Dissertação - ChristianRiedl

Epitaxial Graphene
on Silicon Carbide Surfaces:

Growth, Characterization, Doping and
Hydrogen Intercalation
( Epitaktisches Graphen auf Siliziumkarbid-Oberächen:

Wachstum, Charakterisierung, Dotierung und
Wassersto-Interkalation)
Der Naturwissenschaftlichen Fakultät

der Friedrich-Alexander-Universität Erlangen-Nürnberg
zur
Erlangung des Doktorgrades Dr. rer. nat.
vorgelegt von
Christian Riedl
aus Nürnberg
Als Dissertation genehmigt von der Naturwissen-
schaftlichen Fakultät der Friedrich-Alexander-Universität
Erlangen-Nürnberg
Tag der mündlichen Prüfung: 09.07.2010
Vorsitzender der
Promotionskommission: Prof. Dr. Eberhard Bänsch
Erstberichterstatter: Priv.-Doz. Dr. Ulrich Starke
Zweitberichterstatter: Prof. Dr. Andreas Magerl
Drittberichterstatter: Prof. Dr. Karsten Horn

Contents
1 Introduction 1
2 Epitaxial graphene a new material 5
2.1 Introduction to graphene . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.1 Graphene physics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1.2 Synthesis of graphene . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.2 Silicon carbide as a template for epitaxial graphene . . . . . . . . . . . . . . . . 16
2.2.1 SiC crystal structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 SiC surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3 Experimental background 23
3.1 Investigation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.1.1 Low energy electron diraction (LEED) and microscopy (LEEM) . . . . . 24
3.1.2 Scanning probe microscopy . . . . . . . . . . . . . . . . . . . . . . . . 27
3.1.3 Photoelectron spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . 30
3.1.4 Raman spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.2 Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.3 Experimental setup and sample preparation . . . . . . . . . . . . . . . . . . . . 38
4 Characterization of epitaxial graphene on SiC(0001) 43

4.1 The phase
√ diagram
√ of SiC(0001) . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.2 The (6 3×6 3)R30◦ phase as a precursor of graphitization . . . . . . . . . . . 44
4.2.1 The periodicity of the√interface
√ layer . . . . . . . . . . . . . . . . . . . 44
4.2.2 The nature of the (6 3×6 3)R30 reconstruction . . . . . . .
◦
. . . . . 51
4.3 The growth and identication of graphene layers . . . . . . . . . . . . . . . . . 54
4.3.1 Core level photoelectron spectroscopy . . . . . . . . . . . . . . . . . . . 54
4.3.2 The band structure of epitaxial graphene . . . . . . . . . . . . . . . . . 57
4.3.3 LEED ngerprints for the determination of the number of layers . . . . . 63
4.3.4 STM characterization of epitaxial graphene on SiC(0001) . . . . . . . . 66
4.3.5 Homogeneity of the graphene lms . . . . . . . . . . . . . . . . . . . . 72
4.3.6 Raman Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5 Variation of the graphene growth conditions 89

5.1 Variation of the SiC substrate . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.1.1 Epitaxial graphene on SiC(0001̄) . . . . . . . . . . . . . . . . . . . . . 90
5.1.2 Inuence of the polytype . . . . . . . . . . . . . . . . . . . . . . . . . . 95
iii
5.1.3 Non-basal plane SiC surfaces . . . . . . . . . . . . . . . . . . . . . . . 96
5.2 Variation of the growth technique . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.2.1 Carbon evaporation in UHV conditions . . . . . . . . . . . . . . . . . . 100
5.2.2 Graphitization of SiC(0001) in an argon atmosphere . . . . . . . . . . . 104
5.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6 Tuning the material properties of epitaxial graphene on SiC(0001) 109

6.1 Hole doping of epitaxial graphene . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.1.1 Atomic hole doping by means of Bi and Sb . . . . . . . . . . . . . . . . 111
6.1.2 Molecular hole doping by means of F4-TCNQ . . . . . . . . . . . . . . . 112
6.2 Decoupling epitaxial graphene from the SiC substrate . . . . . . . . . . . . . . 122
6.2.1 Hydrogen intercalation below epitaxial graphene . . . . . . . . . . . . . 122
6.2.2 Gold intercalation below epitaxial graphene . . . . . . . . . . . . . . . . 135
6.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
7 Summary 141
8 Zusammenfassung 145
Bibliography 149
Publications 163
Chapter 1
Introduction
The driving forces of new scientic developments may be based on needs dened by the society
as a whole or on human's curiosity in understanding and solving fundamental questions. The eld
of material sciences has always been addressing both aspects. In the past few years this could be
observed in an impressively clear manner for the material graphene, which for a long time has just
been used as a theoretical toy model and was presumed not to exist in a free state. Graphene
can just be seen as a single atomic plane pulled out of bulk graphite. However, it is the rst truly
two-dimensional material to be experimentally accessed in the year 2004 [1]. Since then it has
provoked an enormous interest in material sciences and lead to the formation of the new research
eld of relativistic condensed matter physics [24]. This development originates from the fact
that graphene gives detailed insight into a large variety of physical phenomena related to quantum
electrodynamics (QED) that were undiscovered so far, in particular in solid state systems. Apart
from this principal scientic background the material graphene exhibits promising mechanical and
electronic properties. In this respect it has to be considered that advances in microelectronics over
the past decades have been governed by Moore's law stating that the number of transistors on
commercially available integrated circuits follows an exponential law with time. This technological
evolution is the basis for the digital revolution, which has been starting at the end of the 20th
century. Moore's law naturally implies that the size of the transistors has to become progressively
smaller soon reaching dimensions in the nm range. As a consequence classical physics is not
yet valid any more whereas the laws of quantum mechanics will play the major role. Future
electronics are therefore named nanoelectronics, which is one of the main branches of nanoscale
science and nanotechnology. Due to its extraordinary electronic properties the material graphene
has very promising prospects for applications in this future market area.
The element carbon is probably one of the most intriguing elements in nature since it is the
basis of all organic chemistry and life. Carbon-based systems exhibit a large variety of dierent
structures and physical properties as the four valence electrons of carbon can form bonds with
surrounding atoms in dierent ways. They are rst of all characterized by dierent hybridizations
of the electron in the s -orbital and the three electrons in the p -orbital. Diamond for instance
has sp 3 -hybridized carbon atoms. The well known graphite, a stack of single graphene layers, on
the contrary shows sp 2 -hybridization whereas the fourth electron stays in the pz -orbital. These
orbitals are forming the π -bands in the solid that are responsible for most of graphene's amazing
properties as they show a linear dispersion at the Fermi level. Since graphene is also the building
block of fullerenes and carbon nanotubes, it can be seen as the mother of all graphitic materials.
R.E. Smalley, who was awarded the Nobel Prize in chemistry in 1996 for the discovery of the
fullerenes, realized the potential of graphene by stating in his Nobel lecture [5]: "Carbon has
2 Chapter 1. Introduction
this genius of making a chemically stable two-dimensional, one-atom thick membrane in a three-
dimensional world. And that, I believe, is going to be very important in the future of chemistry
and technology in general."
The rst pioneering theoretical investigation of graphene's band structure has been performed
by P.R. Wallace already in the year 1947 [6] and is still the starting point for an understanding of
graphene's astonishing properties. The scientic interest at that time was a completely dierent
one than nowadays since the rst nuclear reactors had been built and graphite was used as a
neutron moderator to control the nuclear reactions. In the following decades graphene has for
sure been synthesized in one or the other way. However, the potential of graphene has not been
recognized so that graphitized samples were often regarded as dirty samples. The potential of
graphene as a system to study unique physical phenomena that are based on QED has been
realized only in the 1980s [7]. Induced by the growing interest in nanoelectronics and carbon
nanoelectronics in particular research eorts to isolate graphene have been intensied around ten
years ago mainly by W.A. de Heer and A. Geim. The rst successful report of graphene isolation
and the subsequent characterization has been published in 2004 by Geim's group [1]. Graphene
was produced by so-called mechanical exfoliation which is amazingly simple as the graphene is just
peeled o a graphite crystal by using a sticky tape and then pressed on a SiO2 substrate. Note
that in principle in every pencil drawing graphene akes are produced. It's an irony of history that
after the invention of the pencil in the 16th century and the later naming of graphite from the
greek word graphein (draw/write), the graphite based material graphene could nowadays have
another impact on information technology, this time via nanoelectronics.
For practical applications, however, it is instrumental to produce graphene in large quantities,
which is not possible by the method of mechanical exfoliation. Among others, one of the most
promising alternatives is the graphitization of silicon carbide (SiC) crystals by thermal decomposi-
tion. W.A. de Heer's group was the rst to show experimental evidence of graphene properties on
high temperature treated SiC samples [4, 8]. Owing to the higher vapor pressure Si is evaporating
away and leaves a surface covered with graphene behind. In contrast to exfoliated graphene it
is called epitaxial graphene, although it does not follow the original meaning of epitaxial growth,
which would imply the deposition of carbon in this case. SiC is a technologically well adapted wide
band gap semiconductor with extraordinary electronic properties so that SiC wafers are available
in large quantities. Although never reaching the importance of Si device technology, SiC has
found its niche markets in high power, high frequency and high temperature device applications.
Epitaxial graphene on SiC could therefore directly be implemented in existing device technology
and industrial processes. More details about the material systems graphene and SiC are given in
Chapter 2.
Motivated by the promising prospects of epitaxial graphene on SiC the goal of my thesis is to
oer a detailed insight into the growth of graphene layers on SiC, the precise chracterization of
their structural and electronic properties and eventually to manipulate their qualities according
to technological relevance. Since a two-dimensional crystal such as graphene can be seen as an
ideal surface, the appropriate investigation methods are all surface science related techniques.
These methods are introduced in detail in Chapter 3. The research eld of surface science is
a sub-area of solid state research. The latter evolved in the rst decades of the 20th century
with the discovery of quantum mechanics as the basis of a microscopic description of nature. In
contrast to investigation methods of bulk materials the access to an analysis of surfaces turned
out to be more complicated and started only some decades later, approximately in the 1960s. The
main reason can be found in the necessity to develop ultra-high vacuum (UHV) techniques for the
preparation and analysis of clean surfaces. Within a very short time surface science has become
3
a rapidly developing research area and is of enormous importance for technological applications
such as catalytic reactions, corrosion eects and nanotechnology in general. Such advances in
surface science were awarded by the nobel prize in chemistry to G. Ertl in 2007. The methods
used in the framework of my thesis are low energy electron diraction (LEED) and microscopy
(LEEM), core level (CLPES) and angular resolved photoemission spectroscopy (ARPES), atomic
force microscopy (AFM), scanning tunneling microscopy (STM) and spectroscopy (STS) and
Raman spectroscopy.
The understanding of the structural and electronic properties of clean SiC surfaces is an
important precondition for a dedicated analysis of their graphitization upon high temperature
treatment. In the last 30 years a common understanding has developed with respect to the
so-called basal plane surfaces of SiC, the Si-terminated SiC(0001) side and the C-terminated
SiC(0001̄) side [9]. In surface phase diagrams the atomic structure of the surface is analyzed in
dependence of the temperature of the SiC crystal. The general observation is the existence of
more carbon rich surface structures for higher annealing temperatures. However, it is important to
note that SiC(0001) and SiC(0001̄) exhibit completely dierent properties in terms of the ordering
of the atoms at the surface. Similarly the graphene growth characteristics are dierent for the two
faces of the basal plane surfaces. In my work, the focus will mainly be on epitaxial graphene on
SiC(0001). An extensive analysis is provided in Chapter 4. It concentrates on epitaxial graphene
samples prepared by annealing in UHV conditions, which was the most frequently used preparation
technique throughout my work. Although epitaxial graphene on SiC(0001) can be grown on a
large scale with a well dened number of layers and high structural order, there has been much
controversy about the interface between√graphene√ and◦ SiC. The interface is represented by a
carbon rich surface structure, named (6 3×6 3)R30 reconstruction, and turned out to be a
covalently bound initial carbon layer, which does not yet give access to the typical graphene π -
bands. This precursor phase of graphitization is discussed in Section 4.2, the subsequent graphene
growth on top of this interface layer in Section 4.3.
The rst evidence of graphene properties on SiC has been conducted via electrical transport
measurements on SiC(0001̄) [4]. Here, the graphene is growing with rotational disorder, limited
thickness control and without the presence of an interface layer. However, the rotational stacking
leads to an electronic decoupling of the graphene layers so that even for multilayer lms the
typical properties of single layer graphene can be observed. To account for the importance of
other growth conditions than the one shown in Chapter 4, an overview of alternative approaches
to synthesize epitaxial graphene on SiC is presented in Chapter 5. Section 5.1 is dedicated to the
variation of the substrate. In particular, the initial stages of the graphitization of SiC(0001̄) is
discussed. Furthermore, a short outlook to the graphitization of non-basal plane surfaces, which
are surfaces obtained after a cut of the bulk SiC crystal away from the (0001) surface direction,
is given. Section 5.2 oers two alternative preparation techniques: molecular beam epitaxy using
direct carbon evaporation in UHV and graphitization in an argon atmosphere, which has recently
been shown to yield large terrace graphene surfaces [10].
Epitaxial graphene on SiC(0001) suers from the already mentioned interface layer between
graphene and SiC. The interface layer is one of the primary suspects for the lower electron
mobilities in comparison to graphene exfoliated from graphite. Furthermore the presence of the
interface layer causes n-doping, which is unfavorable for the application in electronic devices.
Chapter 6 provides solutions to these problems. As elaborated in Section 6.1 the deposition of
antimony, bismuth or of the strong electron acceptor molecule F4-TCNQ induces hole doping
similar to a chemical gating technique. Conditions around charge neutrality can be reached for the
deposition of F4-TCNQ. The actual problem given by the presence of the interface layer can be
4 Chapter 1. Introduction
solved by a manipulation of this layer on an atomic level. The elimination of the covalent bonding
at the interface in order to decouple the graphene layers from the substrate would require to break
and saturate the respective bonds. In Section 6.2 it is demonstrated that both hydrogen and
gold intercalation can induce the desired decoupling. Hydrogen indeed passivates the interface
between graphene and SiC whereas the role of the intercalated Au is not a priori clear. This
tuning of the material properties of epitaxial graphene on SiC(0001) paves the way for a bright
future of both fundamental and application based studies of epitaxial graphene on SiC. This result
and the ndings of the previous chapters are nally summarized in Chapter 7.
Chapter 2
Epitaxial graphene a new material
In this chapter, the physical properties of the two material systems, i.e. graphene, a single atomic
layer of graphite, and silicon carbide (SiC) are reviewed. The structural and electronic properties
of graphene as well as dierent approaches to synthesize graphene are presented. For SiC special
attention is drawn to its surfaces as they are the basis for the understanding of the epitaxial
growth of graphene on SiC.
2.1 Introduction to graphene

2.1.1 Graphene physics
Carbon allotropes with sp2 -hybridization
Since the element carbon has the extraordinary property of possessing many appearances in nature
it gives access to study a large variety of technological and physical aspects. Carbon exists in
dierent allotropes, i.e. in dierent structural modications. Those with sp 2 -hybridization are
displayed in Fig. 2.1. The fullerenes have been discovered in 1985 and can be seen as zero-
dimensional objects. The rst experimental report about the one-dimensional carbon nanotubes
dates back to 1952 but only around fourty years later multi-walled (1991) and single-walled
carbon nanotubes (1993) have been brought into the awareness of the scientic community as
a whole. In the three-dimensional world carbon exists in the well known graphite. Pulling out
a single atomic plane out of bulk graphite results in a two-dimensional crystal, which is named
graphene. The notion graphene is derived from the sp 2 -hybridization of the carbon atoms. It has
been ocally named in 1994 by the International Union of Pure and Applied Chemistry (IUPAC)
and has replaced other notions such as graphite layers, carbon layers or carbon sheets. The
notion graphene is typically used for up to ten graphene layers [11], under certain circumstances
it can even be more [12]. Graphite in turn is made out of stacks of graphene layers that are
weakly coupled by van der Waals forces. Graphene is not only the building block of graphite
but also of fullerenes and carbon nanotubes. Carbon nanotubes can be regarded as a rolled-up
graphene sheet. Very recently, it has successfully been demonstrated to unzip carbon nanotubes
into graphene [13, 14]. Used as a theoretical model system for several decades [6], graphene is
the rst truly two-dimensional crystal that has been experimentally isolated [1] in the year 2004.
The exfoliation of highly oriented pyrolytic graphite (HOPG) and the transfer of micrometer
sized akes to a SiO2 wafer was the rst technique to unambiguously isolate and characterize
monolayer graphene [13]. Nearly at the same time, the potential of SiC as a substrate for the
6 Chapter 2. Epitaxial graphene a new material
Figure 2.1: Carbon allotropes with sp 2 -hybridized bond conguration. The building block of the zero-
dimensional fullerenes, the one-dimensional carbon nanotubes and the three-dimensional graphite is the
two-dimensional graphene.
epitaxial growth of graphene has been realized [4, 8]. Graphene can even be produced completely
free-standing by suspending the graphene akes [15]. It should be noted that for a long time
graphene has been presumed to be thermodynamically unstable as a free-standing layer [16, 17].
However, the recent progress in graphene isolation is not a contradiction to theory since the
stability of the graphene is given by crumpling or wrinkling [15].
With the discovery of graphene's spectacular properties, which will be elaborated below, a new
research eld has been dened, which develops incredibly fast and still has enormous potential.
Figure 2.2 compares the amount of literature for fullerenes, carbon nanotubes and graphene for
the last 19 years. The data is based on the Science Citation Index (SCI) and the Conference
Proceedings Citation Index, Science (CPCI-S). For comparison, the development of the total
number of scientic articles covered by the SCI is displayed as a rough measure for the growth
of scientic literature. It is evident that the research topic graphene attracts tremendous interest
and will have a lasting eect on the research landscape. In the year 2009 the number of graphene
related articles has already outperformed the number of fullerene related articles. Furthermore,
the Science magazine elected the recent progress in graphene research as one runner's up for the
breakthrough of the year in 2009.
The electronic structure of monolayer graphene

