Quasi-Free-Standing Graphene on Silicon Carbide

Quasi-free-standing Graphene on Silicon Carbide

Quasi-freistehendes Graphen auf Siliziumcarbid

Der Naturwissenschaftlichen Fakultät der Friedrich-Alexander-Universität Erlangen-Nürnberg zur Erlangung des Doktorgrades Dr. rer. nat.

vorgelegt von Markus Ostler aus Nürnberg Als Dissertation genehmigt von der Naturwissenschaftlichen Fakultät der Friedrich-Alexander-Universität Erlangen-Nürnberg Tag der mündlichen Prüfung: 16.06.2014 Vorsitzender des Promotionsorgans: Prof. Dr. Johannes Barth Gutachter: Prof. Dr. Thomas Seyller Prof. Dr. Heiko B. Weber Parts of this study were published in:

[Ost12] M. Ostler, R. J. Koch, F. Speck, F. Fromm, H. Vita, M. Hundhausen, K. Horn and Th. Seyller, Decoupling the Graphene Buﬀer Layer from SiC(0001) via Interface Oxidation, Mater. Sci. Forum 717- 720, 649 (2012).

[Ost13] M. Ostler, I. Deretzis, S. Mammadov, F. Giannazzo, C. Spinella, Th. Seyller, A. L. Magna, G. Nicotra and A. La Magna, Direct growth of quasi-free-standing epitaxial graphene on nonpolar SiC surfaces, Phys. Rev. B 88(8), 085408 (2013).

[Ost14] M. Ostler, F. Fromm, R. J. Koch, P. Wehrfritz, F. Speck, H. Vita, S. Böttcher, K. Horn and Th. Seyller, Buﬀer layer free graphene on SiC(0001) via interface oxidation in water vapor, Carbon 70, 258 (2014).

Contents

1 Introduction1

2 Basics5 2.1 Silicon carbide (SiC)...... 5 2.2 Graphene...... 10 2.2.1 Crystal structure...... 11 2.2.2 Electronic properties...... 13 2.2.3 Synthesis of graphene...... 15 2.3 Epitaxial graphene...... 18

3 Experimental methods 21 3.1 Photoemission spectroscopy...... 21 3.2 Raman Spectroscopy...... 28 3.3 Low-energy electron microscopy...... 35 3.3.1 Introduction...... 35 3.3.2 The instrument...... 36 3.3.3 Contrast mechanisms...... 39 3.3.4 Imaging modes...... 42 3.3.5 LEEM and epitaxial graphene...... 45

4 Intercalation of oxygen 47 4.1 Experimental details...... 49 4.2 High-temperature treatment...... 51 4.3 Low-temperature treatment...... 57 4.4 Conclusion...... 60

5 Intercalation of water vapor 61 5.1 Experimental details...... 62

v 5.2 Decoupling of the buﬀer layer...... 63 5.3 Investigation of defects...... 66 5.4 Transition of the band structure...... 69 5.5 Microscopic investigation by PEEM and LEEM...... 73 5.6 Conclusion...... 80

6 Quasi-free-standing graphene on non-polar surfaces 83 6.1 Experimental details...... 83 6.2 Investigation of the interface...... 84 6.3 Stacking of the graphene layers...... 87 6.4 The band structure...... 89 6.5 Morphology and thickness distribution of graphene on non- polar SiC...... 90 6.6 Conclusion...... 93

7 Summary 95

8 Zusammenfassung 99

Bibliography 105

List of publications 121

vi 1 Introduction

Graphene is a carbon allotrope with the carbon atoms being sp2 hybridized and densely packed into a honey comb lattice. The strong σ bonds between the sp2 orbitals lead to a true two-dimensional structure, while the remaining pz orbitals govern the electronic behavior. As a result, graphene exhibits extraordinary thermal, mechanical, and electrical properties [Gei09, Gei07]. Due to graphene’s linear band structure near the Fermi energy, the electrons propagating through the layer lose their effective mass. As a consequence, they are rather described by a Dirac-like equation than the Schrödinger equation. In addition, the fact that graphene is two-dimensional makes the electron waves propagating within graphene accessible to a large number of surface science probes. It also offers the possibility to use graphene as a sensor material for gases or liquids. The outstanding electronic quality of graphene leads to a quantum hall effect at room temperature [Nov07] and allows ballistic transport on a micrometer-scale [May11]. The first isolation of graphene was achieved by exfoliation of graphite [Nov04]. Using this method, graphene layers are peeled off from a graphite crystal with the help of a sticky tape. Monolayer graphene obtained in this way exhibits high structural and electronic quality. The graphene flakes produced this way can reach up to millimeter size [Gra13]. Only the graphite, the sticky tape, an oxidized silicon wafer as substrate and an optical microscope to find the monolayer flakes are necessary to exfoliate high quality graphene. Therefore, exfoliation is still the technique of choice for basic research and for making proof-of-concept devices. It is, however, a very time-consuming method with delicate handcraft, which renders it unfeasible to produce graphene by exfoliation on an industrial scale. Among several other methods, growth of epitaxial graphene on silicon carbide (SiC) by thermal decomposition is well suited for a wafer-scale production [Emt09]. It has long been known that SiC surfaces tend to form

1 Chapter 1: Introduction carbon rich reconstructions, if annealed at elevated temperatures [Van75], something that was often regarded as carbon contamination in SiC studies. Later, it was shown that ultra-thin films of a few graphene monolayers [For98, Cha02] or even monolayer graphene [Ber04] can be grown. In most studies, either the SiC(0001) or SiC(0001) surface was used as a substrate for graphene growth. Both surfaces are opposite to each other, so the orientation in the crystal is the same. They are, however, terminated by different elements. SiC(0001), also called Si-face is terminated by silicon atoms, while SiC(0001), the C-face is terminated by carbon atoms. This leads to a completely different graphene growth on those two planes. On the C-face, it is relatively easy to grow a thin film of 10 − 20 graphene monolayers. The graphene sheets inside the film have a high degree of rotational disorder leading to the decoupling of the graphene sheets [Has08]. Thus, the graphene sheets behave like monolayer graphene and exhibit high charge carrier mobilities of up to 250000 cm2/Vs [Orl08]. In order to use graphene in integrated devices, means to manipulate the charge carrier concentration are necessary. Usually, the carrier concentration is varied by shifting the Fermi energy via an electric field by applying a gate voltage. In a multi-layered material such as few layer graphene on SiC(0001), the outer layers screen the inner layers from the electric field. Thus, the charge carrier concentration in multi-layered graphene can not be varied enough by applying a gate voltage. This fact renders multi-layered graphene unsuitable as material for field effect devices. Although, one group claims that monolayer graphene on the C-face can routinely be grown [Wu09], it is still very challenging for other groups to obtain high quality monolayer graphene on this surface. In contrast, preparation of monolayer graphene on the Si-face is easily controllable. The growth was recently improved by annealing the SiC in argon atmosphere [Emt09]. This procedure leads to the formation of about 1 µm wide terraces covered with monolayer graphene and small stripes of bilayer graphene at the step edges. The different growth conditions on the Si-face are connected to the carbon rich reconstruction which forms at the√ interface√ upon annealing of the substrate. The reconstruction exhibits a (6 3 × 6 3)R30° periodicity and resides√ between the substrate and√ the graphene layers. Thus, it is often called 6 3 or buffer layer. The 6 3 possesses the same honey comb structure as graphene, but is covalently bound to the substrate [Emt08].

2 Lacking the π bands of graphene, the buffer layer is electrically inactive. Nevertheless, it reduces the charge carrier mobility in the graphene layers on top of it [Spe11a]. In addition, donor states are introduced by the buffer layer that effectively dope graphene [Der11, Kop10], also affecting the carrier mobility. A popular method to prepare buffer layer free graphene on the Si-face is intercalation of foreign elements between the SiC and the buffer layer. Vari- ous elements such as gold [Pre09], lithium [Vir10], silicon [Xia12], fluorine [Wal11], germanium [Emt11], oxygen [Oid10] and hydrogen [Rie09] have been used as intercalants in recent years. The intercalant modifies the interface, breaking the bonds between the SiC and the buffer layer. Hence, the intercalation results in a decoupling of the buffer layer which is converted into graphene. For a hydrogen intercalated buffer layer the interaction between the graphene and the substrate is negligible, so it is often termed quasi-free- standing graphene. In this work, novel methods to prepare quasi-free-standing graphene are explored. In chapter2, background information on the properties of silicon carbide, graphene and epitaxial graphene growth on SiC is given and discussed in more detail. Subsequently, the surface science methods which were applied in this work to characterize the obtained material are introduced (chapter3). Those methods include X-ray photoelectron spectroscopy (XPS), angle-resolved photoemission spectroscopy (ARPES), Raman spectroscopy and low-energy electron microscopy (LEEM). In chapter4, the intercalation of oxygen is investigated by means of XPS to study the processes at the interface and the buffer layer. The quality of the oxygen intercalated graphene is examined by Raman spectroscopy and ARPES. In chapter5, an alternative way to oxidize the interface by annealing in water vapor is discussed. In addition to XPS, ARPES and Raman spectroscopy, LEEM was performed to determine the effect on morphology. In chapter6, the possibility to directly grow quasi -free-standing graphene without intercalation is explored. The omission of an intercalation step does not only reduce the preparation time, but also eliminates a possible degeneration of the graphene by the intercalation. For that purpose, graphene growth on the non-polar SiC surfaces (1100) and (1120) is investigated by XPS, ARPES and LEEM. Both surfaces are perpendicular to the Si- and C-face and also to

3 Chapter 1: Introduction each other. Finally, chapter7 summarizes the results of this work and o ﬀers an outlook for further investigations.

4 2 Basics

2.1 Silicon carbide (SiC)

Silicon carbide is a compound consisting of silicon (Si) and carbon (C) in equal parts. As shown in fig. 2.1a, every carbon atom is bound to a silicon atom in a tetrahedral configuration and vice versa [Jär98]. A SiC crystal can be regarded as being build up from layers. Since those layers consist of one layer Si and one layer C, they are referred to as SiC bilayers (fig. 2.1b). There are two possibilities to stack SiC bilayers onto each other. The bilayers can either have the same orientation or they are rotated by 60° [Jär98]. In the unrotated case, the layers need to have a small offset to be placed onto each other. Therefore, the bonds of the two bilayers are staggered as depicted in fig. 2.1c. In the rotated case, the orientation of the bonds of the top bilayers coincides with the one of the bottom layer. This configuration is called eclipsed (fig. 2.1d). Stacking bilayers in the staggered arrangement leads to the zincblende structure, and stacking in the eclipsed arrangement to the wurtzite structure [Jär98]. This is equivalent to the possible stacking in the close-packing of equal spheres leading to a face-centered cubic (fcc) or hexagonal close-packed (hcp) structure. In SiC, ordered sequences of mixed stacking are also possible and lead to over 250 different SiC polytypes [Fis90]. The polytypes are named after their symmetry C for cubic, H for hexagonal and R for rhombohedral and a number which gives the length of the stacking sequence [Ram47]. Crystal planes and directions in the crystal are described by Miller indices in round and square brackets, respectively. For directions, the indices are the components of a vector in the basis of the lattice vectors that points in the specified direction. The length of the vector is chosen in a fashion that all indices are integers. In the case of describing planes, the indices are defined by the reciprocal values of the axis intercepts of the specified plane. As a

5 Chapter 2: Basics

Figure 2.1: Schematic illustration of the atomic structure of SiC. (a) The tetrahedral bond configuration, (b) a SiC bilayer, and the two possibilities to stack SiC bilayers: (c) staggered and (d) eclipsed. Silicon atoms are drawn in red, carbon in gray. Partial bonds to atoms which are not drawn are omitted for clarity. consequence, the [hkl] direction is orthogonal to the (hkl) plane in a cubic basis. This is, however, not the case for any given system of base vectors. In hexagonal and rhombohedral structures, commonly, four basis vector are used. The vectors a1,a2,a3 are coplanar as depicted in fig. 2.2a, which of course makes them linear dependent. The vector c is orthogonal to a1,a2,a3. Consequently, also four indices h,k,i,l are used to describe directions and planes in hexagonal and rhombohedral structures. The linearly dependency leads to the relation i = −(h + k). Figure 2.3 shows the stacking sequences for 3C-, 4H- and 6H-SiC. The crystal structure is shown along the [0001] and [1100] direction. The bonds are projected into the (1120) plane. For simplification, 3C-SiC is treated

6 2.1 Silicon carbide (SiC)

Figure 2.2: (a) Illustration of the hexagonal unit cell of 4H-SiC. Scheme of a 4H-SiC crystal viewed (b) in [1120] and (c) in [1100] direction. The polar SiC surfaces (0001) and (0001) are highlighted in red and blue. The nonpolar SiC surfaces (1100) and (1120) are highlighted in green and yellow, respectively. Atom positions from [Vil06]. here as a rhombohedron, with the cubic [111] direction becoming hexagonal [0001]. In this projection, the Si (or C) atoms of one bilayer can sit in one of three sites: A, B or C. The stacking sequence for 3C-SiC is ABC, for 4H-SiC it is ABCB and ABCACB for 6H-SiC.

7 Chapter 2: Basics

Figure 2.3: Atomic structure of the SiC polytypes (a) 3C-SiC, (b) 4H-SiC, and (c) 6H-SiC. Silicon atoms are drawn in red, carbon in gray. The yellow line illustrates the stacking sequence of the SiC bilayers and the square brackets show the height of the unit cell.

In this work, graphene was grown and studied on 6H-SiC(0001), 4H- SiC(1100) and 4H-SiC(1120). Therefore, these surfaces are presented in more detail. Commercially available SiC wafers are cut orthogonal to the c-direction. This means one side of the wafer has the crystal orientation (0001) and the other (0001). While SiC(0001) is terminated with silicon atoms and therefore often referred to as the Si-face, SiC(0001) is carbon terminated and hence referred to as the C-face. Most of the epitaxial graphene is grown either on the Si- or the C-face, because of the commercial availability of wafers with these surface orientations. Of rather minor interest (so far) for graphene growth are the planes (1100) and (1120). They are orthogonal to each other and to the (0001) plane as illustrated in fig. 2.2a. The atomic structure of those surfaces is shown in fig. 2.2b and fig. 2.2c for 4H-SiC. In the top, the (0001) plane is highlighted in red. The termination with Si (dark red balls) is clearly visible. Likewise, one can see the C termination of the (0001) plane on the bottom in blue highlighting. The difference in

8 2.1 Silicon carbide (SiC) the two surfaces means that they are polar, which results in a very different graphene growth on them. The (1100) and (1120) surfaces are nonpolar. This is a consequence of the sixfold symmetry of hexagonal SiC making (1100), (1010), (0110), (1100), (1010) and (0110) equivalent surfaces. These are the surfaces of the SiC hexahedron shown in fig. 2.2a. This example shows that the notation with four indices is useful for symmetry operations. A sixfold symmetry operation is the product of a threefold symmetry and a mirror plane. The threefold symmetry allows cycle permutation of the first three indices. Inverting the first three indices, as required by the mirror planes, produces the other three equivalent surfaces. Of course, this is also true for the (1120), (1210), (2110), (1120), (1210) and (2110) surfaces. The atomic structure of the (1100) plane (and equivalent planes) is shown on the sides of fig. 2.2b highlighted in green. The surface structure of the (1100) plane is determined by the polytype. As shown in fig. 2.4a, there are two types of (1100) surfaces for 2H-SiC. Type I surfaces have one dangling bond per surface atom and are, therefore, energetically more favorable than type II surfaces, which have two dangling bonds per surface atom [Rau99]. For 4H-SiC, there are three possibilities to cut the bulk to create a (1100) surface, which are denoted A, B, and C in fig. 2.4b. The cuts through B and C yield a surface with two type I and two type II surface atoms per unit cell. Theoretical calculations showed that the ideal surface is formed if the crystal is cleaved at position A, which creates a surface with only type I surface atoms [Rau00, Rau01]. Experimental investigations of hydrogen terminated 4H-(1100) also determined surface A to be the ideal one [Sey05]. As shown in fig. 2.4b, the surface is rough and exhibits carbon terminated facets and silicon terminated facets. For 6H-SiC, the size of the facets and therefore also the surface roughness is greater [Rau01]. In contrast, there is only one type of (1120) surface for both hexagonal polytypes, which is shown on the sides of fig. 2.2c highlighted in yellow. The surface consists of parallel chains of Si and C atoms with one dangling bond each. Thus, the relaxed (1120) surfaces are significantly smoother than the (1100) surfaces [Rau01]. From those structural differences, one can expect a different graphene growth on the four presented surfaces. This will be discussed in detail in chapter6.

9 Chapter 2: Basics

Figure 2.4: Schematic side view of the SiC(1100) surface of 2H-SiC and (b) 4H-SiC. (Adapted from [Rau01]). (a) For 2H-SiC, there are two diﬀerent types of surfaces. (b) For 4H-SiC, there are three possibilities (denoted A, B, C) to create a (1100) surface.

2.2 Graphene

Graphene is a 2-dimensional carbon allotrope. All carbon atoms are sp2 hybridized leading to a planar configuration. In theoretical considerations, graphene has been a building block for other carbon allotropes for a long time. Graphite consists of many graphene layers, which are held together by Van der Waals interaction and stacked to form a 3-dimensional structure. Carbon nano tubes can be considered as a rolled up graphene sheet forming a quasi 1-dimensional system. Likewise, one can think of a carbon fullerene to be a graphene sheeted wrapped into a ball. With that in mind, it is not surprising that graphene has long been an interesting subject for theorists. Wallace [Wal47] calculated the electronic structure of graphene already in 1947. However, it took another 57 years until graphene was first isolated by Novoselov et al. [Nov04, Nov05] in 2004. Via a scientific search engine such as Google Scholar [Sch13], the number of hits for the keyword “graphene” can be determined. They are proportional to the amount of published articles and the scientific activity. In fig. 2.5, the number of hits is plotted for each year from 1990 to 2012. On a linear scale (fig. 2.5a), a drastic increase after 2005 can be seen. Plotted on a logarithmic scale, an exponential increase would result in a linear dependency.

10 2.2 Graphene

(a) (b) hits per year hits in 1000 per year

date date

Figure 2.5: Number of hits of “graphene” on Google Scholar [Sch13] per year from 1990 to 2012 plotted with (a) linear and (b) logarithmic axis.

In ﬁg. 2.5b the positive curvature even shows an over exponential growth. Hence, graphene research is still far away from its peak. Many more advances can be expected for the future in graphene research.

2.2.1 Crystal structure The sp2 hybridization of the carbon atoms in graphene means that the 2s and two of the three 2p orbitals (without loss of generality px and py) hybridize to three energy degenerate orbitals. The remaining pz plays an important role in the electronic properties of graphene, which will be discussed below (section 2.2.2). The three planar sp2 orbitals deﬁne the crystal structure by forming σ bonds with their neighbor carbon atoms. The resulting crystal structure is drawn in ﬁg. 2.6a. The bond length amounts to 1.42 Å and there is an angle of 120° between two σ bonds. The base vectors a √ a √ a = ( 3,1) and a = ( 3,−1) (2.1) 1 2 2 2

11 Chapter 2: Basics

Figure 2.6: (a) Crystal structure of graphene. The base vectors a1 and a2 deﬁne the unit cell. The two sublattices corresponding to the two atoms in the unit cell are labeled with A and B. (b) First Brillouin zone of graphene with reciprocal lattice vectors b1 0 and b2. High symmetry points are labeled with Γ, M, K and K .

of this so-called honey comb lattice deﬁne the unit cell (shaded√ in gray in ﬁg. 2.6a). The lattice constant a amounts to a = |a1| = |a2| = 3 · 1.42Å = 2.46Å [Sai98]. The unit cell contains two carbon atoms leading to a crystal structure with hexagonal sublattices A and B. By solving the system of equations

ai · b j = 2πδi j , (i, j ∈ {1,2}) (2.2) the reciprocal lattice vectors 2π 2π 2π 2π b1 = ( √ , ) and b2 = ( √ ,− ) (2.3) 3a a 3a a

can be obtained. Vector b1 is perpendicular to a2, and b2 to a1 as demanded by eq. 2.2. Those reciprocal base vectors describe the lattice translation in k space, i.e. they point from one reciprocal lattice point to an adjacent one. They also define the first Brillouin zone (BZ). The first Brillouin zone is the reciprocal equivalent to the Wigner–Seitz primitive cell. It is constructed by drawing lines from a reciprocal lattice point

12 2.2 Graphene to its closest neighbors. At the midpoint of these connections a perpendicular line is drawn. The smallest area enclosed by the perpendicular lines is the BZ. In the case of graphene, the BZ is hexagonal as displayed in ﬁg. 2.6b. The BZ has high symmetry points, which are labeled: Γ for the middle of the BZ, M for the midpoint of the hexagon and adjacent corners are alternately labeled K and K0. Labeling three corners K0 expresses that they correspond to the other sublattice as the three corners K. They are connected by time reversal symmetry [Cas09]. The high-symmetry axes can be used to reduce the information of the BZ to a few cuts through the BZ. It is common to display a calculated (s. ﬁg. 2.7) or measured (s. section 3.1) band structure along those high-symmetry cuts. Figure 2.7 shows for example the graphene band structure from Γ to M, M to K and K to Γ.

2.2.2 Electronic properties The ﬁrst band structure calculations of graphene were performed by Wallace [Wal47] in 1947. He investigated the electronic properties of graphite. It was, however, easier to consider just one layer of graphene. Later, interactions between graphene layers were included to improve the model of graphite [McC57, Slo58]. Wallace used a tight binding approximation considering only nearest neighbor interactions. A more accurate band structure can be calculated by means of density functional theory (DFT). The calculated band structure of graphene displayed in ﬁg. 2.7[ Lat06] exhibits three σ bands, which are formed by the sp2 orbitals. The three anti-bonding σ∗ are not visible in this energy range. The remaining ∗ pz orbitals overlap forming bonding and anti-bonding states, the π and the π bands. The electron dispersion relation E(k) of the π and π∗ bands obtained from tight binding are given by [Sai98]: ±tω(k) E±(k) = (2.4) 1 ± sω(k) s √ 3k a ky a ky a ω(k) = 1 + 4 cos x cos + 4 cos2 (2.5) 2 2 2

13 Chapter 2: Basics

Figure 2.7: Band structure of graphene calculated by density functional theory. Reprint with permission from [Lat06]. Copyright (2006) by The American Phys- ical Society.

The transfer integral t and the overlap integral s can be ﬁtted to experimental data to obtain t = −3.033eV and s = 0.129. The positive solution describes the bonding π band, the negative the anti-bonding π∗ band. The π and π∗ bands are touching where E±(k) = 0; at the so-called Dirac energy ED. For E±(k) to be zero, ω(k) has to be zero. This is the case in the corners of the BZ 0 at the K and K points. In neutral graphene, the Fermi energy EF lies at ED. For this reason, EF is often referenced to ED to describe doping of graphene in units of eV. In the vicinity of the K and K0 points, the free charge carrier dispersion relation can be approximated by [Cas09]

E± = ±hv¯ F |k|. (2.6)

The positive sign describes electrons and the negative holes. The dispersion relation has a striking resemblance to the photon dispersion relation E = | | 3 ta hc¯ k . For the free charge carriers in graphene, the Fermi velocity vF = 2 h¯ corresponds to the velocity of light c. |k| is the magnitude of the wavevector k with respect to the regarded K or K0 point.

14 2.2 Graphene

For this reason, charge carriers in graphene are called Dirac fermions and can be described by the Dirac equation with the Hamiltonian [Gei07]:    0 k − ik  ˆ  x y · H = hv¯ F   = hv¯ F σˆ k (2.7) kx + iky 0 The wavevector k is referenced to the K point and σˆ is the 2D Pauli matrix. Electronic states near ED are composed of states corresponding to different sublattices. The Pauli matrix describes a pseudospin, which is necessary to index sublattice A and B in the two-component wavefunction. The pseudospin is conserved, which gives rise to a chiral symmetry. This supresses intravalley back scattering k → −k. Therefore, intravalley scattering is only possible if the chiral symmetry is broken by i.e. surface ripples, atomic- sized defects or boundaries [Ki08]. However, for charge carriers further away from the Dirac Energy ED, the Dirac approximation becomes less accurate. This effect is called trigonal warping, because the circular Dirac cone transforms into a three-fold symmetry further away from ED.Effects of trigonal warping have been shown to give rise to a small but nonzero probability of back scattering [And98]. Furthermore, also intervalley scattering is suppressed at room temperature. To scatter a charge carrier from one Dirac cone to the next one, the momentum transfer has to be quite high. At room temperature, phonons cannot provide this momentum transfer. As a consequence of these two effects, graphene has a very high elastic mean free path of up to 600 nm [Ber06] and a high charge carrier mobility of up to 200000 cm2/Vs measured in suspended graphene [Bol08].

