. Phys. Chem. Lett. 2011, 2 (9), 1004−1008. (40) Liang, Y. T.; Hersam, M. C. J. Am. Chem. Soc. 2010, 132 (50), 17661−17663. (41) Green, A. A.; Hersam, M. C. Nano Lett. 2009, 9 (12), 4031− 4036. (42) Vijayaraghavan, A.; Sciascia, C.; Dehm, S.; Lombardo, A.; Bonetti, A.; Ferrari, A. C.; Krupke, R. ACS Nano 2009, 3 (7), 1729− 1734. (43) Wang, Z.; Xu, H.; Zhang, Z.; Wang, S.; Ding, L.; Zeng, Q.; Yang, L.; Pei, T.; Liang, X.; Gao, M.; Peng, L.-M. Nano Lett. 2010, 10 (6), 2024−2030. (44) Bolotin, K. I.; Sikes, K. J.; Hone, J.; Stormer, H. L.; Kim, P. Phys. Rev. Lett. 2008, 101 (9), 096802. (45) Moser, J.; Barreiro, A.; Bachtold, A. Appl. Phys. Lett. 2007, 91 (16), 163513. (46) Bethoux, J. M.; Happy, H.; Dambrine, G.; Derycke, V.; Goffman, M.; Bourgoin, J. P. IEEE Electron Device Lett. 2006, 27 (8), 681−683. (47) Rouhi, N.; Jain, D.; Zand, K.; Burke, P. J. Adv. Mater. 2011, 23 (1), 94−99.

1188

dx.doi.org/10.1021/nl203316r | Nano Lett. 2012, 12, 1184−1188