Besides being the rst one-atom-thick material that has been experimentally investigated, it is
the electronic structure of graphene that stands behind its amazing prospects oered both for
fundamental physics and technology. The graphene lattice structure, despite or because being
quite simple, is the reason for its peculiar electronic structure. The sp 2 -hybridization between
one s orbital and the px - and py - orbitals leads to a hexagonal planar structure. The σ -bonds
2.1. Introduction to graphene 7
Figure 2.2: Time-dependent number of articles dealing with fullerenes, nanotubes and graphene in either
the abstract or the title. The data is based on the Science Citation Index (SCI) and the Conference
Proceedings Citation Index, Science (CPCI-S). The total number of articles covered by the SCI is shown
as a rough measure for the growth of scientic literature. Source: SCI and CPCI-S under Web of Science
(WoS) and STN International (SCI growth). The image is an updated version of the one shown in
Ref. [18] (The help by W. Marx is gratefully acknowledged).
Figure 2.3: Sketch of the graphene lattice structure with its sp 2 -hybridized carbon atoms. The sp 2 -
orbitals form σ -bonds, the unhybridized pz -orbital the π -bonds.
Figure 2.4: (a) Honeycomb lattice structure of graphene with a unit cell comprising a basis of two
atoms, which leads to two equivalent sublattices. The nature of the crystal structure stands behind all
the spectacular properties of graphene. (b) Graphene Brillouin zone in reciprocal space together with
the reciprocal unit vectors.
are formed by the in-plane sp 2 -orbitals as sketched in Fig. 2.3. These bonds are responsible for
the robustness of graphene's lattice structure. As the σ -bands have a lled shell due to the
Pauli principle, they form a deep valence band far away from the Fermi level. The unaected
pz -orbital is oriented perpendicular to the planar structure and forms covalent π -bonds with the
neighbouring carbon atoms. These bonds result in the π -band, which is half lled and responsible
for the conductivity in graphene.
The hexagonal graphene lattice is displayed in top view in Fig. 2.4 (a) and exhibits a basis
of two atoms per unit cell. As a consequence, graphene consists of two equivalent sublattices
marked A (green) and B (red). The lattice vectors of graphene can be written in the following
way:
a √ a √
~a1 = ( 3, −3), ~a2 = ( 3, 3), (2.1)
2 2
with a ≈ 1.42 Å being the C-C bond length. The corresponding reciprocal lattice of graphene
and its Brillouin zone is shown in Fig. 2.4. The reciprocal lattice vectors are
√ √
~b1 = 2π ( 3, −1), ~b2 = 2π ( 3, 1). (2.2)
3a 3a
The corners of the Brillouin zone are of particular importance for the physics of graphene. Due
0
to the two graphene sublattices these corners are the K- and K -point, respectively and are often
designated as dierent "valleys".
The electronic band structure can quite easily be determined from an analytical tight binding
approach. The rst tight binding description was already given by P.R. Wallace in 1947 [6] and
contains all the outstanding properties of graphene's electronic structure. P.R. Wallace considered
nearest- and next-nearest-neighbor interaction for the graphene pz -orbitals, but neglected the
overlap between wave functions centered at dierent atoms. The nowadays probably better known
tight binding approximation by Saito et al. [19] includes this overlap integral, but considers only
interactions between nearest neighbors within the graphene sheet. The dispersion relation E (k)
of the π -bands as calculated from the original description by P.R. Wallace is given by:
q
0
~
E± (k) = ± t 3 + f (~k) − t f (~k), (2.3)
Figure 2.5: (a) The graphene π -band dispersion in three-dimensional representation as calculated from
the original tight-binding approach by P.R. Wallace [6]. According to Ref. [20, 21] the nearest and next
nearest-neighbor interaction parameters are set to t = 2.7 eV and t = 0.2 t , respectively. The inset
0
shows a zoomed region close to the K-point where a linear band structure can be observed. (b) The π -
and σ -bands of graphene as calculated from the tight-binding approach by Saito et al. [19]. The image
is taken from Ref. [19].
√
√ 3 3
with f (~k) = 2 cos ( 3ky a) + 4 cos ( ky a) cos ( kx a), (2.4)
2 2
where the plus sign belongs to the upper conduction (π ∗ -) and the minus sign to the lower
valence (π -) band, the parameter t is the nearest-neighbor hopping energy (hopping between
0
dierent sublattices) and t is the next nearest-neighbor hopping energy (hopping in the same
0
sublattice). The corresponding band structure is plotted in Fig. 2.5 (a) with t = 2.7 eV and t =
0.2 t [20, 21]. If the next nearest-neighbor interaction is neglected, the valence and the conduction
bands become symmetric, a representation, which is also quite often used in visualizations of
graphene's band structure. A non-zero overlap integral similarly breaks the symmetry between
the π - and π ∗ -bands [19]. Both approaches by Wallace and Saito et al. give a quite good approach
for the energy dispersion of the π -bands. For a more accurate description, in particular away from
the high symmetry directions, ab initio calculations are necessary or interactions up to the third
nearest neighbor have to be included in the tight binding descriptions [20]. To have a complete
overview of graphene's valence band structure, Fig. 2.5 (b) shows both the π - and σ -bands of
graphene as determined from the tight binding description by Saito et al.. As mentioned earlier
the three σ -bands are far below the Fermi level EF (EF =0). The π - and π ∗ -bands however meet
0
each other at the Fermi energy at the K- and K -points. This crossing point is called Dirac point
(ED ) for reasons that become clear below. As the involved electrons belong to two dierent
sublattices, an energy gap is not opened. Consequently, graphene can be seen as a zero-gap
semiconductor or as a zero-overlap semimetal. As highlighted by the zoom in Fig. 2.5 (a), the
π -bands exhibit another peculiarity, namely a linear dispersion around the Dirac point. A rst
0
order expansion around the K- (or K -) point of E (K ~ + ~q ) with |~q | |K~ | yields
E± (~q ) ≈ ± h̄vF |~q | , (2.5)

~ denotes the position of
where vF = 3ta/2h̄ is the so-called Fermi velocity and the vector K
the K-point in momentum space with its origin located at Γ. The Fermi velocity amounts to
around 1 · 106 m/s. For condensed matter systems the parabolic dispersion h̄2 q 2 /(2m) would
usually be expected, where m is the electron mass. The velocity in such a case depends on the
energy: v = 2E /m. Before discussing the consequences of the linear dispersion of the π -bands
p
around the Dirac point, it should be noted that a second order expansion of E (K ~ + ~q ) results in
a direction dependence in momentum space. Constant energy maps show a threefold symmetry,
which is called trigonal warping. Section 4.3.2 will give more details in this respect.
"Relativistic condensed matter physics"

The most striking feature of graphene's band structure is the linear dispersion of the π -bands
around the Dirac point, which has rst been shown by P.R. Wallace in 1947 [6]. The conical
shape resembles the light cones in the dispersion relation for photons. Only in 1984 it has been
realized that graphene's energy dispersion resembles the energy of ultrarelativistic particles, which
are described by the massless Dirac-Weyl equation [7]. It should be noted that there is nothing
particular relativistic about the electrons in graphene, it however is their interaction with the
periodic potential of the honeycomb lattice that results in quasiparticles with a formal equivalence
to the (2+1)-dimensional (two spatial dimensions and one time dimension) Dirac-Weyl equation
in quantum electrodynamics (QED). The Dirac like Hamiltonian around the K-point is given by:

0 qx − iqy
Ĥ = h̄vF ~σ · ~q = h̄vF , (2.6)
qx + iqy 0
where ~σ is the two-dimensional Pauli matrix. The corresponding wave function is

−iθ /2
1 e ~q
Ψ±,K~ (~q ) = √ , θ~q = arctan(qx /qy ) , (2.7)
2 ±e iθ~q /2
where the sign ± corresponds to the π ∗ - and π -bands and θ~q is a phase factor. If the phase θ is
rotated by 2π , the wave function changes its sign, which indicates a phase of π . This phase is
called "Berry's phase". Furthermore, it should be noted that time reversal transforms the wave
0
function at the K-point into the wavefunction at the K -point so that the two valleys are related by
time-reversal symmetry. The wave function is a two-component spinor, which takes the relative
contributions of the dierent sublattices A and B into account. This description resembles the
spin in QED. However, in the case of graphene this is just a formal index and is not a real spin
so that the notion "pseudospin" is used. Considering the Dirac cone in Fig. 2.5 (a) an electron
with momentum |~q | and a hole with momentum −|~q | have the same pseudospin underlying
the fact that they originate from the same sublattice. In other words, the same pseudospin is
given for electrons and holes originating from the same branch (with one branch being dened
by Equation 2.5). The physical potential of the experimental isolation of graphene in 2004 by
Novoselov et al. [1] is that graphene allows to study QED eects in a condensed matter system,
which represents a new paradigm in solid state physics. The quasiparticles in graphene, which are
massless Dirac fermions, can be seen as electrons that have lost their rest mass or as neutrinos
that acquired the charge of an electron (neglecting the tiny mass of neutrinos). The occurrence
of phenomena in QED is often inversely proportional to the speed of light c . As vF ≈ 1/300 c
these eects are strongly enhanced in graphene and should dominate those eects related to the
true spin [11].
Another important quantity in QED is the chirality, dened as the projection of ~σ on the
direction of motion ~q . The direction of the quasiparticle pseudospin is either parallel (positive
chirality) or antiparallel (negative chirality) to their momentum, with states in graphene below the
Dirac point having a positive chirality (+1) and states above the Dirac point having a negative
chirality (-1). The quasiparticles around the Dirac point thus have a well dened chirality so
that the chirality represents a good quantum number. Considering again the Dirac cone of
Fig. 2.5 (a) it can be concluded that, for EF = ED , electrons (or equivalently holes) of opposite
momentum direction have opposite pseudospin, but the same chirality. The chirality in graphene
has important implications for the scattering of the charge carriers. As calculated from the two
component wave function, the angular scattering probability in monolayer graphene is proportional
to cos 2 (φ/2). Hence backscattering (φ = π ) is forbidden. Another consequence is the suppression
of weak localization close to the Dirac point [22] and the absence of a quantum-interference
(localization) magnetoresistance. Note that in most metals a negative magnetoresistance can be
observed, i.e. the conductivity increases when a small magnetic eld is applied. The concepts of
chirality and pseudospin have direct consequences for the interpretation of band structure and
scanning tunneling microscopy measurements as discussed in Chapter 4. The shape of constant
energy maps, which can be seen as constant energy cuts through the Dirac cone of Fig. 2.5 (a)
as well as quantum interference patterns in scanning tunneling microscopy are governed by this
physical background.
From a more general point of view, the analogy to QED allows to study relativistic quantum
mechanical eects in a condensed matter system. The Klein paradox and the so-called "Zitter-
bewegung" are such phenomena, which are unobservable in particle physics [11]. Klein tunneling
describes perfect tunneling of relativistic electrons through arbitrarily high and wide barriers [23]
and is related to the suppression of backscattering. The Klein tunneling is obviously counterintu-
itive and in stark contrast to non-relativistic quantum mechanics. It has recently been veried in
experiment [24, 25]. The "Zitterbewegung" describes jittery movements of a relativistic electron
due to interference between parts of its wavepacket belonging to positive (electron) and negative
(positron) energy states [11, 26].
Graphene - a playground for new applications

For the interpretation of experiments and for the application in electronic devices the understand-
ing and the control of the charge carrier density and the density of states is an important issue.
For electron gases in general, the number of electrons at EF for T = 0 is basically given by
deviding the k -space volume Vk of all states by the k -space volume Ωk of one state [27]. In
case of graphene, valley and spin degeneracy contribute each with a factor of 2. Furthermore,
the two-dimensionality of the electron system and the linear dispersion relation E (~q ) = h̄vF q
have to be taken into account. The concentration of electrons at the Fermi energy EF (i.e. the
number of electrons per volume V ) is then straight away given by
4 Vk 4 qF2 π qF2 EF2
n = = = = . (2.8)
V Ωk V ((2π)2 /V ) π h̄2 vF2 π
Hereby qF denotes the electron momentum with respect to the K-point at EF . The density of
states (DOS) as the derivative of the energy dependent concentration n(E ) consequently amounts
to
2E
ρ(E ) = 2 2 . (2.9)
h̄ vF π
The DOS shows a linear behavior. It should be noted that a graphene sheet rolled up to a
carbon nanotube exhibits a dierent density of states due to its one-dimensional nature. The
very rst transport experiments on graphene have shown a strong ambipolar eld eect such
that the charge carrier concentration can be tuned continuously in a wide range from zero up
to ≈ 1013 cm−2 for electrons as well as holes. The variation of the so-called cyclotron mass in
dependence of the charge carrier concentration have shown the rst evidence for massless Dirac
fermions in graphene [2, 3]. For epitaxial graphene on SiC(0001), the graphene is intrinsically
n-doped directly after growth (see Section 4.3.2), which can be overcome by chemical doping
with metals or molecules as will be elaborated in Section 6.1.
One of the most appealing properties of graphene is its excellent electronic quality. Mobilities
up to 200.000 cm2 /Vs have been measured for graphene exfoliated from HOPG [28, 29]. For
comparison, the mobility in Si is around 1.400 cm2 /Vs [30], in GaAs around 8.500 cm2 /Vs [30]
and in InSb around 78.000 cm2 /Vs [31]. Even more important is the fact that the mobility is
rather unaected by temperature and upon changing the carrier concentration by both electrical
or chemical doping [32]. The electronic transport is ballistic on the submicrometer scale [11].
For epitaxial graphene on SiC(0001) mobilities up to 29.000 cm2 /Vs have been measured [33],
for multilayer epitaxial graphene on SiC(0001̄) mobilities even up to 250.000 cm2 /Vs have been
determined from far infrared transmission experiments [34].
It is probably exaggerated to regard graphene nanoelectronics as the successor of silicon
microelectronics. However, for some specic niche markets graphene could outperform other
materials in the near future. One important drawback of monolayer graphene is its gapless
spectrum, which is the reason for low on/o-ratios in graphene eld eect transistors. A solution
is oered by quantum connement as graphene nanoribbons exhibit a band gap in dependence
of their width [35, 36]. Another possibility for the opening of gaps in the π -band system would
be a chemical modication, which has recently been shown for hydrogenated graphene, called
graphane [37].
A further indication of a high quality of an electronic system is the observation of the quantum
Hall eect (QHE). For graphene exfoliated from HOPG, it has been observed quite early [2, 3]
and even at room temperature [38] whereas for epitaxial graphene on SiC(0001) and SiC(0001̄) it
took some more eorts [33, 3941]. For monolayer graphene, a new type of quantum Hall eect
can be observed. Due to the analogy to QED it is a relativistic counterpart of the conventional
integer QHE and is named a "chiral QHE" or an "anomalous QHE". The QHE sequence is
shifted by 1/2 with respect
√ to the standard QHE so that it is called a "half-integer QHE". The
Landau levels have a N dependence, where N is the Landau level index. Very recently, due to
improvements in the sample preparation for exfoliated graphene also the fractional QHE could
be observed [42, 43].
Another important observation is that graphene's conductivity stays nite even if the concen-
tration of charge carriers tends towards zero. The already mentioned "Zitterbewegung" can be
responsible for this observation. Whereas theory suggests a minimum conductivity of 4e 2 /hπ the
experimental data give 4e 2 /h [11]. This so-called "missing pie"-problem is not yet solved up to
now.
Only recently, researchers have started to focus on the non-electronic properties of graphene,
which are equally impressive as the electronic properties. Graphene has a breaking strength of
≈ 40 N/m [44], which is at the theoretical limit. The room-temperature thermal conductivity
(≈ 5000 W/(mK) [44]) and Young's modulus (≈ 1.0 TPa [45]) are also exhibiting record values.
Graphene can be stretched elastically by 20% more than any other crystal [44]. Unlike any other
material, graphene expands with decreasing temperature. The consequence for high temperature
grown epitaxial graphene on SiC is the formation of compressive strain and will be discussed in
Section 4.3.6. Also important is the observation that graphene is impermeable to gases [46] (see
Figure 2.6: The π -bands of ideal (a) monolayer and (b) bilayer graphene in the vicinity of the K-point as
calculated from a tight binding approach. Whereas monolayer graphene shows a linear dispersion, bilayer
graphene exhibits a parabolic spectrum. Both systems are a zero-gap semiconductor or a zero-overlap
semimetal.
also Section 6.2.1) and can even be used as a gas sensor [32].
Bilayer graphene and graphene stacks