2.2.3 Synthesis of graphene In this section, the most commonly used procedures for graphene synthesis are described.

Exfoliated graphene Monolayer graphene was ﬁrst isolated in 2004 by exfoliation from graphite [Nov04]. The starting material is a 3-dimensional graphite crystal in form

15 Chapter 2: Basics

of highly oriented pyrolytic graphite (HOPG), which is a synthetic graphite formed by cracking hydrocarbons at high temperature and subsequent annealing, often under application of pressure [Dre02]. A sticky tape is applied onto the HOPG. By removing the tape, the HOPG sheet crystal is cleaved and a fresh, clean surface is created. Carefully rubbing the surface against a substrate results in exfoliation of graphene layers, which remain on the substrate [Nov05]. Finding the graphene on the substrate is not trivial, as this method does not only produce monolayer graphene but also graphite flakes with more layers, which stay behind on the substrate. To detect the monolayers one needs the means to distinguish monolayer graphene from all the other thicker flakes. To facilitate this, silicon oxide is used as a substrate. In an optical microscope, light passes through the graphene layer and is reflected at the silicon oxide/silicon interface. A part of the reflected light is reflected at the silicon oxide/graphene interface and so forth. In this way, light can travel through the oxide layer various times. The light that is transmitted through the graphene layer interferes with light that traveled a different path length. Adjusting the thickness of the oxide layer accordingly to the wavelength of the incident light, leads to a high contrast in the optical microscope for monolayer graphene. For white light, i.e. without any filters, best contrast is achieved with a silicon oxide thickness of 90 nm or 280 nm [Bla07]. The size of the graphene flakes is usually smaller than 2000 µm2. The largest monolayer flake found so far had an area of 1500000 µm2 [Gra13]. This and the time-consuming searching for monolayer graphene renders the mechanical exfoliation method unfeasible for the electronic industry. For scientific research, however, exfoliation is an excellent starting point, because it produces high quality graphene with very low acquisition costs. Furthermore, chemical exfoliation [Par09] is another method to isolate graphene layers. Within this huge area of chemical research, graphene layers are functionalized while they are still stacked in form of graphite. The functionalized graphene sheets can then be brought into solution creating colloidal suspensions. After deposition of the graphene flakes on a substrate, the functional groups can be removed generating similar results as mechanical exfoliation, albeit with higher defect densities in the graphene layers. Since this is a strictly chemical approach, this work will not go into further detail.

16 2.2 Graphene

Graphene on metals Graphene growth on transition metals [Win09] has been demonstrated for a variety of substrates: ruthenium [Him82, Sut08], iridium [Kho84, Cor08], cobalt [Ham80], rhenium [Gal85], nickel [Eiz79, Obr07, Rei09], platinum [Ham80, ZP87], palladium [Ham80, Lan92], and copper [Li09, Mat11]. De- pending on the substrate, graphene growth is either a chemical vapor deposition (CVD) or a growth by suface segregation of carbon solved in the bulk. In CVD, hydrocarbons, e.g. ethylene [Lan92], methane [Li09] or hexane [Sri10], are decomposed and deposited on the surface. For surface segregation, introduced carbon diﬀuses into the bulk of the substrate at high temperatures. Upon cooling, graphene is formed by the carbon which segregates at the surface. One of the most interesting metal substrates is copper foil. It is relatively cheap, obtainable in pieces with very large areas, and it is thin making it feasible to dissolve the foil in acid. Graphene is grown at temperatures up to 1000 ◦C by CVD in a mixture of methane and hydrogen [Li09]. Being able to dissolve the copper foil afterwards leaves this method as the most promising candidate for large area transferable graphene growth.

Epitaxial graphene Epitaxial graphene is grown by thermal decomposition of silicon carbide (SiC) at elevated temperatures. Already in 1975, Van Bommel et al. [Van75] observed the graphitization of SiC upon annealing in vacuum. Later, it was demonstrated that ultra-thin graphite films of a few graphene monolayers can be grown by this method [For98, Cha02]. Ideally, the SiC is etched in hydrogen prior to graphitization on the one hand to remove polishing damage and on the other hand to obtain a well defined starting point. This leads to an ordered, atomically flat surface [Ram98]. Annealing SiC at T ≥ 1150 ◦C breaks the Si–C bonds at the surface. The carbon forms graphene, while the silicon sublimates due to its higher vapor pressure. Three bilayers of SiC have to be consumed to have sufficient carbon for one graphene layer. Emtsev et al. [Emt09] improved the method by annealing in an argon atmosphere instead of vacuum. This introduces a probability that sublimated silicon

17 Chapter 2: Basics

gets reﬂected back to the surface by argon atoms hindering graphitization. To accomplish the same graphitization rate in argon atmosphere the temperature is raised to about 1650 ◦C. This leads to an increase in the mobility of the particles of the surface. The continuing sublimation and condensation of particles, in particular at the SiC step edges results in step bunching meaning that several steps grow together to one larger step. Thus, the width of the terraces in the SiC substrate that were obtained by the hydrogen etching increases. The samples grown in this way exhibit large areas of monolayer graphene with small stripes of bilayer graphene at the step edges. Areas with three or more layers of graphene are rarely observed. The method is suitable to grow monolayer graphene on a wafer scale and is used throughout this work. The properties of graphene on SiC are discussed in more detail in section 2.3.

2.3 Epitaxial graphene

Epitaxial graphene (EG) is grown by thermal decomposition of SiC at elevated temperatures. Usually, the (0001) or (0001) surfaces of 4H- or 6H-SiC are used as substrates. Graphene growth on (0001) is still very challenging because it is very hard to obtain just one monolayer of EG. Thus, most proof- of-concept electronic applications use EG grown on (0001) [Gu07, Lin11]. For the same reason, the (0001) surface was not used in this work. Graphene was grown on 6H-SiC(0001), 4H-SiC(1100) and 4H-SiC(1120). The growth on the nonpolar surfaces (1100) and (1120) is discussed in chapter6, so the discussion in this section is limited to graphene on (0001) (Si-face). On√ the Si-face,√ the carbon layer that forms closest to√ the substrate is a (6 3 × 6 3)R30° reconstruction (further on termed 6 3). It possesses the same honey comb structure as graphene, but is covalently bound to the substrate [Emt08]. Therefore, this so-called buffer layer lacks the π bands of graphene. When the growth continues, a second carbon layer forms at the interface [Han11]. This second, new layer now becomes the buffer layer, while the first layer√ is relaxed. The outermost layer, which now resides on top of the new 6 3 buffer layer, has the properties of graphene. It is referred to as monolayer graphene (MLG). The next layer would again form at the

18 2.3 Epitaxial graphene interface, becoming the new buffer layer with two graphene layers residing on it, and so on. The buffer layer underneath MLG leads to a strong√ electron doping (n ≈ 1 × 1013 cm−2) of the latter [Der11, Kop10]. The 6 3 is only partially bound to the√ silicon atoms of the substrate. About one third of the carbon atoms of the 6 3 is sp3 hybridized and bound to Si atoms of the substrate. The other two thirds are sp2 hybridized and only bound in a single plane. Due to the lattice mismatch, some of the Si atoms of the surface are expected to be not saturated by the buffer layer. Those Si atoms introduce dangling-bond bands within the SiC bandgap. Moreover the 2pz orbitals of the buffer layer, which originate from C atoms that are not bound to the substrate, give rise to surface states within the SiC bandgap. The strong electron doping is most plausibly explained by electron transfer from the interface to the graphene layer. The intrinsic carrier concentration in MLG affects the carrier mobility µ. Pristine MLG with n ≈ 1 × 1013 cm−2 has a mobility of around 2000cm2/Vs at T = 25K [Emt09, Job10]. Doping MLG with an overlayer of tetrafluoro- tetracyanoquinodimethane (F4-TCNQ) to n ≈ 1 × 1011 cm−2 yields a carrier mobility of 29000 cm2/Vs at T = 25K [Job10]. This is similar to what is observed for exfoliated graphene on SiO2 close to charge neutrality [Nov05, Zha05]. Moreover, the charge carrier mobility of MLG at room temperature is also decreased by its temperature dependency [Emt09]. The mobility at a charge carrier concentration of n ≈ 1 × 1013 cm−2 decreases from around 2000 cm2/Vs at T = 25K to values of around 900 cm2/Vs at room temperature, even when atomically flat Hall bars are prepared [Job10]. It has been observed that this is a direct consequence of the presence of the buffer layer [Spe11a].

3 Experimental methods

3.1 Photoemission spectroscopy

In photoemission spectroscopy (PES), the photoelectric eﬀect is used to investigate the electronic structure. The sample is illuminated by monochromatic photons which cause the sample to emit electrons, the so-called photoelectrons. The photon is fully absorbed, therefore, transferring its entire energy to an electron bound in the sample. If the energy of the photon is high enough, the photoelectron can leave the sample and can be analyzed. The photoelectrons carry information about their former binding energy EB in their kinetic energy Ekin. This is due to the energy conservation of the incident photon energy

h¯ω = Ekin + EB + ΦSpec. (3.1)

The work function Φ is defined as the difference of the Fermi energy EF and the vacuum level Evac. The sample and the spectrometer are connected, so they have the same EF. The kinetic energy is measured in the spectrometer and therefore with respect to Evac of the spectrometer. For this reason the work function of the spectrometer ΦSpec has to be taken into account instead of the work function of the sample. This means that the work function can be calibrated by measuring photoelectrons originating from states with known binding energy, e.g. from the Fermi edge or from the Au4f core level of a gold foil. Figure 3.1 shows the schematic drawing of the photoelectric effect leading to (a) photoelectrons and (b) Auger electrons. By emitting the photoelectron, the photon creates a hole, which can be filled in two ways. Either an electron of a higher orbit fills it, losing its excess energy by emitting a photon, or this energy can be transferred to a third electron, the Auger electron. In the drawn case, the hole is created in the 1s orbital of the K shell. It is filled by

21 Chapter 3: Experimental methods

Figure 3.1: Schematic drawing of the (a) photo effect and (b) Auger effect. Electrons within different orbitals (1s, 2s, 2p) are represented by solid circles, holes by hollow circles.

an electron of the 2p orbital of the L shell. The emitted Auger electron also originates from the L shell. Such a transition is called KLL. Auger electrons are analyzed in Auger electron spectroscopy. This method was, however, not utilized in this thesis. Auger electrons are brieﬂy mentioned here, because they are also visible in the photoemission spectra. In this thesis, two forms of photoemission spectroscopy have been utilized: X-ray photoelectron spectroscopy (XPS) and angle-resolved photoemission spectroscopy (ARPES). In XPS, the photoelectrons of the electron core levels are investigated. High energy photons from X-ray sources have to be used to excite core level electrons with binding energies in the order of keV. In ARPES, the valance band electrons are studied. Here, sources with a lower photon energy such as a helium lamp or a synchotron have to be used. Valance band electrons also have a wave vector k dependence, which is given by their dispersion relation E(k), or band structure. In ARPES, also the emission angle of the photoelectrons is analyzed to determine their wave vector k. Conservation of momentum before and after photon absorption states that 0 kphoton + kelectron = k electron. (3.2)

With kphoton kelectron, the momentum of the photon kphoton can be neglected, which means that the wave vector of the electron is not signiﬁcantly changed during photo excitation. However, the wave vector changes when the electron

22 3.1 Photoemission spectroscopy

Figure 3.2: Due to the potential diﬀerence inside and outside the sample, the photoelectron is re- fracted. The parallel component of the the wave vector kk is preserved, while the perpendicular com- inside outside ponent changes from k⊥ to k⊥ . leaves the sample. In this refraction process, the change occurs only in k⊥, the wave vector component perpendicular to the surface. The parallel component kk is preserved and can be obtained by measuring the kinetic 2 outside 2 energy Ekin = h¯ (k ) /(2m) and the polar angle θ of the photoelectron outside outside the sample (s. ﬁg. 3.2). With kk = k sinθ, the parallel component then amounts to √ 2mEkin kk = sinθ. (3.3) h¯ In ARPES, the binding energy EB of the states in the valence band is plotted against the wave vector kk to obtain the band structure.

Analyzer In XPS and ARPES, the kinetic energy of the photoelectrons is analyzed. In addition, the polar angle θ also needs to be determined in ARPES. Both can be accomplished with a 180° hemispherical energy analyzer. Figure 3.3 shows a schematic drawing of a hemispherical energy analyzer in XPS mode. The photoelectrons from the sample can enter the lens system from a relatively large acceptance area. In the lens system, they get refocused to the entrance slit and decelerated by the retarding potential Uret. The electric ﬁeld in the hemisphere deﬂects the retarded electrons. If their kinetic energy matches the pass energy Epass, they are imaged at the center of the detector. Electrons with a lower Ekin appear towards the inner side, electrons with higher Ekin towards the outer side of the hemisphere. The range of this energy window can be set r by changing Uhem and with that changing Epass = e · d Uhem. A spectrum can r be recorded by scanning Uret and by that scanning Ekin = e · ( d Uhem − Uret). In the second dimension of the detector, usually the angular distribution is imaged. The acceptance angle can be varied by changing the potentials of

23 Chapter 3: Experimental methods

Figure 3.3: Schematic drawing of a hemispherical energy analyzer. Highlighted in yellow are the paths of photoelectrons with the same kinetic energy Ekin originating from different positions on the sample. The electrons enter the lens system under different angles, get refocused and decelerated by the retarding potential Uret. Due to the potential difference (2Uhem) between the outer and inner hemisphere, only r electrons with Ekin = Epass − e · Uret = e · ( d Uhem − Uret) reach the detector.

the electron lenses in the lens system. The result is a 2D data set in Ekin(θ), which can be expressed in EB(kk) using eq. (3.1) and eq. (3.3). EB(kk) is a cut through the Brillouin zone and the sample orientation to the analyzer determines where the measurement cuts the BZ. In some setups it is possible to tilt the sample during the measurement by an angle β. This leads to a 3D data set in Ekin(θ,β), which has to be processed to obtain the band structure in EB(kx,ky).

Detector To detect electrons both energy and angular resolved at the same time, a 2D display detector is used. This is a combination of channel plate, a phosphor

24 3.1 Photoemission spectroscopy screen and a CCD1 image sensor. The channel plate works like an electron multiplier. It has many micro channels where the photoelectrons get multiplied by cascades, so the electrons stay spatially resolved. After ampliﬁcation, the electrons are accelerated toward the phosphor screen where they lead to the emission of light. A CCD camera then records the data by imaging the phosphor screen. Core level photoelectrons have no k dependence except for diﬀraction processes. Therefore, only the energy resolution is needed for XPS. If a 2D detector is used, the second dimension of the detector is integrated to record a spectrum I(Ekin). In a pure XPS setup, an electron multiplier or an array of them is used instead.

Data analysis

The binding energy EB of electrons is determined by the atomic number of the atoms and by the quantum number of the speciﬁc state. As a consequence, the spectrum of photoelectrons exhibits discrete lines at binding energies typical for the elements present in the sample. Figure 3.4 shows the XPS spectrum I(EB) of a graphene sample intercalated by water vapor (s. chapter5). The prominent lines labeled O1s, C1s, Si1s, Si2p originate from the corresponding core levels. The satellite peaks appearing at higher binding energies than the core levels are photoelectrons which lost kinetic energy due to plasmonic excitations. The peak labeled O KLL originates from Auger electrons of a KLL transition of oxygen. The valance band (VB) is hardly visible because the cross section is quite low using a photon energy of h¯ω = 1486.7eV. The photon source was a monochromated Al Kα x-ray gun. The areas under the peaks are analyzed to determine the element composi- tion of the surface. For this, the inelastic background has to be subtracted from the spectrum. Then, the areas can be compared by taking the cross sections of the involved core levels into account. This way, the percentage of oxygen on the surface can for example be measured. A more detailed analysis of a core level line reveals that the binding energy is also inﬂuenced by the so-called chemical shift. This means that the expected

1charge-coupled device

25 Chapter 3: Experimental methods

Figure 3.4: XPS survey spectrum of a graphene sample intercalated by water vapor (s. chapter5). Elements can be identified by the binding energy of their core levels. Additional to the core levels O1s, C1s, Si1s and Si2p, also the Auger peak O KLL and the valance band (VB) are labeled. binding energy for a certain element is shifted due to the bonding configuration of the element. The main cause is the different electronegativity of atoms taking part in the bond. Therefore, by studying the chemical shift, the atomic structure can by analyzed. The chemical shift is often of the same order of magnitude as the resolution limit of the experimental setup meaning that the components are not resolved in the experimental data. In such a case a fit to a physical model can achieve a deconvolution of the components. Thus, all components and the background of inelastic scattered photoelectrons have to be described mathematically. The background increases in steps at the peak positions, due to the increasing number of photoelectrons available for inelastic scattering after each peak. It is referred to as a Shirley background [Shi72]. The symmetric components are described by Voigt lines [Arm67]. The Voigt function V(x) is a convolution of a Lorentzian function L(x) and a Gaussian function G(x), where x describes the energy dependence. Z V(x) = (G ∗ L)(x) = G(τ)L(x − τ)dτ (3.4)

26 3.1 Photoemission spectroscopy

2 2 e−x /(2σ ) γ G(x) = √ , L(x) = (3.5) σ 2π π(x2 + γ2) Here, σ describes the standard deviation of the Gaussian distribution and γ the half width at half maximum of the Lorentzian distribution. The areas of the three functions V(x), G(x), L(x) are normed to one. The Lorentzian term accounts for the lifetime broadening due to the decay processes of the excited hole state during photoemission. The Gaussian term accounts for normal distributed statistic errors of the experimental setup and inhomogeneities of the sample. This includes the limited resolution of both, the incident photons and the analyzer. Inhomogeneities in the sample can lead to differences in the chemical environment of the elements, which affects the binding energy and, therefore, broadens the peaks. In metals, like graphene, the creation of deep localized holes in the photoemission process is followed by a drastic rearrangement of the electrons near the Fermi energy [Gad75]. In this rearrangement electron-hole pairs are generated resulting in an energy loss of the photoelectron. Since the probability of the pair generation decreases for higher energy loss (further away from the Fermi energy), the metallic core levels are asymmetric with an low-energy tail. The asymmetric components are best described by convolution of the Voigt function and a Mahan lineshape [Mah75] Θ(x) ξ α B(x) = e−x/ξ, (3.6) Γ(−α)x x where Γ is a gamma function and Θ is the step function. The fit parameters α and ξ describe the asymmetry. The convolution with the Voigt function Z M(x) = (B∗ V)(x) = B(τ)V(x − τ)dτ (3.7) assures that the broadening effects mentioned above are taken into account. Figure 3.5 shows a high resolution spectrum of the C1s core level of epitaxial graphene on SiC. There are carbon atoms in four different bonding configurations leading to four components in the XPS spectrum. As described in section 2.3, monolayer graphene resides on a buffer layer which is partly bound to the SiC substrate. The experimental data are shown as hollow circles

27 Chapter 3: Experimental methods

Figure 3.5: XPS C1s core level spectrum of a epitaxial graphene sample. The high resolution of the experimental data (circles) enables to distinguish several contributions of carbon atoms in different bonding configurations or environments. The components of the buffer layer (red and yellow), graphene (blue) and bulk SiC (green) are extracted by fitting a model curve (black line) to the data. in fig. 3.5. The model curve is fitted to the data and shown as a black line, which represents the sum of the components contributing to the C1s core level signal. The most prominent contribution (green) stems from carbon atoms from the SiC substrate. The S1 (red)√ and S2 (orange) components are attributed to carbon atoms from the 6 3 buffer layer with S2 being bound only in plane and S1 also being bound to the substrate. The asymmetric component (blue) is based on carbon atoms from the graphene layer.

3.2 Raman Spectroscopy

Raman spectroscopy utilizes inelastic scattering of photons. A sample is illuminated by a laser and the scattered light is analyzed using a suitable spectrometer. In the case of a solid, inelastic scattering involves the creation

28 3.2 Raman Spectroscopy

(Stokes (S) process) or annihilation (anti-Stokes (AS) process) of a phonon. Conservation of energy yields

h¯ωL = h¯ωS ± h¯Ω, (3.8) with the frequency of the incoming photon ωL, of the scattered photon ωS and of the phonon Ω. The (+) corresponds to the Stokes, and the (−) to the anti- Stokes process. In addition, the phonon concentration in a solid is dependent on the temperature, given by the Bose-Einstein statistics. Thus, also the ratio of the Stokes and anti-Stokes intensities is temperature dependent [Kit13] I(AS) = exp(−h¯Ω/k T). (3.9) I(S) B In this way, the sample temperature can be determined by measuring the Stokes and anti-Stokes lines. However, for all other applications in Raman spectroscopy, it is sufficient to only measure the more intense and narrower Stokes line. Therefore, the following concepts are only discussed for the Stokes process. First order Raman scattering can be regarded as a three step process, which is depicted in fig. 3.6. In the general case (fig. 3.6a), a photon of the laser with the energy h¯ωL is absorbed and excites the electron system from the ground state E0 to a virtual, excited state. In the second step, a phonon is created decreasing the energy of the electron system by h¯Ω to a second virtual state. In the third step, the system relaxes to the ground state E0 by emitting a scattered photon with the energy h¯ωS. If the energy of either virtual state matches an actual electron state En, the Raman process is resonant and the scattering cross section is enhanced. This is either called incoming resonance (fig. 3.6b), if h¯ωL = En − E0 or outgoing resonance (fig. 3.6c), if h¯ωS = En − E0. In a Raman spectrum, the intensity of the inelastic scattered light is plotted over its energy loss. For historic reasons the energy scale is not in eV but in −1 1 E wavenumbers ν¯ with the unit cm . With ν¯ = λ = hc , the conversion factor E/ν¯ amounts to 1.24 × 10−4 eV/cm−1. Figure 3.7 shows the Raman spectrum of a hydrogen intercalated buffer layer sample as an example. In addition to the conservation of energy, conservation of momentum

kL = kS + K (3.10)

29 Chapter 3: Experimental methods

Figure 3.6: Schematic drawing of Raman scattering (a) without resonance. The system is excited by an incoming photon h¯ωL to a virtual state, relaxes to a second virtual state by creating a phonon h¯Ω, and reaches the ground state E0 by emitting a scattered photon h¯ωS. (b) Incoming resonance, meaning that there is a real state En with En − E0 = h¯ωL. (c) Outgoing resonance, where the energy of the scattered light h¯ωS = En − E0.

has to be taken into account for the scattering process as well. Here kL is the wave vector of the incoming laser light, kS of the scattered light, and K of the phonon. In a 180° backscattering process (kL = −kS), the transfer of momentum is maximized, and the wave number of the phonon becomes |K| = 2|kL|. For a laser within the visible spectrum, this is in the order of 4π/λ ≈ 0.02nm−1 = 2 × 107 m−1. Compared to the size of the Brillouin zone of graphene 2π/a = (2π/2.46)Å−1 ≈ 3 × 1010 m−1, the momentum of the photon is quite small. Therefore, only phonons with very small K from the vicinity of the Γ point of the phonon dispersion relation are involved in the ﬁrst order Raman scattering. In other words, the expected amplitude in the Raman spectrum is proportional to the density of states in the phonon dispersion relation around the Γ point. Especially for graphene, this is, however, not the case. There are several reasons for that: the selection rule, resonant

30 3.2 Raman Spectroscopy

Raman scattering, and second and third order Raman processes, which are all discussed in the following. In order to understand the contributions to the Raman spectrum of graphene, the dispersion relation of the phonons has to be taken into account. Figure 3.8 shows the calculated phonon dispersion relation of graphene, adapted from [Laz08] and [Mal09]. As discussed in section 2.2, the unit cell of graphene contains two carbon atoms, A and B. This means that there are six phonon branches: three acoustic branches (A) and three optic branches (O). The optic or acoustic excitations can either be parallel or perpendicular the A-B bond. If the excitation is parallel to the A-B bond it is denoted in-plane longitudinal (iL). If the excitation is perpendicular to the A-B bond, it can either be in-plane transversal (iT) or out-of-plane transversal (oT). By only looking at the Γ point of the dispersion relation, one would expect a contribution at around 850 cm−1 from the oTO band. However, this phonon branch is not Raman active, because of the selection rule due to the odd symmetry for the mirror operation on the graphene plane [Sat11].

