High-Performance Charge Transport in Semiconducting Armchair Graphene Nanoribbons Grown Directly on Germanium Robert M. Jacobberger and Michael S. Arnold* Department of Materials Science and Engineering, University of Wisconsin−Madison, Madison, Wisconsin 53706, United States S Supporting Information *
ABSTRACT: The growth of graphene on Ge(001) via chemical vapor deposition can be highly anisotropic, affording the facile synthesis of crystallographically controlled, narrow, long, oriented nanoribbons of graphene that are semiconducting, whereas unpatterned continuous graphene is semimetallic. This bottom-up growth overcomes longstanding challenges that have limited top-down ribbon fabrication (e.g., inadequate resolution and disordered edges) and yields ribbons with long segments of smooth armchair edges. The charge transport characteristics of sub-10 nm ribbons synthesized by this technique (which are expected to have band gaps sufficiently large for semiconductor electronics applications) have not yet been characterized. Here, we show that sub-10 nm nanoribbons grown on Ge(001) can simultaneously achieve a high on/off conductance ratio of 2 × 104 and a high on-state conductance of 5 μS in field-effect transistors, favorably comparing to or exceeding the performance of nanoribbons fabricated by other methods. These promising results demonstrate that the direct synthesis of nanoribbons on Ge(001) could provide a scalable pathway toward the practical realization of high-performance semiconducting graphene electronics, provided that the width uniformity and positioning of the nanoribbons are improved. KEYWORDS: graphene nanoribbon, semiconductor, field-effect transistor, charge transport, chemical vapor deposition, germanium
U
literature. The widespread applicability of sonochemical unzipping is limited, however, as it does not offer control over the edge orientation of the ribbons, results in poor yield, and does not allow for the deposition of ribbons from solution onto substrates with controlled placement and alignment. Duplicating the high-degree of structural precision and excellent charge transport properties afforded by sonochemical unzipping using more controllable and scalable methods has proven difficult. For example, patterning and etching nanoribbons from continuous sheets of graphene via top-down lithography yields relatively wide ribbons that have rough, defective edges,18−20 limiting Gon/Goff to 7 × 102 with a correspondingly low Gon of 7 × 10−2 μS in FETs.21 Wide ribbons can be narrowed to 10 nm) because of limitations in the ribbon transfer methodology utilized and in the FET architecture employed. As a consequence of the relatively large widths of the ribbons measured, Gon/Goff was limited to 101−102. In this work, we use an optimized nanoribbon transfer protocol and shorter channel lengths (Lch) to ensure that nanoribbons narrower than 10 nm can be accessed in FET measurements. We grow a polydisperse mixture of nanoribbons with widths varying from approximately 5 to 30 nm and measure their charge transport characteristics. We show that nanoribbons grown on Ge(001) via CVD can achieve both a high Gon of 5 μS and a high Gon/Goff of 2 × 104 (at room temperature) at the same time, favorably comparing to or exceeding the performance of nanoribbons fabricated by other methods and motivating future research into this promising approach for nanoribbon synthesis.