In this section, only the properties of monolayer graphene have been discussed so far. However,
bilayer, trilayer and few layer graphene can be regarded as graphene as well. Only at around ten
graphene layers the notion "thin lms of graphite" or "graphite" is appropriate [11, 47]. Bilayer
graphene is of particular interest as the nature of its quasiparticles is comparably spectacular to
monolayer graphene. Let us rst consider the electronic structure of AB stacked bilayer graphene
(Bernal type stacking). The (4×4) tight-binding bilayer Hamiltonian close to the K-point gives
the following four band contributions [48, 49]:
γ12 U2 v32 √ 1/2

α (q) = ± + + 2
vF + 2 2 α
h̄ q + (−1) Γ . (2.10)
2 4 2
The band index α=1,2 is due to the fact that valence and conduction band now consist of two
bands each. Γ is given by the following expression:
1 2
Γ = (γ − v32 h̄2 q 2 )2 + vF2 h̄2 q 2 (γ12 + U 2 + v32 h̄2 q 2 ) + 2γ1 h̄3 v3 vF2 q 3 cos(3ϕ) . (2.11)
4 1
The value v3 amounts to: √
v3 = 3aγ3 /2h̄. (2.12)
γ1 is the out-of plane nearest neighbor interaction parameter in eV, γ3 is the out-of plane next-
nearest neighbor interaction parameter in eV. U (eV) is the dierence in on-site Coulomb potentials
of the two layers and is zero for ideal bilayer graphene. The Fermi velocity amounts to vF = 1.07·
106 m/s and has been determined from a t to band structure measurements (cf. Section 6.1.2).
In the same way, the other parameters have been set to the following values: γ1 = 0.4 eV,
γ3 = 0.12 eV. The azimuthal angle ϕ is set to 30◦ and refers to a direction perpendicular to ΓK.
It, however, has only a minor inuence on the shape of the bands in the vicinity of the K-point.
The graphene lattice constant a amounts to 2.46 Å. Fig. 2.6 compares the band structure of
ideal monolayer and bilayer graphene in the vicinity of the K-point as calculated from the tight
binding parametrization. The monolayer band structure is just a two-dimensional projection of
the three-dimensional Dirac cone shown in Fig. 2.5 (a). For bilayer graphene, the two valence and
two conduction bands meet at the Fermi level without overlap or gap formation. The bands are
not linear any more but have a parabolic shape. In fact, the quasiparticles in bilayer graphene are
not massless but massive chiral Dirac fermions [50]. Consequently, a dierent but also new type
of QHE can be observed, where the plateau at the Landau level N = 0 is missing. The Landau
quantization has a N(N − 1) dependence. In contrast to monolayer graphene, backscattering
p
is allowed. Furthermore, for bilayer graphene, the same minimum conductivity as for monolayer
graphene is measured indicating that the minimum conductivity is not related to the linear band
structure but to the chiral character of the quasiparticles instead. Whereas ideal bilayer graphene
does not have a band gap, it is however possible to induce a band gap with a size reaching several
hundred meV by changing the dierence U of the on-site Coulomb potentials of the two layers.
Let us consider the example of epitaxial bilayer graphene on SiC. The already mentioned intrinsic
n-doping of epitaxial graphene on SiC is caused by a dipole at the graphene-SiC interface. Due to
screening, the upper graphene layer experiences a lower charging in comparison to the lower one.
The resulting dierence in the on-site Coulomb potentials between the two layers opens a band
gap of around 120 meV in size. By means of surface transfer doping for deposited n- or p-type
dopants, a precise control over the band gap size can be obtained. This has been demonstrated
for the deposition of the n-dopant potassium [49], or, as elaborated in Section 6.1.2 and Ref. [51],
for the p-dopant molecule F4-TCNQ. In transport measurements, similar results can be obtained
by changing the carrier concentration from a backgate [52], or more sophistically, by changing
the charge carrier concentration for the upper and lower graphene layer independently by top-
and backgate [53].
With increasing number of graphene layers the band structure gets more complicated. Three
valence and conduction bands are present in trilayer graphene, four in four layer graphene and so
on. It has to be considered, however, that the stacking between the layers does not need to be
Bernal type as in graphite, it can also be rhombohedral (ABCABC). This signicantly inuences
the detailed appearance of the band structure [47, 54, 55]. It is furthermore important to note
that slight rotations between the dierent layers with respect to each other result in a decoupling
of the layers [56]. This so-called turbostratic stacking is observed for epitaxial graphene on
SiC(0001̄). In such a case, even multilayer graphene exhibits the linear band structure and
quasiparticle chirality of monolayer graphene [4, 57].
2.1.2 Synthesis of graphene

The graphene synthesis can be devided into two principal routes: the rst one is to split the
strongly layered material graphite into individual atomic planes, and the second one is to epitaxially
grow the graphene on top of other crystals. It has to be emphasized that the production and
isolation of graphene has a longer history in science than suggested by the recent graphene hype.
Indeed, years or even decades before the nowadays called "discovery of graphene" by Novoselov et
al. in 2004 [1], graphene had been synthesized in dierent ways. In this previous graphene related
research, however, either monolayer graphene was not unambiguously isolated or the potential
of graphene was not realized so that graphitized samples were partially even regarded as "dirty"
samples. In that respect the true value of the work in 2004 is not the "discovery of graphene"
but rather the denition of a new research eld by presenting an easy way to produce and identify
graphene as well as by accessing the amazing physical and technological potential of graphene in
the follow-up publications by A. Geim's group and P. Kim's group [2, 3].
The isolation of graphene by Novoselov et al. [1] is based on mechanical exfoliation. It consists
of repeated peeling with an adhesive tape on a graphite crystal such as HOPG. Usually, a graphite
powder with macroscopically thick akes is used instead of a solid graphite crystal. The simplicity
of this technique is remarkable and is often compared to drawing with a pencil. In a next step,
the exfoliated graphene akes on the adhesive tape are transferred to a SiO2 surface by gently
pressing the tape onto an oxidized Si-wafer. Such an insulating substrate is ideal for electrical
transport measurements. The critical point is to distinguish between multilayer graphene akes
and single layer graphene akes as it is evident that only a tiny percentage of the transferred
akes is of monolayer character. It has been found that monolayer graphene can be detected
from the contrast in an optical microscope, provided the right thickness of 300 nm of the SiO2
lm is used. With some improvements of this preparation technique during the past years, the
largest akes now reach millimeter size [58].
Whereas the mechanical exfoliation is still widely used and the technique of choice for basic
research and proof-of-concept devices, it is not suitable for large scale production and applica-
tions in nanoelectronics. Instead of cleaving graphite manually, the process could be automated.
Ultrasonic cleavage, for instance, allows the production on an industrial scale [59]. It can be used
to make polycrystalline graphene lms and graphene composite materials [59, 60]. Chemical ap-
proaches for graphene production indeed seem to be quite promising. Also the chemical reduction
of graphite oxide [61, 62] is suitable for a large scale graphene synthesis. Graphite is rst oxi-
dized, which facilitates the liquid phase exfoliation, and is then chemically reduced to graphene.
It should be noted that a quite similar approach was already presented in 1962 by Boehm et
al. [63]. The thickness of the generated graphene akes was determined by transmission electron
microscopy, and presumably indeed showed graphene monolayers. This report can probably be
seen as the rst isolation of graphene, however without attracting much attention. Incidentally,
the principle of exfoliating graphite is even much older and dates back to the 19th century [64].
Graphene can be grown epitaxially on various metallic substrates by chemical vapor depo-
sition of hydrocarbons. This approach has been well known before the potential of graphene
has been realized and has subsequently been rediscovered. In particular, graphene can be
grown on Pt(111) [6567], Ir(111) [6769], Ni(111) [7072], Rh(111) [67], Ru(0001) [67, 73],
TiC(111) [74], Co(0001) [75] and on Cu [76]. One advantage is that the growth is self-limiting
in many cases so that a dened monolayer of graphene can be grown. For technological appli-
cations, however, this method cannot be used in a straight forward manner as the substrate is
conducting. First attempts are on the way to grow the metal on a cheap and insulating substrate
(such as MgO). After the growth of a graphene layer on the metallic lm the metal is etched
away so that the electronic properties are accessible via the insulating substrate [77, 78].
One of the most promising approaches for large area production of graphene on a (semi-)insu-
lating substrate is the epitaxial growth of graphene by high temperature annealing of SiC wafers.
Due to the higher vapor pressure of silicon, the silicon evaporates and leaves graphene layers
behind. The details of this process as well as the characterization of epitaxial graphene on SiC
and its manipulation according to technological relevance is discussed in the following chapters.
SiC wafers are available in large quantities and the device fabrication can be based on existing
industrial processes. While graphene akes exfoliated from HOPG are already commercially
available (cf. www.grapheneindustries.com), the price for 1 cm2 , however, would amount to
100 million euros. The price for 1 cm2 of epitaxial graphene on SiC, on the contrary, can be
estimated to around 200 euros. In the meantime, the homogeneity of the epitaxial graphene on
SiC has tremendously increased (see Section 5.2.2 and Ref. [10, 79]) and the size of one domain
is only slightly below the size of exfoliated graphene akes. Due to the necessary optimization
in the preparation conditions and due to some uncertainty about the inuence of the substrate
towards the graphene layers, the research on graphene on SiC is a bit behind with respect to
the one for exfoliated graphene but the distance is reduced constantly. It furthermore has to be
considered that the graphene properties and growth mode are completely dierent for dierent
SiC substrates such as SiC(0001) and SiC(0001̄). Epitaxial graphene on SiC(0001̄) can be grown
only with rotational disorder and limited thickness control. Whereas the monolayer band structure
is preserved even for multilayer lms [57], an external electric eld is screened within just a couple
of near-surface layers thus limiting the potential for electronic applications. Epitaxial graphene
on SiC(0001), on the contrary, can be grown on a large scale with a well dened number of layers
and high structural order so that it is ideal for applications in electronics. In terms of history of
epitaxial graphene on SiC, the rst pioneering works with respect to graphitization of SiC crystals
date back to 1962 by D.V. Badami [80] and with respect to SiC surfaces to 1975 by van Bommel
et al. [81]. It should however be mentioned that E. Acheson, who patented the rst method for
SiC production ("carborundum") in 1893, already realized in 1896 that overheating of SiC leads
to the formation of graphite [82]. The potential for graphene based nanoelectronics has been
recognized by W.A. de Heer [12], which lead to the rst publications in 2004 and 2006 concerning
epitaxial graphene on SiC(0001) and SiC(0001̄) [4, 8]. This work was performed independently
from the exfoliation technique rst published by Novoselov et al. in 2004 and 2005 [1, 2].
2.2 Silicon carbide as a template for epitaxial graphene

Silicon carbide (SiC) is a semiconductor composed of silicon and carbon in an equal stoichiometric
ratio. Although both elements are available in large quantities on earth, naturally developed SiC
can only be found in very small amounts, either in some inclusions in minerals and diamond or in
meteorites. H. Moissan was the rst to identify SiC in an meteorite found in 1893 [83], so that this
mineral was called Moissanite. SiC however, can be observed in higher quantities in other parts
of the universe in much larger quantities such as in red giants, in which the light elements have
already been used up in the fusion processes. The rst observation of articial SiC formation can
be dated back to 1824 to J.J. Berzelius [84], the discoverer of silicon. In the following decades,
only non-systematic studies of SiC synthesis in small quantities on a laboratory scale have been
reported (Despretz, 1849; Marsden, 1880; Colson, 1882 [85]). The rst SiC production on an
industrial scale has been initiated by E.G. Acheson in 1893 [86]. By means of furnace heating,
he transformed quartz sand (SiO2 ) and carbon into SiC, which he named "carborundum". It was
initially used to produce whetstones and powdered abrasives.
The realization of the potential of SiC for electronics can be dated back to the discovery
of electroluminescence in 1907 by using SiC [87], which can be seen as an early version of a
light emitting diode (LED). For application in electronics, the reliable production of high quality
SiC crystals is necessary. Important milestones were the Lely growth process of SiC crystals in
1955 [88], the seeded sublimation growth by Tairov and Tsvetkov in 1978 [89], the possibility
to grow single crystal SiC on Si substrates invented by Matsunami et al. in 1981 [90] and the
high-quality epitaxy at low-temperatures on o-axis SiC substrates in 1987 [91]. Today's state
2.2. Silicon carbide as a template for epitaxial graphene 17
Figure 2.7: (a) Si-C tetrahedron with sp 3 -hybridized silicon and carbon atoms, (b) hexagonal SiC bilayer
as the basic building block of the SiC crystal structure, (c) Si-C tetrahedra bonded in identical orientation
(left image) and rotated by 60◦ with respect to each other (right image) leading to cubic (zinc blende)
and hexagonal (wurtzite) SiC crystal structures. The images are adapted from Ref. [92].
of the art size of commercially available SiC wafers is 4 inch (100mm) in diameter and the step
up to 6-inch (150mm) wafers will soon be done.
Despite its success story over the past decades, SiC has never reached the importance that Si
has for device technology. This can mainly be attributed to diculties in the crystal growth and
a low quality of the SiC/SiO2 interface, which is important for the device performance. However,
SiC has found its niche markets in specic areas, where other materials such as Si reach their
limits. SiC is chemically inert, physically robust, it exhibits high carrier mobilities, a large band
gap (2.2 eV - 3.3 eV), a high breakdown voltage, a high thermal conductivity and a high saturation
electron drift velocity. As a consequence, SiC is used in high power, high frequency and high
temperature device applications. Apart from this area of operations, SiC is used as substrate for
LEDs made from GaN, which actually is the largest market for commercial SiC products. Further
applications concern abrasive and cutting tools, break discs in cars and gemstones in jewelry.
Very recently, it has been discovered that SiC shows promising prospects in biocompatibility. It is
already used as a base material for medical implants. One of the future commercial application
areas could be the epitaxial growth of graphene on SiC wafers.
2.2.1 SiC crystal structure

In SiC each atom is bonded to four atoms of the respective other element in tetrahedral con-
guration (cf. Fig. 2.7) (a). As in the case of diamond or silicon the valence electrons in SiC
are sp 3 -hybridized. The Si-C bond length amounts to 1.89 Å, which is slightly below the average
between the Si-Si bond length (2.35 Å) and the C-C bond length (1.54 Å) in the pure materials.
Due to the higher electronegativity of C the covalent bonds have an ionic contribution of around
12%.
Linking the tetrahedra together a layer of planar hexagonal symmetry is formed, in which each
element is exclusively positioned in either the upper or the lower sublayer, as shown in Fig. 2.7 (b).
This SiC bilayer is the basic building block of the SiC crystal structure. The bulk structure is built
up by stacking the SiC bilayers on top of each other in perpendicular direction. The interlayer
distance is 1.89 Å and equals the Si-C bond distance, the SiC bilayer distance determined as the
distance between two silicon (or carbon) atoms projected along the perpendicular axis ("c -axis")
amounts to 2.52 Å.
It has to be considered, however, that there are two ways to arrange two bilayers with respect
to each other as displayed in Fig. 2.7 (c): either the orientation of the Si-C tetrahedra in the
bottom layer coincides with the orientation of the tetrahedra in the top layer (left image of
Fig. 2.7 (c)) or the Si-C tetrahedra of the bottom layer are rotated by 60◦ relative to the top
Figure 2.8: Crystal structure of dierent SiC polytypes displayed parallel to the (112̄0) plane for (a) cubic
3C-SiC (zinc blende structure), (b) hexagonal 4H-SiC and (c) hexagonal 6H-SiC. The layer orientation
and stacking sequence are indicated by the enhanced thick black line parallel to the (112̄0) projection
plane. In the plot, large violet spheres correspond to Si atoms, small green spheres to C atoms. The
images of panel (a)-(c) are taken from Ref. [92]. (d) Hexagonal crystallographic notation, which is used
for SiC crystals independent of the actual lattice symmetry. It is based on four Miller-Bravais indices
a1 ,a2 ,a3 and c with a1 + a2 + a3 = 0. The top view sketch denotes dierent crystal orientations.
Furthermore, the (0001) and the (0001̄) direction are highlighted on the right. The labeling "Si-face"
and "C-face", respectively, is given with respect to panel (a)-(c) and is discussed in Section 2.2.2.
Si GaAs 4H-SiC 6H-SiC 3C-SiC