31 Chapter 3: Experimental methods

Figure 3.8: Phonon dispersion relation of graphene. (Adapted from [Laz08] and [Mal09]). The optical (O) and acoustic (A) phonon branches in the direction in-plane (i) and out-of-plane (o), longitudinal (L) and transverse (T) are shown.

The iLO and iTO branches are degenerate at the Γ point and, therefore, contribute to the Raman spectrum with only one signal at around 1580 cm−1. This signal is called the G band or G line and is present in Raman spectra of all graphene-like materials such as the spectrum of hydrogen intercalated buffer layer shown in fig. 3.7. The G line process involves the excitation of an electron from the cone-shaped electron dispersion relation of graphene at the K point, as depicted in fig. 3.9a. In undoped graphene the lower Dirac cone is fully occupied and the upper cone fully unoccupied. This way, there is always an occupied state Ei and an unoccupied state En with En − Ei = h¯ωL for incoming resonance. Of course, there is also outgoing resonance for other states with Em − E j = h¯ωS. However, the G line signal is not manly determined by the resonant contributions since they interfere destructively with each other. This was demonstrated by blocking part of the resonant

32 3.2 Raman Spectroscopy pathways through high doping of the graphene, which resulted in an increase of the G line intensity [Che11]. The most important Raman processes in graphene are schematically shown in fig. 3.9. The incoming laser light is indicated by a blue arrow and the scattered light in red. Inelastic scattering under creation of a phonon is displayed by dashed arrows while elastic scattering at defects by dotted arrows. The processes (b)-(e) involve phonons with K , 0 and can not be deduced from the Γ point of the phonon dispersion relation. Therefore, second order processes are needed to fulfill the conservation of momentum. In the D band process (fig. 3.9b), a phonon of the iTO branch around the K point is created [Mal09]. This can only happen if the charge carriers can be scattered intervalley from one Dirac cone to the adjacent one by defects in the graphene. For this reason a D band only appears in defective samples and its intensity compared to the G band is a measure for the defect density [Can11]. Since the electron-hole pair creation and the elastic scattering at the defect are resonant, this is a double resonant process. Very similar to the D band is the D0 band (fig. 3.9c). A iLO phonon is created and the transfer of momentum is compensated by resonant scattering at a defect. This time the scattering is intravalley, e.g. within a Dirac cone. This means that the defect has to break chiral symmetry and by comparing the D and D0 peak intensities, the nature of the defect can be studied [Eck12]. The overtone of the D band is the 2D band (fig. 3.9d, e). It is sometimes also called G0 band, because it is also present in pristine graphene, while the D band is associated with defects. However, the involved phonons are the same as in the D band, so 2D is the more accurate nomenclature. In contrast to the D band, the conservation of momentum is achieved by creating a second phonon with K2 = −K1. Thus, no defects are necessary for the activation of the 2D band. The process can either be double resonant (fig. 3.9d), if the electron is scattered by both phonons and then recombines. Or the process can be triple resonant (fig. 3.9e), if the electron and the hole are both scattered by each of the two phonons in a resonant scattering [Mal09]. Then, electron and hole recombine at the adjacent Dirac cone. The triple resonance may explain the high 2D intensity in graphene compared to the G peak. In this work, Raman spectroscopy was employed to investigate possible defects in graphene. Due to the resonant nature of the Raman processes

33 Chapter 3: Experimental methods

Figure 3.9: Raman processes in graphene. (Adapted from [Mal09]). Blue arrows represent the incoming light which creates an electron-hole pair. Red arrows indicate the scattered light. The dashed arrows show inelastic scattering under creation of a phonon, dotted arrows indicate elastic scattering at a defect. More detailed information can be found in the text.

34 3.3 Low-energy electron microscopy in graphene, Raman spectroscopy does not only probe phonon states, but also enables to probe the electronic system of graphene. Variations in the electronic properties due to strain, doping, magnetic ﬁelds or defects all aﬀect the positions, widths or intensities of the Raman peaks [Fer13].

3.3 Low-energy electron microscopy

3.3.1 Introduction Low-energy electron microscopy (LEEM) uses elastically backscattered electrons with energies in the order of 10 eV for imaging [Bau07]. The development was driven by the wish to combine low-energy electron diffraction (LEED) with an imaging system, e.g. transmission electron microscopy (TEM). The main difference between TEM and LEEM is the energy of the electrons used for imaging. In TEM, typical electron energies range from 100 to 300 keV. Those high-energy electrons are transmitted through thin samples. The low energy (10 eV) in LEEM makes even thin samples nontransparent for the imaging electrons. Consequently, a backscattering geometry must be used like it is applied in LEED. There, the backscattering geometry is realized by an electron gun that emits electrons through a hole in a fluorescent screen used for detecting the diffraction pattern. With this setup, it is possible to image diffraction. For imaging the sample surface, however, the incident and the emitted electron beams need to be separated. Depending on the setup, a beam seperator deflects both the incoming and outgoing electrons by 60° [Tel85] or 90° [Tro98] separating the two beams by 120° or 180°, respectively. What TEM and LEEM have in common is the ability to display diffraction spots and use them for imaging. For simplification, the principle is shown in fig. 3.10a for a TEM instrument. If the sample is illuminated by a plane wave, the diffraction image is located in the back focal plane of the objective lens. An intermediate lens images the back focal plane into the object plane of the projective lens which produces the final diffraction pattern on the detector screen. In imaging mode (fig. 3.10b), a contrast aperture is inserted in the back focal plane of the objective, limiting the angular acceptance to normal emission. In addition, the intermediate lens is turned off, so that the image

35 Chapter 3: Experimental methods

Figure 3.10: Principle of (a) diffraction and (b) imaging in a transmission electron microscope. (Adapted from [Bau98]). By turning off an intermediate lens and adding a contrast aperture in the back focal plane of the objective, the microscope is switched from diffraction to imaging mode. of the sample is produced on the detector screen by the projective lens. By moving the contrast aperture, tilting the incoming electron beam, or tilting the sample, it is also possible to use diffraction spots other than the specular (0,0) spot for imaging. This is referred to as dark field imaging, as opposed to bright field imaging where the specular (0,0) reflection is used for imaging.

3.3.2 The instrument The following section describes the setup of a LEEM instrument using the example of the Specs FE-LEEM P90, which was used in this work. Figure 3.11

36 3.3 Low-energy electron microscopy shows the path of the electrons from the electron gun via the sample to the viewing screen. The electrons originate from a cold field emission gun with a low energy spread of ≈ 0.3eV. The potential difference between the electron gun and the ground level is usually kept at −15kV to accelerate the electrons into the condenser column. The gun lens and gun stigmator make sure that the electrons reach the sample as a plane wave. This can be checked by looking at the diffraction spots. They are only imaged as sharp points if the sample is illuminated by a parallel beam. The condenser lenses and deflectors are used to adjust the shape, position and incident angle of the electron beam. To select a different diffraction spot than the specular (0,0) spot for dark field imaging, best practice is to use the condenser lenses and deflectors to tilt the electron beam and shift the desired diffraction spot into the optical axis. The beam separation is accomplished by a magnetic prism array, which deflects the electron beam by 90° towards the sample. The sample is kept at a potential difference of −15kV + US to the ground. Thus, the electrons are decelerated from 15 keV to eUS between the objective lens and the sample. The high potential difference and the small distance of the objective lens and the sample results in high electric field in the order of 100 kV/cm. This requires very flat and clean sample surfaces to avoid the degradation of the sample by arc discharge. The backscattered electrons are accelerated by the objective lens. As shown in fig. 3.10 for TEM, the diffraction image is located in the back focal plane of the of the objective lens. However, it would be impractical to insert an aperture at this position because the illuminating electrons would also have to pass through there. Instead, a transfer lens images this plane to the energy filter and through the beam separator into the projector column. Here, the diffraction image is reproduced in the plane of the contrast aperture, which selects one diffraction spot for imaging. Of course, the aperture can also be completely removed to view the whole diffraction image. For dark field imaging, it is also possible to move the aperture to the desired diffraction spot instead of tilting the illumination electron beam. However, moving the aperture away from the optical axis enhances lens errors in the projector column, so it is better to tilt the electron beam. For materials with small lattice constant, and therefore, large reciprocal vectors like graphene,

37 Chapter 3: Experimental methods

Figure 3.11: Schematic layout of a SPECS FE-LEEM P90. (Adapted from [SPE11b]). the electron beam cannot be tilted far enough. Then, moving the aperture and tilting the beam can accomplish dark field imaging. The lenses of the projector column control the magnification of the sample image. By changing the voltage of the electrostatic electron lenses, the magnification can be varied. In the schematic illustration (fig. 3.11), the intermediate lens is turned off so that the image of the sample is produced on

38 3.3 Low-energy electron microscopy the viewing screen. Just like in the example of the TEM, the lens can be used to produce the diﬀraction image on the detector. The viewing screen is very similar to an ARPES 2D detector described in section 3.1. It is a combination of a channel plate to multiply the electrons, a ﬂuorescent screen to make them visible, and a CCD camera to record them and allow a computer analysis of the measured data.

3.3.3 Contrast mechanisms A step-free homogeneous sample surface would be unattractive to study in a microscope, as it has no contrast, which results from areas with diﬀerent LEEM intensities. The mechanisms that determine the intensity and therefore the contrast are manifold and are discussed in the following.

Reflectivity contrast In TEM, the strong forward scattering of fast electrons can be described by the first Born approximation. At low energies the approximation fails and has to be replaced by more accurate method like the partial wave analysis [Bau98]. The incoming plane wave and the scattered wave are expanded into spherical harmonics centered at the atom. The phase differences ηl between the incident and the scattered partial waves are calculated. The scattering amplitude in the nonrelativistic case is given by

∞ 1 X f (θ,k) = (2l + 1)exp(2iη − 1) P (cosθ), (3.11) 2ik l l l=0 where k is the wave number and Pl are the Legendre polynomials. The 2 intensity distribution of the scattered electrons√ is ∝ | f (θ,k)| and depends on the scattering angle θ and the energy E ∝ k. In condensed matter, where electrons are not scattered by single atoms but also by the surrounding atoms, the 180° backscattering amplitude is determined by the band structure E(k) perpendicular to the surface [Bau07]. A high backscattering amplitude or electron reflectivity is obtained if the electron is reflected before it is attenuated significantly by inelastic scattering.

39 Chapter 3: Experimental methods

Figure 3.12: Sketch of the relation of the electron reﬂectivity in normal incidence to the band structure perpendicular to the surface. (Adapted from [Bau07]). A higher density of states (DOS) creates dips, a lower DOS creates peaks in the reﬂectivity curve.

Of course, this is the case for electrons with E < Evac because they are reflected before they enter the sample. Figure 3.12 sketches the relationship between the band structure and the electron reflectivity. If an electron matches the energy of a band gap, there are no allowed states in the crystal and it forms an evanescent wave. Since the extinction length of this electron wave is much smaller than the inelastic mean free path, it is reflected with high probability. The situation is similar where the bands are steep and therefore the density of states (DOS) is small. Consequently, energies where the DOS is high can be observed as dips in the reflectivity curve. Thus, the reflectivity intensity in bright field imaging is dependent on the electronic structure. Everything that influences the electronic structure can create a contrast, e.g. an overlayer, different materials or adsorbates.

Step contrast Another contrast mechanism is the phase contrast [Alt98]. It originates from the interference of electron waves with diﬀerent path lengths and, therefore, diﬀerent phases. Phase contrast includes the step contrast [Chu98] and the quantum size contrast [Alt01].

40 3.3 Low-energy electron microscopy

The step contrast arises from the interference of electron waves which are reﬂected from the lower and the upper terraces next to a step. For specular reﬂection, the phase shift φ is given by [Chu98]

φ = kd = (2π/λ)2a0, (3.12) where k = 2π/λ is the electron wave vector, λ is the electron wavelength, d = 2a0 is the path length difference, and a0 is the step height. In the low energy regime of LEEM the electron wave length is in the order of the step height, so a wide range of phase shifts can be observed by making small changes in the electron energy. However, this simplistic approach fails to explain the appearance of a step contrast in LEEM. Therefore, more sophisticated calculations where made by Chung and Altman [Chu98]. In their model, they calculate the step contrast as the interference of the Fresnel diffracted waves from terrace edges which meet at a step. It predicts a rich interference phenomenon caused by the surface step with oscillations of LEEM intensity extending over the upper and lower terrace. However, only the strongest features of the oscillations located immediately adjacent to a step are observed in the experiment. This is due to the limited beam coherence as a result of energy spread of the source and lens aberrations. Their calculations also predict that the interference pattern is asymmetric in step sense for a defocused image. This offers the possibility to identify which terrace is the lower and which terrace is the upper one.

Quantum size contrast The second phase contrast, the quantum size contrast, arises from interference inside a thin film. Similar to a Fabry-Perot interferometer, the electron waves reflected at the top surface interfere with electron waves that passed the thin film. The phase shift φ is given by the path length difference,

φ = k0d = k0(2t), (3.13) where k0 is the wave vector in the thin film, d = 2t is the path length difference, and t is the film thickness. For a thin film with a constant inner potential V0,

41 Chapter 3: Experimental methods the free electron dispersion can be adopted and the phase shift may be written as 2t p φ = 2m(E + V ), (3.14) h¯ 0 with E being the electron energy of the illumination beam. This simplistic model demonstrates the key feature of the quantum size effect (QSE), that the reflected intensity oscillates as a function of both energy and film thickness. A film with n atomic layers shows exactly (n − 1) QSE interference peaks between consecutive Bragg peaks. Or in other words, it shows n dips for n atomic layers. This allows to study the thin film structure by measuring the reflectivity in dependence of the electron energy. The thickness of the film can then be determined by counting the dips.

3.3.4 Imaging modes Brigt field and dark field The imaging modes bright and dark field differ in the electrons which are used for imaging. In bright field, only electrons from the specular (0,0) diffraction spot can pass through the aperture and are imaged onto the viewing screen. The contrast is given by the mechanism described above (section 3.3.3). If any other diffraction spot is used, the method is called dark field imaging. As explained in section 3.3.2, this is usually not realized by moving the aperture, but by tilting the illumination beam. This way, the chosen diffraction spot stays on the optical axis, which minimizes spherical aberration in the projector column. In dark field imaging, the contrast is mainly governed by the azimuthal properties of the sample. A good example is the (2 × 1) reconstruction of the Si(100) surface. The reconstruction alternates form (2 × 1) to (1 × 2) on consecutive steps. This is sketched in fig. 3.13a as rows and columns. The resulting LEED pattern shown in fig. 3.13b exhibits spots which originate from the “row” domains (blue), spots from the “column” domains (red) and spots which are present in both (gray). If a gray spot is used for imaging such as the specular (0,0), the only visible contrast originates from the steps (fig. 3.13c). In the dark field image fig. 3.13d, the domains corresponding to

42 3.3 Low-energy electron microscopy

Figure 3.13: The two domains of the (2 × 1) reconstruction of the Si(100) surface are sketched in (a); ordered in rows or columns. The diffraction pattern (b) is a superposition of the patterns from the two domains. Spots indicated in blue (red) originate from the row (column) formation, spots indicated in gray are present in both. LEEM (c) bright field images exhibiting only step contrast and dark field images using 1 1 (d) the (0, 2 ) and (e) the ( 2 ,0) diffraction spot. All images were recorded at the same sample position. the used diffraction spot are bright, while the other domains are dark. As a consequence, the contrast is inverted for the other diffraction spot (fig. 3.13e).

µ-LEED In conventional low energy electron diffraction (LEED), the area of the illuminated spot is in the order of 1 mm2. Consequently, the diffraction image is always a superposition of different domains present on the surface. In LEEM, the area illuminated by the electron beam is in the order of 1000 µm2 or 10−3 mm2. In µ-LEED, the spotsize is even smaller. An aperture in the

43 Chapter 3: Experimental methods

magnetic prism array limits the field of view. This way, depending which hole in the aperture is used, the probed area is in the order of 1 µm2. This is often small enough to image the diffraction pattern of a single domain on the surface. The low intensity of the diffraction image due to the small area is compensated by the channel plate voltage and longer exposure times of the CCD camera.

PEEM Photoemission electron microscopy (PEEM) is actually a method of its own. It can, however, be easily realized in a LEEM instrument and it is invaluable for aligning the LEEM projector column. In PEEM mode, the electron beam is turned off or blanked by a shutter. Instead, the sample is illuminated by an UV source such as a Hg lamp. Due to the photoelectric effect, photoelectrons are emitted, as described in section 3.1. The objective lens accelerates those photoelectrons and they are imaged just like the backscattered electrons in LEEM. With the broad spectrum of a Hg lamp, the PEEM contrast is governed mainly by differences in the surface potential due to work function differences. The alignment of the Hg lamp is relatively easy and it illuminates the sample over a wide area. This makes PEEM extremely useful when aligning the projector column of a LEEM instrument [Tro93].

Low-energy electron reflectivity spectroscopy With the help of the computer based data acquisition, it is possible to auto- matically record a series of n LEEM images with different electron energies. Usually, this is done for bright field images. Such a data set of n images each having x · y pixels can also be regarded as xy low-energy electron reflectivity (LEER) spectra with n data points each. This means, if a series of images is taken, the LEER spectrum of every pixel in the field of view is recorded. This is a huge advantage compared to scanning probe microscopy, where scanning probe spectroscopy is only performed at selected points of the microscopy image.

44 3.3 Low-energy electron microscopy

3.3.5 LEEM and epitaxial graphene LEEM is a powerful technique to investigate graphene [Man12]. The LEER spectra of epitaxial graphene contain n dips for graphene with a thickness of n monolayers [Oht08, Hib08]. This effect is similar to the quantum sized effect (QSE) discussed above. The difference is that the QSE occurs in a material with a constant inner potential V0. Of course, this is not the case in multi layered graphene, where the inner potential is different inside and between the layers. The commonly accepted explanation of the dips in the reflectivity spectra of graphene was proposed by Hibino et al.[Hib08]. As previously discussed in section 3.3.3, the scattered electron is reflected with high probability, if there are no matching states at the corresponding energy creating peaks in the reflectivity curves. Dips form, if the scattered electron couples to a state, so the electron can be attenuated by inelastic scattering. In the interpretation of Hibino et al.[Hib08], these states can be derived from the graphite band structure in ΓA direction. Epitaxial graphene only has a few layers, so the band splits into discrete energy levels. Incident electrons with a matching energy can couple to those states, which increases the transmission probability creating dips in the reflectivity curve. This way,√ n minima are obtained for n layers of graphene without counting the 6 3, because of its modified band structure [Emt08]. This interpretation has recently been challenged by Feenstra, Srivastava et al. [Fee13a, Fee13b, Sri13]. They claim that inter-layer states between graphene layers are responsible for the dips. The interpretation√ gives rise to n − 1 dips for n graphene layers, however, this time the 6 3 is counted, because it is graphene-like. For graphene on SiC(0001), the new interpretation does not change the graphene thickness. The results are the same if n dips are matched to n graphene√ layers, or if n − 1 dips are matched to n carbon layers including the 6 3. After conversion of the buffer layer to graphene, however, the two interpretations predict a differing number of dips for the same graphene thickness. LEER spectra of water vapor treated graphene samples are discussed in section 5.5 to test both interpretations.

4 Intercalation of oxygen

The intercalation of foreign atoms between a graphene layer and its substrate is known for a long time. Kholin et al. [Kho84] intercalated cesium and potassium underneath graphene on iridium, which was still called monolayer graphite instead of graphene at that time. Other examples are the intercalation of cesium, potassium, and sodium [Nag94] or gold [Shi00] underneath monolayer graphite on nickel. All three studies observed that the electronic properties of the topmost graphite layer became similar to those of bulk graphite upon intercalation. Or in other words, the intercalants weakened the interaction between the monolayer graphite and the substrate. In analogy, the intercalation of foreign atoms is also a viable route to create buffer layer free graphene on the SiC(0001) surface. In recent years, various elements such as gold [Pre09], lithium [Vir10], silicon [Xia12], fluorine [Wal11], germanium [Emt11], oxygen [Oid10, Mat12, Oli13] and hydrogen [Rie09, Spe11a, Spe10, Rob11, Tan12] have been used as intercalants. Fig- ure 4.1 illustrates the intercalation of hydrogen under MLG. By modifying the SiC interface, the buffer layer is decoupled and converts into graphene. Therefore, monolayer graphene, which resides on the buffer layer is converted into bilayer graphene. By starting with one carbon layer less, i.e. just the buffer layer, monolayer graphene is obtained after the intercalation process. To preserve the intrinsic electronic properties of graphene, it is desirable that the interaction of the intercalated elements and the graphene layer is minimal. Therefore, the non-metallic intercalants such as hydrogen, oxygen, and fluorine are of special interest. They form strong covalent bonds to the Si atoms of the SiC surface, leading to bonding and anti-bonding states far away from the Fermi energy. Thus, the substrate stays insulating and the charge carrier transport is limited to the graphene layer. In the case of hydrogen, the intercalation works exceptionally well. It has been shown that hydrogen-treated buffer layer transforms into monolayer

47 Chapter 4: Intercalation of oxygen

Figure 4.1: A schematic picture of the intercalation of hydrogen. (Adapted from [Spe11a]). (a) In pristine√ monolayer graphene (MLG), the graphene√ layer resides on the buﬀer layer (6 3). The lattice mismatch between the 6 3 and the SiC substrate leads to the formation of dangling bonds (db). (b) After hydrogen intercalation, the√ SiC interface is passivated by hydrogen, which leads to the decoupling of the 6 3 resulting in quasi-free-standing bilayer graphene (QFBLG).

√ graphene [Rie09]. Eliminating the 6 3 in this way improves the electrical properties when compared to normal MLG which resides on the buffer layer. Hydrogen intercalation leads to graphene with almost no temperature dependence of the mobility [Spe11a], a pronounced quantum hall effect [Tan12], and an overall enhancement in the performance of high frequency graphene transistors [Rob11]. For these reasons, hydrogen-treated buffer layer is often referred to as quasi-free-standing monolayer graphene (QFMLG). In this chapter, the intercalation of oxygen is studied. It was first observed with graphene on ruthenium by Sutter et al. [Sut10]. The graphene was grown by carbon segregation from Ru(0001) after exposition to ethylene at high temperature [Sut08]. LEEM and micro-sized spot ARPES (µ-ARPES) indicated a decoupling effect of the oxygen treatment. At the same time, Oida et al.√[Oid10] have investigated the intercalation of molecular oxygen under the 6 3 of SiC. They oxidized the SiC interface in ◦ molecular oxygen in a low-temperature√ process (T = 250 C, p = 1atm, t = 5s). A 3 Å thick oxide layer below the 6 3 was evident from ion-scattering and XPS. In low-energy electron diffraction (LEED), the diffraction spots related to the buffer layer weakened, while the first order graphene spots prevailed. Electron energy loss spectroscopy (EELS) showed π-plasmons at E = 6.5eV

48 4.1 Experimental details after the√ oxygen treatment, which are related to graphene and do not occur in the 6 3. XPS also showed a transition in the C1s core level components without changing the over all area under the C1s core level spectrum. Oida et al. [Oid10] concluded√ that no carbon has been lost due to etching and that the carbon layer of the 6 3 turned into undamaged graphene. In the following, oxygen intercalation was carried out to perform additional studies by means of XPS, ARPES and√ Raman spectroscopy. ARPES can unambigously show the transistion of 6 3 to graphene via the band structure, while Raman spectroscopy is well suited to investigate oxygen induced defects in graphene-like materials. The procedure and the main results were published in [Ost12].