RESULTS AND DISCUSSION Arrays of randomly distributed graphene nanoribbons aligned roughly along Ge⟨110⟩ are grown on Ge(001) via CVD. The nanoribbons are transferred onto SiO2 (15 nm) on Si or HfO2 (15 nm) on Si substrates using a dry transfer method adapted from Lee et al.,35 which is schematically shown in Figure 1a−e. Briefly, the nanoribbons are peeled off of the Ge(001) surface 8925
DOI: 10.1021/acsnano.7b03220 ACS Nano 2017, 11, 8924−8929
Article
ACS Nano
Figure 2. (a, b) Histogram of width (a) and aspect ratio plotted against width (b) measured by SEM for 327 ribbons from a representative sample used to fabricate FETs. (c) Probability of occurrence of ribbons in FETs plotted against ribbon width. Top axis provides the expected Gon/Goff based on data from Li et al.15
ribbons of widths 1 cm2), resulting in arrays of ribbons with minimal structural distortion. Thus, this synthesis and subsequent dry transfer offer a simple and reproducible method to fabricate arrays of aligned nanoribbons on arbitrary substrates, including technologically relevant dielectrics. After transfer of the nanoribbons, FETs are fabricated. Source and drain electrodes with Lch of 25−120 nm are defined with electron-beam lithography, and Cr/Pd/Au (0.7/10/8.3 nm) contacts are thermally evaporated. The contacts are patterned at random locations across the substrate so that the ribbons are perpendicular to the source and drain electrodes. This random placement of devices results in FETs containing nanoribbons of varying width in which the contact lengths (Lc) between the ribbon and the metal electrodes are of varying distance. Figure 1h shows a schematic of the FET architecture, and Figure 1i shows a scanning electron microscopy (SEM) image of an FET with a nanoribbon channel with apparent width of ∼7 nm and Lch of ∼25 nm. The width distribution (Figure 2a) of the nanoribbons is characterized via SEM. Most of the nanoribbons vary in width from 5 to 30 nm. Approximately 1.5, 12, and 59% of the ribbons have widths of 40 at a Vds of 0.1 V on 15 nm of SiO2 on Si (black circles) and 15 nm of HfO2 on Si (orange diamonds). Before patterning the source and drain electrodes, the ribbons are annealed using various conditions, which does not significantly impact the FET performance (see Figure S4 and Supporting Information). (b) Expected (black squares) and measured (red circles) cumulative probability of occurrence of nanoribbons in FETs is plotted against Gon/Goff. (c) Comparison of Gon versus Gon/Goff at room temperature for nanoribbons synthesized on Ge(001) via CVD (black and gray stars) compared to those reported in the literature with Gon/Goff > 30. The plot does not include an FET with an armchair nanoribbon channel synthesized via polymerization in Bennett et al.,31 in which Gon/Goff is 103 and Gon is 7.5 × 10−5 μS.
encapsulation,39 by surface treatment,40 or by using hydrophobic substrates.41 The threshold voltage in the forward (negative to positive) Vgs sweep is likely more representative of the threshold voltage in the absence of hysteresis than in the reverse (positive to negative) Vgs sweep.41 The Ids versus Vgs characteristics often display small features, such as local maxima and minima. Some of these features are reproducible in sequential Vgs sweeps of the same device, while others are not (see Figure S2 and Supporting Information). Figure 3d plots the Ids versus Vgs characteristics of a nanoribbon with a Gon/Goff of 1 × 104 on 15 nm of HfO2, which can be compared to the FET on 15 nm of SiO2 with similar Gon/Goff in Figure 3c. The HfO2 gate dielectric increases electrostatic control over the ribbon channel due to its higher dielectric constant of ∼16, making it a more promising dielectric for more aggressively scaled, high-performance FETs. The Igs versus Vgs curves for the FETs on SiO2 and HfO2 in Figure 3 are plotted in Figure S3. Figure 4a compares Gon versus Gon/Goff for 178 FETs with Gon/Goff > 40 measured at a Vds of 0.1 V on both SiO2 and HfO2 substrates at room temperature. For FETs with a Gon/Goff > 103, which are obtained with an Lch of 25−65 nm, Gon is relatively high and varies from 0.3 to 8 μS. Similarly, for FETs with Gon/Goff < 103, which are obtained with an Lch of 25−120 nm, Gon is also high and varies from 0.4 to 27 μS. The nanoribbons exhibit large variability in both Gon/Goff and Gon. The variability in Gon/Goff can be largely attributed to polydispersity in width. We analyze the expected versus the measured variability in Gon/Goff in Figure 4b. The expected variability is determined from the distribution in Figure 2c, in which the expected Gon/Goff is determined as a function of nanoribbon width, w, according to an empirical relationship found by Li et al.,15 Gon/Goff = a exp(Eg/kBT), where Eg = b/w is the band gap of the ribbon, a and b are constants, kB is Boltzmann’s constant, and T is the temperature of 300 K. Li et al. justify this relationship because Goff is thermally activated over a Schottky barrier to the conduction band on the order of ∼Eg15 and first-principles calculations have predicted a similar dependence of Eg on w.1 On the basis of data from Li et al., we use a = 0.24 and b = 1.6 eV nm, which estimates Gon/Goff of 1, 5, 15, 102, 103, and 106 for widths of 50, 20, 15, 10, 7.5, and 4 nm, respectively. The resulting expected distribution in Gon/ Goff closely matches the measured distribution (Figure 4b). We expect roughly 1.0, 11, and 53% of the FETs to have a Gon/Goff > 103, 102, and 15, respectively, which agrees well with our measured yield of 1.2, 10, and 30%, respectively.