Lattice parameters a, c [Å] 5.43 5.65 3.08/10.08 3.08/15.12 4.36
Energy gap EG [eV] 1.11 1.43 3.26 3.02 2.2
Thermal conductivity [W/cmK] 1.5 0.5 5.0 5.0 5
Relative dielectric constant s 11.8 12.8 9.7 9.7 9.7
Saturation drift velocity vs [cm/s]·107 1 2 2.7 2 2.7
Electron mobility µn [cm2 /Vs] 1400 8500 900 450 1000
Hole mobility µp [cm2 /Vs] 600 400 100 50
Breakdown eld EB [MV/cm] 0.3 0.4 3.0 3.0
Max. operating temperature T [ ◦ C] 300 460 1240 1240 1240
Table 2.1: Comparison of the physical properties of 4H-SiC, 6H-SiC and 3C-SiC to those of Si and GaAs
at 300K (taken from Ref. [30]).
layer (right image of Fig. 2.7 (c)). The rst stacking type leads to the cubic zinc blende structure,
the latter one to the hexagonal wurtzite structure. Whereas ionic compounds often crystallize in
the wurtzite structure such as ZnO or in the NaCl structure such as PbS, covalent compounds
show a tendency to the zinc blende structure such as GaAs or the diamond structure such as C,
Si or Ge. Comparable with the observation that the Si-C bond includes both covalent and ionic
contributions, no preferential crystal structure exists for SiC. In fact, dierent bilayer stacking
types are practically energetically degenerate so that mixed hexagonal and cubic sequences along
the c -axis can be realized. This phenomenon is called "polytypism", the one-dimensional form of
polymorphism. In the case of SiC the polytypism is strongly pronounced, around 170 polytypes
are known [93]. Preferentially in older literature, the cubic SiC crystal structure is also called
β -SiC whereas all other congurations, hexagonal and rhombohedral structures, are referred to
as α-SiC.
The most common polytypes are shown in Fig. 2.8 (a)-(c) in a side view perspective, i.e. par-
allel to the (112̄0) plane. Hereby, the designation of crystal orientations follows the hexago-
nal notation with four Miller-Bravais indices (a1 a2 a3 c) as introduced in panel (d). Note that
a1 + a2 + a3 = 0. The vector ~c is oriented perpendicular to the plane of the vectors a~1 , a~2 and
a~3 and denes the (0001) orientation. Other common crystal orientations are also indicated in
Fig. 2.8 (d). Now let us have a closer look at the crystal structure of the polytypes 3C, 4H and
6H. The layer orientation and stacking sequence are indicated by the enhanced thick black line
in panel (a)-(c) of Fig. 2.8. Bilayer rotations as in the case of 4H- and 6H-SiC can therefore
easily be identied. In order to distinguish dierent kinds of polytypes, dierent notations exist,
among which the most common one is the Ramsdell notation [94]. It is based on the choice of a
hexagonal unit cell independently of the actual lattice symmetry. The number of bilayers within
this hexagonal unit cell and a letter indicating the true lattice symmetry ("C" for cubic, "H"
for hexagonal or "R" for rhombohedral) are the two parameters necessary to identify dierent
polytypes. Correspondingly, panel (a) of Fig. 2.8 shows a cubic structure with three bilayers per
unit cell (3C-SiC), panel (b) a hexagonal structure with four bilayers per unit cell (4H-SiC) and
panel (c) a hexagonal structure with six bilayers per unit cell (6H-SiC). For further clearness,
each bilayer can be indicated by a letter according to the lateral position of its upper atoms with
respect to the same type of atoms in a reference bilayer named A, which is chosen to be the
topmost bilayer of the surface.
Dierent polytypes exhibit dierent physical properties as shown in Table 2.1. Most impor-
Figure 2.9: Surface termination layer sequences on 4H-SiC(0001): Congurations S2 and S1 diering
by the number of identically oriented bilayers directly at the surface. The layer orientation and stacking
sequence are indicated by the enhanced thick black line parallel to the (112̄0) projection plane. In the
plot, large violet spheres correspond to Si atoms, small green spheres to C atoms. Long bonds drawn
within the projection plane represent single bonds, shorter double lines indicate two bonds directed by
60◦ into and out of the plane. The image is adapted from Ref. [92].
tantly, the stacking sequence aects the band gap of SiC. It ranges from 2.2 eV for 3C-SiC
to 3.26 eV for 4H-SiC. Although the dierent bilayer stacking sequences do not inuence the
lattice parameter a, it signicantly aects the band gap size. In principle, this would oer the
unique possibility to fabricate layered structures of periodically changing band gap. These elec-
tronic heterostructures would not suer from any strain caused by a varying lattice parameter in
the dierent components. However, until now the growth of dierent polytypes can not yet be
controlled suciently. Despite lower electron mobilities in comparison to Si and GaAs, all SiC
polytypes have considerable advantages in terms of breakdown voltage, thermal conductivity and
maximum operating temperature.
2.2.2 SiC surfaces

The bulk structure of SiC has important implications for the surfaces of SiC crystals. The most
common SiC surfaces are the so-called basal plane surfaces developing by cutting the hexagonal
SiC crystal along the (0001) plane, i.e. perpendicular to the c -axis. As indicated in Fig. 2.8 (d)
two dierent surface terminations are possible: the Si-terminated SiC(0001) surface, often called
"Si-face" (looking on top in Fig. 2.8 (a)-(c)), and the C-terminated SiC(0001̄) surface, often
named "C-face" (looking from the bottom in Fig. 2.8 (a)-(c)). The basal plane surfaces are
therefore polar surfaces. Note that (111) oriented 3C-SiC corresponds to the (0001) orientation
in hexagonal polytypes.
To interpret the experimental results from surface analysis techniques, in particular from low
energy electron diraction (LEED), a closer look at the surface termination layer sequences is
instructive. Figure 2.8 has shown that dierent polytypes have a dierent stacking of the near
surface layers. However, even for a single polytype, dierent congurations for the topmost layers
can be present as displayed in Fig. 2.9 for 4H-SiC(0001). It becomes evident that the surface
does not have to be terminated by a full unit cell as drawn in Fig. 2.8 but any other layer in
the sequence of the unit cell might be the topmost layer at the surface. Consequently, for 4H-
SiC(0001) there are four dierent surface terminations possible: two congurations with either
one or two identically oriented bilayers at the very top of the surface, which are denoted S1 and S2,
respectively. Due to a glide plane symmetry in hexagonal polytypes the other two congurations
have the same structure except for a 60◦ rotation of the whole semi-innite crystal so that they
are denoted S1* and S2*. For 6H-SiC(0001), there are even six dierent congurations possible,
named S1, S2, S3 and S1*, S2*, S3*. Dierences in the stacking arrangement have to be taken
into account when comparing calculated and measured LEED intensity spectra. However, already
the interpretation of a standard LEED pattern of 4H- or 6H-SiC(0001) needs the consideration of
Sx and Sx* terminations. Whereas a LEED pattern of a single Sx or Sx* domain is of threefold
symmetry, a sixfold symmetry is observed in a LEED experiment of a hexagonal SiC crystal. It has
to be considered that Sx and Sx* domains are energetically degenerate so that they are usually
equally distributed on the surface1 . The domain size of Sx and Sx* terminations is typically larger
than the transfer width of the LEED electron beam. For distances larger than the transfer width,
which amounts to around 5-10 nm for a conventional LEED optics [96], interference is suppressed
due to a missing phase relation. Therefore, an incoherent superposition of the LEED patterns
takes place resulting in a sixfold symmetry of the observed LEED pattern. On the contrary, LEED
on a microscopic surface area (so-called "micro-LEED" as explained in Section 3.1.1) can probe
domains of one surface termination only and indeed exhibits a threefold symmetry as shown in
Fig. 6.16 (a) of Section 6.2.1 or in Ref. [97, 98]. Incidentally, for the cubic 3C-SiC crystal the
situation is more simple, the threefold symmetry of the crystal can be observed in a normal LEED
pattern since the SiC bilayers are always equally oriented.
The Si or C atoms in the top layer have three bonds to the underlying atoms within the
SiC bilayer, whereas the fourth bond, which is pointing upwards, is lled by one electron only.
Such an unsaturated bond is called "dangling bond". Dangling bonds make the formation of
such a surface energetically unfavorable. As a consequence, these surfaces cannot be observed in
experiments. There are two principle possibilities to circumvent this unfavorable situation. The
rst one is to saturate the dangling bonds e.g. by hydrogen atoms, which has indeed been shown
by Seyller [99]. The second one, which is chosen by nature in the absence of any foreign atoms,
is the atomic rearrangement of the near surface layers. As such a reorganization is accompanied
by the formation of new bonds, it is named "reconstruction". A surface reconstruction or an
adsorption of foreign atoms may induce a dierent periodicity at the surface, which can most
easily be identied in LEED (cf. Section 3.1.1). In most cases, surface reconstructions can
be described by the Wood notation indicating the relation of the surface reconstruction unit
vectors to the primitive unit cell vectors of the bulk terminated SiC surface [100]. For substrate
unit vectors ~a1 and ~a2 and reconstruction unit vectors ~b1 and ~b2 the surface reconstruction is
usually denoted as (p×q )RΦ where p = b1 /a1 , q = b2 /a2 and Φ being the angle between the
reconstruction and substrate unit vectors. If Φ = 0 the angle√ is√omitted. A typical example of a
very complex surface reconstruction is the carbon rich (6 3×6 3)R30◦ structure, which will be
discussed in detail in Section 4.2. The Wood notation cannot be used if the angle between ~b1
and ~b2 is not the same as the angle between ~a1 and ~a2 . In that case, the transformation matrix
between reconstruction and substrate unit vectors is used to designate the surface reconstruction.
It has to be emphasized that the SiC(0001) and the SiC(0001̄) surface are completely dierent
surfaces with dierent characteristics. The growth, oxidation, surface reconstruction and graphi-
tization behavior is dierent for the two polarities. Even the properties of the unreconstructed
1 However,it has also been shown that stepped SiC surfaces may exhibit step heights of exactly one unit cell
over a macroscopic surface area and therefore exhibit a threefold symmetry in LEED [95]. Such a step structure
can be generated under certain preparation conditions in the ex-situ SiC surface treatment of hydrogen etching,
a method which is introduced in Section 3.2.
surfaces dier. Ab initio calculations allowing relaxation of the SiC bilayers in these idealized
surfaces show dierent intra- and interlayer distances [101]. Charge densities of the dangling
bond states show a stronger localization for the C-terminated side than for the Si-terminated
side. Therefore, the asymmetry of the Si-C bonds, already present in the bulk structure as men-
tioned above, is enhanced at the surface. The dierent surface reconstructions, which exist for
SiC(0001) and SiC(0001̄), are in the meantime quite well understood [9]. The analysis of these
surface phases is of fundamental importance for the growth of SiC crystals, for the stability with
respect to decomposition at high temperatures and for the physical understanding of oxide/SiC-
and metal/SiC-interfaces in electronic devices. It turns out that the nature of the reconstruction
does not depend on the polytype (4H, 6H, 3C) but only on the surface orientation (Si-face or
C-face). In surface phase diagrams, the atomic structure of the surface is analyzed in dependence
of the temperature of the SiC crystal [9]. In the scope of the discussion of the graphitization
characteristics of SiC(0001) and SiC(0001̄) in Chapter 4 and Section 5.1.1, respectively, the
corresponding surface phase diagrams are introduced. The general trend is that a higher tem-
perature leads to the evaporation of Si and therefore to C-rich surface reconstructions, whereas
the external supply of Si allows the access to Si-rich surface reconstructions. In addition, well
ordered epitaxial oxide layers can exist on both faces.
There is no principal restriction to study the basal plane surfaces of SiC only. On hexagonal
polytypes, several non-basal plane surface orientations have attracted interest due to the possibil-
ity of faster growth rates for bulk material and homoepitaxial lms. Furthermore, non-basal plane
surfaces directly exhibit the long period of the SiC bulk polytype stacking in the direction of the
c -axis. Strongly anisotropic atomic arrangements are present at the surface giving rise to elec-
tron correlation, which is interesting for one-dimensional electronic states [102, 103]. The most
interesting non-basal plane surfaces of SiC are SiC(112̄0) ("a-plane"), SiC(11̄02), SiC(1̄102̄)
and SiC(11̄00) ("m-plane"), which will be introduced later in Fig. 5.6 of Section 5.1.3. Re-
cently, atomic chains with a one-dimensional semiconducting surface state have been observed
on SiC(11̄02) [104]. However, there is still little knowledge accumulated, mainly in terms of the
detailed analysis of dierent surface reconstructions on non-basal plane SiC surfaces. A recent
review by U. Starke gives an overview of the current research status [105]. In particular, the
graphitization of SiC surfaces has so far been concentrated on the basal plane surfaces only.
Section 5.1.3 investigates the possible graphene evolution on non-basal plane surfaces.
Chapter 3
Experimental background
In this chapter, the experimental background information necessary for the understanding of
the subsequent chapters is given. Section 3.1 oers a short introduction into the measurement
techniques used in my thesis. The goal is not a complete description of these methods. Therefore,
it will often be referred to the literature for a deeper understanding. It should be considered that
the scope of this thesis has intentionally been set up in the framework of a multidisciplinary
surface science work so that all techniques are of relevance. After Section 3.2 has given some
details about the samples used in this work, Section 3.3 presents an overview of the experimental
setup. The work has been performed at various kinds of preparation and measurement systems.
The aim is to give an impression of the dierent approaches, which have been applied to get
insight into the properties of graphene epitaxially grown on SiC.
3.1 Investigation methods

Since the goal of this work is to analyze ultra-thin graphitic layers on a SiC substrate, the
appropriate investigation methods are all surface science related techniques. One of the key
concepts for surface sensitive measurement techniques is the inelastic mean free path λinel of
electrons in the near surface area. The inelastic mean free path is determined as the mean
distance electrons can travel without energy loss due to inelastic scattering. Within λinel the
intensity of elastically scattered electrons has decreased to 1/e . Figure 3.1 displays the inelastic
mean free path of electrons in dierent materials in dependence of their kinetic energy together
with a theoretical calculation [106]. For dierent materials the same behavior is observed so that
this curve is often called a "universal curve". An explanation for this universality is given by the
fact that the inelastic scattering of electrons in this energy range is mainly due to excitations
of conduction electrons, which have more or less the same density in all elements. The two
main processes that determine the curve are: rst, single-particle excitations involving valence
electrons and to a much lesser extent ionisation of core levels of the atoms; second, the excitation
of plasmons, which sets in at energies higher than a few tens of eV. The curve has a broad
minimum around a kinetic energy of about 50-100 eV. Here, the inelastic mean free path is less
than 10 Å meaning that an electron, leaving the solid with elastic scattering processes only, must
originate from the rst few layers. The degree of attenuation for electrons in a depth d is given by
exp(−d/λ). The elastically scattered electrons are probed in methods such as low energy electron
diraction (LEED) and microscopy (LEEM) as well as core level and angle-resolved photoelectron
spectroscopy (CLPES and ARPES). In the case of the latter two techniques the electrons have
24 Chapter 3. Experimental background
Figure 3.1: The inelastic mean free path of electrons in a solid as a function of their kinetic energy. The
dots are experimental data, the dashed curve a calculation (taken from Ref. [106]).
been photoexcited before by an incident photon beam. The electrons contain valuable information
about the surface physics of the analyzed material. A variation of the energy of these electrons,
e.g. by changing the energy of the incident electron or photon beam allows to access dierent
surface sensitivities.
3.1.1 Low energy electron diraction (LEED) and microscopy

