Place a Graphene Sheet Where You Want It Literature Review

Robert McCorkell, May 2012

This project aims to develop a substrate transfer process for graphene in order to fabricate graphene based devices that use substrates benecial to the properties required by the device. In this review graphene and its interesting electrical and physical properties are introduced, and the methods to be used for manufacturing and transferring graphene are explained in detail. Possible devices to be fabricated to demonstrate and test the transfer process are then discussed, along with the methods to be used for electrical characterisation.

Introduction

these devices can then be determined using various techniques, which will be discussed. Graphene is a two dimensional monolayer of carbon atoms arranged in a hexagonal lattice (Figure 1,a). It has two atoms per unit cell, and can be viewed as two interpenetrating triangular lattices [1]. Due to this structure, a graphene sheet can be seen to exhibit armchair and zigzag edges shown in Figure 1,b [2]. Graphene sheets are the basic building blocks of graphitic materials of all other dimensionalities. They can make 1D fullerenes, 2D nanotubes, or simply be stacked to form 3D graphite [2]. Graphene has recently become very popular as a eld of research primarily because of its remarkable properties, often superior to other materials. Its charge carriers are massless Dirac fermions, it is extremely electrically and thermally conductive, very elastic and the worlds strongest material. It is the rst known two-dimensional atomic crystal, the thinnest known object and im1

The goal of this project is to develop a mechanical transfer process for graphene between substrates. To do so, graphene will rst need to be manufactured using the mechanical exfoliation technique (popularly called the scotch tape technique) on a silicon oxide substrate which aids in identication of single layer graphene. Reasons for this will be discussed. Once the graphene can be transferred between substrates with a reasonable success rate, simple electronic devices can be made as a practical test and demonstration of the process. Devices will mainly be produced using hexagonal boronnitride (hBN) as a substrate since it interferes with graphenes electronic properties less than other substrates like SiO2 . Mechanical exfoliation will be used to manufacture single layer graphene and hBN because there is currently no way of making wafer-scale graphene of extremely high quality. The electrical characteristics of

permeable to any other molecules [1]. In order to create graphene akes using micromechanical cleavage a substrate of SiO2 will be used. The purpose of this substrate is to make the process of nding single layer graphene akes among graphitic akes of varying thickness easier, by making any akes of the correct thickness optically visible [3]. The reason this is possible is discussed later. A problem with SiO2 as a substrate is it has a high surface roughness and a large density of charge traps and defects. This can have an adverse eect on the electrical properties of a graphene ake situated on it. Therefore in order to build electrical devices from graphene akes a dierent substrate is necessary. Hexagonal boron nitride (hBN) has a similar lattice structure to graphene, and an ultra at surface (Figure 2) that is free of dangling bonds and charge traps [3, 4]. Used as a substrate, hBN would allow graphene to retain its high carrier mobility, similar to the levels of freestanding graphene [3] and would therefore be a nearly ideal surface. From this it is clear that in order to make eective use of graphenes electrical properties in novel devices, a method for transferring graphene from one substrate to another is very important.

Figure 1: a, the honeycomb crystal structure of graphene [1]. b, Scanning electron micrograph of a graphene crystal. Zigzag and armchair edges are represented by blue and red lines respectively [2].

2
2.1

Electrical Characteristics
Band Structure

Figure 2: Atomic force microscopy images As mentioned earlier, graphene has two showing the surface roughness of hBN and atoms per unit cell organised as two in285nm of thermally grown SiO2 [3]. terpenetrating triangular lattices forming a 2

Figure 3: The low-energy band structure of graphene [1]. hexagonal lattice. Due to this hexagonal symmetry, graphenes conduction and valence bands touch at two points in the Brillouin zone, K and K (referred to as the Dirac Points) (Figure 3) [5]. This means that graphene can be regarded as a zerooverlap semimetal with a gapless spectrum and linear dispersion relation [1, 6]. Conversely, boron-nitride for example (which plays a signicant role in this project as a substrate) has no symmetry between the sublattices, causing a large gap in the electronic spectrum making it a good insulator [1].