4.1 Experimental details

Buffer layer samples were grown by annealing of SiC(0001) at elevated temperatures in argon atmosphere. The 6H-SiC wafers were purchased from SiCrystal and cut in 5 by 10 mm samples. To remove possible organic and metallic contaminations, the samples were cleaned by a wet chemical procedure, as described in [Sie02]. This includes subsequential baths for 10 min in acetone and iso-propanol, during sonication, followed by four steps. Step A consists of 10 min in a 4:1 mixture of sulfuric acid (H2SO4) and hydrogen ◦ peroxide (H2O2) at 180 C, step B of 5 min in 5 % hydrofluoric acid (HF) at room temperature (RT), step C of 10 min in a 4:1:1 mixture of water, H2O2 and hydrochloric acid (HCl) at 80 ◦C, followed by step D of 5 min in 5 % HF at RT. Each step is followed by rinsing the samples in deionized water. In order to remove polishing damage and to create atomically flat terraces in the SiC, the samples were etched in hydrogen after chemical cleaning. In this process, the samples are heated to about 1500 ◦C in 1 bar of hydrogen atmosphere under a constant flow of 0.5 slm1 hydrogen. This procedure was performed in the same furnace as the graphene growth in argon atmosphere. Previously, the optimal etching conditions in the furnace were investigated and a computer automation was programmed. The details of the hydrogen etching

1standard-liter per minute

49 Chapter 4: Intercalation of oxygen and the furnace automation are described in [Ost09, Ost10]. Graphitization was carried out in the same furnace in an argon atmophere of 1 bar. Annealing the SiC substrate for 15 min at 1475 ◦C led to the formation of the buffer layer. Oxygen treatment was realized in two different approaches: (a) in situ annealing at higher temperatures (about 750 ◦C) and lower oxygen pressure (about 10−4 mbar), or (b) ex situ annealing at lower temperatures (about 270 ◦C) and high oxygen pressure (100-1000 mbar). In the high-temperature approach, the sample was heated to about 750 ◦C for 30 min under constant oxygen flow. The molecular oxygen was inserted trough a leak valve and guided to the sample via a tube, resulting in a partial oxygen pressure at the sample which is about two orders of magnitude higher than the chamber pressure. This offers the possibility to carry out the oxygen treatment under ultra high vacuum (UHV) conditions in a chamber with a base pressure of 2 × 10−10 mbar. During the oxygen treatment the chamber pressure increases to 7 × 10−7 mbar by the incoming oxygen, so the partial oxygen pressure at the sample can be estimated to be about 7 × 10−5 mbar or 5 × 10−5 torr. Since the sample is annealed in oxygen for 30 min, this results in an oxygen dose of 5 × 10−5 torr · 30 · 60 × 106 µs = 9 × 104 L (Langmuir). The low-temperature oxygen treatment was carried out in a furnace which can be evacuated to about 10−2 mbar and was subsequently filled with oxygen gas. The furnace consists of a hot wall glass reactor which is heated by halogen lamps. Both, the temperature and the oxygen pressure was varied to optimize conditions. Based on the ARPES spectra, the best results were obtained at a temperature of 270 ◦C and a 200 mbar oxygen atmosphere. The oxygen-treated buffer layer samples were investigated by means of ARPES, XPS, and Raman spectroscopy. ARPES measurements were carried out at the synchrotron BESSY II with a toroidal analyzer at beamline U125-2/SGM and with a SPECS PHOIBOS 100 analyzer at beamline UE56- 2/PGM-2. XPS data were also taken with these two setups at BESSY II. Additional XPS measurements were carried out with a SPECS PHOIBOS 150 analyzer and a monochromated Al-Kα light source. Micro-Raman spectra were measured under ambient conditions using a Jobin Yvon T64000 triple spectrometer with a liquid N2 cooled CCD detector. For excitation, a frequency doubled Nd:YVO4 laser with a wavelength of 532 nm was focused on the sample by a 100× objective.

50 4.2 High-temperature treatment

4.2 High-temperature treatment

Figure 4.2 shows the transition of the buffer layer to graphene due to the oxygen treatment at increasing temperatures.√ On the bottom (red), the XPS C1s core level spectrum of the pristine 6 3 is shown. Using a photon energy of 1486.7 eV, the most prominent feature is the bulk SiC component at 283.8 eV binding energy. The feature at the higher binding energies is due to the buffer layer and is discussed below in more detail. After 10 min −5 ◦ of annealing the sample in 7 × 10 mbar of oxygen√ atmosphere at 500 C, a signal between√ the bulk component and the 6 3 feature appears. At the same time, the 6 3 feature decreases in intensity. With increasing temperature in subsequent oxygen treatments of 10 min each, the new component increases while the buffer layer feature further decreases. This trend continues up to a temperature of 750 ◦C, where the new component has its maximum intensity. For 800 and 850 ◦C, the bulk peak broadens and the new component decreases in intensity with increasing temperature, indicating that the layer leading to the new component is etched away at temperatures above 750 ◦C. Furthermore, the position of the bulk component shifts towards lower binding energies for increasing√ process temperature. This is due to a different band bending for 6 3 on SiC, graphene on oxide on SiC, and oxide on SiC after etching of the graphene. √ ARPES of the buffer layer (6 3) samples was measured prior and directly after√ the oxygen treatment. Figure 4.3a shows the ARPES data of the pristine 6 3 along the ΓKM and ΓMΓ directions in the graphene Brillouin zone. It exhibits graphene-like σ bands and discontinous√ states in the regions of the π band, a typical observation for the 6 3 [Emt08]. Additionally, a weak graphene-like π band can be observed. This indicates that a small fraction of the surface is covered by graphene due to slight overheating during sample preparation. The graphene coverage can not be detected by XPS, so it has to be very small (< 5%) and can be neglected. After the high-temperature oxygen treatment at a temperature of 750 ◦C, the band structure changes significantly (fig. 4.3b). A prominent π band with a linear dispersion at the K point is clearly visible. This closely resembles the band structure of uncoupled monolayer graphene. However, a much sharper√π band is expected for undisturbed graphene. The features typical for the 6 3

51 Chapter 4: Intercalation of oxygen

√ Figure 4.2: XPS C1s core level spectra of pristine 6 3 and after high-temperature oxygen treatment for 10 min at various temperatures. Spectra are offset in intensity. disappear, leading to the suggestion that the buffer layer is transformed into graphene. For a more detailed view, a different sample was prepared under similar conditions at the synchrotron facility BESSY II, where it is possible to investigate the sample with lower photon energies leading to an enhanced surface sensitivity. Figure 4.4 shows XPS C1s core level spectra taken (a) before and

52 4.2 High-temperature treatment

Figure 4.3: Band structure measured by ARPES at hν = 65eV of (a) pristine buﬀer layer and (b) after in situ oxygen treatment.

(b) after oxygen treatment at 750 ◦C with a photon energy of 384 eV. Decon- volution of the spectrum by fitting a model curve to the data, as described√ in section 3.1, reveals three Voigt-shaped components in the pristine 6 3 spectrum. With the enhanced surface sensitivity of the synchrotron radiation, the bulk component at 283.8 eV appears the weakest here. The other two components S1 at 285.8 eV and S2 at 285.1 eV can be attributed to carbon atoms of the buffer layer. Atoms which are only bound in plane to other carbon atoms contribute to the S2 signal, atoms which are also covalently bound to the topmost silicon atoms of the substrate contribute to S1 [Emt08]. As above, the buffer layer components are drastically weakened after oxygen treatment and a new asymmetric component at 285.3 eV appears, repre- senting graphene. The graphene component is best fitted by a Mahan line [Mah75], taking the metallic character of graphene into account. In contrast, the buffer layer and bulk components are√ symmetric and are therefore best described by Voigt lines. The remaining 6 3 components after oxygen treatment indicate that the buffer layer was not completely transformed. Again, treatment at temperatures above 750 ◦C results in significant etching of the graphene. There are no CO groups detectable in the XPS spectra, which signals would be expected to appear above 288 eV.

53 Chapter 4: Intercalation of oxygen

Figure 4.4: XPS C1s core level spectra of (a) pristine buﬀer layer, (b) after oxygen treatment at 750 ◦C, and (c) Si2p core level spectrum before and after oxygen treatment.

Figure 4.4c shows the corresponding Si2p core level spectrum before and after oxygen treatment. The spectrum of the pristine buﬀer layer exhibits its maximum at 101.7 eV and shows a double peak structure, which is due to the spin-orbit splitting of the 2p orbital. After the oxygen treatment a broadening of the signal is clearly visible. This indicates that the silicon at the interface is oxidized and in this way the Si–C bonds are replaced by Si–O bonds. The buﬀer layer, which as a result, no longer binds to the substrate transforms

54 4.2 High-temperature treatment

into graphene. This results in a decoupled graphene layer on top of a SiOx interface. The decreasing intensity of the graphene component in the C1s core level spectrum of the oxygen-treated samples at 800 ◦C and above indicates etching of the graphene at these temperatures. In order to investigate if this also hap- pens at 750 ◦C or lower temperatures on a smaller scale, Raman spectroscopy was applied. As described before (section 3.2), Raman spectroscopy is more sensitive to defects than XPS. √ Figure 4.5a shows the measured Raman spectra of the pristine 6 3 and after a high-temperature treatment at 600 ◦C. Both spectra are dominated by the contributions from the SiC substrate. Therefore, it is common practice to subtract the signal of the bare SiC. The difference spectrum is shown on the bottom in fig. 4.5b. The most prominent feature is the D band at 1316 cm−1 followed by the G band at 1565 cm−1. As discussed in section 3.2, the D band originates from a process that creates phonons from the iTO branch around the K point of the phonon dispersion relation. To satisfy the conservation of momentum, the process involves intervalley electron scattering at defects and is, therefore, a good measure for the defect density in graphene. For point- like defects, the mean distance between defects LD can be estimated from the intensity ratio ID/IG of the D and G band [Can11]. This mean distance LD amounts to about 7 nm for buffer layer treated with oxygen at 600 ◦C. Additional defect-induced peaks are the D0 peak at 1595 cm−1 and the D + D0 peak at 2896 cm−1. In the D0 process, a phonon from the iLO branch is created. Similar to the D band, the momentum of the phonon has to be compensated by elastic scattering at defects. For the D0 process, the defect scattering has to be intravalley. The I(D)/I(D0) ratio can be used to classify the type of defect. In the work of Eckmann et al.[Eck12] an I(D)/I(D0) ratio 3 of 7 for vacancy-like defects and 13 for sp hybridization√ is reported. From the Raman spectrum of the oxygen-treated 6 3, an I(D)/I(D0) ratio of 3.3 is deducted. This means that the defects are most likely vacancy-like, e.g. consisting of missing carbon atoms in the graphene lattice. This is in line with the XPS C1s spectrum in fig. 4.4b, which shows no signs for sp3 hybridization of carbon atoms in the form of C–O or Si–C bonds, of course not taking the substrate into account. The D + D0 is a two phonon process which requires also two elastic scattering events, one intervalley and one intravalley. Those

55 Chapter 4: Intercalation of oxygen

Figure 4.5: Raman spectra (a) as measured before and after oxygen treatment and (b) SiC background corrected spectra of oxygen treated sample (lower spectrum) and hydrogen intercalated sample (upper spectrum).

56 4.3 Low-temperature treatment peaks were also observed in with argon ions bombarded exfoliated graphene [Can11]. The 2D peak at 2629 cm−1 is very weak. In pristine monolayer graphene, the 2D peak is the most prominent feature and originates from the double or triple resonant processes in the Dirac cone. Its weakness indicates the degeneration of the transformed graphene by the high defect density. For comparison,√ ﬁg. 4.5b shows the Raman spectrum of a hydrogen intercalated 6 3 sample, oﬀset in intensity. The hydrogen intercalated sample was prepared as described in ref. [Spe10] and exhibited a rather small D peak, a sharp G peak and a prominent 2D peak. The high temperatures of around 750 ◦C used in the processes described this section may be responsible for the introduction of defects. The high temperature was, however, essential for the intercalation process, if the partial oxygen pressure was limited to about 10−4 mbar. Due to the UHV compliant procedure used in this section the oxygen pressure could not be increased. In the experiments described in the next section, the oxygen treatment was carried out in an external furnace to overcome this obstacle.

4.3 Low-temperature treatment

A low-temperature (250 ◦C) and high-pressure oxidation process was proposed by Oida et al. [Oid10] to minimize the introduction of defects to the graphene layer. They state that in their oxidation process in 1 atm of O2 for 5 s, the buﬀer layer is signiﬁcantly etched at 300 ◦C, but undamaged and transformed at 250 ◦C. In our experiments, various temperatures and oxygen pressures were applied to optimize suitable parameters. Judging by ARPES spectra, best results were obtained at a temperature of 270 ◦C and 200 mbar oxygen atmosphere. The low-temperature treatment was carriered out in an external furnace in Erlangen, while the ARPES measurements were conducted at BESSY√ II in Berlin. Samples can be prepared in advance as the oxygen-treated 6 3 is stable under ambient conditions. The ex-situ approach, however, made it unfeasible to measure the exact same samples before and after√ oxygen treatment with ARPES. Nevertheless, the quality of the pristine 6 3 was

57 Chapter 4: Intercalation of oxygen checked by XPS prior to oxygen√ treatment. Figure 4.6a shows the ARPES data of the oxygen-treated 6 3. A graphene-like π band is clearly visible with a local maximum at around 2 eV at the M point and a linear dispersion in the vicinity of the K point. As above in the√ case of the high-temperature process, this indicates decoupling of the 6 3 and its transformation into graphene. However, no improvement in terms of sharpness can be observed in comparison with the high-temperature process. XPS data (not shown) also support the oxidation of the interface and the transformation of the buﬀer layer in graphene. √ Figure 4.6b shows the Raman spectrum of the 6 3 sample after the low- temperature oxygen treatment. The amplitude of the D peak in relation to the G peak is smaller compared to the high-temperature treatment. This means that the intensity ratio ID/IG is smaller, so one would expect the defect density to be smaller than in the high-temperature process. However, the full width at half-maximum (FWHM) of the D peak is much larger as for the high temperature process. As described in ref. [Can11], the ID/IG ratio has its maximum at a mean defect distance LD of 5 nm. If the defect density is higher and the mean distance smaller, the graphene is so degenerated that the D peak broadens and the ID/IG decreases again. From the Raman spectrum in ﬁg. 4.6b, the mean defect distance can be estimated to be around 2 nm. To gain a better understanding of the amount of introduced defects, a defect density can be calculated from the mean distance. Figure 4.7 shows the densest packing of defects with a distance of exactly LD. The area that one defect occupies equals the area of the primitive unit cell of the hexagonal√ 2 lattice. If the defect distance is LD, then one defect would occupy LD 3/2. This means that a mean defect distance of 2 nm results in a defect density of 13 −2 ρD = 3 × 10 cm . The carbon density in graphene is ρC = 2/A, where A is the area of the unit cell (s. section 2.2.1). With a2 √ (2.46Å)2 √ A = |a × a | = 3 = 3 = 5.24Å2, (4.1) 1 2 2 2 ρC becomes −2 15 −2 ρC = 2/A = 0.38Å = 3.8 × 10 cm . (4.2) If we consider the defects to be carbon vacancies, then only about 1 % of the carbon atoms have to be missing to create a mean defect distance of 2 nm.

58 4.3 Low-temperature treatment

√ Figure 4.6: (a) ARPES spectra of the 6 3 after low-temperature oxygen treatment taken at hν = 120eV and (b) corresponding Raman spectrum.

Oida et al. [Oid10] compared the areas of the C1s components in XPS before and after the oxygen treatment. They speciﬁed an uncertainty of 5 % for this procedure. Our Raman spectroscopy measurements show that 1 % carbon loss already degenerates the graphene signiﬁcantly.

59 Chapter 4: Intercalation of oxygen

Figure 4.7: Sketch of a hexagonal lattice, the densest of all possible circle packings. For circles with a radius of L√D/2, the area of the primitive unit cell amounts to 2 LD 3/2.

4.4 Conclusion

Two methods for oxygen intercalation were presented: a high-temperature and a low-temperature approach. To achieve the same degree of decoupling in the low-temperature process, a higher partial oxygen pressure was necessary. ARPES and XPS showed that both oxygen treatments decouple the buﬀer layer. This is not achieved by functionalization of the buﬀer layer√ with oxygen groups, but by oxidation of the SiC interface. Since the 6 3 can no longer bind to the substrate, it is then transformed into graphene. Those results demonstrate that the oxygen intercalation works very similar to the well established intercalation by hydrogen. Raman spectroscopy showed that oxygen intercalation was not possible without introducing a large amount of defects in the graphene. The defects are most likely vacancies which are created due to etching by oxygen. Unfor- tunately, the defect density did not decrease in the low-temperature process as proposed by Oida et al. [Oid10]. As shown by Raman spectroscopy measurements, the introduction of defects can not be excluded by XPS data alone. The creation of defects leaves the oxygen treatment inferior to the already well established hydrogen intercalation process. The ARPES and XPS measurements, however, showed that oxidation of the SiC interface leads to a decoupled graphene layer. Therefore, an alternative way to oxidize the interface is investigated in the next chapter.

60 5 Intercalation of water vapor

In the previous chapter oxygen was intercalated to decouple the buffer layer from the substrate by oxidizing the interface. Since this work was published, two other groups reported on the decoupling of the buffer layer by oxidizing the interface. Mathieu et al.[Mat12] studied oxygen intercalation under monolayer graphene. They used a relative high temperature and lower oxygen pressure in combination with a very long process duration (T = 500 ◦C, p = 10−4 torr, t = 3h). From LEEM, µ-ARPES and µ-LEED measurements, they concluded that the buffer layer is partially decoupled from the√ substrate. Since they started with monolayer graphene residing on the 6 3, oxygen intercalation led to bilayer graphene on an oxidized SiC(0001) surface. In the work of Oliveira et al. [Oli13], MLG was annealed in air at 600 ◦C for 40 min to decouple the buffer layer. The decoupling was unambiguously demonstrated using ARPES, XPS and Raman spectroscopy. They showed that this leads the formation of bilayer graphene with an almost negligibly small density of defects. Efforts to carry out the same process with buffer layer samples, however, lead to a decoupled graphene layer with a much larger defect density.√ They concluded that MLG is much more inert to oxidizing agents than 6 3. Furthermore, the air used in the process was not dried, so the question about the role of water arises. Hence, in this chapter, pure water vapor is used as an oxidant, eliminating gaseous oxygen. In analogy to the previous√ chapter, XPS was performed to determine the decoupling of the 6 3 (section 5.2). Raman spectroscopy was performed (section 5.3) to investigate a possible introduction of defects.√ ARPES was carried out (section 5.4) to study the transformation from 6 3 into graphene via the band structure. In addition, photoemission electron microscopy (PEEM) and low-energy electron microscopy (LEEM)√ were utilized (section 5.5) to investigate the effect of the transition of 6 3 and MLG on the morphology. The main results of this chapter were published in [Ost14].

61 Chapter 5: Intercalation of water vapor

Figure 5.1: Photographs of the vacuum chamber used for the water vapor treatment. In (a) a petri dish ﬁlled with deionized water is placed below the sample heater. (b) By evacuating with high pumping speed the water is frozen and residual gases can be pumped. (c) Sample is annealed with closed pumping valve in water vapor atmosphere.

5.1 Experimental details

The sample preparation prior to oxidation was the same as described in the previous chapter. The SiC wafer was cut, chemically cleaned and etched in a hydrogen atmosphere. In order to grow the buffer layer and MLG, the substrates were annealed for 15 min in 1 bar of argon atmosphere at temperatures of 1475 ◦C and 1675 ◦C, respectively. Please check section 4.1 for more details. Annealing in water vapor was carried out in a dedicated vacuum chamber, which is equipped with a rotary pump that achieves a base pressure of ∼ 10−3 mbar. An open reservoir with deionized water was placed inside the chamber, as depicted in fig. 5.1. The water was frozen by evacuating with high pumping speed. This slowed down the evaporation/sublimation rate of the water enabling the pump to efficiently evacuate the chamber. After about 15 min, the base pressure was reached and the valve to the pump was closed. This allowed the ice to melt and warm up to room temperature leading to a water partial pressure of ≈ 28mbar. The sample was then heated with the help of a conductive heater. The temperature of the sample was held constant for 30 min at 500 ◦C and 650 ◦C for the treatment of buffer layer and MLG samples, respectively.

62 5.2 Decoupling of the buﬀer layer

The samples were investigated by XPS, Raman spectroscopy, ARPES and LEEM. XPS measurements were carried out with a SPECS PHOIBOS 150 analyzer and a monochromated Al-Kα light source. Micro-Raman spectra were measured under ambient conditions using a Jobin Yvon T64000 triple spectrometer with a liquid N2 cooled CCD detector. For excitation, a frequency doubled Nd:YVO4 laser with a wavelength of 532 nm was focused on the sample by a 100× objective. ARPES measurements were carried out at the synchrotron BESSY II with a SPECS PHOIBOS 100 analyzer at beamline UE56-2/PGM-1. PEEM images, LEEM bright ﬁeld images and reﬂectivity curves were recorded by a SPECS FE-LEEM P90 system.

5.2 Decoupling of the buffer layer √ XPS measurements of the buffer layer (6 3) and MLG before and after annealing√ in water vapor are shown in fig. 5.2. The C1s core level spectrum of the 6 3 in fig. 5.2a can be deconvoluted in three symmetric Voigt-shaped components [Emt08]. The most prominent component at (283.85 ± 0.05) eV originates from√ the carbon atoms of the SiC substrate. There are two carbon species in the 6 3, which create the two signals S1 and S2 at (285.0 ± 0.1) eV and (285.65 ± 0.05) eV, respectively. S2 stems from atoms bound only in- plane and S1 from atoms additionally bound to the SiC substrate [Emt08]. For MLG, the carbon atoms of the graphene layer reside on top of the buffer layer, as discussed in section 2.3. This is also reflected in the C1s core level spectrum of MLG shown in fig. 5.2b. Deconvolution√ of the spectrum reveals the same components as in the spectrum of the 6 3 plus an additional, asymmetric component at (284.70 ± 0.05) eV which is labeled G and originates from the graphene layer. After the water vapor treatment, the spectrum of the buffer layer shown in fig. 5.2c exhibits only two components. The larger one at (282.55 ± 0.05) eV can be attributed to the substrate. The smaller one at (284.25 ± 0.05) eV is asymmetric and has similar shape and intensity as the G component of MLG. Thus, it is also labeled G. With the buffer layer components missing and a graphene-like component√ arising after water vapor treatment, this is a strong indication that the 6 3 becomes decoupled from the substrate. The

63 Chapter 5: Intercalation of water vapor

√ Figure 5.2: XPS C1s core√ level spectra of (a) pristine buﬀer layer (6 3), (b) pristine MLG, (c) water-treated√ 6 3, and (d) water-treated MLG. Si2p core level spectra of (e) water-treated 6 3 and (f) water-treated MLG. Spectra are oﬀset in intensity for clarity. Spectrum (f) was taken at at 60° emission angle, all other spectra at normal emission (α = 0°).

64 5.2 Decoupling of the buﬀer layer

√ decoupling leads to the transformation of the 6 3 into graphene as it is the case when annealing the buffer layer in hydrogen [Rie09, Spe10]. ◦ Annealing MLG in water vapor for√30 min at temperatures below 600 C resulted in an only partial decoupled 6 3. Thus, a higher temperature (600 ◦C) was used for MLG than for buffer layer samples. The reason that the intercalation of water under MLG requires a somewhat higher annealing temperature is most likely due to the diffusive nature of the process. The spectrum of water-treated MLG in fig. 5.2d also exhibits two components. The substrate signal lies at (282.90 ± 0.05) eV and the second, asymmetric component G at (284.25 ± 0.05)√ eV. The graphene component is larger compared to the√ water-treated 6 3 sample. The reason for that is that the atoms of the 6 3 transformed into graphene add to the signal of the graphene atoms√ already present before treatment. The XPS C1s core level spectra of 6 3 and MLG annealed in water vapor have a strong resemblance to those of buffer layer and MLG annealed in hydrogen, respectively [Spe10]. √ The Si2p core level spectrum in fig. 5.2e of the 6 3 after water vapor treatment contains two components. To account for the spin-orbit splitting in the Si2p core level, each component is modeled as a Voigt doublet with a splitting of 0.6 eV and an area ratio of 1:2. Deconvolution places the substrate component SiC at (100.31 ± 0.05) eV and shows an additional component at (100.8 ± 0.1) eV. This means that the additional component has a chemical shift of 0.5 eV and can therefore be identified as Si+ [Sie01]. In the case of water-treated MLG, the Si2p core level spectrum shown in fig. 5.2f exhibits three Voigt doublets. In addition to the substrate component at (100.63 ± 0.05) and the Si+ component at (101.1 ± 0.1) eV, there is a component at (102.6 ± 0.1) eV. Although, the component is visible at normal emission, for better visibility, the spectrum measured in a 60° emission geometry is shown. The chemical shift amounts to 2.0 eV, which can be 4+ assigned√ to Si [Sie01]. The wide energy spectra (survey scans) for water- treated 6 3 and MLG exclusively show signals from C, Si, and O. Therefore, + Si can be assigned to C3–Si–O, i.e. to a silicon atom of the topmost SiC bilayer bound to a single oxygen atom. The Si4+ signal can be attributed to O–Si–O or in other words SiO2. From the area ratio of the subtrate and the 4+ Si component the thickness of the SiO2 layer can be determined. It is very thin (≤ 1 monolayer) and is apparently only formed in the case of treated

65 Chapter 5: Intercalation of water vapor

MLG. This effect is probably based on the increased process temperature which was needed to intercalate the water under two carbon layers. So far, only the difference of the shifted components to the bulk component have been considered. The substrate components, however, also shift in position towards lower binding energy after the water vapor treatment. The shift is the same for the bulk components√ in the C1s and Si2p core level spectra and amouts to 1.2 eV for 6 3 and 0.9 eV for MLG. Since the bulk of the substrate is not affected by the surface modification of the water vapor treatment, the shift of the bulk components is not considered a chemical shift. The water treatment causes a variation of the Fermi level position with respect to the SiC band edges, i.e. it changes the surface band bending. This affects the photoelectrons when they leave the sample and changes their apparent binding energy. Thus, the shift is the same for the photoelectrons originating√ from the C or Si of the bulk. The difference in the amount of the shift for 6 3 and MLG indicates that changes in the band bending are√ different. This is likely due to the different interfaces for water-treated 6 3 and MLG, which is reflected by the absence and presence of SiO2, respectively.