(LEEM)
LEED is the method of choice to analyze the periodicity of an ordered surface and can even be
used for complete surface structure determinations. The principle of this technique can be dated
back to 1927 when Davisson and Germer observed that the intensity distribution of low energy
electrons backscattered from a surface follows the laws of diraction. This observation conrmed
the theoretical prediction by de Broglie from 1924 that electrons
√ with momentum p , energy E
and mass m can be attributed a wavelength λ = h/p = h/( 2mE ) ≈ 150/E [eV ] Å with h
p
being Planck's constant. The range of 25 - 600 eV for the electron energy E corresponds to
wavelengths of about 0.5 - 2.5 Å, which are on the scale of atomic distances. For their ndings
de Broglie and Davisson were awarded the Nobel prize in physics in 1929 and 1937, respectively.
The rst LEED experiments have been developed in the early 1960s. A schematic LEED setup is
shown in Fig. 3.2 (a). The primary electron beam of dened electron energy from an electron gun
impinges on the surface under normal incidence to get the maximum symmetry of the diraction
pattern. The elastically backscattered electrons are projected on a uorescent screen where the
LEED diraction pattern is monitored by a charge-coupled device (CCD) camera. Conventionally,
a four-grid-optics is used to eciently select these elastically backscattered electrons with high
energy resolution [96]. A typical LEED pattern is displayed in panel (b) of Fig. 3.2 exhibiting a
3.1. Investigation methods 25
Figure 3.2: (a) Experimental setup of LEED (left) together with the theoretical description (right). The
intersection of the Ewald sphere with the reciprocal lattice rods denes the directions of the dirac-
tion spots on the LEED screen. Since the electrons do not probe only one, but several near surface
layers, the lattice rods have a modulated intensity with respect to dierent√ electron
√ energies (taken
from Ref. [95, 107]). (b) Exemplary LEED pattern at 140 eV exhibiting a (6 3×6 3)R30◦ periodicity.
(c) Multiscattering processes in LEED, which are of importance in contrast to X-ray diraction.
√ √
(6 3×6 3)R30◦ periodicity, which will be discussed in detail in Chapter 4.
The position of the diraction spots in a LEED pattern is determined by the periodicity of the
two-dimensional surface lattice and is given by the Laue equation known from X-ray diraction:
~k − ~k0 = G
~ (3.1)
~ = (G
with G ~ k , Gz ) being a reciprocal surface lattice vector and ~k0 = (0, 0, k0z ) and ~k = (~kk , kz )
according to the choice in Fig. 3.2 (a). ~k0 and ~k are the wave vectors for the incident and
diracted electron beam. Due to the two-dimensionality of a surface, Gz is not dened by the
diraction conditions so that the reciprocal lattice consists of rods and not of points. With the
help of the Ewald construction, also known from X-ray diraction, the position of the LEED spots
on the screen according to the Laue equation can easily be determined as shown in Fig. 3.2 (a).
The interpretation of a LEED experiment is, however, more complex as the intensity of the
diraction spots is energy dependent. The intensity of the diracted electron beam is given by
I (~k, ~k0 ) = |F (~k, ~k0 )|2 · |S(~k − ~k0 )|2 . (3.2)
The lattice factor S(~k − ~k0 ) contains the Laue condition. The intensity of the thereby selected
diraction maxima is however modulated with the scattering or structure factor F (~k, ~k0 ). F (~k, ~k0 )
depends on the atomic scattering factors of the atoms and on the positions of the atoms within
the unit cell. For instance, it can happen that certain diraction spots are suppressed due to
the internal symmetry of the unit cell. Another aspect to be considered is that the intensity
of the diraction spots is modulated in dependence of the energy since LEED is not probing
the topmost surface layer but also deeper lying layers. Due to the necessity to full the Laue
equation also in the third dimension the reciprocal lattice can now be seen as reciprocal lattice
rods with modulated intensity as sketched in Fig. 3.2 (a). A further inuence on the intensity
of the diraction spots is multiple scattering as shown in Fig. 3.2 (c). Whereas up to here only
kinematic scattering has been discussed, a dynamical scattering theory for the structure factor
F (~k, ~k0 ) is needed, which takes into account that one electron can be scattered several times.
The scattering cross sections for low energy electrons are by orders of magnitude larger than in
X-ray diraction. Including all these aspects of scattering theory it is possible to conduct a surface
structure analysis based on LEED intensity spectra [108, 109]. LEED is still the dominating tool
Figure 3.3: (a) Schematic illustration of a low energy electron microscope (LEEM). The sample is imaged
with electrons that are reected at the surface. These electrons are emitted from the electron gun on the
lower left. It is also possible to use electrons that are photoexcited at the surface with the help of a light
source displayed on the upper right (PEEM). The image shows the original setup from E. Bauer [110].
(b) Typical bright eld LEEM image taken at 5.1 eV showing the morphology of epitaxial graphene on
SiC(0001) with a dierent number of graphene layers exhibiting a dierent contrast.
in surface structure determinations. However, even without performing a structure analysis a

careful inspection of LEED intensity spectra can already be quite helpful in many cases.
Low energy electron microscopy (LEEM), invented by Ernst Bauer in 1962 and rst operational
in 1985 [110], is quite similar to LEED. However, the electron beam diracted at the sample is now
used to image the surface of the sample. The original setup from 1985 is shown in Fig. 3.3 (a).
The imaging capability is accompanied by a relatively complex setup. A LEEM instrument is much
more expensive and more dicult to operate than a LEED instrument. One of the main tasks is
to separate the illuminating beam from the diracted beam, which is accomplished by means of
a magnetic deection eld. Two important modes of operation have to be distinguished: in the
bright eld mode, the most common mode, the directly reected beam ((0,0) diraction order) is
used to image the surface area, whereas in the dark eld mode an arbitrary diraction order spot
can be chosen for imaging. LEEM has a resolution of approximately 10 nm. It can also be used
as a conventional LEED on a microscopic surface region (micro-LEED). Using an incident photon
beam such as the one from the lamp indicated in Fig. 3.3 (a), the emitted photoelectrons can
be taken to image the surface area (photoemission electron microscope (PEEM)) or to measure
core levels on a microscopic area (micro-CLPES). LEEM and PEEM are very powerful techniques
in terms of real-time observations of dynamic processes at surfaces such as adsorption, reaction,
segregation, thin lm growth, etching, or sublimation. For epitaxial graphene on SiC(0001), it
is in particular possible to spatially resolve surface areas with a dierent number of layers [111]
and to monitor transformations between them (see Section 4.3.5 and Section 6.2.1 as well as
Ref. [112]). An exemplary bright eld image of such a graphene sample is shown in Fig. 3.3 (b):
a dierent number of graphene layers is reected by dierent intensity vs. energy characteristics
and thus by a dierent contrast in the LEEM image.
Figure 3.4: (a) Illustration of the AFM principle: a laser beam is reected from the cantilever to a
photodetector, giving information about the vertical displacement of the cantilever. Scanning over the
surface yields AFM images. The illustration is taken from Ref. [114]. (b) Typical AFM image of a
SiC(0001) surface with a regular array of atomically at terraces.
3.1.2 Scanning probe microscopy
Atomic force microscopy
Often used to display the real space morphology of samples on a micrometer area as in the case
of this thesis, atomic force microscopy (AFM) is a very powerful and versatile surface analysis
technique, which reaches atomic resolution and is widely used from biology and chemistry to
physics. The atomic force microscope (AFM) has been developed in 1986 [113] on the basis of a
scanning tunneling microscope (STM). It does not require ultra-high vacuum (UHV) conditions
and, in contrast to STM, it is applicable to conducting as well as insulating samples. The sample
is positioned on a movable sample holder. A cantilever with an integrated sharp tip out of silicon
or silicon nitride for instance, is situated above the sample as shown in Fig. 3.4 (a). During
measurements, the tip is scanning over the surface. However, for practical reasons the scanning
is usually accomplished with a xed tip and a movable sample holder. The interactions of the
tip with the surface can be monitored by an optical deection scheme with a laser beam being
reected at the cantilever and subsequently detected with a photodetector. Three dierent AFM
modi can be distinguished according to the shape of the potential between tip and sample. In
the contact mode the image contrast in AFM arises from variations of the repulsive forces due to
the Pauli repulsion. A feedback loop keeps the force between the tip and the surface constant.
The exemplary AFM image in panel (b) of Fig. 3.4 has been taken in this mode and shows a
regular array of atomically at terraces on a SiC(0001) crystal. The non-contact mode makes
use of attractive long-range forces such as van der Waals forces. Eventually, the tapping mode
exploits the (attractive) minimum in the tip-sample potential and leads to a periodic tapping of
the tip at the surface. The latter two are dynamic modi in the sense that the cantilever is driven
to oscillate around its resonance frequency. During scanning, the interaction of tip and sample
provokes a change in the amplitude, resonance frequency and phase of the cantilever, which are
used for feedback control.
Figure 3.5: (a) The principle of STM with a metallic tip scanning in close proximity over the surface.
(b) Schematic illustration of the tunneling process between tip and sample. (c) The rst real space image
with atomic resolution showing the (7×7) reconstruction of Si(111) [116].
Scanning tunneling microscopy (STM) and spectroscopy (STS)
The scanning tunneling microscope provides a real space image of the surface with atomic reso-
lution. It has been developed by G. Binnig and H. Rohrer in 1982 [115], who were awarded the
Nobel prize in physics in 1986. The concept is based on the tunneling eect known from quantum
mechanics. As displayed in Fig. 3.5 (a), a sharp metallic tip is brought into close distance to the
surface such that a tunneling current between sample and tip can be measured, which would be
forbidden in classical physics. The sample has to be either metallic or semiconducting. If a small
voltage (bias) is applied to the tip in the range of a few mV to a few V and if the tip is at a
distance of around 1-5 Å to the sample surface, a tunneling current of around 0.05-10 nA can be
measured. The tunneling current decreases exponentially with the tip-sample distance as shown
below. By scanning the tip laterally across the surface, a topographic image can be recorded
whereupon two major imaging modi have to be distinguished: in the constant height mode the
tip-sample distance and the bias is kept constant while an image of the tunneling current is
recorded. In the constant current mode a feedback loop keeps the tunneling current constant
by adjusting the tip-sample distance. The variations of the tip positions are used to record the
STM micrograph. As for the constant height mode, the bias voltage remains unchanged. STM
allows to get a three-dimensional real space image of the surface, which reects the local density
of states (LDOS) at the chosen bias as will be shown in this subsection. Another operation mode
of STM is scanning tunneling spectroscopy (STS). Here, the derivative of the tunneling current
with respect to the bias is measured in dependence of the bias. In this way, direct access to the
LDOS is obtained.
It should be noted that in STM images it is not the atomic positions that are imaged but
rather the electronic states at the tip position. In general, the situation is even more complicated
as the convolution of the electronic states from the tip and the sample is displayed (see below).
However, even if the structural arrangement of the atoms cannot directly be deduced, STM oers
useful structural information such as periodicity, symmetry or the presence of defects. For a more
complete picture, dierent bias voltages have to be applied in order to probe dierent electronic
states. For semiconductor surfaces, for instance, STM images for lled and empty states facilitate
the interpretation of the structural arrangement at the surface. The rst real space image with
atomic resolution recorded by G. Binnig and H. Rohrer [116] is shown in Fig. 3.5 (c) and displays
the lled states of the (7×7) reconstruction of Si(111), a reconstruction, which was controversially
interpreted at that time.
For a more detailed understanding of the explanations given so far, Fig. 3.5 (b) shows an
elementary model for the STM tunneling process between a metallic sample positioned at z = 0
and the tip positioned at z = d (see Ref. [117] for instance). The tunneling barrier can be
approximated by the averaged work function Φ between tip and sample: Φ = 1/2(Φtip + Φsample ),
which usually amounts to around 5 eV. Applying a bias potential eU allows electrons of energy En
in the energy interval EF − eU ≤ En ≤ EF to tunnel through the barrier with EF denoting the
Fermi energy. The tunneling current is then proportional to the sum of the tunneling probabilities
of these electrons. These tunneling probabilities can be derived from the Schrödinger equation.
Assuming that the bias potential eU applied to the sample is much smaller than the work function
Φ, the tunneling current is
EF
X √
I (d, U) ∝ |ψn (z = 0)|2 e −2κd with κ = 2mΦ/h̄ , (3.3)
En =EF −eU
m being the electron mass and ψn being the wavefunction for an electron with energy En . By
using the denition of the LDOS ρ(z, E ) for a suciently small energy interval
E
1 X
ρsample (z, E ) := |ψn (z)|2 (3.4)
En =E −
the tunneling current amounts to
I (d, U) ∝ eUρsample (z = 0, EF )e −2κd ∝ eUρsample (z = d, EF ) . (3.5)
As noted already above, the current is therefore proportional to the LDOS of the sample at the
Fermi energy at a distance d , i.e. the position of the tip. The current decreases exponentially
with the tip-sample distance d : for a work function of around 5 eV a change of 1 Å in the distance
causes a change of nearly one order of magnitude in the current.
So far, it is already conceivable that the measurement of dI /dV in tunneling spectroscopy
yields information about the LDOS. A better understanding is given by the so-called Bardeen
approach, which makes use of time dependent perturbation theory (see Ref. [117] for instance).
The probability of an electron in the state Ψ with energy EΨ to tunnel from the sample to the
tip into a state χ with energy Eχ is given by Fermi's Golden Rule
2π
wΨχ = |M|2 δ(EΨ − Eχ ) (3.6)
h̄
with M being the tunneling matrix element, which is given by the overlap of the surface wave-
functions of the states Ψ and χ at an arbitrary separation surface between sample and tip. The
δ function indicates that only states with the same energy level are allowed to tunnel into each
other (elastic tunneling). Applying a bias voltage U and approximating the Fermi distribution by
a step function (i.e. T := 0K ), the tunneling current is:
Z eU
4πe
I = ρsample (EF − eU + )ρtip (EF + )|M|2 d . (3.7)
h̄ 0
Usually, the matrix element can be set to be constant so that the tunneling current is proportional
to the convolution of the LDOS of the sample and of the tip. Tip and sample contribute equally to
the tunneling current and are therefore interchangeable. For a negative potential on the sample,
the occupied states generate the current, whereas in case of a positive sample bias the electrons
tunnel from the tip into the unoccupied states of the sample. Consequently, STM images at
dierent bias contain dierent information of the sample surface. The Bardeen approach also
explains that in STS the LDOS can directly be obtained. Assuming a free-electron metal tip
leads to a constant LDOS and from Equation 3.7 we get:
dI /dV ∝ ρsample (EF − eU) . (3.8)
Evidently, the dynamic tunneling conductance measured in STS is proportional to the LDOS of
the sample.
3.1.3 Photoelectron spectroscopy

Photoelectron spectroscopy (PES) is based on the photoelectric eect, which has been discovered
by H. Hertz and W. Hallwachs in 1886. A theoretical explanation of this eect has been given
A. Einstein in 1905 by postulating the quantum hypothesis for light, for which he was awarded
the Nobel prize in physics in 1921. Depending on the energy of the monochromatized incident
photon beam, two dierent PES main techniques have developed, which will be explained in the
following two subsections: the use of X-rays gives access to the core-levels of a sample, the use
of ultraviolet light probes the valence bands of the solid. Several comprehensive reviews and
monographs exist in the literature (see e.g. Ref. [96, 118, 119]).
A schematic representation of the photoemission process in the case of core level photoelectron
spectroscopy (CLPES) and valence band or angle resolved photoelectron spectroscopy (ARPES)
is shown in Fig. 3.6 (a) and 3.7 (a) of the corresponding subsections. Photoemission from
solids can be considered within three steps (three step model) [96, 118, 119]: (1) electron
excitation, (2) electron transport to the surface and (3) escape through the surface into the
vacuum. Provided the incident photon energy is higher than the binding energy of the electron plus
the work function of the sample, the electron reaches the vacuum level. The ux of photoemitted
electrons is subsequently detected in a spectrometer, where, in general, the kinetic energy of the
electrons and their escape angle can be measured. The kinetic energy contains the information
about the energy of the initial state, and the escape angle contains the information about the
momentum ~k of the initial state in the solid. The measured kinetic energy Ekin is given by the
following relation:
Ekin = hν − EB − ΦA , (3.9)
with ΦA being the work function of the analyzer (but not that of the sample) and EB the binding
energy. Since EB is referenced to the Fermi level EF , Equation 3.9 is independent of the sample's
work function. EF is the same for the analyzer and the sample provided they are in electrical
contact with each other. It has to be emphasized that only elastically scattered electrons are of
interest since only these electrons contain the information of the initial state in the valence band
or the core-level region.
Photoelectron spectroscopy is a surface sensitive technique since the inelastic mean free path
of the photoexcited electrons is in the range of a few Å as shown previously in this chapter
in Fig. 3.1. This holds both for CLPES and ARPES. Often it is dicult to distinguish surface
related eects from bulk related eects. By changing the incident photon energy, electrons
with dierent kinetic energies and thus dierent inelastic mean free paths are generated. This
procedure results in dierent surface sensitivities allowing to discriminate between surface and
bulk related phenomena. Another possibility to change surface sensitivity besides varying the
Figure 3.6: (a) Schematic illustration of the photoemission process in CLPES (b) A typical survey scan
in CLPES for the determination of the elemental composition of the sample.
incident photon energy, is to change the angle of the incident photon beam with respect to
the surface normal. Under grazing incidence the eective penetration depth is much smaller
than for normal incidence so that the surface sensitivity is increased. In the home laboratory,
it is not possible to continuously change the excitation energy. X-ray sources rely on the Kα
core-level radiation of Mg or Al for instance, UV sources make use of He I or He II excitation.
However, the acceleration of charged particles such as electrons results in a continuous radiation
spectrum. This principle is used at synchrotron light sources where the incident photon beam
in the photoemission experiments can be chosen according to the needs. Furthermore dierent
polarizations can be chosen and the beam spot size goes down do around 0.005 mm2 . A home
laboratory UV source illuminates an area of around 1 mm2 , whereas a conventional X-ray source
illuminates a few mm2 .
Core level photoelectron spectroscopy (CLPES)

