Sie sind auf Seite 1von 4

Analysis of ballistic monolayer and bilayer graphene field-effect transistors

Yijian Ouyang, Paul Campbell, and Jing Guo Citation: Appl. Phys. Lett. 92, 063120 (2008); doi: 10.1063/1.2841664 View online: http://dx.doi.org/10.1063/1.2841664 View Table of Contents: http://apl.aip.org/resource/1/APPLAB/v92/i6 Published by the American Institute of Physics.

Additional information on Appl. Phys. Lett.


Journal Homepage: http://apl.aip.org/ Journal Information: http://apl.aip.org/about/about_the_journal Top downloads: http://apl.aip.org/features/most_downloaded Information for Authors: http://apl.aip.org/authors

Downloaded 02 May 2013 to 142.150.190.39. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://apl.aip.org/about/rights_and_permissions

APPLIED PHYSICS LETTERS 92, 063120 2008

Analysis of ballistic monolayer and bilayer graphene eld-effect transistors


Yijian Ouyang,1 Paul Campbell,2 and Jing Guo1,a
1

Department of Electrical and Computer Engineering, University of Florida, Gainesville, Florida 32611, USA 2 Naval Research Laboratory, Washington, District of Columbia 20375, USA

Received 23 October 2007; accepted 17 January 2008; published online 15 February 2008 We examine and compare ballistic performance limits of metal-oxide-semiconductor eld-effect transistors with monolayer and bilayer graphene channels. Under low source-drain biases and cryogenic temperatures, the leakage current of the bilayer device is orders of magnitude smaller than that of the monolayer device. The advantage lowers at raised temperatures and source-drain biases. The bilayer device, however, still has qualitatively different and more favorable I-V characteristics. We nd the ballistic on-state channel conductance and the minimum channel conductance have distinctly different dependences on the channel length. 2008 American Institute of Physics. DOI: 10.1063/1.2841664 Graphene has been intensively studied for potential electronics applications due to the excellent transport properties1,2 since its successful isolation.3,4 Recent experiments demonstrate monolayer and bilayer graphene eldeffect devices with doped source and drain extensions.5,6 The devices have a similar structure to the technologically important silicon-on-insulator SOI metal-oxide-semiconductor eld-effect transistors MOSFETs, but have a graphene channel. The experiments and potential technological importance of graphene MOSFETs motivate this theoretical study on ballistic monolayer and bilayer graphene MOSFETs. Although scattering may dominate in the preliminary experiments,5,6 the characteristics7 of ballistic device establish the performance limits and are important for nanoscale graphene MOSFETs, which are preferred for better device performance. In this letter, device characteristics of graphene MOSFETs are simulated by numerically solving the Dirac equation using nonequilibrium Greens function NEGF formalism. Compared to monolayer graphene FETs, the minimum source-drain channel conductance of bilayer graphene FETs is orders of magnitude smaller at cryogenic temperature and low drain biases. The advantage in terms of a smaller minimum leakage current signicantly decreases at raised temperature and drain biases, but the bilayer device still has a qualitatively different I-V characteristics. The channel length scaling characteristics of both devices are also examined. The cross section along the channel of the modeled monolayer bilayer graphene MOSFETs is schematically shown in Fig. 1a. The graphene channel is intrinsic, and the source and the drain extensions are n-type doped with a doping density of 1 1013 / cm2. To achieve charge neutrality at this doping density, the equilibrium Fermi level is 370 meV above the neutral point in the monolayer graphene, and 220 meV above the neutral point in the bilayer graphene. We consider a wide graphene channel, for which the periodic boundary condition is applied and the wave modes in the transverse direction are plane waves. For a transverse mode with a wave vector ky, the carriers in the monolayer graphene are described by the Dirac Hamiltonian,
a

H m k y =

iF i + iky x

Ux

iF i

iky x

Ux

, 1

where x is the direction along the channel, as shown in Fig. 1a, and Ux is the band prole of the neutral point as a function of the position. The Hamiltonian of the bilayer graphene near the neutral point is described by6,8

H b k y =

Ux + /2

t0 i + iky x

t0 i
2

iky x

Ux /2

where t0 = 2 / 2m*, m* 0.054m0 is the effective mass extracted from an interlayer coupling of 0.4 eV, m0 is the free electron mass, and is the bandgap that can be opened by generating electric eld between two layers. The retarded Green function of the mode is computed as9

Electronic mail: guoj@u.edu.