2.2

Dirac Fermions

An integral part of what makes graphenes electronic properties so unique is the fact that at low energies graphenes charge carriers act like massless, chiral, Dirac fermions and the chemical potential exactly crosses 3

the Dirac point. This behaviour mimics the physics of massless fermions in quantum electrodynamics (QED) apart from that these fermions move at the Fermi velocity vf = c/300 rather than the speed of light [5, 6]. These massless Dirac fermions are quasiparticles, which are simply a convenient description of the behaviour of the electrons due to their interaction with the periodic potential of graphenes hexagonal lattice, and can be viewed as electrons with no rest mass or as neutrinos with an electric charge e [2]. Another parallel with QED is the way in which electrons and holes in graphene are described by wavefunctions. In conventional condensed matter physics, electrons and holes must be described by separate Schrdinger equations, whereas in graphene o quasiparticles are more easily described by the Dirac equation [2]. States in the valence and conduction bands must be described by two-component wavefunctions representing the relative contributions of each of graphenes sublattices. This description is very similar to that of spinor wavefunctions in QED, apart from that the spin index in graphenes case refers to the index of the sublattices, and is usually referred to as pseudospin [2, 6]. As a result, electron and hole states are linked, with properties analogous to charge-conjugation symmetry in QED [6]. From this we can essentially say that electrons and holes are described by the same spinor wavefunction and are linked by charge-conjugation [1]. The introduction of pseudospin gives an explanation for charge carriers in graphene being referred to as chiral. Chirality is a quantity that again originates from parallels with QED, and can be described as the

projection of pseudospin on the wavevector k, which is positive for electrons and negative for holes. Chirality can be said to represent the fact that electron and hole states are connected because they originate from the same sublattices. Many electronic properties in graphene can be explained using the conservation of pseudospin and chirality [2]. A distinctive feature of graphene in a magnetic eld due to the presence of Dirac fermions is the anomalous integer quantum hall eect (IQHE). In the classical Hall effect charge carriers moving through a conductor experience a force due to an applied magnetic eld. This force causes the charge carriers to accumulate at one side of the conductor, causing a measurable transverse potential dierence. In the quantum hall eect, when a magnetic eld is applied to a two-dimensional electronic system it causes electrons in the system to form circular orbits. The magnetic ux enclosed by one of these orbits is one magnetic ux quantum of h/e, and the electron wavefunction gains a phase shift of 2. In quantum mechanics these cyclotron orbits are quantised, and therefore charged particles in the material occupy discrete energy values called Landau levels. This is the cause of quantisation of electrical conductivity at low temperatures known as the quantum hall eect. In normal materials the integer quantum hall effect takes place at low temperatures, and gives a hall conductivity of xy = N e2 /h (Figure 4a.) where N is an integer representing the Landau Level index factor [7]. In contrast to the IQHE in other materials, the sequence of steps in the hall conductivity of graphene are shifted by 1/2 (Figure 4b.) due to a Berrys phase shift of 4

in quantum oscillations. This gives a new hall conductivity of xy = 4e2 /h(N + 1/2) where the 4 accounts for double degeneracy for both dispersion valley and electron spin [2]. The high cyclotron energies of Dirac fermions also allow this phenomenon to be observed at room temperature, unlike other materials such as Si [5]. Under specic conditions, electronic interactions in graphene can also give rise to composite fermions that carry a fraction of electronic charge. This means only a fraction of the Landau level lling factor is present causing the fractional quantum hall eect to occur [8].

2.3

Klein Paradox

Another characteristic feature of the Dirac fermions chiral symmetry in graphene is the Klein paradox [1, 5]. This paradox refers to the fact that for potential barriers that exceed energies of 2m0 c2 , the probability of tunnelling for relativistic particles approaches unity and is equal to one for massless particles. This can be explained by the fact that massless relativistic particles move at a constant velocity of the speed of light, but any form of backscattering is impossible since at the boundary the particles velocity must be equal to zero. Therefore the particle is forced to tunnel through the barrier [1]. This process is is very counterintuitive in comparison to classical systems because the tunnelling probability for the particle is weakly dependant on the barrier height, and increases for high barriers. Another explanation for this phenomena is that whilst the barrier is repulsive for electrons, it is attractive for positrons, causing positron states to be present within the