CLPES is also known as X-ray photoemission spectroscopy (XPS) or electron spectroscopy for
chemical analysis (ESCA). The name ESCA is the original notion and has been introduced by
K. Siegbahn, who started to develop this technique in the 1950s and 1960s [96]. The full potential
as a method for surface sensitivity was recognized in the early 1970s. For his contributions to
advances in electron spectroscopy, K. Siegbahn was awarded the Nobel prize in physics in 1981.
Since the binding energies of core level electrons typically lie in the range of a few tens of
eV to one or several keV, the photon energy of the incident beam has to be high enough for the
photoemission process, which is displayed in Fig. 3.6 (a). The corresponding excitation energies
for the laboratory sources based on Mg Kα and Al Kα radiation are 1253.6 eV and 1486.6 eV,
respectively. As binding energies of core level electrons have characteristic and well dened values
for dierent elements, the name ESCA describes the practical use of this technique in the best
way: it serves to determine the chemical composition of the sample surface. A typical survey
scan of a SiC surface is shown in Fig. 3.6 (b). The sharp peaks can be associated with core level
excitations of silicon, carbon and oxygen. Furthermore the holes created in the atoms give rise
to the emission of so-called Auger electrons, which can be detected in the spectrum, too.
CLPES however, is much more versatile than just a ngerprint technique to determine the
types of elements present in the near surface layers. Let us consider the photoemission process for
one atom in the solid. The absorption of a photon with energy hν excites the system consisting
of N electrons and initial state energy E i (N) into a nal state ion with energy E f (N − 1, k) plus a
photoelectron with kinetic energy Ekin . The parameter k labels the level, from which the electron
has been removed. The binding energy with respect to the vacuum level can then be expressed
by
EB (k) = E f (N − 1, k) − E i (N) . (3.10)
From Equation 3.10 it is clear that the binding energy depends on the initial state and the nal
state and can be inuenced by so-called initial and nal state eects, which are discussed in the
following.
Within an one-electron picture, the binding energy is equal to the negative orbital energy k
of the emitted electron, which can be calculated from the non-relativistic Hartree-Fock approach:
EB (k)K .T . = −k . (3.11)
This approximation is called Koopmans' theorem. However, the nal state (N-1) electron system
undergoes relaxation eects due to the screening of the positive charge of the photo-hole by the
remaining electrons. This causes a reduction of the binding energy of the ejected electron by the
relaxation energy ∆relax > 0 so that Equation 3.11 has to be written in the following form:
EB (k) = −k − ∆relax . (3.12)
In addition to the relaxation energy, it has to be taken into account that the ion created in the
ionization process does not necessarily have to be in its ground state, giving rise to so-called
shake-up or shake-o satellite peaks. These two eects and the relaxation process are called nal
state eects. They can cause a shift in the binding energy up to several eV.
Initial state eects on the contrary, are already present before the ionization process. The
core-level binding energies of a certain element are subject to variations depending on the chemical
state of the atom such as oxidation state or chemical environment. This phenomenon is called
core level chemical shift. Chemical shifts can be described in terms of variations of the initial
state orbital energies plus a possible dierence in the relaxation energies. The latter one is called
interatomic relaxation energy in contrast to intraatomic relaxation within one atom as considered
in the previous paragraph. Using Equation 3.12 the chemical shift amounts to
∆EB = −∆k − ∆(∆relax ) . (3.13)
Chemical shifts have signicant contributions to the binding energy. As explained in Section 2.2.1,
the element C has a higher electronegativity than the element Si so that in a Si-C bond the Si
atom will experience an increased Coulomb potential in comparison to an unbonded Si atom.
Si 2p core levels in SiC are indeed shifted to binding energies that are around +1.2 eV higher
than for pure elemental Si [120]. Due to the high electronegativity of oxygen, the shift of the
Si 2p binding energy in SiO2 with respect to pure Si is even +3.9 eV [121].
This discussion emphasizes the necessity to distinguish dierent components in the core level
spectrum of one element by means of a tting routine. Before tting, a proper background
subtraction has to be performed assuming a linear or a so-called Shirley background for instance.
For SiC core level peaks, a Voigt line shape can be used, which is a convolution of Lorentzian
and Gaussian functions. The Lorentzian function describes intrinsic broadening due to the nite
lifetime τ of the core hole. According to the Heisenberg's uncertainty principle, the corresponding
width amounts to h̄/τ = 6.6·10−16 /τ (eV). The Gaussian broadening arises from the instrumental
resolution of the analyzer and the light source, from inhomogeneity of the sample and from phonon
broadening. Core level peaks of metals but also of graphitic SiC surfaces exhibit an asymmetric
line shape with additional energy losses on the higher binding energy side. The reason can be
found in the screening process of the photoexcited core level hole state, which produces electron-
hole pair excitations in the Fermi sea of conduction electrons [118]. For these metallic or graphitic
core level peaks, a Doniach-Sunjic prole can be used in the tting procedure. For Si 2p core
levels, it has to be considered that spin-orbit interaction causes a splitting of the Si 2p peak into
two peaks Si 2p1/2 and Si 2p3/2 . Due to the occupancy with two and four electrons, respectively,
the intensity ratio, called branching ratio, is 1:2. The energy dierence (spin-orbit-splitting)
amounts to 0.6 eV.
A quantitative surface analysis by means of CLPES is based on the fact that the ionization
probability (cross section) of a core level is practically independent of the valence state of the
respective element. This implies that the intensity, i.e. the integrated peak area, is proportional
to the number of atoms in the detected sample area. The peak area depends in general on the
following factors: the photon ux density, the illuminated surface area, the atomic photoionization
cross section of the corresponding core level, the concentration of the corresponding element in
the sample, the instrumental detection eciency and the probability that the photoelectrons
leave the sample surface, which usually is of the form exp(−d/λ) with d being the escape depth
and λ being the inelastic mean free path. With these ingredients, it is possible to determine
the atomic concentration of dierent elements in the surface region. Apart from curve tting,
a quantitative evaluation of CLPES data in this work was only necessary for the determination
of the thickness x of a thin graphitic overlayer on top of the SiC substrate surface. As both
substrate and overlayer consist of carbon, the consideration of this problem is rather simple, since
in particular the photoionization cross sections and the energy dependent instrumental detection
eciency are practically identical. Also the inelastic mean free paths can be assumed to be the
same. The overlayer thickness x for normal incidence of the photon beam is then given by

Ix nbulk
x = λ · ln 1 + , (3.14)
Ibulk nx
with Ix ,Ibulk being the measured peak areas of the core level of overlayer and substrate (C 1s in
this case) and nx ,nbulk being the atomic concentrations of the corresponding element (C in this
case). Note that the main uncertainty is coming from the inaccurate knowledge of the inelastic
mean free path of the photoexcited electrons.
Angle resolved photoemission spectroscopy (ARPES)

The goal in angle resolved photoemission spectroscopy (ARPES) is to gain information about
the angular and energy distribution of photoelectrons. ARPES is usually associated with the
photoemission of valence band electrons. The initial state of these electrons is characterized
by their energy and by their wavevector so that the valence band structure E (~k) of a solid can
be determined. In the case of He I (21.21 eV) or He II (40.8 eV) excitation, ARPES is often
called angle resolved ultraviolet photoelectron spectroscopy (ARUPS). However, for clearness only
the notion ARPES is used throughout this thesis. The token UPS describes the corresponding
angle integrated technique and has rst been experimentally established in 1971 by Eastman and
Cushion [96].
As shown in Fig. 3.7 (a), the principle of the photoionization process is the same as for CLPES
so that the same considerations hold for ARPES. However, since ARPES gives information about
Figure 3.7: (a) Schematic representation of the photoemission process in ARPES. (b) Schematic view of
an experimental geometry in ARPES. The angle θ contains the information about the parallel momentum.
All momenta displayed in the sketch are nal state momenta after transmission through the surface.
(c) Upon transmission into vacuum, only the parallel momentum of the photoelectrons is conserved, the
perpendicular momentum does not follow any lattice periodicity condition. All momenta shown in the
sketch are nal state momenta. (d) Typical band structure image of monolayer epitaxial graphene on
SiC(0001) around the K-point of the graphene Brillouin zone, obtained with a two-dimensional detector
with CCD camera. Throughout this thesis, the band structure is plotted in inverse grayscale like in this
image. (e) Schematic view of a surface projected bulk band structure. The dispersion of possible surface
states (dashed lines) and surface resonances (short dotted lines) are indicated. The momenta in this
sketch are initial state momenta. Panel (b),(c) and (e) are taken from Ref. [119].
the valence band dispersion, special attention has to be taken to the wave vector ~k . Upon
photoexcitation, the following relation holds for the momentum of the initial state and the nal
state:
~kf = ~ki + G
~B , (3.15)
~ B being a reciprocal bulk lattice vector. When the photoexcited electron passes through
with G
the surface, only the parallel momentum of the nal state electron is conserved so that
~kf k = ~k ex − G
~S , (3.16)
fk
~ S being a reciprocal lattice vector of the surface. The perpendicular momentum ~k ex does
with G f⊥
not follow such a lattice periodicity condition. Using Equation 3.16, the parallel momentum part
of Equation 3.15 transforms into:
~kik = ~k ex − G
~S − G
~ Bk . (3.17)
fk
According to Fig. 3.7 (b) and (c), the kinetic energy of the electron amounts to Ekin,f = Ekin,i =
Ekin = h̄2 /(2m) (kik 2 2
+ ki⊥ ) = h̄2 /(2m) kik
2
sin−2 (θ). Note that the momentum relations at the
surface resembles Snell's refraction law. In the reduced zone scheme the parallel momentum of
the initial state is given by
~kik = 1/h̄ 2mEkin sin(θ) = 0.512 · Ekin [eV ] · sin(θ) [Å−1 ] .

p p
(3.18)
With the help of Equation 3.9 and 3.18, the band structure E (~kik ) is directly obtained. State of
the art spectrometers have a two-dimensional detector with CCD camera that directly records the
band structure, provided the analyzer is operated in the angular dispersion mode. Such a band
structure map is demonstrated in Fig. 3.7 (e) for a monolayer of epitaxial graphene on SiC(0001)
with its typical linear π -bands at the K-point of the graphene Brillouin zone, an image that the
reader will encounter many times throughout this thesis.
For a two-dimensional material such as graphene, ki⊥ does not exist so that E (~kik ) represents
the initial state band dispersion. For three-dimensional materials, however, E (~kik ) gives access
only to the surface projected band structure as shown in Fig. 3.7 (e). The dashed lines indicate
surface states, which exist only in the band gap of the surface projected band structure and
do not have any dispersion with kf ⊥ . If a surface state mixes with bulk bands, it is called a
surface resonance. The initial state band dispersion along the surface normal E (ki⊥ ) is much
more complicated to determine [96, 118, 119]. It can in principle be done by taking a series of
PES spectra in normal emission (i.e. kik = 0) for dierent photon energies and an assumed nal
state dispersion such as a free-electron like band.
3.1.4 Raman spectroscopy

In Raman spectroscopy, photons of an incident laser beam are inelastically scattered in the solid
or in molecules, whereby information about the phononic properties of the measured crystal
or the vibrational properties of the molecule is obtained. The technique is named after Sir
C.V. Raman who rst observed Raman scattering in 1928 by means of sunlight. In 1930 he
therefore was awarded the Nobel prize in physics. The Raman scattering is not a typical surface
science technqiue as it also probes bulk properties. For epitaxial graphene samples for instance,
a superposition of bulk related SiC signals and surface related graphene signals will be expected.
Figure 3.8: (a) Schematic view of a typical setup in Raman spectroscopy. (b) Energy level diagram
for the explanation of the Rayleigh, Stokes and anti-Stokes signal. (c) Raman spectrum showing the
intensity vs. the Raman shift of the Rayleigh, Stokes and anti-Stokes lines.
A typical experimental setup is shown in Fig. 3.8 (a). The laser wavelength is in the range of
visible light (e.g. 488 nm in our case), the diameter of the laser spot on the sample is around
400 nm.
The scattering of monochromatic photons at molecules or solids comprises elastic scattering
and inelastic scattering. As sketched in Fig. 3.8 (b) for the energy levels in a molecule, the
incident photon creates a virtual energy state. After deexcitation the system may either be in its
initial state or in a vibrational state. If initial and nal state coincide, the scattering is elastic and
is called Rayleigh-scattering. If the nal vibrational state of the molecule (or the solid) is more
energetic than the initial state, the photon wavelength is shifted towards lower values, which gives
rise to the appearance of the so-called Stokes signal. If the nal vibrational state is less energetic,
the photon wavelength is shifted towards higher values giving rise to the so-called anti-Stokes
signal. It can only be observed if light is scattered at molecules or solids that are initially in an
excited state. In a Raman spectrum, in which the intensity is plotted versus the wavelength shift
upon Raman scattering (so-called Raman shift), the Rayleigh signal as well as the Stokes and
anti-Stokes lines can be identied (see sketch in Fig. 3.8 (c)). The Raman shift is usually given in
units of inverse cm. It amounts from some hundred to some thousand cm−1 , which corresponds
to a wavelength shift of some tens of nm. The Rayleigh signal is the most intense part in the
detected light, whereas the intensity of the inelastically scattered light is orders of magnitude
lower. As a consequence, the inelastically scattered photons have to be carefully ltered out in
an experimental setup by using several monochromators in series.
In solid state physics, Raman spectroscopy allows to get detailed insight into crystallinity,
defects, strain and relaxation, temperature, charge carrier concentration and doping, phonon
modes as well as other low frequency excitations such as plasmons or magnons. For the growth
of SiC crystals for instance, it is dicult to grow a single polytype only. Raman spectroscopy helps
to identify mechanical stress, stacking faults or polytype inclusions in the grown crystals [122].
Raman spectroscopy has historically played an important role in the structural characterization
of graphitic materials, and has also become a powerful tool for understanding the behavior
of electrons and phonons in graphene [123]. Section 4.3.6 gives more details about Raman
spectroscopy on graphene in general and epitaxial graphene on SiC in particular.
3.2. Samples 37
3.2 Samples
The SiC(0001) and SiC(0001̄) samples were cut from wafers of dierent polytype (4H and 6H).
The sample size is typically in the mm2 regime and ranges from 6 x 6 mm2 to 6 x 13 mm2 depending
on the sample holder specications of the dierent characterization techniques. The wafers of two
inch or three inch size have been purchased from SiCrystal AG, Cree Research Inc. or Norstel AB.
All wafers were on-axis wafers, which means that the surface orientation is along the basal plane
direction within about typically 0.5◦ . In most cases the wafers were n-doped with nitrogen with
a concentration range of around 5 · 1017 to 1 · 1019 cm−3 range corresponding to a resistivity of
around 0.1 to 0.01 Ωcm. Sometimes also semi-insulating samples have been used, which exhibit a
resistivity larger than 105 Ωcm. The wafers have undergone either an "optical polishing" procedure
or "chemical mechanical polishing" (CMP) procedure. Note that all SiC(0001̄) samples have been
optically polished.
As received optically polished SiC wafers usually have many residual polishing scratches that
are up to 10 nm deep and a few hundred nm wide. They cannot be removed, as usual in surface
science experiments, by sputter and annealing cycles in UHV due to the dierent vapor pressure
of Si and C. A solution is oered by hydrogen etching. Owing to the slight misorientation of
the wafer of around 0.5◦ , this procedure generates a regular array of atomically at terraces of
approximately 0.5 µm width [124] as exemplarily displayed by an atomic force microscopy (AFM)
image in Fig. 4.1 (a) of Section 4.1. The hydrogen etching has been performed in a hot-wall
chemical vapor deposition (CVD) reactor operated by C.L. Frewin, C. Locke, C. Coletti and S.E.
Saddow at the University of South Florida in Tampa/USA [124, 125]. Before hydrogen etching,
all the samples were cleaned using either the so-called "piranha" clean [126] followed by an HF
dip or the so-called "RCA" clean [127] in order to remove any metallic or organic contamination.
The CVD reactor was loaded with a dedicated H2 etching hot-zone. During hydrogen etching,
the samples are inductively heated at around 1500 ◦ C for around 10 minutes under a ow of
puried molecular hydrogen with a ow rate of around 5 standard liters per minute (slm). The
hydrogen pressure can range from 150 Torr to 760 Torr. The specics for the etching processes are
slightly dierent for dierent sample types and are summarized in Ref. [125]. During the hydrogen
etching process, approximately several hundred nm of SiC material are removed. Depending on
the process quality, the sample surface can be completely passivated with hydrogen. Incidentally,
for the purpose of hydrogen passivation without etching, already an annealing temperature of
around 1000 ◦ C is enough [99].
SiC wafers that have undergone a CMP polishing procedure usually have a roughness with
a root mean square (RMS) value of 1 Å and show atomically at steps. Before any preparation
procedure, the samples also have to undergo an "RCA" clean [127] or a "piranha" clean [126]
followed by an HF dip. For CMP polished samples, hydrogen etching can be applied but it is
not necessary. As hydrogen etching is often accompanied by step bunching, which for some
purposes is of disadvantage, CMP polished samples are a good starting material. It should also
be considered that high temperature treatment and/or silicon deposition sometimes may lead
to a smoothening of the surface and a reduction of scratches originating from optically polished
samples.
In Section 5.1.3, a short outlook to graphitization of non-basal plane surfaces is given. The
4H-SiC((1̄102̄)) and 4H-SiC((11̄02)) samples originated from wafers that have been provided
by W.J. Choyke from the University of Pittsburgh. The wafers were cut from an n-doped 4H-
SiC bulk crystal with an inclination of 62.1◦ . The samples, which have been cut out of the
mechanically polished wafer, were "RCA" or "piranha" cleaned and then hydrogen etched [105].
The 4H-SiC((112̄0)) samples have been provided by R.F. Davis from the North Carolina State
University and have also been hydrogen etched. 3C-SiC/Si(111) samples, consisting of a 3C-SiC
lm on top of a Si(111) substrate, are briey discussed in Section 5.1.2 and have been provided
by C.L. Frewin, C. Locke and S.E. Saddow of the University of South Florida in Tampa/USA.
3.3 Experimental setup and sample preparation