FIG. 1. Color online a The cross section of the modeled monolayer bilayer graphene MOSFET. The source and drain extensions are doped to n type, and the channel is intrinsic. b The sketches of the equilibrium potential proles, Ux, and the local bandstructure with the transverse wave vector ky = 0. The dotted and solid lines in the band structure indicate the binding states and the antibinding states, respectively. 2008 American Institute of Physics

0003-6951/2008/926/063120/3/$23.00

92, 063120-1

Downloaded 02 May 2013 to 142.150.190.39. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://apl.aip.org/about/rights_and_permissions

063120-2

Ouyang, Campbell, and Guo

Appl. Phys. Lett. 92, 063120 2008

GE, ky = E + i0+I Hm,bky 1E, ky 2E, ky1 , 3 where the subscript m b denotes the monolayer bilayer graphene, and 12E , ky is the self-energy of the semiinnite doped graphene source drain extension, which is computed for every transverse mode ky at the energy E using the SanchoRubio iterative method.9 The source-drain transmission or current is computed by summing over all transverse modes. At zero temperature, T = 0 K, the channel conductance is Gch = G0 T k E = E F , W ky y 4
FIG. 2. Color online The source-drain conductance versus the gate voltage at T = 0 K. The dashed curve is for the monolayer MOSFET, and the solid curve is for the bilayer MOSFET. The channel length is Lch = 40 nm.

which is proportional to the transmission coefcient at the Fermi level EF, TkyE = trace1E, kyGE, ky2E, kyGE, ky+ , 5 where W is the channel width, G0 = 4q2 / h is the quantum conductance constant that includes spin degeneracy and val+ ley degeneracy, and 12 = i12 1 2 is the contact broadening. The channel conductance in the unit of S / m is independent of W when it is large. The ballistic source-drain current at a nite temperature is ID = G0 qW

dE TkyE f 1E f 2E ,
ky

where f 12E is the Fermi-Dirac distribution function of the source drain contact. We focus our attention on quantum transport in the monolayer and bilayer graphene devices. The band prole, Ux, is determined by extending an electrostatic potential formula for single-gated, ultra-thin-body SOI MOSFETs, to the graphene MOSFETs.10 The gate voltage relates to the surface potential at the beginning of the channel S using the gradual channel approximation, VG = S + qn/Cox , 7 where n is the charge at the beginning of the channel computed by the NEGF formalism,9 and Cox is the gate insulator capacitance. The modeled device has a high- HfO2 gate insulator r = 20 with a thickness of 4.5 nm, which results in an effective oxide thickness of about 0.9 nm. Figure 1b sketches the potential prole and the local band structures at ky = 0 in the source extension and in the channel for the monolayer device on the left and the bilayer device on the right. The Hamiltonian of the monolayer graphene in Eq. 1 shows that for a band structure calculation, the positive-going states E = Fkx are binding states of two sublattices of the graphene and the negative-going states E = Fkx are antibinding states, as indicated by the solid lines and the dashed lines in Fig. 1b, respectively. Similarly, Eq. 2 indicates that for a band structure calculation of the bilayer graphene, the conduction band states are binding states, and the valence band states are antibinding states. Compared to the perfect matching of the wave functions for electrons injected normally with ky = 0 from the source contact to the channel in the monolayer device, the mismatch of the wave functions in the bilayer device, as shown in Fig. 1b, results in block of normally injected electrons when the graphene channel is modulated to p type by the gate, which lowers the channel conductance as shown