Figure 4a also shows that Gon at a given Gon/Goff can vary by over an order of magnitude. We attribute this variation, in large part, to variations in Lc between the nanoribbon and the source and drain electrodes. If Lc is less than the transfer length (∼100 nm for graphene),42−44 the contact resistance is expected to increase roughly linearly with decreasing Lc.45 The length of the source and drain electrodes used here varies from 30 to 500 nm. In FETs in which the length of the source and drain electrodes is 40, which are fabricated via lithographic patterning (orange diamonds),21,46−48 narrowing of wider ribbons via edge etching (green upward triangles), 22,23,25 polymerization (blue squares),31,32 sonochemical unzipping of graphite (red circles),15−17 and unzipping of carbon nanotubes (magenta downward triangles).49,50 In addition to CVD on Ge(001), the only methods that yield ribbons with Gon/Goff > 103 in FETs are polymerization,31,32 sonochemical unzipping of graphite,15−17 and edge etching of wider ribbons.22,23 Importantly, the Gon of nanoribbons grown on Ge(001) at a given Gon/Goff are among the highest reported in the literature. For the nanoribbons with the highest Gon/Goff, Gon is likely at least partially limited by a Schottky barrier at the nanoribbon/ contact interface, which is indicated by the nonlinear Ids versus Vds characteristics at low Vds (see Figure S5 and Supporting Information).51 From the relationship of Li et al.,15 we can estimate that the ribbons with Gon/Goff of 102, 103, and 104 have band gaps of 0.16, 0.22, and 0.28 eV, respectively. These data confirm that nanoribbons grown on Ge(001) have large band gaps ≫ kBT at room temperature and indicate that the crystallinity and edge morphology of the ribbons33,34 are sufficiently well ordered to realize high on-state conductance and promising charge transport characteristics.
CONCLUSIONS In conclusion, the development of nanoribbon technologies has been hindered by difficulties in the scalable synthesis of ribbons with sub-10 nm widths, controlled crystallographic orientation, 8927
DOI: 10.1021/acsnano.7b03220 ACS Nano 2017, 11, 8924−8929
Article
ACS Nano and smooth armchair edges, making it difficult to realize both high Gon/Goff and high Gon in FETs. The direct synthesis of aligned, narrow, and semiconducting graphene nanoribbons with predominantly smooth armchair edges on Ge(001) via CVD overcomes many of these synthetic challenges. Consequently, nanoribbons synthesized by this technique can simultaneously exhibit a high Gon/Goff of 2 × 104 and a high Gon of 5 μS in FETs, favorably comparing to or exceeding the performance of ribbons fabricated by other techniques and validating the high quality of nanoribbons grown on Ge(001). If the polydispersity in ribbon width and length is reduced and if the location of the ribbons can be controlled, large-area arrays of narrow semiconducting nanoribbons grown using CVD may enable significant advances in state-of-the-art semiconductor electronics.
AUTHOR INFORMATION
METHODS
REFERENCES
Corresponding Author
*E-mail:
[email protected]. ORCID
Robert M. Jacobberger: 0000-0001-5947-5308 Michael S. Arnold: 0000-0002-2946-5480 Notes
The authors declare no competing financial interest.
ACKNOWLEDGMENTS This work was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences, under Award #DESC0016007.