barrier which align in energy with the electrons outside [6]. Matching between electron and positron wavefunctions across the barrier then leads to these high tunnelling probabilities [10]. Graphenes chiral quasiparticles lead to perfect electron to hole conversion at a potential barrier, and with a tunnelling probability equal to one for normal incidence the result is a guaranteed nite minimum conductivity [1]. With only one extra layer, bilayer graphene has quite dierent properties to its monolayer cousin. The positioning of the layers exhibit AB (Bernal) stacking, and the material has low energy bands that cross at zero energy, but unlike monolayer graphene are parabolic. The symmetry between layers in bilayer graphene still gives rise to chiral symmetry between electrons and holes, but here we have a dierent type of quasiparticle, the massive chiral fermion which does not have an analogy in QED. In this material however, it is possible to break this chirality by targeting one of the two layers with an electric eld with the use of a gate voltage, thus breaking the symmetry [1, 2]. Figure 4: Schematic illustration of the integer QHE with blue and orange peaks representing full Landau levels. a) Integer QHE found in conventional 2D semiconductor systems and b) half-integer QHE behaviour for massless Dirac fermions in single-layer graphene. Adapted from reference [9].

3
3.1

Manufacturing Graphene
Mechanical Exfoliation

An extremely popular method for manufacturing graphene is the mechanical cleavage technique. Originally developed by the group led by K. Novoselov and A. Geim [11], this technique is most commonly used because it is very easy to perform, cheap, and it is currently not possible to create 5

wafer-scale graphene of very high quality [1]. Other methods of graphene manufacture include chemical reduction of graphite oxide, graphitisation of single crystal SiC and chemical vapor deposition (CVD) to grow graphene on metal substrates, none of which provide graphene with mobilities as high as that of graphene exfoliated from natural graphite. In the case of graphene grown by CVD, substrate transfer methods can reduce the quality further by introducing cracks and tears [12, 13]. In the simplest versions of the mechanical cleavage method described in reference [11] a piece of adhesive tape is pressed against a high quality graphite crystal, removing the top layer. The tape is then pressed against a substrate leaving behind thin akes of graphite [1] typically with a thickness of less than 10nm which attach strongly to the substrate due to van der Waals or capillary forces [11]. In this case the substrate of choice is a Si wafer with a layer of 300nm thick SiO2 on the surface. Of the many small graphite akes that can be found on the Si wafer at the end of the exfoliation process, there is a reasonable probability that some of them will be small enough in thickness to be considered graphene. The most dicult part is nding them within the mass of larger akes. Currently the easiest method for identifying a single-layer crystal is Raman spectroscopy, but at high resolution it would be impossible to scan an entire substrate for 2D crystals or even to nd them by scanning the surface at random [14]. The quickest way to nd such crystals would be optically with the use of the human eye, but monolayer graphene appears transparent to the eye. However, on our SiO2 substrate 6

the added optical path from a graphitic lm under violet light shifts the interference colours so that a lm appears blue whilst a 300nm SiO2 wafer appears violet-blue [11]. This allows for quick and easy identication of single layer to few layer graphene between the much larger graphite crystals on the SiO2 surface.

3.2

Conrmation with Raman Spectroscopy

Once graphene sheets are identied optically their thickness and subsequent number of layers must be conrmed. This can be done through analysis of the materials Raman spectrum. This process essentially illuminates a sample with a laser beam, and measures how the wavelength is shifted relative to the source due to phonons. The relative shapes and intensities shown on a plot of the resulting spectrum can give us information on the thickness of the graphene. In graphite there are three main characteristic peaks in the Raman spectrum (Figure 5,a). The G peak located at approximately 1600cm1 appears due to a resonance of the carbon bonds. The 2D peak is located at approximately 2700cm1 and is caused by electron interactions with the incident photons, and subsequent interactions with phonons moving it into an adjacent Brillouin zone and back again. Finally the D peak located at approximately 1350cm1 is similar to the 2D peak but the electron only interacts with one phonon, moving it into the adjacent Brillouin zone where it can decay if a hole is available in the same zone. This is only possible if the sample contains defects, or if the measurement was taken

near the edge of the sample [15, 16]. The intensity of the G peak shown in Figure 5,a and the shape of the 2D peak shown in Figure 5,b, where the 2D peak is more symmetric for single layer graphene, can be used to identify the thickness of a suspected graphene sheet. The presence of a D peak can also be used to measure the purity of the sample [15].