UHV system for room temperature STM
This UHV system is equipped with a preparation chamber, an STM chamber, a sample load-
lock station and a sample magazine [128]. The chambers are constructed of stainless steel with
quartz windows and are equipped with a complex pumping system consisting of oil-free fore-
vacuum pumps (membrane pumps or scroll pumps), turbo molecular pumps, ion getter pumps
and Ti sublimation pumps. The base pressure for the STM chamber is in the 10−11 mbar regime,
for the preparation chamber in the low 10−10 mbar regime. The dierent chambers are connected
via a sample transfer system in UHV.
After clamping the samples in a Mo sample holder, they were loaded into the UHV chamber.
The heating of the samples is performed by electron bombardment with electrons being emitted
from a tungsten lament and then accelerated with high voltage. In contrast to other UHV
systems, the heating lament is mounted inside the manipulator end piece. Sample cooling is
accomplished with a water ow through the manipulator. A four grid LEED optics and Auger
electron spectroscopy (AES) allow for a continuous monitoring of the preparation process in the
preparation chamber. In order to avoid external magnetic inuence for the low energy electrons,
the chamber is surrounded by a set of Helmholtz coils. Since a CCD camera is mounted behind
the LEED screen, images and movies for the analysis of LEED intensity spectra can be recorded
with the software EE2000 from M.F. Opheys. The Auger electrons are excited with a 3 keV
electron gun and are detected with a 150◦ spherical sector analyzer of the type VG CLAM 100.
Auger electron spectroscopy is not explicitly discussed in this thesis so that the reader should
refer to Ref. [95, 96, 129, 130].
Subsequent to the initial outgassing, the samples were prepared by Si deposition and annealing
as explained in Section 4.1 for SiC(0001), or, for other sample types in the corresponding sections,
respectively. Si deposition is carried out using an electron beam evaporator (EFM3T) with a ux
rate of approximately 1ML/min. The evaporant is a rectangular rod cut out of 2 mm thick Si
wafers. This Si rod is annealed by electron bombardment until evaporation sets in, with the
electrons being emitted from a tungsten lament and then accelerated using high voltage. By
monitoring the ion ux, which is directly proportional to the ux of the evaporated atoms, the
evaporation rate can be controlled and to some extent estimated. For a precise determination of
the deposition rate, however, a quartz crystal microbalance has to be used. Generally speaking,
Si deposition on SiC(0001) samples (at 800 ◦ C sample temperature) is an extremely useful tool to
clean the samples from oxygen, one or two graphene overlayers and even from deposited metals.
The experiments with carbon evaporation on SiC, as discussed in Section 5.2.1, were performed
together with A. Al-Temimy with a commercial carbon source of the company MBE-Komponenten
GmbH with a ux of around 1ML/min evaporated from a highly oriented pyrolytic graphite
(HOPG) lament. The sublimation of carbon requires very high temperatures up to 2300 ◦ C.
Therefore, the hot zone around the lament is completely shielded with pyrolytic graphite. The
ux rate has been determined by carbon deposition on top of Si(111). The subsequent analysis
with AES and a SiC reference sample allow for the thickness determination of the evaporated
3.3. Experimental setup and sample preparation 39
carbon lm [129].

A short remark concerning sample temperatures might be instructive. The sample temper-
atures were determined by an infrared pyrometer. However, it should be noted that the exact
temperature measurement in the case of electron bombardment heating is hampered by the glow
of the lament, which is visible due to the large band gap of SiC. In this thesis, temperature
values for UHV experiments are all related to the experiments using resistive heating, since the
sample heating in the UHV system for photoelectron spectroscopy and at the synchrotron beam-
lines was done via resistive heating. However, it appears that even then the pyrometer values
for the same preparation step may dier for dierent samples, which can be ascribed to subtle
inuences of dierent doping levels or the polytype dependent band gap. Absolute temperature
values can therefore practically not be given. Furthermore, it should not be neglected that the
quartz windows in preparation chambers are often covered with unknown evaporator materials.
The STM chamber contains a Besocke-type room temperature STM [131], which is compat-
ible with the sample holders used in the preparation chamber. The sample holder is positioned
on top of three piezoelectric elements with helical ramps on the sample holder, enabling a coarse
tip-sample approach. The lateral positioning with respect to the tip is also accomplished by these
three piezos. In the center of the piezos, a fourth piezo carries the tip and realizes the scanning as
well as the vertical positioning with respect to the sample surface. The STM tip has been made
by electrochemically etching of a tungsten wire. Unfortunately, the tip cannot be exchanged in
vacuum and cannot be prepared by external means such as sputtering. Air cushions are used to
isolate the whole UHV system from vibrations. For the analysis of STM images, the software
WSxM [132] was used.
UHV system for photoelectron spectroscopy

The electronic structure was analyzed in a multiuser UHV system with a load-lock station, a
sample magazine and a sample distribution chamber connecting an angle resolving electron spec-
trometer chamber with a preparation chamber. The latter is equipped with LEED and dierent
ports for evaporators. Sample heating is performed by resistive heating with Mo contacts that are
also used to x the sample on the sample holder. Both analysis and preparation chamber allow for
cooling the samples down to liquid nitrogen temperatures. The experiments involving the depo-
sition of Bi, Sb, and Au, as discussed in Section 6.1.1 and 6.2.2, were performed in collaboration
with I. Gierz and C. Ast of the department Kern. The evaporation of Bi and Sb was achieved by
using an electron beam evaporator, the evaporation of Au by using a commercial Knudsen cell.
In a Knudsen cell, a lament heats a ceramic crucible with the evaporant inside, which rst starts
to melt and then to evaporate. The measurements, including the evaporation of TCNQ- and
F4-TCNQ-molecules in UHV, are discussed in Section 6.1.2 and have been performed together
with C. Coletti using the same preparation chamber. An evaporator, designed specically for the
evaporation of molecules, was used, which, however, also consists only of conventional ceramic
crucibles that are heated by a lament.
The ARPES experiments were carried out at room temperature using He II radiation (40.8 eV)
from a monochromatized UV source (He gas discharge lamp by Scienta). The spot size on the
sample is around 1 mm2 . The sample manipulator allows for an accurate alignment of both
the azimuthal and polar surface orientation to about 0.1◦ . A hemispherical analyzer (SPECS
PHOIBOS 150) equipped with a 2D detector and a CCD camera was used to map the band
structure. The energy resolution is around 10 meV, and the angular resolution amounts to 0.3◦ ,
−1
which corresponds to a momentum resolution at the Fermi level of around 0.014 Å at 40.8 eV
incident photon energy. The dispersion scans were measured perpendicular to the ΓK-direction
of the graphene Brillouin zone. The spectrometer was also used for CLPES experiments with a
conventional unmonochromatized Mg Kα source (1253.6 eV), which illuminates several mm2 of
a sample.
UHV system for STS at 6K
The STS spectra and some of the topographic STM images shown in Section 4.3.4 have been
taken in collaboration with L. Vitali and R. Ohmann of the department Kern with a home-
built low-temperature UHV STM [133]. The STM UHV-system consists of two chambers: a
preparation chamber for sample cleaning and preparation and a measurement chamber, which
contains the liquid helium cryostat with the STM. On the sample holder, the sample is clamped
with Mo plates. Note that the graphene samples to be measured in this low temperature STM
have been transported through air from the UHV system for room temperature STM (see above).
It turned out that the graphene integrity remained intact apart from adsorbed water or impurities,
which could be removed by annealing the sample in the STM UHV-system up to 750 ◦ C. The
whole STM UHV-chamber is eciently isolated from any vibrations or noise. It consists rst of
one active damping stage of the company TMC, second of two passive damping stages of the
company Newport Corp. and third of a measurement hutch that is isolated against acoustic noise.
The operation of the STM is performed outside the hutch. Typical measurement temperatures
of around 6K are achieved by the liquid helium cryostat. Similar to the previously discussed room
temperature STM, a Besocke type microscope is used [131]. However, the tip is being mounted
on a disc with helical ramps that are positioned on three piezoelectric actuator elements. The tip
can thus be approached towards the sample, which is situated below. The dI /dV -spectra and
the energy resolved conductance maps, reported in Section 4.3.4, have been recorded by means
of a lock-in technique applying a sample bias modulation in the range of 10-15 mV.
Atmospheric pressure graphitization system
The atmospheric pressure graphitization has been adopted from Ref. [10] to improve the sample
quality as discussed in Section 5.2.2. The sample is placed on a graphite susceptor with the
shape of a beaker. The susceptor is positioned in a vertically mounted quartz glass tube, which
is cooled with water. After loading the sample into the reactor the glass tube is pumped down
to around 1 · 10−5 mbar. Next, the Ar ow is started at an Ar pressure of around 950 mbar.
In order to generate an Ar ow, a fore-vacuum pump is used to pump the Ar gas. Due to
the lack of a mass ow controller, monitoring of the ux is impossible. The heating is carried
out with a radio frequency (RF) induction furnace. Typical annealing and cooling rates were
around 150 ◦ C per minute, the annealing time was 10 min. The temperatures are measured with
a pyrometer and are around 1430 ◦ C for zerolayer graphene on SiC(0001) and around 1520 ◦ C
for monolayer graphene on SiC(0001). Owing to the dierent setup in comparison to a UHV
system and due to the cooling water owing in the quartz tube, a precise comparison with the
temperatures of the UHV preparation technique is not possible. The goal of this experimental
setup was to have a preliminary atmospheric pressure graphitization apparatus. Currently, a
combined hydrogen etching, graphitization and hydrogenation machine is set up with a better
control of the measurement parameters.
3.3. Experimental setup and sample preparation 41
Hydrogenation of graphene samples

The hydrogen treatment or hydrogenation of epitaxial graphene samples on SiC(0001) is discussed
in Section 6.2.1. The graphene samples were transported through air from the graphitization setup
(UHV or atmospheric pressure graphitization) to a CVD reactor for the hydrogen treatment. The
experimental setup is shown in Fig. 6.12 (a) of Section 6.2.1 and is similar to the one proposed
in Ref. [99]. The samples were mounted on a Mo sample holder with a capacity of two to three
samples. The sample holder can be transferred from the load-lock to the horizontally aligned
quartz tube and is then placed on a graphite susceptor. After pumping down to a pressure of
1 · 10−7 mbar, the ow of hydrogen is started with the help of a fore-vacuum pump. The hydrogen
is cleaned in a palladium diusion cell from 5.0 to 8.0 purity grade. A mass ow controller sets
the ux to 1.5 slm at a pressure of around 950 mbar. Before starting the annealing process, the
hydrogen is kept owing for around ve minutes to remove any impurities from the chamber or
the hydrogen lines. The samples were annealed for 10 min at temperatures between 600 ◦ C and
1000 ◦ C with a heating and cooling rate of around 100 ◦ C per minute. Note that the temperature
was measured with a thermocouple at the backside of the graphite susceptor so that the real
temperature of the samples might be higher. The annealing was accomplished with a light oven
with ve halogen lamps and a total possible heating power of maximum 5 kW.
Beamline I311 of the MAX radiation synchrotron laboratory

High resolution CLPES experiments, as shown in Chapter 4 and 6, were carried out using syn-
chrotron radiation at beamline I311 [134] at the 1.5 GeV MAX II storage ring of the MAX
radiation laboratory (Lund, Sweden). The beamline is equipped with an undulator that provides
soft X-ray and UV radiation in the energy range of 30 - 1500 eV. The energy resolution is 4 meV at
30 eV up to 1.4 eV at 1500 eV. A modied SX-700 monochromator is used in order to select the
desired energy values. The endstation consists of a separate preparation and analysis chamber on
top of each other, with the same manipulator being used for both chambers. LEED and several
ports for evaporators are available in the preparation chamber. The samples were loaded via a
home-built small preparation chamber, at which evaporators can be mounted and where samples
can be heated. On the sample holder, the sample is clamped on Mo contacts via small Mo plates
allowing for resistive heating. The analysis chamber is equipped with a Scienta SES200 analyzer.
Due to the experimental setup, normal electron emission translates into a photon incidence angle
of 55◦ . In order to calibrate the core level spectra, the binding energies are referred to the Fermi
level, which is determined from a Ta foil. Note that the design of the manipulator and sample
holder are practically not suitable for ARPES band structure measurements.
The LEEM (and micro-LEED) experiments were performed with a LEEMIII instrument from
the company Elmitec. The LEEM chamber is directly connected with the CLPES chamber so
that it can be used for PEEM (and micro-CLPES), too. The microscope has a spatial resolution
better than 10 nm in the LEEM mode and 30 nm in the PEEM mode. Since the LEEM instrument
is a rather complex system, it has to be operated by a sta scientist. The data presented in this
thesis have been acquired in collaboration with A.A. Zakharov.
Surface and Interface Spectroscopy (SIS) beamline of the Swiss Light Source (SLS)
Some of the ARPES results shown in Section 4.3.2 and 6.1.2 were acquired at beamline X09LA
(Surface and Interface Spectroscopy (SIS)) at the Swiss Light Source (SLS) [135]. The energy
range of the beamline is 10-800 eV, the spot size on the sample amounts to 50 x 100µm2 .
Dierent polarizations such as left circular, right circular and horizontal linear polarization are
possible. Note that the beamline setup allows only a relative and not an absolute denition of right
and left circular polarization. The beamline serves two endstations that can be exchanged via a
modular rotating platform. Our experiments were performed at the high-resolution photoemission
spectroscopy (HR-PES) endstation. It consists of two preparation chambers, one load-lock system
and the analysis chamber. The chambers are connected with a sample transfer system in UHV.
The preparation chambers have two LEED optics and several ports for evaporators. On the
sample holder, the sample is clamped on Mo contacts via small Mo plates allowing for resistive
heating. The analysis chamber is equipped with a Scienta SES2002 analyzer. Although samples
in the analysis chamber can be cooled down to liquid helium temperature, the sample temperature
in the experiments, presented in this thesis, was kept at 80 K by means of counter heating. The
manipulator is designed for a sample holder movement in polar and azimuthal direction, the
maximum tilt amounts to 30◦ . For the graphene samples, the dispersion scans were measured
parallel to the ΓK-direction of the Brillouin zone. The energy resolution is below 10 meV, the
angular resolution amounts to 0.1◦ , which corresponds to a momentum resolution at the Fermi
−1 −1
level of around 0.0035 Å for 30 eV incident photon energy and 0.0077 Å for 90 eV incident
photon energy.
AFM system
The AFM, that was used in this work, is the model Autoprobe CP of the (former) company
ThermoMicroscopes, now supported by Veeco Instruments. The AFM is equipped with a 5 µm
and a 100 µm scanner. For the measurements a conventional Si tip on a triangular shaped
cantilever was used. For SiC and epitaxial graphene on SiC, contact, non-contact and tapping
mode are all convenient modes of operation. The AFM images were analyzed with the software
WSxM [132].
Raman spectroscopy
For the Raman measurements, an "NTEGRA Spectra Epi setup" system of the company NT-
MDT was used. The Raman spectra were measured under ambient conditions using an Ar+ laser
with a wavelength of 488 nm. The incident laser power was 12 mW, the spot size ≈ 1 µm in
diameter, and the exposure time for epitaxial graphene samples was set to 30 min. Because the
Raman signal of graphene is much smaller than that of SiC, it is necessary to use high power
levels and long acquisition times. The Raman measurements, presented in Section 4.3.6, 5.1.3
and 6.1.2, have been acquired in collaboration with D.S. Lee, B. Krauss and J.H. Smet of the
department von Klitzing.
Chapter 4
Characterization of epitaxial graphene
on SiC(0001)
In this chapter, the basic properties of epitaxial graphene on SiC(0001) will be discussed in
detail. For clarity, it should be noted here that the epitaxial graphene samples are grown in
UHV conditions using on-axis 4H- or 6H-SiC with an initial nitrogen doping level ranging from
1015 cm−3 (semi-insulating) to 1019 cm−3 (strong n-doping). A discussion of dierent graphene
growth conditions and the inuence of the nature of the chosen SiC substrate is given in Chapter 5.
For SiC(0001), the graphene growth is dominated by the fact that the rst carbon layer is
still covalently bound √to the√substrate and therefore does not yet have the typical properties
of graphene. This (6 3×6 3)R30◦ reconstructed interface layer is often named "zerolayer
graphene" and acts as a buer layer for the graphene layers that subsequently grow on top.
After the introduction of √ the surface
√ phase diagram of SiC(0001) in Section 4.1, a detailed
characterization of the (6 3×6 3)R30◦ reconstruction is given in Section 4.2. The growth,
identication and characterization of graphene layers on SiC(0001) is described in Section 4.3.
4.1 The phase diagram of SiC(0001)