(1) Yang, L.; Park, C. H.; Son, Y. W.; Cohen, M. L.; Louie, S. G. Quasiparticle Energies and Band Gaps in Graphene Nanoribbons. Phys. Rev. Lett. 2007, 99, 186801. (2) Fang, T.; Konar, A.; Xing, H.; Jena, D. Mobility in Semiconducting Graphene Nanoribbons: Phonon, Impurity, and Edge Roughness Scattering. Phys. Rev. B: Condens. Matter Mater. Phys. 2008, 78, 205403. (3) Liang, G. C.; Neophytou, N.; Nikonov, D. E.; Lundstrom, M. S. Performance Projections for Ballistic Graphene Nanoribbon FieldEffect Transistors. IEEE Trans. Electron Devices 2007, 54, 677−682. (4) Yu, J.; Liu, G. X.; Sumant, A. V.; Goyal, V.; Balandin, A. A. Graphene-on-Diamond Devices with Increased Current-Carrying Capacity: Carbon sp2-on-sp3 Technology. Nano Lett. 2012, 12, 1603−1608. (5) Bae, M. H.; Li, Z. Y.; Aksamija, Z.; Martin, P. N.; Xiong, F.; Ong, Z. Y.; Knezevic, I.; Pop, E. Ballistic to Diffusive Crossover of Heat Flow in Graphene Ribbons. Nat. Commun. 2013, 4, 1734. (6) Ouyang, Y.; Yoon, Y.; Fodor, J. K.; Guo, J. Comparison of Performance Limits for Carbon Nanoribbon and Carbon Nanotube Transistors. Appl. Phys. Lett. 2006, 89, 203107. (7) Guo, J. Modeling of Graphene Nanoribbon Devices. Nanoscale 2012, 4, 5538−5548. (8) Fiori, G.; Iannaccone, G. Simulation of Graphene Nanoribbon Field-Effect Transistors. IEEE Electron Device Lett. 2007, 28, 760−762. (9) Choudhury, M. R.; Yoon, Y.; Guo, J.; Mohanram, K. Graphene Nanoribbon FETs: Technology Exploration for Performance and Reliability. IEEE Trans. Nanotechnol. 2011, 10, 727−736. (10) Tsuchiya, H.; Ando, H.; Sawamoto, S.; Maegawa, T.; Hara, T.; Yao, H.; Ogawa, M. Comparisons of Performance Potentials of Silicon Nanowire and Graphene Nanoribbon MOSFETs Considering FirstPrinciples Bandstructure Effects. IEEE Trans. Electron Devices 2010, 57, 406−414. (11) Zhang, Q.; Fang, T.; Xing, H. L.; Seabaugh, A.; Jena, D. Graphene Nanoribbon Tunnel Transistors. IEEE Electron Device Lett. 2008, 29, 1344−1346. (12) Ouyang, Y.; Yoon, Y.; Guo, J. Scaling Behaviors of Graphene Nanoribbon FETs: A Three-Dimensional Quantum Simulation Study. IEEE Trans. Electron Devices 2007, 54, 2223−2231. (13) Yan, Q. M.; Huang, B.; Yu, J.; Zheng, F. W.; Zang, J.; Wu, J.; Gu, B. L.; Liu, F.; Duan, W. H. Intrinsic Current-Voltage Characteristics of Graphene Nanoribbon Transistors and Effect of Edge Doping. Nano Lett. 2007, 7, 1469−1473. (14) Liang, G. C.; Neophytou, N.; Lundstrom, M. S.; Nikonov, D. E. Ballistic Graphene Nanoribbon Metal-Oxide-Semiconductor FieldEffect Transistors: A Full Real-Space Quantum Transport Simulation. J. Appl. Phys. 2007, 102, 054307. (15) Li, X. L.; Wang, X. R.; Zhang, L.; Lee, S. W.; Dai, H. J. Chemically Derived, Ultrasmooth Graphene Nanoribbon Semiconductors. Science 2008, 319, 1229−1232. (16) Wang, X. R.; Ouyang, Y. J.; Li, X. L.; Wang, H. L.; Guo, J.; Dai, H. J. Room-Temperature All-Semiconducting Sub-10-nm Graphene
Nanoribbon Synthesis. Ge(001) (Wafer World) substrates are loaded into a horizontal quartz tube furnace with an inner diameter of 34 mm, and the system is evacuated to