5.3 Investigation of defects

As discussed in the previous chapter, Raman spectroscopy is an appropriate tool to investigate defects in graphene-like materials. In this section it was performed to investigate possible damage introduced by the water vapor treatment. Furthermore, it also allows√ the unambiguous identification of graphene. The Raman spectra of the 6 3 (bottom) and MLG (middle) after annealing in water vapor are shown in fig. 5.3. In the presented data, the spectrum of the bare SiC substrate is already subtracted. This is necessary, because SiC contributes strongly to the Raman spectrum in the region of the G peak. For comparison, there is also shown a difference spectrum of MLG annealed in hydrogen (top), which is known as quasi-free-standing bilayer graphene (QFBLG). √ −1 The water-treated 6√ 3 exhibits a G peak at 1571 cm and a 2D peak at 2647 cm−1. Pristine 6 3 has a small and broad feature in the D and G region and no contributions in the 2D region [Fro13]. Therefore, the arising G and

66 5.3 Investigation of defects

√ Figure 5.3: Raman spectra of water-treated MLG (center, red) and treated 6 3 (bottom, blue) and MLG after annealing in hydrogen (top, black) is given for comparison. The wavelength of the excitation laser was 532 nm. For clarity spectra are oﬀset in intensity.

√ 2D peaks in the water-treated 6 3 indicate that the buﬀer layer transformed into graphene. There are, however, also defect induced contributions in the Raman spectrum. The most prominent peak is the D peak at 1321 cm−1, which indicates intervalley scattering of charge carriers at defects in the graphene. The D0 peak at 1600 cm−1 shows that also intravalley scattering occurs at the defects. The D + D0 peak at 2915 cm−1 originates from a second order process involving inter- and intravalley scattering at defects. The signal is always visible in graphene if the D and D0 contributions are high enough. The mean distance between defects LD can be estimated from the ratio of√ the G and D intensities (I(D)/I(G))[Can11]. In the case of water-treated 6 3, it amounts

67 Chapter 5: Intercalation of water vapor

to 7 nm. As discussed in the previous chapter, the I(D)/I(D0) ratio is used to classify the type of defect [Eck12]. Eckmann et al.[Eck12] determined an 0 3 I(D)/I(D ) ratio of 7 for vacancy-like defects and√ 13 for sp hybridization. From the Raman spectrum of the water-treated 6 3, an I(D)/I(D0) ratio of 8.4 was deducted. This means that the defects are vacancy-like, which is consistent with the XPS C1s spectrum in fig. 5.2c. The spectrum shows no signs for sp3 hybridization of carbon atoms in the form of C–O or Si–C bonds, of course not taking the substrate into account. √ The Raman spectroscopy√ results for the water-treated 6 3 are very similar to the oxygen-treated 6 3 discussed in chapter4. The mean defect distance of LD = 7nm after water treatment is the same as after the high-temperature oxygen treatment. Also the I(D)/I(D0) ratios indicate that the defects are vacancy-like in both cases. The Raman spectrum of water-treated MLG (center of fig. 5.3) is significantly different. The most prominent contribution is a sharp G peak at 1588 cm−1 followed by the 2D peak at 2693 cm−1. Figure 5.4 shows the Ra- man spectrum in the 2D region in more detail. The 2D peak is asymmetric and can not be fitted by one Voigt-shaped component. It can, however, be modeled by four components, which indicates that the water-treated monolayer graphene behaves like bilayer graphene in Raman spectroscopy [Röh08]. Furthermore, the spectrum exhibits almost no defect related contributions. The spectrum is very similar to the one of high quality QFBLG (top of fig. 5.3).

◦ The√ monolayer graphene was annealed in water vapor at 650 C, while the 6 3 was annealed at 500 ◦C. The damaging effect of water or oxygen dissolved in the water are expected to be greater at higher temperatures. Therefore, it is remarkable that the water√ -treated MLG exhibits√ almost no defects compared to the water-treated 6 3. The observation that 6 3 is more receptive√ to defects than MLG was also made by Oliveira et al. upon annealing 6 3 and MLG in air [Oli13]. In all cases, the SiC interface next to the buffer layer is oxidized by water or oxygen, but the defect densities suggest that the pathway of the oxidizing√ agents is different for buffer layer samples and MLG samples. The 6 3 exhibits a strong corrugation [Mal07, Lau08], which possibly makes it more reactive to water and oxygen than a smoother√ sheet of graphene. In the case of the oxygen or water treatment of 6 3 samples, the

68 5.4 Transition of the band structure

Figure 5.4: 2D region of the Raman spectrum of water-treated MLG shown in the center of fig. 5.3. The measured data are shown as circles, the four Voigt components as red lines, and their sum as a black line. bare buffer layer is exposed to the oxidizing agents, which results in a high defect density. In the case of MLG, the buffer layer is covered by a more inert graphene layer. To oxidize the SiC interface, the oxidizing agents have to transmit the graphene layer either at defects or SiC step edges and the buffer layer is only exposed from the side. This diffusive process is possibly the reason that the buffer layer is not significantly damaged by annealing in water or air, if it is covered by a monolayer graphene.

5.4 Transition of the band structure √ The changes in the electronic structure of 6 3 and MLG after water vapor treatment can be witnessed√ in ARPES. The measured valence band features for water-treated 6 3 are shown in ﬁg. 5.5a and for water-treated MLG in

69 Chapter 5: Intercalation of water vapor

fig. 5.5b. The valence band is shown along the two high symmetry directions: ΓMΓ on the left√ side and ΓKM on the right. The ARPES spectrum of the water-treated 6 3 exhibits a sharp π band with a local maximum at the M point and a linear dispersion in the vicinity of the K point. This dispersion of the π band is typical of monolayer graphene and indicates the transformation of the buffer layer into decoupled graphene. The decoupling is consistent with the XPS and Raman spectra described above (fig. 5.2 and fig. 5.3). More detailed results can be gained from the zoomed in region in the vicinity of the K point shown in fig. 5.6a. The spectrum is measured perpendicular to the ΓKM direction, in the so-called KK direction. This has the advantage that both sides of the Dirac cone are visible. In ΓKM direction, one branch of the π band is not visible in ARPES due to interference of the photoelectron amplitudes arising from the two sublattices in graphene or graphite [Gie11, Shi95]. In fig. 5.6a, the branches of the π band can be extrapolated, which places their intersection, the Dirac point (0.25 ± 0.03) eV above the Fermi energy. This doping is in line with the position of the C1s core level of (284.25 ± 0.05) eV measured with XPS. The doping corresponds to a p-type carrier concentration in the order of 4.5 × 1012 cm−2. Electronic transport measurements carried out with the help of Hall bars determined a hole concentration of 8 × 1012 cm−2, which is somewhat larger but still consistent with the photoemission data. The Hall bar measurements also determined a charge carrier mobility of 420 cm2/Vs at room temperature. This is lower than the 900 cm2/Vs of pristine MLG at room temperature [Emt09, Job10], but reasonably high if the high defect density observed in Raman spectroscopy is considered. Note that those defects apparently have only little impact on the sharpness of the bands measured in ARPES at room temperature. Figure 5.5b shows the ARPES measurements of water-treated MLG. As above, a graphene-like π band is clearly visible with a local maximum at the M point and a linear dispersion in the vicinity of the K point up to the Fermi energy. The bands are, however, much broader and there is√ a considerably higher background than in the case of the water-treated 6 3. The background is espeacially high in the energy window between about 5 eV and 10 eV. In this energy range, different O2p states have been observed in previous studies of oxidized SiC surfaces [Hol99, Hol00, Sie00, Vir02]. A

70 5.4 Transition of the band structure

Figure 5.5: ARPES band structure measured at 70 eV photon energy in ΓMΓ and ΓKM direction of (a) buﬀer layer and (b) monolayer graphene after water vapor treatment.

71 Chapter 5: Intercalation of water vapor

Figure 5.6: Zoomed in region of the ARPES band structure√ maesurements in the vicinity of the K point in KK direction of water-treated (a) 6 3 and (b) MLG, corresponding to the overview measurement shown in fig. 5.5. The blue dotted line illustrates the extrapolation of the π bands to the Dirac point. high density of states between 5 eV and 10 eV binding energy was reported by Virojanadara and Johansson [Vir02] for oxidized SiC(0001). Hollering, Sieber et al.[Hol99, Hol00, Sie00] studied silicate adlayer structures on hexagonal SiC{0001} surfaces, where they observed O2p-Si(3s,3p) bonding states at a binding energy of around 9 to 11 eV with respect to the SiC valence band maximum, while the O2p lone pair states were found at around 6 eV. Therefore, the diffuse intensity in the ARPES data between 5 and 10 eV can be attributed to valence band states of the thin oxide layer formed during water vapor treatment of MLG. As shown by XPS in section 5.2, the average thickness of the oxide layer is less than a monolayer which translates into a partial coverage of the surface. The inhomogeneous coverage with SiO2 most likely results in a inhomogeneous doping of the bilayer graphene on top. Averaging over areas with differently doped graphene could explain the broadening of the π band observed in fig. 5.5b. This is consistent with the water-treated buffer layer, where no SiO2 was observed and as a result the bands are narrower. Apparently,

72 5.5 Microscopic investigation by PEEM and LEEM the broadening of the bands cannot be directly taken as evidence for defects, because water-treated MLG has the lower D peak intensity in the Raman spectrum, but broader bands in ARPES. The zoomed in region in the vicinity of the K point of water-treated MLG is shown in ﬁg. 5.6b. Oliveira et al.[Oli13] were able to resolve the individual π bands in ARPES arising from the formation of a bilayer graphene upon annealing MLG in air. This was not possible for the water-treated MLG probably due to the limited resolution induced by the inhomogeneous doping as discussed above. The exact position of the Dirac point is also hard to determine. In XPS, the C1s core level was observed at (284.25 ± 0.05) eV suggesting a hole concentration in the order of 1013 cm−2. Hall transport measurements carried out under ambient conditions show a p-type carrier concentration of 2 × 1013 cm−2. Residual photo resist used for patterning of the hall bar structures may induce additional p-type doping. This would be a plausible reason to explain the discrepancy to the photoemission results performed in UHV. Hall transport measurements determined the mobility to be 790 cm2/Vs, which is quite promising if the high carrier concentration is taken into account.

5.5 Microscopic investigation by PEEM and LEEM

The spectroscopic methods presented above all average over their respective spot size. This requires homogeneous samples to rule out that the observed results are not a superposition from different positions on the sample. Ideally, the spectroscopic methods are combined with microscopic techniques. As discussed in section 3.3, LEEM and PEEM are valuable tools to study the morphological properties of epitaxial graphene.√ Figure 5.7 shows PEEM images of a 6 3 sample (a) before and (b) after water treatment. The images were recorded from the same sample position under UV illumination from a Hg-lamp and without the use of an electron energy filter. Since the Hg-lamp has a broad energy spectrum, the contrast in this setup is mainly governed by differences in the work function Φ [Bau89].

73 Chapter 5: Intercalation of water vapor

Figure 5.7: PEEM images of (a) pristine and (b) water-treated buffer layer taken without an energy filter under Hg-lamp excitation to image work function differences.

The image of the sample in fig. 5.7a is mostly light gray√ except for lines with a darker shade of gray. The growth parameters of the√ 6 3 were chosen in a way to avoid areas of SiC which are not covered by 6 3. As√ a consequence, it is possible that MLG forms at the SiC step edges during 6 3 growth. Buffer layer has a lower Φ of 3.75 eV [Mat07] and MLG a higher Φ of 4.33 eV [Mat07]. Thus the buffer layer areas appear brighter and the areas covered with MLG darker in the PEEM image. In the lower part of the image, the SiC step structure is dominated by two screw dislocations in the SiC substrate. Since the MLG preferable forms at the step edges, the screw dislocations are visible as hexagonal structures in the PEEM image. The feature in the top right corner is also due to an increased graphitization, where a mark was carved into the sample surface before the wet chemical cleaning to assist in sample positioning. After the water vapor treatment the same structures are visible in the PEEM image (fig. 5.7b). The contrast of the image is, however, inverted. Considering the low process temperature of 500 ◦C, it is obvious that the surface morphology does not change by the water treatment. Previous XPS measurements suggest that MLG is not transformed at a process temperature

74 5.5 Microscopic investigation by PEEM and LEEM of 500 ◦C. This leads to the areas which appear in light gray in fig. 5.7b still being assigned to MLG. If the work function of the MLG stays the same upon water√ treatment, then the inverted contrast means that the work function of the 6 3 areas changes from a value lower than MLG to a value√ higher than MLG. Thus, water treatment increased the work function of 6 3 at least by 0.6 eV. In addition, bright field LEEM images were recorded before (fig. 5.8a) and after (fig. 5.8b) water treatment. The displayed images were acquired with an electron energy of 1.0 eV and 6.2 eV, respectively. Each of the images is part of a larger set of images which were taken with energies spanning a wider range. The set of images allows the extraction of reflectivity spectra for every pixel of the image. In fig. 5.8c, the spectra averaged over the dashed and solid rectangles√ were plotted with corresponding lines. The reflectivity curve of the 6 3 (dashed line) is relatively flat. This was reported√ before by Hibino et al.[Hib08]. In contrast, the spectrum of water-treated 6 3 exhibits two dips; one at 1.4 eV, the other at 6.0 eV, which will be discussed below in more detail. √ Additional to the 6 3 sample, a nominal MLG sample was investigated before and after water treatment. The corresponding bright field images from the same position of the MLG sample before and after water treatment are shown in fig. 5.9a and fig. 5.9b at an electron energy of 4.6 eV. As above, a series of 50 images each was recorded at different electron energies and the reflectivity spectra were extracted for every pixel. Analysis of the spectra showed that there are four significantly different spectra to be found in each set of images. Those typical (standard) spectra, labeled A-D for pristine MLG and E-H for water-treated MLG, are plotted in fig. 5.10a and fig. 5.10b, respectively. In a second step, the spectrum of each pixel was fitted to the corresponding four standard spectra. The pixel was then assigned to the standard spectrum, where the fit resulted in the smallest standard deviation. By coloring the pixel in the color of the assigned standard spectrum, false color images were generated which are shown in fig. 5.9c for the pristine MLG and in fig. 5.9d for the water-treated MLG. Comparison of the two false color images with each other shows that the reflectivity curves before and after treatment can unambiguously be matched; A, B, C, D to E, F, G, H, respectively.

75 Chapter 5: Intercalation of water vapor

Figure 5.8: Bright ﬁeld LEEM images show buﬀer layer (a) before and (b) after water vapor treatment at 1.0 and 6.2 eV, respectively. Spectra recorded within the dashed and solid rectangles are plotted in (c) with corresponding lines.

As discussed in section 3.3.5, the reflectivity curves can be used to identify the number of graphene layers. For the pristine MLG sample, it makes no difference if the interpretation by Hibino et al.[Hib08] or the one by Feenstra et al.[Fee13a] is used, because there is still a buffer layer. In both explanations the number of dips corresponds to the number of graphene layers ontop of the buffer layer. Thus, the colored areas in fig. 5.9c can unambiguously√ be matched to a graphene thickness. The purple areas (A) are assigned to 6 3,

76 5.5 Microscopic investigation by PEEM and LEEM

Figure 5.9: Bright field LEEM images recorded at 4.6 eV of (a) pristine and (b) water- treated MLG. Letters A-H indicate where the corresponding reflectivity spectra shown in fig. 5.10 were taken. False color images (c) and (d) were generated by comparing the spectrum of every pixel with the standard spectra shown in fig. 5.10. The length of the scale bar is 1 µm. orange areas (B) to MLG, regions in light red (C) to bilayer graphene (BLG), and the small region in dark red (D) to trilayer graphene (TLG). From the false color image the total coverage of 1.3 ML can be extracted. The coverage

77 Chapter 5: Intercalation of water vapor

Figure 5.10: LEEM reflectivity spectra of (a) pristine and (b) water vapor treated MLG taken at the indicated positions in the bright field images shown in fig. 5.9. for this particular sample is higher than for a normal nominal MLG sample. Nevertheless, the regions with different coverage allow studying the behavior of LEEM reflectivity√ spectra for a variety of thicknesses at the same time. Similar to the 6 3 sample, the surface morpholgy is also unchanged by the ◦ water vapor treatment. Even at the√ higher process temperature of 650 C, this is not surprising. However, the 6 3 areas of the pristine sample, indicated in purple in fig. 5.9c, appear larger than the corresponding areas after water treatment, indicated in grey in fig. 5.9d. This√ is an artifact caused by the large difference between the work function of 6 3 and MLG. The difference is so high that it is not possible to focus on both, the buffer layer and MLG at the same time. As a consequence, the buffer layer regions are out of focus and appear larger if the system is focused on MLG. After water treatment, when

78 5.5 Microscopic investigation by PEEM and LEEM

√ the 6 3 is converted into graphene, the work function difference is reduced and both regions can be sharply imaged at the same time. The fact that there are buffer layer regions on the nominal MLG sample√ creates the opportunity to compare those regions√ with the nominal 6 3 sample. The reflectiviy curve of water-treated 6 3 (E) is similar to the one plotted in fig. 5.8c. They both exhibit two dips with the first one at 1.4 eV in both spectra.√ The second dip, however, occurs at different positions. For ◦ a nominal 6 3 sample processed in water vapor at 500 C (fig. 5.8c), the√ second dip lies at 6.0 eV. Whereas, the spectra from the water-treated 6 3 regions found on a nominal MLG sample processed at 650 ◦C (fig. 5.10b, curve E) show the dip at 4.2 eV. As discussed above, the different process temperatures lead to a different interface reflected by the presence or absence of a SiO2 layer. Thus, the shift in energy of this dip strongly suggests that it originates from the interface.√ In the studies of 6 3 intercalated by H [For11] and by Ge [Emt11], there were also two dips evident in the reflectivity curves. The authors drew the same conclusion that the second dip originates from the interface. In both references, the lower energy dip was associated with the formation of monolayer graphene. This is also consistent with the explanation√ of the dips by Hibino et al.[Hib08]. In this interpretation the pristine 6 3 does not contribute to the reflectivity curves but upon decoupling by H or Ge intercalation, it transforms into graphene and adds a dip. In contrast, both dips can also be explained by the interpretation of Feenstra et al. [Fee13a]. Here, the dips are√ caused by inter-layer states between√ carbon layers which include the 6 3. Thus, transformation of the 6 3 into graphene does not add an extra dip to the reflectivity curve. As discussed in detail in ref. [Sri13], it is, however, possible that inter-layer√ states form between the substrate and the carbon layers. For pristine 6 3 the coupling to the substrate is so strong that the energy of the inter-layer state is too large to show up as dip in the reflectivity curve. If the buffer layer becomes decoupled by intercalation, the distance to the substrate increases and the energy of the inter-layer state decreases,√ creating a dip in the reflectivity curve. In the case of water-treated 6 3, the distance between the substrate and the graphene layer seems to be large enough that two inter-layer states are formed in the energy range of the reflectivity spectrum, potentially leading to two dips.

79 Chapter 5: Intercalation of water vapor

The reflectvity curve of water-treated MLG (F), also shows two dips; a dip at about 2.8 eV which was also present before water treatment and an additional dip at 5.8 eV. In the spectrum of the treated BLG (G), this additional feature is also observed as a shoulder at the same energy. At the same time, the two dips of pristine BLG at 1.5 eV and 4.4 eV are still present after the water treatment, although somewhat shifted to 1.5 eV and 4.0 eV. At this stage, the interpretation of Hibino et al.[Hib08] fails. If the dips would be created by states in the graphene layers, then the reflectivity curve of treated MLG should resemble the one of pristine BLG and treated BLG the one of pristine TLG. In other words, the treated BLG would exhibit three equidistant dips for the three graphene layers and maybe a fourth one denoted to the interface. This is, however, not the case. The observed results can rather be explained based on the interpretation√ by Feenstra et al. [Fee13a]. Transforming monolayer graphene√ on top of 6 3 into two graphene layers, or bilayer graphene on top of 6 3 into three graphene layers does not change the inter-layer states between the carbon layers. For this reason, the dips typical of monolayer and bilayer graphene are√ preserved. The additional dip is again attributed to the decoupling of the 6 3 from the substrate. This gives reason why the additional dip occurs at the same energy for treated MLG and treated BLG. In the case of trilayer graphene, there is no signifcant difference in the reflectivity curves of the pristine sample (D) and after the water vapor treatment (H). The relative amplitudes of the dips change a bit, but the positions√ of the dips stay the same. However, no statement can be made if the 6 3 below the trilayer regions is even decoupled. Since those regions amount to only a small fraction of the surface, XPS cannot determine if intercalation occurred in these regions.

5.6 Conclusion

The XPS measurements demonstrate that water vapor is suitable to oxidize SiC within a controlled process in the absence of molecular oxygen gas. If the process temperature is chosen accordingly,√ the√ oxidation is successful even if the SiC is protected by the 6 3 or the 6 3 and several graphene

80 5.6 Conclusion layers. Depending on the process temperature, the SiC interface is either just passivated by oxygen or at higher temperatures a thin SiO2 layer is observed. ARPES, XPS, and Raman spectroscopy√ show that the oxidation of the interface√ leads to a decoupling of the 6 3 which transforms it into graphene. If 6 3 is used as the starting material, this leads to quasi-free- standing monolayer graphene, albeit containing a significant amount of defects. Those results are very similar to the high-temperature annealing in oxygen from the previous chapter. Raman spectroscopy shows that also the high defect densities introduced by the water-treatment are comparable with the high√ defect densities after oxygen treatment. This means that annealing√ the 6 3 in water vapor does not offer a milder way to decouple the 6 3 by oxidation of the SiC interface. √ If MLG is used as the starting material, decoupling of the underlying 6 3 leads to the formation of quasi-free-standing bilayer graphene. For the intercalation to take place, this process has to be carried out at higher temperatures. Nevertheless, Raman spectroscopy shows no significant amounts of defects in the QFBLG. Apparently,√ graphene is more inert to water vapor and can even protect the underlying 6 3. Similar results were also observed by Oliveira et al.[Oli13] upon annealing MLG in air. Since they did not use dried air, the results presented here potentially indicate that water vapor might have played an important role in their experiments. After water treatment of MLG, ARPES cannot resolve the two π bands of the two graphene layers. Most likely this is due to the underlying SiO2 layer, which introduces inhomogeneities into the doping of the graphene layers, limiting the resolution. The LEEM and PEEM measurements complete the investigations of the water-vapor treatment. They support√ the observations made by XPS, ARPES and Raman spectroscopy that the 6 3 decouples and transforms into graphene. Furthermore, the reflectivity curves of the water-treated samples are used to test two competing interpretations of the formation of the dips in the spectra. The commonly accepted interpretation by Hibino et al.[Hib08] fails to explain the measured spectra for water-treated MLG and BLG. The observations, however, support the recent re-interpretation by Feenstra et al. [Fee13a, Fee13b] and Srivastava et al. [Sri13]. Finally, the intercalation of water vapor to decouple the buffer layer is not considered an improvement over the treatment in pure oxygen gas. However,

81 Chapter 5: Intercalation of water vapor annealing MLG in water vapor leads to the formation of QFBLG with promising properties. The low defect densities rival QFBLG produced by annealing MLG in hydrogen.