Substrate-Transfer Method

Whilst SiO2 is an excellent material for nding exfoliated graphene akes, it is not ideal for preserving graphenes unique electrical characteristics. With thermally grown SiO2 , graphenes ultra-high mobility can be inhibited by high surface roughness and a large density of charge traps as seen in Figure 2 [3]. Hexagonal boron nitride on the other hand has a similar 2D planar atomic structure to graphene, with an ultra-at surface relatively free of dangling bonds and charge traps. The only known environment better for preserving mobility is to have graphene suspended, but this leaves the sheet fragile, and imposes limitations on the design of graphene based devices [3, 4]. In order to transfer exfoliated graphene akes between SiO2 and hBN substrates a similar transfer process will be performed to that presented in Figure 6. First the hBN will be exfoliated from single crystals onto silicon wafers coated in thermal oxide. Graphene is then exfoliated onto a layer of SiO2 stacked on a water soluble layer and PMMA and suspended in a water bath (Figure 6,i). The thickness of the 7

Figure 5: a) Raman spectra of graphite and graphene clearly showing the G and 2D peaks at approximately 1600cm1 and 2700cm1 respectively. No D peak is visible at 1350cm1 because this measurement was taken at the centre of a graphene sample with few defects. b) Dierences in shapes between the 2D peaks of various graphite thicknesses. Single layer graphene has a clear symmetric peak and peaks corresponding to subsequent layers are made up of multiple peaks causing the whole to be asymmetrical. Adapted from Reference [15].

PMMA can be chosen to give similar optical identication advantages to that of SiO2 described above, thus allowing conrmation that monolayer graphene is present on the PMMA polymer. The thickness of the graphene akes can be further conrmed with the use of Raman spectroscopy. The water soluble layer can then dissolve allowing the layer of silicon to sink, leaving the PMMA layer, which is extremely hydrophobic, oating on the surface (Figure 6,ii). With the silicon wafer removed, the PMMA layer along with the graphene akes can be adhered to a glass slide attached to a micromanipulator arm on an optical microscope and precisely positioned onto the boron-nitride substrate (Figure 6,iii). Water can then be driven o the substrate, PMMA and graphene akes by heating, and nally the PMMA is dissolved in acetone leaving the graphene successfully transferred to the new hBN substrate (Figure 6,iv). An alternative method to this process is to use polyvinyl alcohol (PVA) in place of the water soluble layer, which allows the PMMA to be mechanically peeled from the original substrate without exposing the PMMA or graphene to water [8].

Figure 6: Schematic illustration of transfer process. i) Graphene is exfoliated onto polymer stack suspended in water, ii) soluble layer dissolves allowing silicon to sink, iii) PMMA layer is adhered to glass slide and aligned over exfoliated hexagonal boron nitride (hBN) akes, iv) PMMA layer is dissolved in acetone leaving graphene on hBN substrate [4]. phy, and the whole sample including electrical leads can be formed into a desired shape through the process of etching [3, 4, 8]. The process of depositing electrical contacts with e-beam lithography involves depositing a shadow mask made from PMMA onto a graphene sample, and where the PMMA has not been deposited, metal will be able to come into contact with the graphene during metal evaporation. The PMMA layer which blocks any unwanted metal, can then be washed away during lift-o, leaving behind the desired electrical contacts. If necessary reactive ion etching can be used to shape a device, for example a Hall bar where the a specic geometry is required to perform quantum hall eect experiments. A Hall bar (rst reported with graphene in 2004 [11]) is one of the simplest devices 8