below. For a transverse mode with ky 0, the transmission of a bilayer channel can be close to 1 when the resonance condition is satised, but the average transmission over all transverse modes of a bilayer graphene is smaller than that of a monolayer graphene.8 Figure 2 plots the channel conductance GD = ID / VDVD=0, as a function of the gate voltage at zero temperature for both devices with a channel length of Lch = 40 nm. We nd that the channel conductance of the bilayer device is smaller than that of the monolayer device when the channel is modulated to p type, with the minimum channel conductance more than one order of magnitude smaller. The result is consistent with the picture of wave function matching in Fig. 1b. In contrast to the nearly symmetric conductance curve of the monolayer device with regard to the minimum conductance bias point, the conductance curve of the bilayer device is highly asymmetric, with the conductance of the n-type conduction branch about an order of magnitude larger than that of the p-type branch. In terms of the minimum channel conductance, the advantage of the bilayer MOSFET is signicant compared to the monolayer device. Figure 3 plots the minimum and the on-state VG = 0.7 V channel conductances as a function of the channel length for both the monolayer and the bilayer MOSFETs. We nd the on-state conductance nearly independent of the channel length for both devices because of the ballistic transport of the propagating electron waves when the devices are turned on. In contrast, the minimum channel conductance signicantly decreases as the channel length increases. For

FIG. 3. Color online The minimum conductance the unmarked curves and the on-state conductance the curves with circles vs the channel length Lch for the monolayer MOSFET the dashed curves and the bilayer MOSFET the solid curves.

Downloaded 02 May 2013 to 142.150.190.39. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://apl.aip.org/about/rights_and_permissions

063120-3

Ouyang, Campbell, and Guo

Appl. Phys. Lett. 92, 063120 2008

FIG. 4. Color online The source-drain currents vs the gate voltage at VD = 0.3 V and T = 300 K. The channel length is Lch = 40 nm. The current of the monolayer device saturates at VG 0.55 V, due to source exhaustion, in which the current is limited by the nite doping density in the source. The band prole of the monolayer MOSFET at VD = 0.3 V and VG = 0.28 V EFS ETop = 0.18 eV is shown in the inset.

the monolayer device, the minimum channel conductance occurs when the Fermi level is at the neutral point, at which the density of states is zero, and the conductance is contributed by quantum-mechanical tunneling. The transmission for a transverse mode ky is Tky exp2kyL and the minimum channel conductance is nearly inversely proportional to the channel length,11 as shown in Fig. 3. Figure 3 also shows that the minimum channel conductance of the bilayer device also decreases as the channel length increases. Figure 4 shows the simulated source-drain current as a function of the gate voltage for monolayer and bilayer MOSFETs at VD = 0.3 V and room temperature. The result indicates that the I-V characteristic of the bilayer device is qualitatively different from that of the monolayer device. Compared to the nearly symmetric n-type and p-type currents and the U shape I-V characteristic of the monolayer MOSFET, the n-type branch of the bilayer MOSFET I-V has a much stronger dependence on the gate voltage than the p-type branch. The minimum source-drain current of the bilayer device is still smaller than that of the monolayer device, but the advantage in terms of smaller source-drain leakage is signicantly reduced. The reason is explained as follows. Under a nite drain bias, the source-drain current is determined by the average transmission in an energy range approximately between the source and drain Fermi levels, in which the difference between the source and drain Fermi functions, f 1E f 2E, is approximately 1, as indicated by Eq. 6. The channel conductance at zero temperature is proportional to the transmission by Eq. 4. Because the split between the source and drain Fermi levels, equal to qVD = 0.3 eV, is much larger than the periods of oscillations of the p-type conductance of the bilayer device, as shown in Fig. 2, the average of the transmissions smoothes the oscillations. An applied source-drain voltage of 0.3 V results in a signicantly decreased ratio of the on current to the minimum leakage current, and the disappearance of the oscillations of the currents. The international technology roadmap for semiconductors ITRS calls for an on current of 2200 A / m at a power supply voltage of 0.7 V for a double gate silicon MOSFET with an effective oxide thickness of 9 , in year 2020 at the end of the roadmap.12 The modeled device has a gate insulator of the same effective oxide thickness. The current at VG = 0.7 V and VD = 0.3 V is 4600 A / m for the ballistic monolayer device and 4000 A / m for the bilayer