82 6 Quasi-free-standing graphene on non-polar surfaces

Commonly, epitaxial graphene is grown by thermal decomposition on (0001) or (0001) surfaces of the hexagonal polytypes 4H- and 6H-SiC [Emt09, Hee11, Vir08]. As discussed in section 2.1, those surfaces are single- element terminated and are, therefore, referred to as Si-face or C-face, respectively. Graphene growth on the C-face is still very challenging, as it is very hard to obtain just one monolayer of graphene. On the Si-face, where the graphene growth is more uniform, a buffer layer always√ forms between the SiC and the graphene. As discussed before, the 6 3 buffer layer is known to be detrimental for the charge carrier√ mobility in the graphene layers on top of it [Spe11a]. In addition, the 6 3 efficiently dopes graphene by introducing donor states [Der11, Kop10]. In the previous chapters, intercalation of oxygen and water vapor was used to transform the buffer layer into graphene. This leads to buffer layer free graphene on top of an oxidized SiC interface. Although buffer layer intercalation and subsequent formation of quasi-free- standing graphene is a promising approach, it is still unclear if this is the only process that can lead to quasi-free-standing graphene. In this chapter, a method is demonstrated to directly grow quasi-free-standing graphene without the need of intercalation. This is achieved by using the non-polar, low-index (1120) and (1100) planes as substrates.

6.1 Experimental details

Graphene was grown on 4H-SiC(1120) and 4H-SiC(1100) using sublimation growth in Ar atmosphere at elevated temperatures [Emt09]. The SiC substrate material was purchased from INTRINSIC Semiconductor. Graphene growth

83 Chapter 6: Quasi-free-standing graphene on non-polar surfaces

was performed very similar to the procedure on 6H-SiC(0001), described in section 4.1. Prior to graphitization, the samples were cleaned by a wet chemical procedure, as described in [Sie02] and etched in hydrogen to remove polishing damage. For the graphene growth in Ar atmosphere, the annealing time and temperature was adjusted in order to obtain a coverage of 1-2 monolayers. Best results were achieved at a temperature of 1600 ◦C for 10 min. Note, that this is a lower temperature and shorter time than the 1675 ◦C for 15 min, which were applied to grow MLG on SiC(0001). After growth, the samples were investigated using various surface sensitive techniques such as XPS, ARPES, LEED, and LEEM. XPS measurements were carried out using a SPECS PHOIBOS 150 analyzer. As X-ray source was a SPECS FOCUS 500 X-ray monochromator which provides monochromated Al Kα radiation with a photon energy of 1486.7 eV was applied. The total energy resolution of the experiment was 0.3 eV. As previously described in chapter4 and section 5.2, XPS is very well suited to identify a potential buﬀer layer [Emt08, Ost10]. ARPES data were taken at the synchrotron facility BESSY II at beamline UE56/2-PGM-1 with a SPECS PHOIBOS 100 analyzer. LEEM and µ-LEED images were acquired with a SPECS FE-LEEM P90 system. The LEEM images were used to determine the morphology of the sample and the graphene layer thickness distribution. In addition to µ-LEED, conventional LEED recorded by SPECS ErLEED optics, were acquired to investigate rotational disorder of the graphene layers.

6.2 Investigation of the interface

XPS measurements are well suited to identify the bonding configuration of various elements, e.g. carbon. This fact is utilized to investigate the interface between the non-polar surfaces SiC(1120) and SiC(1100) and the overlaying graphene. Figure 6.1 shows C1s core level spectra of graphene on (a) 4H- SiC(1120) and (b) 4H-SiC(1100). For comparison, additional spectra are also plotted in fig. 6.1: (c) quasi-free-standing monolayer graphene (QFMLG) on H-terminated 6H-SiC(0001) [Spe10], (d) monolayer√ graphene (MLG) on 6H-SiC(0001) [Ost10], and (e) buffer layer (6 3) on 6H-SiC(0001) [Ost10]. The (0001) surfaces of 6H- and 4H-SiC are identical with respect to the

84 6.2 Investigation of the interface properties of epitaxial graphene and the interface structure, so the spectra can be compared. Deconvolution of the spectra shows symmetric and asymmetric components. As above, Voigt line shapes [Arm67] were used for the symmetric components and Mahan line shapes [Mah75] were used for the asymmetric components. For the non-polar surfaces 4H-SiC(1120) and 4H-SiC(1100), deconvolution reveals two components. The asymmetric component labeled with G originates from the graphene layer. The asymmetry reflects the metallic character of graphene. The peak position was determined to (284.50 ± 0.05) eV for both 4H-SiC(1120) and 4H-SiC(1100). The larger component labeled with SiC at (283.11 ± 0.05) eV for 4H-SiC(1120) and (282.95 ± 0.05) eV for 4H- SiC(1100) is caused by emission from C atoms in the SiC substrate. The C1s core level spectrum of MLG on SiC(0001) shown in fig. 6.1d is comprised of four different components [Emt08, Ost10]. Additional to the asy- metric graphene component G at (284.70 ± 0.05) eV and the bulk component SiC at (283.85 ± 0.05) eV, there are two components S1 at (285.0 ± 0.1) eV and S2 at (285.65 ± 0.10) eV, which originate from the buffer layer. As shown before in section 5.2, the same components were observed, if the buffer layer is prepared on its own (fig. 6.1e). After conversion into QFMLG via intercalation of hydrogen [Rie09, Spe10], the components S1 and S2 disappear. As depicted in fig. 6.1c, the spectrum consists of the component G at (284.23 ± 0.05) eV and the component SiC at (282.61 ± 0.05) eV. Note that the SiC component appears at different binding energies for the different samples. This variation is due to the different positions of the surface Fermi level with respect to the SiC band edges in the presented samples. So that, the resulting variations in the surface band bending are shifting the binding energy of the photoelectrons originating from the bulk. The important observation here, however, is the lack of interface layer related components in the C1s spectra of the non-polar surfaces. No components similar to S1 or S2 occur, which would indicate a strongly bound carbon layer as the one observed on SiC(0001). After growing graphene on the SiC(0001) surface, also no buffer layer is observed in XPS [Emt08]. In this case, however, a (2 × 2)C structure exists below the graphene sheet [Hie08]. The (2 × 2)C structure passivates the SiC(0001) so that no strong covalent interactions are possible between

85 Chapter 6: Quasi-free-standing graphene on non-polar surfaces

Figure 6.1: XPS C1s core level spectra of (a) graphene on 4H-SiC(1120), (b) graphene on 4H-SiC(1100), (c) quasi-free-standing monolayer graphene√ on H-terminated SiC(0001), (d) monolayer graphene on SiC(0001), and (e) the 6 3 on SiC(0001). The component denoted SiC is due to emission of C atoms in the SiC substrate. The G component originates from the graphene layer. The components S1 and S2 are characteristic for the buﬀer layer. All spectra are normalized to the same height of the SiC component and the curves are oﬀset from each other for clarity. The vertical dashed line gives the C1s core level position of graphite.

86 6.3 Stacking of the graphene layers the surface and the graphene sheet [Mag09]. As a consequence the graphene layers grown on SiC(0001) contain rotational stacking faults [Has08]. In the next section the stacking of the graphene layers on the non-polar SiC surfaces is investigated.

6.3 Stacking of the graphene layers

Figure 6.2a shows a conventinal LEED image of graphene on 4H-SiC(1120) acquired at an electron energy of 140 eV. Analysis of the diffraction spots reveals that there are three different kinds of spots visible in the image. Most diffraction spots belong to the rectangular lattice of 4H-SiC(1120) with its unit cell being indicated by a red rectangle. In addition, there are six spots arranged in a hexagon which can be described by the reciprocal lattice vectors g1 and g2 of graphene. The third kind of diffraction spots cannot be described by either the SiC nor the graphene lattice alone. Two of those spots are indicated by small blue arrows. Their positions can only be described by linear combinations of both, SiC and graphene base vectors. Thus, they originate from second order diffraction processes in the SiC/graphene system. Those spots indicate that the observed LEED pattern is not a superposition of graphene and bare SiC. Instead, LEED imaged the diffraction spots of the graphene layers as well as the underlying SiC substrate. The fact that only six graphene spots are visible demonstrates that there is no significant rotational disorder over the whole macro-sized LEED spot. For the conventional LEED which was applied, the spot size is in the order of mm2. Of course, small rotational disorder as recently observed for graphene on SiC(0001) [Wal13] cannot be ruled out. The LEED image revealed that all graphene layers are aligned in a way that the ΓK direction coincides with the [0001] direction of SiC. This represents the vertical direction in fig. 6.2a. Figure 6.2b-d show µ-LEED images of graphene on 4H-SiC(1100). All of them were recorded in a LEEM instrument at 46 eV with a 1 µm aperture at different sample positions. This setup significantly reduces the spot-size compared to the conventional LEED. In fig. 6.2b and fig. 6.2c, the SiC spots are visible along a vertical line through the center of the diffraction image. The distortion of the line is due to a slight misalignment of the magnetic lens

87 Chapter 6: Quasi-free-standing graphene on non-polar surfaces

Figure 6.2: (a) Conventional LEED image of graphene on 4H-SiC(1120) taken at 140 eV and (b)-(d) µ-LEED images of graphene on 4H-SiC(1100) taken at 46 eV with an 1 µm aperture at various sample positions.

system of the LEEM instrument. At other sample positions (diffraction image, fig. 6.2d), the graphene layer was too thick for the substrate to be seen at the given electron energy. The graphene spots are located at the outer edge of the µ-LEED images. In fig. 6.2b and fig. 6.2c, at least five different graphene orientations can be identified; in fig. 6.2d there are two. Although, the µ- LEED probes only about 0.8 µm2 of the sample, all of the recorded diffraction images show more than one graphene orientation at once. This indicates that graphene on 4H-SiC(1100) has a high degree of rotational disorder.

88 6.4 The band structure

6.4 The band structure

The band structure measured by ARPES can unambigously identify graphene by its linear π band in the vicinity of the K point. In ARPES, the measured spectrum is averaged over the illuminated spot. The spot size depends on the photon source and averages in the order of 10−3 mm2 for the used beamline UE56/2-PGM-1 at BESSY II. Thus, it is very challenging to measure a sample with rotational stacking faults as graphene on 4H-SiC(1100). On the other hand, graphene on 4H-SiC(1120), which shows no rotational disorder qualiﬁes as a good system for ARPES measurements. Figure 6.3a depicts the band structure of graphene on 4H-SiC(1120) in the vicinity of the K-point measured at a photon energy of 65 eV. The spectrum is displayed perpendicular to the ΓK direction in the KK direction, in which both branches of the Dirac cone are visible. The linear π bands cross each other 0.25 eV below the Fermi energy EF. This corresponds to n-type doped graphene with a carrier concentration of about 4 × 1012 cm−2. Figure 6.3b shows the circular Fermi surface above the Dirac point. Due to interference of the photoelectron amplitudes arising from the two sublattices in graphene, there are no states visible towards the Γ point. Those results are expected for n-type doped graphene [Oht07] and indicate that the graphene layer is not strongly coupled to the underlying substrate.

89 Chapter 6: Quasi-free-standing graphene on non-polar surfaces

Figure 6.3: (a) Band structure of graphene on 4H-SiC(1120) in the vicinity of the −1 K-point measured in KK direction with ky = 1.7Å . (b) Corresponding Fermi surface showing the Dirac cone above the Dirac point.

6.5 Morphology and thickness distribution of graphene on non-polar SiC

The morphology and the thickness distribution of graphene on non-polar surfaces is investigated by means of LEEM (see section 3.3). Similar to the LEEM measurements described in section 5.5, bright field images were taken at different electron energies. This procedure enables the extraction of reflectivity curves for every pixel in the image, which can then be used to generate false color images containing the thickness information. In fig. 6.4a a bright field LEEM image of graphene on 4H-SiC(1120) at 4.4 eV is displayed. Five significantly different reflectivity curves were obtained upon extracting the spectra of every pixel. Those five typical spectra are plotted in fig. 6.4b. The minima in the reflectivity curves are not as pronounced as they are on SiC{0001} surfaces [Hib08]. The unusual shape

90 6.5 Morphology and thickness distribution of graphene on non-polar SiC

Figure 6.4: (a) LEEM bright field image at ELEEM = 4.4eV of graphene on 4H- SiC(1120). (b) Typical reflectivity spectra offset from each other for clarity. (c) False color image of same area as described in (a) and colored corresponding to the spectra shown in (b). The length of the scale bar is 1 µm. of the spectra complicates the assignment of them to a number of layers. Nevertheless, it was possible to assign the typical reflectivity curves to 1, 2, 3 and 4 monolayers (ML) of graphene by analyzing the width of the minima structure in the spectra. As in the previous chapter5, false color images were generated by fitting the spectra of every pixel in the image to each of the typical reflectivity curves. The pixel can then be assigned to the best fitting typical reflectivity curve and the corresponding color is chosen for the pixel. The so generated false color image is displayed in fig. 6.4c. By adding the number of graphene layers of every pixel and dividing by the number of pixels, an average graphene thickness of (2.0 ± 0.1) ML was calculated. Alternatively,

91 Chapter 6: Quasi-free-standing graphene on non-polar surfaces the graphene thickness can also be determined from the XPS C1s spectrum shown in fig. 6.1a. Since the graphene layer attenuates the bulk SiC signal, the average thickness can be calculated from the intensity ratio of the surface and bulk components. This resulted in a thickness of (2.0 ± 0.1) ML, which fully supports the LEEM results. This clearly indicates that the assignment of the reflectivity curves to a number of graphene layers is reasonable. The false color image shows 1 µm wide terraces with continuous monolayer graphene. However, large areas of bi- and trilayer graphene and also some areas with 4 ML were also observed. The same measurements and data analysis was performed for graphene on 4H-SiC(1100). A bright field image taken at 2.6 eV is shown in fig. 6.5a. For this sample, six typical reflectivity curves are observed upon extracting the spectra of every pixel, which were plotted in fig. 6.5b. The false color image generated in the same fashion as fig. 6.4c, is depicted in fig. 6.5c. The minima of the reflectivity curves recorded from graphene on 4H-SiC(1100) are more distinctive, so the assignment of the spectra to 1-4 ML is obvious. In addition, there is a flat curve, which is plotted in black in fig. 6.5b. From its shape, it could potentially be attributed either to a buffer layer or to uncovered SiC. Since no evidence for a buffer layer was observed in XPS, this possibility is less likely. Instead, it is attributed to bare, uncovered SiC. The false color image revealed that there are larger areas of mono-, bi-, and trilayer graphene and smaller areas of uncovered SiC and four ML graphene.

92 6.6 Conclusion

Figure 6.5: (a) LEEM bright field image at ELEEM = 2.6eV of graphene on 4H- SiC(1100). (b) typical reflectivity spectra offset from each other for clarity. (c) false color image of same area as described in (a) and colored corresponding to the spectra shown in (b). The length of the scale bar is 1 µm.

6.6 Conclusion

In this chapter, XPS, ARPES, LEED and LEEM analysis demonstrated that quasi-free-standing graphene directly grows on the non-polar (1120) and (1100) surfaces of hexagonal SiC. XPS shows only two different bonding configurations of carbon; the one from the SiC substrate and the other from the graphene layers. This excludes a strongly bound buffer layer as observed on SiC(0001). On SiC(1100), LEED and LEEM suggest that the growth resembles the one on SiC(0001) exhibiting large rotational disorder. Addition- ally, the graphene layer thickness distribution is similarly difficult to control,

93 Chapter 6: Quasi-free-standing graphene on non-polar surfaces resulting in samples with small areas of monolayer graphene. Generation of false color images from LEEM data also revealed that areas of uncovered SiC exist alongside of areas with up to 4 monolayers graphene. This means that the growth can not easily be improved by increasing or decreasing the temperature. In the case of graphene on SiC(1120), LEED shows that graphene grows without rotational disorder, even though no buffer layer is present. The false color LEEM images revealed that the graphene coverage is more uniform consisting of larger areas of monolayer graphene. These results could provide alternative ways of wafer-scale graphene production. Thermal decomposition on the non-polar SiC surfaces allows the direct growth of quasi-free-standing graphene without the need of an additional intercalation step. The main findings of this chapter were published in [Ost13] together with theoretical calculations based on the density functional theory, which also predict the absence of a buffer layer for graphene on both non-polar surfaces. Thus, the theoretical results are clearly in line with the experimentally determined findings.

94 7 Summary

The exceptional electronic transport properties of graphene promise a bright future for high frequency, graphene-based electronic devices which could outperform conventional semiconductor materials. Growing graphene on SiC wafers is a promising approach, as this procedure produces graphene on top of an insulating substrate without the need of transferring it to another substrate. Thus, thermal decomposition of SiC at elevated temperatures may be considered as the most suitable way to produce graphene wafers for electronic applications. Usually, one of the two opposing (0001) or (0001) surfaces of hexagonal SiC is used as substrate. The SiC(0001) surface is silicon terminated and, therefore, also called Si-face, while the carbon terminated SiC(0001) surface is also called C-face. Graphene growth on the C-face is challenging to control and often results in films with several layers of graphene. This is undesirable for the production of graphene-based field effect devices, as the outer graphene layers screen the inner layers from the electric field applied by the gate electrode. In contrast, the graphene growth on the Si-face is slower and more uniform.√ √ The process is governed by a carbon rich reconstruction with a (6 3 × 6 3)R30° periodicity, acting as a bu√ffer layer between the substrate and the graphene layers. This buffer layer (6 3 for short), exhibits the same honey comb structure as graphene, but is covalently√ bound to the substrate and, therefore, electronically inactive. The 6 3, however, lowers the charge carrier mobility of the graphene layers residing on top of it and significantly dopes them. The aim of this work was to investigate procedures to improve epitaxial graphene on SiC which would be suited for graphene-based electronic devices. Three different approaches were studied to produce graphene on SiC without a buffer layer impairing its electronic properties. First, the transformation of the buffer layer into graphene by means of oxygen intercalation√ was discussed. The oxidation of the SiC interface decoupled the 6 3 and resulted in a buffer

95 Chapter 7: Summary

layer free graphene residing on an oxide layer. However, this treatment led to the formation of substantial defects in the graphene structure. Second, the oxidation of the interface was conducted by annealing the samples in water vapor instead of oxygen. This approach promised to have a less degrading effect on the graphene layer, while achieving the same degree of decoupling. Third, the direct growth of buffer layer free, quasi-free-standing graphene on SiC was demonstrated, using the low-index, non-polar surfaces (1100) and (1120) as substrates. Oxygen intercalation was performed by two different methods: a high- temperature treatment at 750 ◦C in oxygen atmosphere with a pressure of 10−4 mbar and a low-temperature treatment at 270 ◦C in a 200 mbar oxygen atmosphere. The higher partial oxygen pressure compensated for the lower temperature, so that both treatments resulted in a similar degree of oxidation. XPS analysis revealed that the oxidation occurred at the SiC interface, ter- minating the topmost silicon atoms from the substrate. It was also evident from XPS data, that the binding configuration of the carbon atoms from the√ buffer layer was changed by the oxygen intercalation, suggesting that the 6 3 was decoupled from the substrate and transformed into graphene. ARPES measurements showed√ graphene-like π bands after oxidation, corroborating the transition of the 6 3 into graphene. Those results indicated that the decoupling effect of the oxygen intercalation is very similar to the well established intercalation by hydrogen. Raman spectroscopy, however, indicated that the oxygen intercalation introduced a large amount of defects in the graphene. The low-temperature treatment promised to be less aggressive to the buffer layer. However, this treatment led to even higher defect densities than the high-temperature treatment. Raman spectroscopy showed that the defects were carbon vacancies with a mean distance between defects of about 7 nm for the high- and about 2 nm for the low-temperature treatment. The mean distance of 2 nm translated to about 1 % missing carbon atoms in the graphene lattice. Next, an alternative approach was studied to oxidize the SiC interface by annealing in water vapor. XPS√ data showed√ that the SiC was oxidized even tough it was covered by the 6 3 or the 6 3 and several graphene layers. Similar to the intercalation of oxygen, the buffer layer was decoupled and transformed into graphene. This transformation was observed by a change of

96 the carbon binding configuration in XPS and in addition by the appearance of graphene-like π bands in the ARPES band structure measurements. Ra- man spectroscopy, however, again indicated an introduction of large defect densities in the buffer layer upon water vapor treatment. The mean distance between defects was determined to be 7 nm, which is similar to the amount of defects introduced by the high-temperature oxygen treatment. Thus, water vapor as an assumedly milder oxidation agent did not improve the quality of the transformed√ buffer layer. Furthermore, monolayer graphene residing on the 6 3 was used as the starting material for the water treatment leading to the formation of quasi-free-standing bilayer graphene. XPS measurements revealed that the process had to√ be carried out a higher temperatures for a complete decoupling of the 6 3 below the graphene layer. Nevertheless, Raman spectroscopy showed no significant amount of defects,√ indicating that the more inert graphene layer on top protects the underlying 6 3. In addition, the surface was investigated by microscopy using LEEM and PEEM to com- plement the spectroscopy measurements. The morphology was unchanged by the water vapor treatment. The oxidation of the interface manifested in√ a change of the work function observed by PEEM. The decoupling of the 6 3 and transformation into graphene was also observed in the LEEM reflectivity curves. The commonly accepted interpretation of the LEEM reflectivity curves failed to explain the measured spectra, but the observations support the recent re-interpretation by Feenstra et al. [Fee13a, Fee13b] and Srivastava et al. [Sri13]. Third, the direct growth of quasi-free-standing graphene on the non-polar (1120) and (1100) surfaces of hexagonal SiC was investigated. For both surfaces, XPS analyses showed carbon atoms in only two different bonding configurations: the SiC substrate and graphene. These results excluded a strongly bound buffer layer as observed on SiC(0001). LEED and LEEM data demonstrated that the graphene growth on the two surfaces differs substantially. On the SiC(1100), large rotational disorder was observed and the thickness distribution was difficult to control. On the SiC(1120), graphene grew without rotation disorder and the graphene coverage was more uniform consisting of larger areas of monolayer graphene. ARPES measurements of graphene on SiC(1120) showed linear, graphene-like π bands and revealed a small n-type doping.

97 Chapter 7: Summary

Finally, it was possible to decouple and transform the buffer layer into graphene by oxidizing the SiC interface using both oxygen or water vapor intercalation. However, both methods introduced a significant amount of defects. The intercalation of water vapor to oxidize the interface was not considered an improvement over the treatment in pure oxygen gas, as similar high defect densities were created. Nevertheless, annealing monolayer graphene in water vapor led to the formation of quasi-free-standing bilayer graphene with promising properties. The low defect densities rival QFBLG which was obtained by annealing MLG in hydrogen. Preliminarily transport measurements were carried out in this system, which showed promising charge carrier mobilities. Further research could be conducted in the future to pattern field effect devices from water vapor intercalated QFBLG. Thermal decomposition on the non-polar SiC surfaces allowed the direct growth of quasi-free-standing graphene without the need of an additional intercalation step. Especially the growth on SiC(1120) provided promising results, which should be further improved by optimizing the growing parameters, potentially leading to alternative ways of wafer-scale graphene production.