Possible Devices

With the mechanical exfoliation and substrate transfer processes developed to a reliable success rate, it will then be possible to construct simple graphene based devices in order to test their electrical characteristics. It will therefore be necessary to deposit electrical contacts made from Cr/Au onto the graphene ake using e-beam lithogra-

Figure 7: A simple hall bar device on a BN substrate. Adapted from reference [8]. that can be created and investigated for this project, since it merely involves a sheet of graphene either freestanding or on a substrate. A standard Hall bar device measures current across the length of the bar (horizontally in Figure 7), and the potential dierence across two degrees of freedom of the charge carriers. The change in electrical characteristics can then be investigated after applying a magnetic eld. With this device it will be possible to investigate the various forms of the quantum hall eect through simple low temperature electrical transport measurements. Variations to the standard hall bar experiments could include applying the magnetic eld at dierent angles, applying an electric eld, or applying the current in dierent directions on the device to search for asymmetries. This would allow us to probe deeper insights into the quantum hall eect and the underlying effects of Landau levels. A common device that has the potential to be greatly improved with the introduction of graphene is the eld-eect transistor (FET). Currently digital logic devices rely 9

heavily on metal-oxide-semiconductor FET (MOSFET) performance, and in semiconductor electronics the scalability of these devices is key to increasing the complexity and speed of integrated circuits. The fundamental structure of an FET consists of source and drain electrodes connected by a channel region, and a gate electrode separated from the channel region by a barrier. The conductivity of the channel region is then controlled by applying a voltage between the gate and the source. In order for FETs to quickly respond to variations of this voltage, short gates and fast charge carriers are required, but short gates often suer from various forms of electrical degradation collectively known as short-channel eects. A FET with a thin barrier and thin gate control-led-region (vertically in Figure 8) should be resistant to these short-channel eects down to short gate lengths (horizontally in Figure 8) [17]. Therefore this would allow for extremely thin channel regions, for which graphene is a perfect candidate with its thin planar structure and high mobility. The rst graphene based FET was reported in 2007 [18], and has since been improved with cuto frequencies reaching 1GHz in 2008 [19] and 100GHz in 2010 [20]. These device were all fabricated on a silicon substrate, so it would be of interest to test how the performance of similar devices is aected with an increased mobility provided by a hBN substrate, depending on the availability of equipment to characterise high frequency devices. A more complex device that can be made using a hBN substrate is a BN/Graphene/BN transistor in analogy to reference [3]. In this paper bilayer graphene

to have a mobility no higher than that of a silicon channel [17]. This hurdle would need to be overcome for graphene channels to be benecial in digital logic applications, and could be investigated further in this project.

References
Figure 8: A cross section of an n-channel silicon metal-oxide semiconductor eld effect transistor (MOSFET). Thin barrier and gate-control-led region measured vertically allows for thin a channel region down to short gate lengths measured horizontally. In this case graphene is an excellent candidate for a thin channel region [17]. [1] K Novoselov. Nobel Lecture: Graphene: Materials in the Flatland. Reviews of Modern Physics, 83(3):837849, 2011. [2] A K Geim and K S Novoselov. The rise of graphene. Nature Materials, 6(3):183191, 2007. [3] Han Wang, Thiti Taychatanapat, Allen Hsu, Kenji Watanabe, Takashi Taniguchi, P Jarillo-Herrero, and Tomas Palacios. BN/Graphene/BN Transistors for RF Applications. IEEE Electron Device Letters, 2(9):24, 2011. [4] C R Dean, A F Young, I Meric, C Lee, L Wang, S Sorgenfrei, K Watanabe, T Taniguchi, P Kim, K L Shepard, and J Hone. Boron nitride substrates for high-quality graphene electronics. Nature Nanotechnology, 5(October):722 6, 2010. [5] A Castro Neto, F Guinea, N Peres, K Novoselov, and A Geim. The electronic properties of graphene. Reviews of Modern Physics, 81(1):109 162, 2009. [6] M. I. Katsnelson, K. S. Novoselov, and a. K. Geim. Chiral tunnelling and the Klein paradox ingraphene. Nature Physics, 2(9):620625, August 2006. [7] C. Kittel. Introduction To Solid State Physics. Wiley, 2005.