device. The on current exceeds the roadmap requirement. The problem of the graphene MOSFETs, however, is the large minimum source-drain leakage current around 450 A / m for bilayer device and 1150 A / m at VG = 0 V and VD = 0.3 V, so that off-current target cannot be met, but they could be attractive for radio-frequency electronics applications. The analysis is based on several assumptions. First, ballistic transport is assumed. The carrier mean free path in graphene is long, but optical phonon emissions can occur at high source-drain biases. The assumption of ballistic transport sets the performance limits, and we also notice that a carbon nanotube FET can deliver a nearly ballistic dc current even in the presence of optical phonon scattering.13 Future studies are needed for exploring the effect of scattering in graphene FETs. Second, a simple parabolic band structure of the bilayer graphene is used. The bandstructure gradually becomes nonparabolic away from the neutral point.14 Finally, the Dirac Hamiltonian assumes a band structure that is independent of the channel orientation in the graphene plane. The bandstructure can become nonisotropic, which could lead to orientation dependence of the channel conductance. Nonparabolic and nonisotropic band structure effects could add quantitative perturbations to our results, but we expect the qualitative conclusions to remain the same because for the modeled device parameters electron transport occurs near the neutral point. A single gate can also result in a potential drop between the two layers of a bilayer channel and open a bandgap. The simulation results of including the bandgap opening using Eq. 2 indicate that even a bandgap opening of 120 meV, which could only be reached at a high charge density of 1013 / cm2 as numerically computed in the presence of the screening effect,15 reduces the minimum leakage current by a factor of about 3 at VD = 0.3 V. The relatively small effect of the bandgap on the minimum source-drain current is because that the bandgap is small compared to the applied source-drain voltage. It is our pleasure to thank Dr. Carter White and Dr. Daniel Gunlycke of Naval Research Lab, and Professor David Goldhaber-Gordon of Stanford University for stimulating discussions.
C. Berger, Z. Song, X. Li, X. Wu, N. Brown, C. Naud, D. Mayou, T. Li, J. Hass, A. N. Marchenkov, E. H. Conrad, P. N. First, and W. A. de Heer, Science 312, 1191 2006. 2 D. Gunlycke, H. M. Lawler, and C. T. White, Phys. Rev. B 75, 085418 2007. 3 K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos, I. V. Grigorieva, and A. A. Firsov, Science 306, 666 2004. 4 Y. B. Zhang, Y. W. Tan, H. L. Stormer, and P. Kim, Nature London 438, 201 2005. 5 B. Huard, J. A. Sulpizio, N. Stander, K. Todd, B. Yang, and D. GoldhaberGordon, Phys. Rev. Lett. 98, 236803 2007. 6 J. B. Oostinga, H. B. Heersche, X. Liu, A. F. Morpurgo, and L. M. K. Vandersypen, arXiv:0707.2487v2. 7 F. Miao, S. Wijeratne, Y. Zhang, U. C. Coskun, W. Bao, and C. N. Lau, Science 317, 1530 2007. 8 M. I. Katsnelson, K. S. Novoselov, and A. K. Geim, Nat. Phys. 2, 620 2006. 9 S. Datta, Quantum Transport: Atom to Transistor Cambridge University Press, Cambridge, UK, 2005. 10 K. K. Young, IEEE Trans. Electron Devices 36, 399 1989. 11 M. I. Katsnelson, Eur. Phys. J. B 57, 157 2006. 12 ITRS roadmap http://www.itrs.net. 13 J. Guo and M. Lundstrom, Appl. Phys. Lett. 86, 193103 2005. 14 I. Snyman and C. W. J. Beenakker, Phys. Rev. B 75, 045322 2007. 15 E. McCann, Phys. Rev. B 74, 161403R 2006.
1

Downloaded 02 May 2013 to 142.150.190.39. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://apl.aip.org/about/rights_and_permissions

Das könnte Ihnen auch gefallen