98 8 Zusammenfassung

Die außergewöhnlichen elektronischen Eigenschaften von Graphen eröffnen einen zukünftigen Markt für Hochfrequenzbauteile, die auf Graphen basieren und die in der Lage sind effizienter zu arbeiten als bisherige Bauteile aus herkömmlichen Halbleitermaterialien. Das direkte Züchten von Graphen auf SiC Wafern ermöglicht dessen Produktion auf einem isolierenden Substrat und vereinfacht damit die Herstellung, da kein Transferieren der Graphen Schicht auf ein anderes Substrat nötig ist. Somit ist die Graphenzüchtung mittels thermischer Zersetzung von SiC bei hohen Temperaturen als Methode der Wahl anzusehen um Graphen für elektronische Anwendungen herzustellen. Gewöhnlich wird eine, der zwei sich gegenüberliegenden (0001) oder (0001) Oberflächen von hexagonalem SiC als Substrat genutzt. Dabei handelt es sich zum einen um die SiC(0001) Oberfläche, welche mit Silizium terminiert ist und daher auch Si-Seite genannt wird. Zum anderen wird die mit Kohlenstoff terminierte SiC(0001) Oberfläche verwendet, die als C-Seite bezeichnet wird. Das Graphenwachstum auf der C-Seite zu kontrollieren ist bisher eine große Herausforderung, da man oft einen dünnen Film aus mehreren Graphenlagen erhält, diese jedoch bei der Herstellung von auf Graphen basierenden Feldef- fekt Bauteilen unerwünscht sind. Der Grund dafür ist die Abschirmung durch die äußeren Graphenlagen, welche das elektrische Feld, das durch die Gates- pannung induziert wird, abschirmt, sodass die Ladungsträgerkonzentration in den inneren Lagen nicht variiert werden kann. Im Gegensatz dazu veläuft das Graphenwachstum auf der Si-Seite sowohl langsamer wie auch√ gleichmäßiger,√ da es von einer Kohlenstoff reichen Rekonstruktion mit (6 3 × 6 3)R30° Periodizität bestimmt wird, die sich wie eine Pufferlage zwischen dem SiC Substrat√ und dem Graphen verhält. Diese Pufferlage, oder im Folgenden kurz 6 3 genannt, besitzt ebenfalls die gleiche Honigwabenstruktur wie Graphen, ist jedoch durch ihre kovalente Bindung√ an das Substrat elektron- isch nicht aktiv. Nichtsdestotrotz verringert die 6 3 die Beweglichkeit der

99 Chapter 8: Zusammenfassung

freien Ladungsträger in den darauf liegenden Graphenlagen und dotiert die Graphenlagen zudem stark. Das Ziel der vorliegenden Arbeit war es Verfahren zu entwickeln, die die Herstellung von epitaktischem Graphen auf SiC verbessern und damit letztendlich eine besser geeignete Möglichkeit zu schaffen das Material für Graphen basierte elektronische Bauteile einzusetzen. Um Graphen ohne eine Pufferlage herzustellen, welche die elektronischen Eigenschaften von Graphen negativ beeinflusst, wurden drei unterschiedliche Herange- hensweisen angewendet. Zunächst wurde eine Transformation der Pufferlage zu Graphen mittels der Interkalation von Sauersto√ ff untersucht. Das Oxidieren der SiC Grenzschicht entkoppelt die 6 3 und resultiert in Graphen ohne Pufferlage, welches sich dann auf einer Oxidschicht befindet. Des Weiteren wurde die Oxidation der SiC Grenzschicht auch durch eine Behandlung in Wasserdampf statt in Sauerstoff durchgeführt. Der Wasserdampf versprach dabei zunächst die Graphenlage weniger stark anzugreifen und es dennoch im gleichen Maße zu entkoppeln. Drittens wurde das direkte Wachstum von quasi-freistehendem Graphen ohne Pufferlage demonstriert, indem die unpolaren (1100) und (1120) Oberflächen von SiC als Substrat genutzt wurden. Die Interkalation von Sauerstoff wurde hier auf zwei verschiedene Arten durchgeführt: eine Hochtemperaturbehandlung bei 750 ◦C in einer Sauerstoff- atmosphäre mit einem Druck von 10−4 mbar und in einer Niedertemperaturbe- handlung bei 270 ◦C in einer 200 mbar Sauerstoffatmosphäre. Bei letzterem Ansatz kompensierte der höhere Partialdruck des Sauerstoffs die niedrigere Temperatur, sodass beide Behandlungen die SiC Oberfläche im gleichen Maße oxidierten. Die Analyse mit Hilfe von XPS zeigte, dass eine Oxidation der SiC Grenzfläche stattgefunden hat, wobei die oberste Substratschicht durch Sauerstoffatome terminiert wurde. Die XPS Daten weisen ebenfalls darauf hin, dass die Bindungskonfiguration der Kohlenstoffatome in der Pufferlage durch die Sauerstoffinterkalation√ verändert wurde. Diese Änderung lässt darauf schließen, dass die 6 3 vom Substrat entkoppelt und zu Graphen transformiert wurde. ARPES Messungen zeigten nach der Oxidation graphenartige π-Bänder, was ebenfalls auf eine Umwandlung der Pufferlage zu Graphen hin- weist. Diese Ergebnisse deuten an, dass die Interkalation von Sauerstoff einen ähnlich entkoppelnden Effekt hatte wie die bereits bewährte Interkalation von Wasserstoff. Die Ergebnisse der Raman Spektroskopie zeigten jedoch,

100 dass die Sauerstoffinterkalation eine hohe Defektdichte in das Graphen ein- bringt. Die für die Pufferlage zunächst weniger aggressiv angenommene Niedertemperaturbehandlung erwies sich bei der Untersuchung mittels Ra- man Spektroskopie nicht als schonender als die Hochtemperaturbehandlung und erzeugte sogar eine höhere Defektdichte in der Graphenlage. Bei den De- fekten handelte es sich um Kohlenstoffleerstellen mit einem mittleren Abstand von etwa 7 nm bei der Hoch- und etwa 2 nm bei der Niedertemperaturbehand- lung. Ein mittlerer Abstand von 2 nm entspricht dabei einem Verlust von etwa 1 % der Kohlenstoffatome in der Graphenstruktur. Als nächstes wurde ein alternatives Verfahren untersucht um die SiC Grenz- schicht zu oxidieren, dazu wurden die Proben in Wasserdampf√ angelassen. XPS√ Messungen zeigten, dass das SiC, auch wenn es von 6 3 oder auch 6 3 und Graphen bedeckt war, oxidiert wurde. Analog zur Interkalation von Sauerstoff wurde die Pufferlage entkoppelt und es fand eine Transformation zu Graphen statt, was wiederum als Veränderung der Bindungskonfiguration im Kohlenstoff mit Hilfe des XPS und durch die Ausprägung graphenartiger π-Bänder mittels ARPES beobachtet werden konnte. Raman Spektroskopie Messungen deuteten jedoch daraufhin, dass die entkoppelte Pufferlage ebenfalls eine hohe Dichte von Defekten aufwies, die aufgrund der Wasserdampf- behandlung entstanden sind. Der mittlere Abstand der Defekte wurde auf 7 nm bestimmt, ein Wert, der dem der Hochtemperatur-Sauerstoffbehand- lung entspricht. Somit konnte Wasserdampf als vermeintlich schonenderes Oxidationsmittel nicht bestätigt werden, da die Qualität der transformierten Pufferlage nicht verbessert wurde. In einem weiteren Ansatz wurde Monola- gengraphen als Ausgansmaterial für die Wasserdampfbehandlung verwendet, was zur Bildung von quasi-freistehendem Bilagengraphen führte. XPS Mes- sungen zeigten, dass eine√ höhere Prozesstemperatur notwendig war um die von Graphen bedeckte 6 3 zu entkoppeln. In diesem Fall waren keine Defekte in signifikantem Ausmaß mit Hilfe von Raman Spektroskopie√ zu entdecken, ein Indiz dafür, dass die Graphenlage die darunter liegende 6 3 schützt. Um die spektroskopischen Methoden zu komplementieren wurden ergänzend mikroskopische Untersuchungen mit LEEM und PEEM durchge- führt, welche auf keine Veränderungen der Oberflächenmorphologie durch die Wasserdampfbehandlung hinwiesen. Die Oxidation der SiC Grenzschicht führte zu einer Erhöhung der Austrittsarbeit, die mit PEEM beobachtet wer-

101 Chapter 8: Zusammenfassung

√ den konnte. Das Entkoppeln der 6 3 und deren Umwandlung zu Graphen konnte ebenfalls an Hand der LEEM Reflektionsspektren nachvollzogen werden. Die gemeinhin akzeptierte Interpretation zur Entstehung der Minima in den Reflektionsspektren konnte dabei die hier gemessenen Spektren nicht erklären. Die Ergebnisse unterstützen jedoch eine vor kurzem veröffentlichte Reinterpretation von Feenstra et al. [Fee13a, Fee13b] und Srivastava et al. [Sri13]. Als dritter Themenkomplex wurde das direkte Wachstum von quasi-freistehendem Graphen auf den unpolaren (1120) und (1100) Oberflächen von hexagonalem SiC untersucht. Die XPS Analyse zeigte für beide Oberflächen, dass die Kohlenstoffatome nur in zwei verschiedenen Bindungskonfiguratio- nen vorlagen: entweder als SiC Substrat oder als Graphen. Die vorhandene Bindungskonfiguration schließt die Anwesenheit einer stark gebunden Puffer- lage, wie sie auf SiC(0001) gebildet wird, aus. LEED und LEEM Messun- gen zeigten, dass das Graphenwachstum auf den beiden Oberflächen stark voneinander abwich. Auf der SiC(1100) Oberfläche wurde zum einen eine große Fehlordnung in der Rotation der Graphenlagen zueinander beobachtet und zudem war die Graphenschichtdicke meist ungleichmäßig. Im Gegensatz dazu gab es auf der SiC(1120) Oberfläche keine Fehlordnung in der Rotation und die Graphenbedeckung war gleichmäßiger und mit größeren Flächen aus Monolagengraphen bedeckt. ARPES Bandstrukturmessungen von Graphen auf SiC(1120) zeigten lineare, graphenartige π-Bänder und eine niedrige n-Typ Dotierung der Ladungsträger. Zusammenfassend sei also anzumerken, dass die Pufferlage zwar durch die Oxidation der SiC Grenzschicht entkoppelt und zu Graphen transformiert werden konnte, dieser Prozess führte allerdings zur Entstehung einer signifikanten Anzahl von Defekten. Die Oxidation durch Interkalation von Wasserdampf konnte dabei nicht als Verbesserung gegenüber der Interkalation von reinem Sauerstoffgas angesehen werden, da ähnlich hohe Defektdichten entstanden. Das Anlassen von Monolagengraphen in Wasserdampf führte jedoch zu der Bildung von quasi-freistehendem Bilagengraphen mit vielversprechenden Eigenschaften. Die niedrige Defektdichte weist dabei auf eine hohe Qualität hin, die vergleichbar mit QFBLG ist, welches durch Anlassen in Wasserstoff erhalten wurde. Es wurden bereits erste Transportmessungen an diesem Sys- tem durchgeführt, welche in zukünfigen Studien weiter ausgebaut werden

102 sollten um letztendlich auch Graphen-Feldeﬀektbauteile aus Wasserdampf interkaliertem QFBLG einsetzen zu können. Die thermische Zersetzung der unpolaren SiC Oberﬂächen erlaubte das direkte Wachstum von quasi- freistehendem Graphen ohne einen zusätzlichen Interkalationsschritt. Vor allem die Züchtung von Graphen auf SiC(1120) lieferte vielversprechende Ergebnisse. Mit Hilfe einer Optimierung der Züchtungsparameter könnte die Qualität weiter verbessern werden und diese Methode damit eine potentielle, neue Alternative zur Produktion von Graphenwafern darstellen.

103

Bibliography

[Alt98] M. S. Altman, W. F. Chung and C. H. Liu, LEEM Phase Contrast, Surf. Rev. Lett. 05(06), 1129 (1998).

[Alt01] M. Altman, W. Chung, Z. He, H. Poon and S. Tong, Quantum size effect in low energy electron diffraction of thin films, Appl. Surf. Sci. 169-170, 82 (2001). [And98] T. Ando, T. Nakanishi and R. Saito, Berry’s Phase and Absence of Back Scattering in Carbon Nanotubes, J. Phys. Soc. Japan 67(8), 2857 (1998).

[Arm67] B. H. Armstrong, Spectrum line proﬁles: The Voigt Function, J. Quant. Spectrosc. Ra. 7(1), 61 (1967). [Bau89] E. Bauer, M. Mundschau, W. Swiech and W. Telieps, Surface studies by low-energy electron microscopy (LEEM) and conventional UV photoemission electron microscopy (PEEM), Ultramicroscopy 31(1), 49 (1989). [Bau98] E. Bauer, LEEM Basics, Surf. Rev. Lett. 05(06), 1275 (1998). [Bau07] E. Bauer, LEEM and SPLEEM, in P. W. Hawkes and J. C. Spence (editors), Science of Microscopy, 605–656, Springer New York (2007).

[Ber04] C. Berger, Z. Song, T. Li, X. Li, A. Y. Ogbazghi, R. Feng, Z. Dai, A. N. Marchenkov, E. H. Conrad, P. N. First and W. A. de Heer, Ultrathin Epitaxial Graphite: 2D Electron Gas Properties and a Route toward Graphene-based Nanoelectronics, J. Phys. Chem. B 108(52), 19912 (2004).

105 Bibliography

[Ber06] 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. de Heer, Electronic conﬁnement and coherence in patterned epitaxial graphene, Science 312(5777), 1191 (2006). [Bla07] P. Blake, E. W. Hill, A. H. Castro Neto, K. S. Novoselov, D. Jiang, R. Yang, T. J. Booth and A. K. Geim, Making graphene visible, Appl. Phys. Lett. 91(6), 063124 (2007). [Bol08] K. I. Bolotin, K. J. Sikes, Z. Jiang, M. Klima, G. Fudenberg, J. Hone, P. Kim and H. L. Stormer, Ultrahigh electron mobility in suspended graphene, Solid State Commun. 146(9-10), 351 (2008). [Can11] L. G. Cançado, A. Jorio, E. H. M. Ferreira, F. Stavale, C. A. Achete, R. B. Capaz, M. V. O. Moutinho, A. Lombardo, T. S. Kulmala and A. C. Ferrari, Quantifying defects in graphene via Raman spectroscopy at diﬀerent excitation energies, Nano Lett. 11(8), 3190 (2011). [Cas09] A. H. Castro Neto, N. M. R. Peres, K. S. Novoselov and A. K. Geim, The electronic properties of graphene, Rev. Mod. Phys. 81(1), 109 (2009).

[Cha02] A. Charrier, A. Coati, T. Argunova, F. Thibaudau, Y. Garreau, R. Pin- chaux, I. Forbeaux, J.-M. Debever, M. Sauvage-Simkin and J.-M. Themlin, Solid-state decomposition of silicon carbide for growing ultra-thin heteroepitaxial graphite ﬁlms, J. Appl. Phys. 92(5), 2479 (2002).

[Che11] C.-F. Chen, C.-H. Park, B. W. Boudouris, J. Horng, B. Geng, C. Girit, A. Zettl, M. F. Crommie, R. A. Segalman, S. G. Louie and F. Wang, Controlling inelastic light scattering quantum pathways in graphene, Nature 471(7340), 617 (2011). [Chu98] W. Chung and M. Altman, Step contrast in low energy electron microscopy, Ultramicroscopy 74(4), 237 (1998).

106 [Cor08] J. Coraux, A. T. N’Diaye, C. Busse and T. Michely, Structural coherency of graphene on Ir(111), Nano Lett. 8(2), 565 (2008). [Der11] I. Deretzis and A. La Magna, Role of covalent and metallic intercalation on the electronic properties of epitaxial graphene on SiC(0001), Phys. Rev. B 84(23), 235426 (2011). [Dre02] M. S. Dresselhaus and G. Dresselhaus, Intercalation compounds of graphite, Adv. Phys. 51(1), 1 (2002).

[Eck12] A. Eckmann, A. Felten, A. Mishchenko, L. Britnell, R. Krupke, K. S. Novoselov and C. Casiraghi, Probing the nature of defects in graphene by Raman spectroscopy, Nano Lett. 12(8), 3925 (2012). [Eiz79] M. Eizenberg and J. Blakely, Carbon monolayer phase condensation on Ni(111), Surf. Sci. 82(1), 228 (1979).

[Emt08] K. V. Emtsev, F. Speck, T. Seyller, L. Ley and J. Riley, Interaction, growth, and ordering of epitaxial graphene on SiC{0001} surfaces: A comparative photoelectron spectroscopy study, Phys. Rev. B 77(15), 155303 (2008). [Emt09] K. V. Emtsev, A. Bostwick, K. Horn, J. Jobst, G. L. Kellogg, L. Ley, J. L. McChesney, T. Ohta, S. A. Reshanov, J. Röhrl, E. Rotenberg, A. K. Schmid, D. Waldmann, H. B. Weber and T. Seyller, Towards wafer-size graphene layers by atmospheric pressure graphitization of silicon carbide, Nat. Mater. 8(3), 203 (2009). [Emt11] K. V. Emtsev, A. A. Zakharov, C. Coletti, S. Forti and U. Starke, Ambipolar doping in quasifree epitaxial graphene on SiC(0001) controlled by Ge intercalation, Phys. Rev. B 84(12), 125423 (2011). [Fee13a] R. M. Feenstra, N. Srivastava, Q. Gao, M. Widom, B. Diaconescu, T. Ohta, G. L. Kellogg, J. T. Robinson and I. V. Vlassiouk, Low- energy electron reﬂectivity from graphene, Phys. Rev. B 87(4), 041406 (2013).

107 Bibliography

[Fee13b] R. M. Feenstra and M. Widom, Low-energy electron reﬂectivity from graphene: First-principles computations and approximate mod- els, Ultramicroscopy 130, 101 (2013).

[Fer13] A. C. Ferrari and D. M. Basko, Raman spectroscopy as a versatile tool for studying the properties of graphene, Nat. Nanotechnol. 8(4), 235 (2013). [Fis90] G. R. Fisher and P. Barnes, Towards a uniﬁed view of polytypism in silicon carbide, Philos. Mag. Part B 61(2), 217 (1990).

[For98] I. Forbeaux, J.-M. Themlin and J.-M. Debever, Heteroepitaxial graphite on 6H-SiC(0001): Interface formation through conduc- tion-band electronic structure, Phys. Rev. B 58(24), 16396 (1998). [For11] S. Forti, K. V. Emtsev, C. Coletti, A. A. Zakharov, C. Riedl and U. Starke, Large-area homogeneous quasifree standing epitaxial graphene on SiC(0001): Electronic and structural characterization, Phys. Rev. B 84(12), 125449 (2011). [Fro13] F. Fromm, M. H. Oliveira, A. Molina-Sánchez, M. Hundhausen, J. M. J. Lopes, H. Riechert, L. Wirtz and T. Seyller, Contribution of the buﬀer layer to the Raman spectrum of epitaxial graphene on SiC(0001), New J. Phys. 15, 043031 (2013). [Gad75] J. Gadzuk and M. Šunjic,´ Excitation energy dependence of core- level x-ray-photoemission-spectra line shapes in metals, Phys. Rev. B 12(2), 524 (1975). [Gal85] N. Gall’, S. Mikhailov, E. Rut’kov and A. Tontegode, Nature of the adsoption binding between a graphite monolayer and rhenium surface, Sov. Phys.-Solid State 27(8), 2351 (1985). [Gei07] A. K. Geim and K. S. Novoselov, The rise of graphene, Nat. Mater. 6(3), 183 (2007).

[Gei09] A. K. Geim, Graphene: Status and Prospects, Science 324(5934), 1530 (2009).

108 [Gie11] I. Gierz, J. Henk, H. Höchst, C. R. Ast and K. Kern, Illuminating the dark corridor in graphene: Polarization dependence of angle- resolved photoemission spectroscopy on graphene, Phys. Rev. B 83(12), 121408 (2011). [Gra13] Graphene Industries: Products, http://grapheneindustries. com/?Products (2013), retrieved on 19.07.2013. [Gu07] G. Gu, S. Nie, R. M. Feenstra, R. P. Devaty, W. J. Choyke, W. K. Chan and M. G. Kane, Field eﬀect in epitaxial graphene on a silicon carbide substrate, Appl. Phys. Lett. 90(25), 253507 (2007). [Ham80] J. Hamilton and J. Blakely, Carbon segregation to single crystal surfaces of Pt, Pd and Co, Surf. Sci. 91(1), 199 (1980). [Han11] J. B. Hannon, M. Copel and R. M. Tromp, Direct Measurement of the Growth Mode of Graphene on SiC(0001) and SiC(0001), Phys. Rev. Lett. 107(16), 166101 (2011). [Has08] J. Hass, F. Varchon, J. Millán-Otoya, M. Sprinkle, N. Sharma, W. de Heer, C. Berger, P. First, L. Magaud and E. Conrad, Why Multilayer Graphene on 4H-SiC(0001) Behaves Like a Single Sheet of Graphene, Phys. Rev. Lett. 100(12), 125504 (2008).

[Hee11] W. A. de Heer, C. Berger, M. Ruan, M. Sprinkle, X. Li, Y. Hu, B. Zhang, J. Hankinson and E. Conrad, Large area and structured epitaxial graphene produced by confinement controlled sublimation of silicon carbide, Proc. Natl. Acad. Sci. 108(41), 16900 (2011). [Hib08] H. Hibino, H. Kageshima, F. Maeda, M. Nagase, Y. Kobayashi and H. Yamaguchi, Microscopic thickness determination of thin graphite films formed on SiC from quantized oscillation in reflectivity of low-energy electrons, Phys. Rev. B 77(7), 075413 (2008). [Hie08] F. Hiebel, P. Mallet, F. Varchon, L. Magaud and J.-Y. Veuillen, Graphene-substrate interaction on 6H-SiC(0001): A scanning tunneling microscopy study, Phys. Rev. B 78(15), 153412 (2008).

109 Bibliography

[Him82] F. Himpsel, K. Christmann, P. Heimann, D. Eastman and P. J. Feibel- man, Adsorbate band dispersions for C on Ru(0001), Surf. Sci. Lett. 115(3), L159 (1982).

[Hol99] M. Hollering, F. Maier, N. Sieber, M. Stammler, J. Ristein, L. Ley, A. Stampﬂ, J. Riley, R. Leckey, F. Leisenberger and F. Netzer, Elec- tronic states of an ordered oxide on C-terminated 6H–SiC, Surf. Sci. 442(3), 531 (1999). [Hol00] M. Hollering, N. Sieber, F. Maier, J. Ristein, L. Ley, J. Riley, R. Leckey, F. Leisenberger and F. Netzer, Electronic and Atomic Structure of an Ordered Silicate Adlayer on Hexagonal SiC, Mater. Sci. Forum 338-342, 387 (2000). [Jär98] K. Järrendahl and R. F. Davis, Materials Properties and Charac- terization of SiC, in Y. S. Park (editor), SiC Materials and Devices, vol. 52 of Semiconductors and Semimetals, Academic Press (1998). [Job10] J. Jobst, D. Waldmann, F. Speck, R. Hirner, D. K. Maude, T. Seyller and H. B. Weber, Quantum oscillations and quantum Hall eﬀect in epitaxial graphene, Phys. Rev. B 81(19), 195434 (2010). [Kho84] N. Kholin, E. Rut’kov and A. Tontegode, The nature of the adsorption bond between graphite islands and iridium surface, Surf. Sci. 139(1), 155 (1984). [Ki08] D.-K. Ki, D. Jeong, J.-H. Choi, H.-J. Lee and K.-S. Park, Inelastic scattering in a monolayer graphene sheet: A weak-localization study, Phys. Rev. B 78(12), 125409 (2008).

[Kit13] C. Kittel, Einführung in die Festkörperphysik, Oldenbourg Wis- senschaftsverlag, 15 ed. (2013). [Kop10] S. Kopylov, A. Tzalenchuk, S. Kubatkin and V. I. Fal’ko, Charge transfer between epitaxial graphene and silicon carbide, Appl. Phys. Lett. 97(11), 112109 (2010).

110 [Lan92] T. Land, T. Michely, R. Behm, J. Hemminger and G. Comsa, STM investigation of single layer graphite structures produced on Pt(111) by hydrocarbon decomposition, Surf. Sci. 264(3), 261 (1992). [Lat06] S. Latil and L. Henrard, Charge Carriers in Few-Layer Graphene Films, Phys. Rev. Lett. 97(3), 036803 (2006). [Lau08] P. Lauffer, K. V. Emtsev, R. Graupner, T. Seyller and L. Ley, Atomic and electronic structure of few-layer graphene on SiC(0001) studied with scanning tunneling microscopy and spectroscopy, Phys. Rev. B 77(15), 155426 (2008). [Laz08] M. Lazzeri, C. Attaccalite, L. Wirtz and F. Mauri, Impact of the electron-electron correlation on phonon dispersion: Failure of LDA and GGA DFT functionals in graphene and graphite, Phys. Rev. B 78(8), 081406 (2008). [Li09] X. Li, W. Cai, J. An, S. Kim, J. Nah, D. Yang, R. Piner, A. Vela- makanni, I. Jung, E. Tutuc, S. K. Banerjee, L. Colombo and R. S. Ruoff, Large-area synthesis of high-quality and uniform graphene films on copper foils, Science 324(5932), 1312 (2009). [Lin11] Y.-M. Lin, A. Valdes-Garcia, S.-J. Han, D. B. Farmer, I. Meric, Y. Sun, Y. Wu, C. Dimitrakopoulos, A. Grill, P. Avouris and K. A. Jenkins, Wafer-scale graphene integrated circuit, Science 332(6035), 1294 (2011). [Mag09] L. Magaud, F. Hiebel, F. Varchon, P. Mallet and J.-Y. Veuillen, Graphene on the C-terminated SiC (0001) surface: An ab initio study, Phys. Rev. B 79(16), 161405 (2009). [Mah75] G. Mahan, Collective excitations in x-ray spectra of metals, Phys. Rev. B 11(12), 4814 (1975). [Mal07] P. Mallet, F. Varchon, C. Naud, L. Magaud, C. Berger and J.-Y. Veuillen, Electron states of mono- and bilayer graphene on SiC probed by scanning-tunneling microscopy, Phys. Rev. B 76(4), 041403 (2007).