is sandwiched between two layers of hBN with additional contacts made from Cr/Au attached forming the source, gate and drain. This kind of device would serve as a good test of the capabilities of the mechanical transfer process developed in this project. Finally, it would also be possible to fabricate an FET for digital applications. However large-scale graphene is a gapless semiconductor, which means that devices using graphene for a channel can not be switched o. This would make the material unsuitable for logic applications, but there are methods to open a bandgap in graphene. For example in bilayer graphene a bandgap can be opened by applying a bias voltage to one of the layers. However in general electron mobility decreases as the bandgap of a material increases, and in graphene opening a bandgap similar to that of silicon causes it 10

[8] C R Dean, A F Young, P CaddenZimansky, L Wang, H Ren, K Watanabe, T Taniguchi, P Kim, J Hone, and K L Shepard. Multicomponent fractional quantum Hall eect in graphene. Nature Physics, 7(May):58, 2011. [9] K. S. Novoselov, E. McCann, S. V. Morozov, V. I. Falko, M. I. Katsnelson, U. Zeitler, D. Jiang, F. Schedin, and a. K. Geim. Unconventional quantum Hall eect and Berrys phase of 2 in bilayer graphene. Nature Physics, 2(3):177180, February 2006. [10] P Krekora, Q Su, and R Grobe. Klein paradox in spatial and temporal resolution. Physical Review Letters, 92(4):040406, 2004. [11] K S Novoselov, A K Geim, S V Morozov, D Jiang, Y Zhang, S V Dubonos, I V Grigorieva, and A A Firsov. Electric Field Eect in Atomically Thin Carbon Films. Science, 306(5696):1 12, 2004. [12] Xuesong Li, Weiwei Cai, Jinho An, Seyoung Kim, Junghyo Nah, Dongxing Yang, Richard Piner, Aruna Velamakanni, Inhwa Jung, Emanuel Tutuc, Sanjay K Banerjee, Luigi Colombo, and Rodney S Ruo. Large-area synthesis of high-quality and uniform graphene lms on copper foils. Science (New York, N.Y.), 324(5932):13124, June 2009. [13] Xuesong Li, Yanwu Zhu, Weiwei Cai, Mark Borysiak, Boyang Han, David Chen, Richard D Piner, Luigi Colombo, and Rodney S Ruo. Transfer of large-area graphene lms for high-performance transparent conductive electrodes. Nano letters, 9(12):435963, December 2009.

[14] K S Novoselov, D Jiang, F Schedin, T J Booth, V V Khotkevich, S V Morozov, and A K Geim. Twodimensional atomic crystals. Proceedings of the National Academy of Sciences of the United States of America, 102(30):104513, July 2005. [15] a. C. Ferrari, J. C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri, S. Piscanec, D. Jiang, K. S. Novoselov, S. Roth, and a. K. Geim. Raman Spectrum of Graphene and Graphene Layers. Physical Review Letters, 97(18):1 4, October 2006. [16] R Vidano, D B Fischbach, L J Willis, and T M Loehr. Observation of Raman band shifting with excitation wavelength for carbons and graphites. Solid State Communications, 39(2):341344, 1981. [17] Frank Schwierz. Graphene transistors. Nature Nanotechnology, 5(7):487496, 2010. [18] M C Lemme, T J Echtermeyer, M Baus, and H Kurz. A Graphene Field-Eect Device. IEEE Electron Device Letters, 28(4):12, 2007. [19] Inanc Meric, Natalia Baklitskaya, Philip Kim, and Kenneth L Shepard. RF performance of top-gated, zerobandgap graphene eld-eect transistors. 2008 IEEE International Electron Devices Meeting, (c):14, 2008. [20] Y M Lin, C Dimitrakopoulos, K A Jenkins, D B Farmer, H Y Chiu, A Grill, and Ph Avouris. 100-GHz Transistors from Wafer-Scale Epitaxial Graphene. Science, 327(5966):662, 2010.

11