111 Bibliography

[Mal09] L. Malard, M. Pimenta, G. Dresselhaus and M. Dresselhaus, Raman spectroscopy in graphene, Phys. Rep. 473(5–6), 51 (2009). [Man12] K. L. Man and M. S. Altman, Low energy electron microscopy and photoemission electron microscopy investigation of graphene, J. Phys. Condens. Matter 24(31), 314209 (2012). [Mat07] A. Mattausch and O. Pankratov, Ab Initio Study of Graphene on SiC, Phys. Rev. Lett. 99(7), 076802 (2007).

[Mat11] C. Mattevi, H. Kim and M. Chhowalla, A review of chemical vapour deposition of graphene on copper, J. Mater. Chem. 21(10), 3324 (2011). [Mat12] C. Mathieu, B. Lalmi, T. O. Mente¸s,E. Pallecchi, A. Locatelli, S. Latil, R. Belkhou and A. Ouerghi, Effect of oxygen adsorption on the local properties of epitaxial graphene on SiC (0001), Phys. Rev. B 86(3), 035435 (2012). [May11] A. S. Mayorov, R. V. Gorbachev, S. V. Morozov, L. Britnell, R. Jalil, L. A. Ponomarenko, P. Blake, K. S. Novoselov, K. Watanabe, T. Taniguchi and A. K. Geim, Micrometer-scale ballistic transport in encapsulated graphene at room temperature, Nano Lett. 11(6), 2396 (2011). [McC57] J. McClure, Band Structure of Graphite and de Haas-van Alphen Effect, Phys. Rev. 108(3), 612 (1957). [Nag94] A. Nagashima, N. Tejima and C. Oshima, Electronic states of the pristine and alkali-metal-intercalated monolayer graphite/Ni(111) systems, Phys. Rev. B 50(23), 17487 (1994). [Nov04] 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 effect in atomically thin carbon films, Science 306(5696), 666 (2004).

112 [Nov05] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, M. I. Kat- snelson, I. V. Grigorieva, S. V. Dubonos and A. A. Firsov, Two- dimensional gas of massless Dirac fermions in graphene, Nature 438(7065), 197 (2005). [Nov07] K. S. Novoselov, Z. Jiang, Y. Zhang, S. V. Morozov, H. L. Stormer, U. Zeitler, J. C. Maan, G. S. Boebinger, P. Kim and A. K. Geim, Room-temperature quantum Hall effect in graphene, Science 315(5817), 1379 (2007). [Obr07] A. Obraztsov, E. Obraztsova, A. Tyurnina and A. Zolotukhin, Chem- ical vapor deposition of thin graphite films of nanometer thickness, Carbon 45(10), 2017 (2007). [Oht07] T. Ohta, A. Bostwick, J. McChesney, T. Seyller, K. Horn and E. Rotenberg, Interlayer Interaction and Electronic Screening in Multilayer Graphene Investigated with Angle-Resolved Photoemis- sion Spectroscopy, Phys. Rev. Lett. 98(20), 206802 (2007). [Oht08] T. Ohta, F. El Gabaly, A. Bostwick, J. L. McChesney, K. V. Emtsev, A. K. Schmid, T. Seyller, K. Horn and E. Rotenberg, Morphology of graphene thin film growth on SiC(0001), New J. Phys. 10(2), 023034 (2008). [Oid10] S. Oida, F. R. McFeely, J. B. Hannon, R. M. Tromp, M. Copel, Z. Chen, Y. Sun, D. B. Farmer and J. Yurkas, Decoupling graphene from SiC(0001) via oxidation, Phys. Rev. B 82(4), 041411 (2010). [Oli13] M. H. Oliveira, T. Schumann, F. Fromm, R. Koch, M. Ostler, M. Ramsteiner, T. Seyller, J. M. J. Lopes and H. Riechert, Formation of high-quality quasi-free-standing bilayer graphene on SiC(0001) by oxygen intercalation upon annealing in air, Carbon 52, 83 (2013). [Orl08] M. Orlita, C. Faugeras, P. Plochocka, P. Neugebauer, G. Martinez, D. K. Maude, A.-L. Barra, M. Sprinkle, C. Berger, W. A. de Heer and M. Potemski, Approaching the Dirac Point in High-Mobility Multilayer Epitaxial Graphene, Phys. Rev. Lett. 101(26), 267601 (2008).

113 Bibliography

[Ost09] M. Ostler, Atomic Layer Deposition von Aluminiumoxid auf Gra- phen, Diploma thesis, Friedrich-Alexander-Universität Erlangen- Nürnberg (2009). [Ost10] M. Ostler, F. Speck, M. Gick and T. Seyller, Automated preparation of high-quality epitaxial graphene on 6H-SiC(0001), Phys. Status Solidi B 247(11-12), 2924 (2010). [Ost12] M. Ostler, R. J. Koch, F. Speck, F. Fromm, H. Vita, M. Hundhausen, K. Horn and T. Seyller, Decoupling the Graphene Buffer Layer from SiC(0001) via Interface Oxidation, Mater. Sci. Forum 717-720, 649 (2012). [Ost13] M. Ostler, I. Deretzis, S. Mammadov, F. Giannazzo, C. Spinella, T. Seyller, A. L. Magna, G. Nicotra and A. La Magna, Direct growth of quasi-free-standing epitaxial graphene on nonpolar SiC surfaces, Phys. Rev. B 88(8), 085408 (2013). [Ost14] M. Ostler, F. Fromm, R. J. Koch, P. Wehrfritz, F. Speck, H. Vita, S. Böttcher, K. Horn and T. Seyller, Buffer layer free graphene on SiC(0001) via interface oxidation in water vapor, Carbon 70, 258 (2014). [Par09] S. Park and R. S. Ruoff, Chemical methods for the production of graphenes, Nat. Nanotechnol. 4(4), 217 (2009). [Pre09] B. Premlal, M. Cranney, F. Vonau, D. Aubel, D. Casterman, M. M. De Souza and L. Simon, Surface intercalation of gold underneath a graphene monolayer on SiC(0001) studied by scanning tunneling microscopy and spectroscopy, Appl. Phys. Lett. 94(26), 263115 (2009). [Ram47] L. Ramsdell, Studies on silicon carbide, Am. Mineral. 32, 64 (1947). [Ram98] V. Ramachandran, M. Brady, A. Smith, R. Feenstra and D. Greve, Preparation of atomically flat surfaces on silicon carbide using hydrogen etching, J. Electron. Mater. 27(4), 308 (1998).

114 [Rau99] E. Rauls, J. Elsner, R. Gutierrez and T. Frauenheim, Stoichiometric and non-stoichiometric (1010) and (1120) surfaces in 2H–SiC: a theoretical study, Solid State Commun. 111(8), 459 (1999).

[Rau00] E. Rauls, Z. Hajnal, P. Deák and T. Frauenheim, (1010)– and (1120)–Surfaces in 2H–, 4H– and 6H–SiC, Mater. Sci. Forum 338- 342, 365 (2000). [Rau01] E. Rauls, Z. Hajnal, P. Deák and T. Frauenheim, Theoretical study of the nonpolar surfaces and their oxygen passivation in 4H- and 6H-SiC, Phys. Rev. B 64(24), 245323 (2001). [Rei09] A. Reina, X. Jia, J. Ho, D. Nezich, H. Son, V. Bulovic, M. S. Dressel- haus and J. Kong, Large area, few-layer graphene ﬁlms on arbitrary substrates by chemical vapor deposition, Nano Lett. 9(1), 30 (2009). [Rie09] C. Riedl, C. Coletti, T. Iwasaki, A. A. Zakharov and U. Starke, Quasi- Free-Standing Epitaxial Graphene on SiC Obtained by Hydrogen Intercalation, Phys. Rev. Lett. 103(24), 246804 (2009). [Rob11] J. A. Robinson, M. Hollander, M. Labella, K. A. Trumbull, R. Cav- alero and D. W. Snyder, Epitaxial graphene transistors: enhancing performance via hydrogen intercalation, Nano Lett. 11(9), 3875 (2011). [Röh08] J. Röhrl, M. Hundhausen, K. V. Emtsev, T. Seyller, R. Graupner and L. Ley, Raman spectra of epitaxial graphene on SiC(0001), Appl. Phys. Lett. 92(20), 201918 (2008). [Sai98] R. Saito, M. S. Dresselhaus and G. Dresselhaus, Physical Properties of Carbon Nanotubes, World Scientiﬁc Publishing Company (1998). [Sat11] K. Sato, J. S. Park, R. Saito, C. Cong, T. Yu, C. H. Lui, T. F. Heinz, G. Dresselhaus and M. S. Dresselhaus, Raman spectra of out-of- plane phonons in bilayer graphene, Phys. Rev. B 84(3), 035419 (2011).

115 Bibliography

[Sch13] Google Scholar, http://scholar.google.com (2013), retrieved on 05.07.2013. [Sey05] T. Seyller, R. Graupner, N. Sieber, K. Emtsev, L. Ley, A. Tadich, J. Riley and R. Leckey, Hydrogen terminated 4H-SiC(1100) and (1120) surfaces studied by synchrotron x-ray photoelectron spectroscopy, Phys. Rev. B 71(24), 245333 (2005). [Shi72] D. A. Shirley, High-Resolution X-Ray Photoemission Spectrum of the Valence Bands of Gold, Phys. Rev. B 5(12), 4709 (1972).

[Shi95] E. Shirley, L. Terminello, A. Santoni and F. Himpsel, Brillouin-zone- selection eﬀects in graphite photoelectron angular distributions, Phys. Rev. B 51(19), 13614 (1995). [Shi00] A. Shikin, G. Prudnikova, V. Adamchuk, F. Moresco and K.-H. Rieder, Surface intercalation of gold underneath a graphite monolayer on Ni(111) studied by angle-resolved photoemission and high- resolution electron-energy-loss spectroscopy, Phys. Rev. B 62(19), 13202 (2000). [Sie00] N. Sieber, M. Hollering, J. Ristein and L. Ley, Photoemission Study of the Silicate Adlayer Reconstruction on Si-terminated 6H-SiC (0001), Mater. Sci. Forum 338-342, 391 (2000). [Sie01] N. Sieber, T. Seyller, R. Graupner, L. Ley, R. Mikalo, P. Hoﬀmann, D. Batchelor and D. Schmeiß er, PES and LEED study of hydrogen- and oxygen-terminated 6H-SiC(0001) and (0001) surfaces, Appl. Surf. Sci. 184(1-4), 278 (2001).

[Sie02] N. Sieber, T. Seyller, R. Graupner, L. Ley, R. Mikalo, P. Hoﬀmann, D. Batchelor and D. Schmeisser, Wet-Chemical Preparation of Sili- cate Adlayer Reconstructed SiC(0001) Surfaces as Studied by PES and LEED, Mater. Sci. Forum 389-393, 717 (2002). [Slo58] J. Slonczewski and P. Weiss, Band Structure of Graphite, Phys. Rev. 109(2), 272 (1958).

116 [Spe10] F. Speck, M. Ostler, J. Röhrl, J. Jobst, D. Waldmann, M. Hundhausen, L. Ley, H. B. Weber and T. Seyller, Quasi-Freestanding Graphene on SiC(0001), Mater. Sci. Forum 645-648, 629 (2010).

[Spe11a] F. Speck, J. Jobst, F. Fromm, M. Ostler, D. Waldmann, M. Hund- hausen, H. B. Weber and T. Seyller, The quasi-free-standing nature of graphene on H-saturated SiC(0001), Appl. Phys. Lett. 99(12), 122106 (2011). [SPE11b] SPECS GmbH, Berlin, User manual: FE-LEEM P90 Low Energy Electron Microscope (2011). [Sri10] A. Srivastava, C. Galande, L. Ci, L. Song, C. Rai, D. Jariwala, K. F. Kelly and P. M. Ajayan, Novel Liquid Precursor-Based Facile Syn- thesis of Large-Area Continuous, Single, and Few-Layer Graphene Films, Chem. Mater. 22(11), 3457 (2010).

[Sri13] N. Srivastava, Q. Gao, M. Widom, R. M. Feenstra, S. Nie, K. F. McCarty and I. V. Vlassiouk, Low-energy electron reflectivity of graphene on copper and other substrates, Phys. Rev. B 87(24), 245414 (2013). [Sut08] P. W. Sutter, J.-I. Flege and E. A. Sutter, Epitaxial graphene on ruthenium, Nat. Mater. 7(5), 406 (2008). [Sut10] P. Sutter, J. T. Sadowski and E. A. Sutter, Chemistry under cover: tuning metal-graphene interaction by reactive intercalation, J. Am. Chem. Soc. 132(23), 8175 (2010). [Tan12] S. Tanabe, M. Takamura, Y. Harada, H. Kageshima and H. Hibino, Quantum Hall Effect and Carrier Scattering in Quasi-Free-Standing Monolayer Graphene, Appl. Phys. Express 5(12), 125101 (2012). [Tel85] W. Telieps and E. Bauer, An analytical reflection and emission UHV surface electron microscope, Ultramicroscopy 17(1), 57 (1985).

[Tro93] R. Tromp and M. Reuter, Imaging with a low-energy electron microscope, Ultramicroscopy 50(2), 171 (1993).

117 Bibliography

[Tro98] R. M. Tromp, M. Mankos, M. C. Reuter, A. W. Ellis and M. Copel, A New Low Energy Electron Microscope, Surf. Rev. Lett. 05(06), 1189 (1998).

[Van75] A. J. Van Bommel, J. E. Crombeen and A. Van Tooren, LEED and Auger electron observations of the SiC(0001) surface, Surf. Sci. 48(2), 463 (1975). [Vil06] P. Villars and K. Cenzual (editors), Structure Types. Part 4: Space Groups (189) P-62m - (174) P-6, vol. 43A4 of Landolt-Börnstein - Group III Condensed Matter, Springer-Verlag, Berlin/Heidelberg (2006). [Vir02] C. Virojanadara and L. Johansson, Oxidation studies of 4H-SiC (0001) and (0001), Surf. Sci. 505, 358 (2002). [Vir08] C. Virojanadara, M. Syväjarvi, R. Yakimova, L. Johansson, A. Za- kharov and T. Balasubramanian, Homogeneous large-area graphene layer growth on 6H-SiC(0001), Phys. Rev. B 78(24), 245403 (2008). [Vir10] C. Virojanadara, S. Watcharinyanon, A. A. Zakharov and L. I. Jo- hansson, Epitaxial graphene on 6H-SiC and Li intercalation, Phys. Rev. B 82(20), 205402 (2010).

[Wal47] P. R. Wallace, The Band Theory of Graphite, Phys. Rev. 71(9), 622 (1947). [Wal11] A. L. Walter, K.-J. Jeon, A. Bostwick, F. Speck, M. Ostler, T. Seyller, L. Moreschini, Y. S. Kim, Y. J. Chang, K. Horn and E. Rotenberg, Highly p-doped epitaxial graphene obtained by ﬂuorine intercalation, Appl. Phys. Lett. 98(18), 184102 (2011). [Wal13] A. L. Walter, A. Bostwick, F. Speck, M. Ostler, K. S. Kim, Y. J. Chang, L. Moreschini, D. Innocenti, T. Seyller, K. Horn and E. Rotenberg, Small scale rotational disorder observed in epitaxial graphene on SiC(0001), New J. Phys. 15(2), 023019 (2013).

118 [Win09] J. Wintterlin and M.-L. Bocquet, Graphene on metal surfaces, Surf. Sci. 603(10-12), 1841 (2009). [Wu09] X. Wu, Y. Hu, M. Ruan, N. K. Madiomanana, J. Hankinson, M. Sprinkle, C. Berger and W. A. de Heer, Half integer quantum Hall eﬀect in high mobility single layer epitaxial graphene, Appl. Phys. Lett. 95(22), 223108 (2009). [Xia12] C. Xia, S. Watcharinyanon, A. A. Zakharov, R. Yakimova, L. Hult- man, L. I. Johansson and C. Virojanadara, Si intercalation/ deinter- calation of graphene on 6H-SiC(0001), Phys. Rev. B 85(4), 045418 (2012). [Zha05] Y. Zhang, Y.-W. Tan, H. L. Stormer and P. Kim, Experimental observation of the quantum Hall eﬀect and Berry’s phase in graphene, Nature 438(7065), 201 (2005).

[ZP87] H. Zi-Pu, D. Ogletree, M. Van Hove and G. Somorjai, Leed theory for incommensurate overlayers: Application to graphite on Pt(111), Surf. Sci. 180(2-3), 433 (1987).

119

List of publications

2014 23. E. M. Reinisch, T. Ules, P. Puschnig, S. Berkebile, M. Ostler, Th. Seyller, M. G. Ramsey and G. Koller, Development and character of gap states on alkali doping of molecular ﬁlms, New J. Phys. 16(2), 023011 (2014).

22. M. Ostler, F. Fromm, R. J. Koch, P. Wehrfritz, F. Speck, H. Vita, S. Böttcher, K. Horn and Th. Seyller, Buﬀer layer free graphene on SiC(0001) via interface oxidation in water vapor, Carbon 70, 258 (2014).

21. N. Ostler, N. Britzen-Laurent, A. Liebl, E. Naschberger, G. Lochnit, M. Ostler, F. Forster, P. Kunzelmann, S. Ince, V. Supper, G. J. K. Prae- fcke, D. W. Schubert, H. Stockinger, C. Herrmann and M. Stürzl, IFN- γ-induced Guanylate Binding Protein-1 is a novel Actin Cytoskeleton Remodeling Factor, Mol. Cell. Biol. 34(2), 196 (2014).

2013 20. J. Chen, M. L. Nesterov, A. Yu. Nikitin, S. Thongrattanasiri, P. Alonso- González, T. M. Slipchenko, F. Speck, M. Ostler, Th. Seyller, I. Crassee, F. Koppens, L. Martin-Moreno, F. J. García de Abajo, A. B. Kuzmenko and R. Hillenbrand, Strong plasmon reﬂection at nanometer-size gaps in monolayer graphene on SiC, Nano Lett. 13(12), 6210 (2013).

19. D. S. Wastl, F. Speck, E. Wutscher, M. Ostler, Th. Seyller and F. J. Giessibl, Observation of 4 nm Pitch Stripe Domains Formed by Exposing Graphene to Ambient Air, ACS Nano 7(11), 10032 (2013).

121 Chapter :List of publications

18. M. Ostler, I. Deretzis, S. Mammadov, F. Giannazzo, C. Spinella, Th. Seyller, A. L. Magna, G. Nicotra and A. La Magna, Direct growth of quasi-free-standing epitaxial graphene on nonpolar SiC surfaces, Phys. Rev. B 88(8), 085408 (2013).

17. B. Birkner, D. Pachniowski, A. Sandner, M. Ostler, Th. Seyller, J. Fabian, M. Ciorga, D. Weiss and J. Eroms, Annealing-induced magnetic mo- ments detected by spin precession measurements in epitaxial graphene on SiC, Phys. Rev. B 87(8), 081405 (2013).

16. A. L. Walter, A. Bostwick, F. Speck, M. Ostler, K. S. Kim, Y. J. Chang, L. Moreschini, D. Innocenti, Th. Seyller, K. Horn and E. Rotenberg, Small scale rotational disorder observed in epitaxial graphene on SiC(0001), New J. Phys. 15(2), 023019 (2013).

15. M. H. Oliveira, T. Schumann, F. Fromm, R. Koch, M. Ostler, M. Ram- steiner, Th. Seyller, J. M. J. Lopes and H. Riechert, Formation of high- quality quasi-free-standing bilayer graphene on SiC(0001) by oxygen intercalation upon annealing in air, Carbon 52, 83 (2013).

2012 14. M. Orlita, I. Crassee, C. Faugeras, A. B. Kuzmenko, F. Fromm, M. Ostler, Th. Seyller, G. Martinez, M. Polini and M. Potemski, Clas- sical to quantum crossover of the cyclotron resonance in graphene: a study of the strength of intraband absorption, New J. Phys. 14(9), 095008 (2012).

13. R. J. Koch, M. Weser, W. Zhao, F. Viñes, K. Gotterbarm, S. M. Kozlov, O. Höfert, M. Ostler, C. Papp, J. Gebhardt, H.-P. Steinrück, A. Görling and Th. Seyller, Growth and electronic structure of nitrogen-doped graphene on Ni(111), Phys. Rev. B 86(7), 075401 (2012).

12. M. Ostler, R. J. Koch, F. Speck, F. Fromm, H. Vita, M. Hundhausen, K. Horn and Th. Seyller, Decoupling the Graphene Buﬀer Layer from SiC(0001) via Interface Oxidation, Mater. Sci. Forum 717-720, 649 (2012).

122 11. I. Crassee, M. Orlita, M. Potemski, A. L. Walter, M. Ostler, Th. Seyller, I. Gaponenko, J. Chen and A. B. Kuzmenko, Intrinsic terahertz plasmons and magnetoplasmons in large scale monolayer graphene, Nano Lett. 12(5), 2470 (2012).

2011 10. J. Karch, C. Drexler, P. Olbrich, M. Fehrenbacher, M. Hirmer, M. M. Glazov, S. A. Tarasenko, E. L. Ivchenko, B. Birkner, J. Eroms, D. Weiss, R. Yakimova, S. Lara-Avila, S. Kubatkin, M. Ostler et al., Terahertz Radiation Driven Chiral Edge Currents in Graphene, Phys. Rev. Lett. 107(27), 276601 (2011). 9. P. Puschnig, E.-M. Reinisch, T. Ules, G. Koller, S. Soubatch, M. Ostler, L. Romaner, F. S. Tautz, C. Ambrosch-Draxl and M. G. Ramsey, Or- bital tomography: Deconvoluting photoemission spectra of organic molecules, Phys. Rev. B 84(23), 235427 (2011). 8. F. Speck, J. Jobst, F. Fromm, M. Ostler, D. Waldmann, M. Hundhausen, H. B. Weber and Th. Seyller, The quasi-free-standing nature of graphene on H-saturated SiC(0001), Appl. Phys. Lett. 99(12), 122106 (2011). 7. A. L. Walter, A. Bostwick, K.-J. Jeon, F. Speck, M. Ostler, Th. Seyller, L. Moreschini, Y. J. Chang, M. Polini, R. Asgari, A. H. MacDonald, K. Horn and E. Rotenberg, Eﬀective screening and the plasmaron bands in graphene, Phys. Rev. B 84(8), 085410 (2011). 6. A. L. Walter, K.-J. Jeon, A. Bostwick, F. Speck, M. Ostler, Th. Seyller, L. Moreschini, Y. S. Kim, Y. J. Chang, K. Horn and E. Rotenberg, Highly p-doped epitaxial graphene obtained by ﬂuorine intercalation, Appl. Phys. Lett. 98(18), 184102 (2011).

2010 5. I. Crassee, J. Levallois, A. L. Walter, M. Ostler, A. Bostwick, E. Roten- berg, Th. Seyller, D. van der Marel and A. B. Kuzmenko, Giant Faraday rotation in single- and multilayer graphene, Nat. Phys. 7(1), 48 (2010).

123 Chapter :List of publications

4. M. Ostler, F. Speck, M. Gick and Th. Seyller, Automated preparation of high-quality epitaxial graphene on 6H-SiC(0001), Phys. status solidi B 247(11-12), 2924 (2010).

3. J. Ristein, W. Zhang, F. Speck, M. Ostler, L. Ley and Th. Seyller, Char- acteristics of solution gated ﬁeld eﬀect transistors on the basis of epitaxial graphene on silicon carbide, J. Phys. D. Appl. Phys. 43(34), 345303 (2010).

2. F. Speck, M. Ostler, J. Röhrl, J. Jobst, D. Waldmann, M. Hundhausen, L. Ley, H. B. Weber and Th. Seyller, Quasi-Freestanding Graphene on SiC(0001), Mater. Sci. Forum 645-648, 629 (2010).

1. F. Speck, M. Ostler, J. Röhrl, K. V. V. Emtsev, M. Hundhausen, L. Ley and Th. Seyller, Atomic layer deposited aluminum oxide ﬁlms on graphite and graphene studied by XPS and AFM, Phys. status solidi C 7(2), 398 (2010).

124