On the SiC(0001) surface, the stable phases under both Si-rich and C-rich conditions have been
described rst by R. Kaplan [136] in 1989. The present understanding of the phase diagram is
sketched in Fig. 4.1 and comprises three well ordered phases in addition to a variety of metastable
structures [9, 92, 138141]. Arrows with encircled process numbers indicate dierent preparation
steps, which are explained in detail in the following. The starting point is typically a hydrogen
etched sample with atomically at terraces as shown in Fig. 4.1 (a). Alternatively, CMP polished
samples can be taken (cf. Section 3.2). On the Si-rich side of the phase diagram, a well ordered
(3×3) reconstruction develops. This (3×3) phase can be prepared by annealing the sample to
about 800 ◦ C under simultaneous Si deposition with a ux in the regime of about 1ML/min as
indicated by "A1" in Fig. 4.1. As already mentioned in Section 3.3, Si deposition is an extremely
useful tool to clean the samples from oxygen or, to some extent, even from metallic contamination.
An earlier structure analysis by a combined eort of LEED and density functional theory (DFT)
revealed that this phase is characterized by a Si adlayer with an additional Si adcluster [142, 143].
As shown in Fig. 4.2 (a) and (b), in this phase the SiC substrate is covered with a nearly planar Si
adlayer with sp 2 type Si bond coordination (less then 0.27 Å buckling within the layer as indicated
in panel (b) of the gure). On top of this Si layer, a Si adcluster in (3×3) periodicity is formed by
44 Chapter 4. Characterization of epitaxial graphene on SiC(0001)
Figure 4.1: Phase diagram of the SiC(0001) surface:

√ √ (a) AFM micrograph √ of √ a 4H-SiC(0001) after H-
etching. LEED images of the (b) (3×3), (c) ( 3× 3)R30◦ and (d) (6 3×6 3)R30◦ reconstructions
on 4H-SiC(0001). Arrows with encircled process numbers indicate dierent preparation steps. The image
is taken from Ref. [9, 137].
three Si atoms supporting an additional Si adatom in a three-fold coordination [142, 143]. The
arrows in panel (a) indicate the optimization of the Si-Si bond lengths. The corresponding LEED
pattern of the (3×3) structure is displayed in Fig. 4.1 (b). Heating this surface
√ to√approximately
950 C for about 30 min (with or without Si) results in the development of a ( 3× 3)R30◦ phase
◦
(process "A2" in Fig. 4.1). During this procedure, the surface undergoes
√ √ a complicated phase
transformation with several metastable phases [144], until a sharp ( 3× 3)R30 LEED pattern
◦
appears as shown in Fig. 4.1 (c). The structure of this phase has been determined to be a Si
adatom with three bonds to the Si atoms of the topmost SiC bilayer [144]. The close coordination
to a carbon atom in the lower sublayer of this SiC bilayer leads to a signicant buckling in the
substrate, cf. Fig. 4.2 (c) and (d). It should be noted
√ that
√ hydrogen etched samples, directly after
insertion into the√ UHV√system, do also exhibit a ( 3× 3)R30◦ periodicity. This, √ however,
√ is an◦
oxygen induced ( 3× 3)R30 reconstruction and completely dierent from the ( 3× 3)R30
◦
structure with Si adatoms. A further annealing step √ without

√ additional Si deposition to about
1100 C (process "A3" in Fig. 4.1) leads to the (6 3×6 3)R30◦ reconstruction with the LEED
◦
pattern shown in Fig. 4.1 (d). This phase serves as the starting point for the graphene growth, as
will be discussed in the next section. Growth of one to three layers of graphene can be induced
by further heating the sample to 1200 ◦ C - 1400 ◦ C. Note that, for graphitized samples up to
around bilayer graphene, it is possible to move back to the Si-rich (3×3) surface reconstruction
by performing one or several Si-depositions at 800 ◦ C as already mentioned in Section 3.3.
√ √
4.2 The (6 3×6 3)R30◦ phase as a precursor of graphi-
tization
4.2.1 The periodicity of the interface layer
Already in the rst annealing experiments of SiC surfaces in
√ the √1970s, the periodicity of the
precursor phase of graphitization has been designated as (6 3×6 3)R30◦ as determined from
LEED [81]. However, dierent explanations for the observed LEED pattern have been given
in the meantime [4, 9, 81, 145156]. Using LEED and STM, the following two subsections
try√to shed
√ light◦ into the role
√ of√dierent observed periodicities, which, in particular, include
(6 3×6 3)R30 , (6×6), ( 3× 3)R30 and (5×5).
◦
√ √
4.2. The (6 3×6 3)R30◦ phase as a precursor of graphitization 45
Figure 4.2: (a) Top and (b) side

√ view
√ of the Si-rich (3×3) reconstruction on SiC(0001). (c) Top and (d)
side view of the SiC(0001)-( 3× 3)R30◦ Si-adatom reconstruction. Large spheres represent Si atoms,
small spheres C atoms. (after Ref. [138])
√As a √ guideline for the following discussion, Fig. 4.3 displays a typical LEED pattern of the
(6 3×6 3)R30◦ reconstruction together with √a sketch
√ highlighting the√positions
√ of the dierent
diraction spots, which originate from the (6 3×6 3)R30 , (6×6), ( 3× 3)R30◦ and (5×5)
◦
mesh. Only the brightest spots of the LEED pattern √ in panel

√ (a) are designated. Due to
kinematic suppression, the majority of the spots on the (6 3×6 3)R30◦ grid is not √ visible.
√ The◦
(1/3,1/3) spot and the √ (2/3,2/3)
√ spot, which are marked in blue, belong to the ( 3× 3)R30
reconstruction. These ( 3× 3)R30√ ◦
domains
√ disappear with increasing annealing√temperature
√
and do not belong to the actual (6 3×6 3)R30 √
◦
reconstruction
√ (even if the ( 3× 3)R30◦
related spot positions naturally belong to the (6 3×6 3)R30◦ grid). The brightness of the (5×5)
diraction spots (marked with open black circles) and thus the amount of (5×5) domains depends
on√the detailed
√ nature of the preparation procedure as discussed in the following paragraph. The
(6 3×6 3)R30◦ reconstruction already contains the integer order spots of graphene (marked in
red) even without showing the typical properties of graphene. Furthermore, the reciprocal unit
cell of graphene is rotated by 30◦ with respect to the reciprocal unit cell of the SiC substrate
(marked in green).
The role of (5×5) domains

The preparation dependent amount of (5×5) domains is investigated by means of LEED and
STM. This work is published in Ref. [156]. Such an analysis does of course have an impact on
the graphene growth as the homogeneity of the graphene lms depends on the homogeneity of the
interface layer. The process numbers given in Fig. 4.4 indicate three dierent possible procedures
for the in-situ (that is in the UHV chamber) generation of epitaxial graphene on 4H-SiC(0001): the
rst possible pathway (process "A"), although apparently the most complicated one, is to follow
the development of the entire phase diagram as outlined in Section 4.1, which corresponds to the
consecutive steps "A1",√"A2" √ and "A3", i.e. √
to start
√with a◦Si rich surface and prepare one after
the other the (3×3), ( 3× 3)R30 and (6 3×6 3)R30 reconstructions. In a shortcut, the
◦
√ √
Figure 4.3: (a) LEED pattern of the (6 3×6 3)R30◦ reconstruction at 140 eV. (b) Sketch of the LEED
pattern shown in panel (a),√highlighting
√ the positions of the dierent√diraction
√ spots originating from
four dierent grids: the (6 3×6 3)R30◦ grid, the (6×6) grid, the ( 3× 3)R30◦ grid and the (5×5)
grid. Only the brightest spots of the LEED pattern in panel (a) are designated.
Figure
√ √4.4: LEED-patterns √ √of the◦ mixed
( 3× 3)R30 and (6 3×6 3)R30 recon-
◦
struction of 4H-SiC(0001) for three dier-

ent preparation procedures labelled A, B
and C (see text for details), after anneal-
ing at a temperature of about 1130◦ C. Us-
ing procedure A, the Si-rich (3×3) struc-
ture
√ is√ annealed, using procedure B, the
( 3× 3)R30◦ structure is heated up, and
during procedure C, the preparation is con-
ducted by
√ annealing
√ an ex-situ sample. Spots
on a (6 3×6 3)R30◦ grid as well as on a
(5×5) grid can be observed exhibiting dier-
ent relative intensities.
√ √
√ √
C-rich (6 3×6 3)R30◦ reconstruction can also be obtained by heating √ the
√ (3×3) phase directly
up to 1150 ◦ C. We note that between 1100√◦ C and √ 1150 ◦
C the (6 3× 6 3 )R30 ◦
√reconstruction
√
begins to develop but coexists with the ( 3× 3)R30◦ structure. The pure (6 3×6 3)R30◦
reconstruction is accomplished at 1150 ◦ C. Alternatively (process "B"), it is possible √to start
√ with
an initial annealing step at 950 ◦ C in a ux of Si (≈ 1ML/min), by which the √ ( 3×√3)R30◦
structure develops. By further heating of this surface without Si addition, the (6 3×6 3)R30◦
phase can be realized immediately. The simplest method and still quite √ frequently
√ applied is
to directly anneal the ex-situ hydrogen etched sample to reach the (6 3×6 3)R30◦ phase and
subsequently grow graphene layers [156]. This method corresponds to process "C" in Figure 4.4.
√ √ √ √
Figure 4.4 displays LEED-patterns of the coexisting ( 3× 3)R30◦ and (6 3×6 3)R30◦
reconstructions of 4H-SiC(0001) corresponding to the above described preparation procedures
A, B and C and after the sample √ had been
√ annealed at around 1130◦ C. All three patterns are
usually interpreted as having a (6 3×6 3)R30◦ periodicity. Yet, a detailed inspection of the
spots in the vicinity of the 13 diraction order, as displayed in the enlarged sections on the right
of the three LEED patterns (shown both in normal and reverse contrast for clarity), reveals a
more complex scenario. The spot indicated by the √ arrow
√ in the reverse contrast image is at the
( 13 , 31 )-position, which is characteristic for the ( 3× 3)R30◦ structure. It gradually disappears
with increasing temperature as the corresponding domains disappear, too. The spots on the
triangle marked in green have distances of 61 of the substrate's reciprocal surface unit-mesh
vector. However, since the ( 13 , 13 ) position in the center of the triangle is part of a (6×6) grid
on the SiC(0001) surface, it is clear that the√spots √ on the green triangle are shifted with respect
to the (6×6) grid and belong √ to√a true (6 3× 6 3)R30◦ grid (see also Fig. 4.3). It has to
be emphasized that this (6 3×6 3)R30◦ reconstruction indeed has its designated periodicity.
It is not of (6×6) periodicity, which could be assumed from the (6×6) corrugations that can
be observed in STM (see below). However, √ in addition diraction spots can be observed that
√
are not precisely positioned on the (6 3×6 3)R30◦ grid, namely the spots indicated by the
triangles marked in red in the case of preparation procedure B and C. These spots √ have√a larger◦
distance than the spots on the green triangle and thus cannot belong to the (6 3×6 3)R30
grid. They rather have to be attributed to a (5×5) grid as was conrmed from LEED data at
dierent energies. Notably, the absolute intensity of the spots on the (5×5) grid is dierent for
the dierent preparation procedures, and for preparation procedure A, they even seem too faint
to deduce the existence of (5×5) domains at the surface. Of course, spots at dierent diraction
order on the (5×5) grid have dierent intensities, due to multiple scattering. However, the
relative spot intensities on the red triangle are equal indicating the origin from the same kind of
surface structure within the (5×5) domains for the dierent preparation procedures.
The role of the (5×5) periodicity in the dierent preparation√procedures √ can be elucidated
by STM. Figure 4.5 shows a lled state STM image of the (6 3×6 3)R30◦ reconstruction
exhibiting two dierent periodicities, a (6×6) honeycomb structure on the left side and a (5×5)
structure on the right side of the image. The (5×5) structure is characterized by clusters with
a varying number of atoms. We observe this structure √ √for all three preparation procedures
independent of the annealing temperature of the (6 3×6 3)R30◦ reconstruction. The number
and fraction of (5×5) domains, however, is larger for the ex-situ prepared sample (procedure
C), in full agreement with the stronger intensity level of the (5×5) LEED-spots. In our STM
measurements, this (5×5) structure can be seen only rather rarely. Furthermore, the surface
quality of the annealed ex-situ sample is not good enough to obtain a large number of STM
images producing good statistics. A (5×5) reconstruction
√ √ found in LEED and STM was also
previously reported in the framework of the (6 3×6 3)R30◦ phase [145]. In that case, an ex-
Figure
√ 4.5:
√ Typical STM-images of the
(6 3×6 3)R30◦ structure on 4H-SiC(0001)
usually show a (6×6) corrugation as visible on
the left-hand side of the gure (Utip = 1.7 V).
However, they often appear in coexistence with
(5×5)-periodic patches (right-hand side) that Figure 4.6:
√ Atomically
√ resolved STM image
are formed by somewhat disordered structural of the (6 3×6 3)R30◦ reconstruction of 4H-
elements. The latter√are responsible
√ for the de- SiC(0001) (Utip = 2.0 V). It shows appar-
viations from the (6 3×6 3)R30◦ periodicity ent (6×6) corrugations which are composed of
in the LEED-patterns shown in Fig. 4.4, and many additional atoms or atomic clusters. Also
their amount depends on the preparation pro- disordered features such as trimers can be ob-
cedure. served.
situ sample was annealed without simultaneous Si deposition. The fact that there have been no
further reports on a (5×5) reconstruction recently might be due to the circumstance that, during
the last years, it has become quite common to use Si-deposition in the preparation procedure.
Nevertheless, some (5×5) patches are present also when starting the preparation from the Si rich
(3×3) phase.
With respect to the preparation of graphene surfaces, it can be concluded √that details
√ of
the preparation procedure applied during the development of the intermediate (6 3×6 3)R30◦
phase can strongly inuence their quality as well as the atomic arrangement at the surface.
A homogeneous development of the layer can be greatly enhanced by preparation under Si rich
conditions. It should however be noted that the number of (5×5) domains does not solely depend
on the preparation procedure. A dependence on the SiC substrate material has to be considered,
too, as samples cut out from a dierent wafer have shown a low amount of (5×5) domains even
for direct annealing of the samples without Si deposition.
√ √
(6×6) or (6 3×6 3)R30◦ periodicity?
As√already
√ seen in the left part of Fig. 4.5 the domains on the surface that belong to the
(6 3×6 3)R30◦ reconstruction and not to the (5×5) strucutre are√often√displayed as honey-
combs with (6×6) periodicity. The crystallographic structure of the (6 3×6 3)R30◦ reconstruc-
tion is not completely resolved up to now so that the question arises why there is a discrepancy
between the periodicity observed in LEED and the periodicity observed in STM.
To get more detailed insight Fig. 4.6 shows the lled states of the apparent (6×6) periodicity
in more detail [156]. The image demonstrates the complexity of the reconstruction. Dierent
apparent features can be seen that lack an obvious long-range order. Inside the honeycomb rings,
√ √
4.2. The (6 3×6 3)R30◦ phase as a precursor of graphitization
√ √
Figure 4.7: Atomically resolved STM images for the same surface area of the (6 3×6 3)R30◦ reconstruction measured at 0.3 nA and dierent tip
bias voltages as indicated in the gure. Distinctive features are marked with arrows.
49
Figure 4.8: (a),(b) Atomically

√ resolved
√ STM images at Utip =0.2 V, which exhibit rings of two dierent
sizes thus leading to a (6 3×6 3)R30 unit cell. The lower parts display the same surface area as the
◦
upper parts and show a sketch of the atoms or atomic bumps together with the rings of dierent size.
(c) Arrangement of the atoms or atomic clusters together with the corresponding unit cells. Panel (b)
and panel (c) are taken from Ref. [147, 156].
a dierent number of √ atoms√ with a dierent arrangement of the atoms is observed. The structural
appearance of the (6 3×6 3)R30◦ phase strongly depends on the tunneling voltage.
This observation is conrmed in Fig. 4.7 by an image series for tip bias values ranging from
+2.0 V to -2.0 V showing practically the same surface area. These micrographs emphasize
the diculty to resolve the atomic arrangement at the surface in an impressively clear manner.
Arrows indicating specic surface areas such as defects facilitate the comparison of the dierent
STM images. The (6×6) periodicity is mainly pronounced at high tunneling bias, whereas at
low bias it is dicult to observe a clear long range periodicity. Some of the disorder could arise
from excess Si atoms that have not yet desorbed despite the elevated annealing temperatures.
The trimer features, which are mainly pronounced at 1.0 V bias according to Fig. 4.7, have
for example recently been speculated to be Si adatoms [157]. These features could also arise
from Si dangling bonds. With respect to the growth of graphene layers, it is by all means
conceivable that the disorder and height dierences in the interface layer could alter the electronic
quality
√ of√the graphene layers growing on top. It is furthermore evident that STM images of the
(6 3×6 3)R30◦ reconstruction do not only contain topological information as claimed by Chen
et al. [151] but also contain information about the electronic structure.
As seen in Fig. 4.7, low bias STM images hardly allow to identify a√periodic√ arrangement of
the surface atoms. However, under certain tip conditions, the true (6 3×6 3)R30◦ structure
can indeed be resolved as shown in Fig. 4.8. The STM images in panel (a) and (b) at a tunneling
voltage of 0.2 V (panel (b) from Ref. [147, 156]) reveal two types of rings with slightly dierent
size
√ thus√forming◦ a unit cell larger than that of the (6×6) periodicity, namely a unit cell of
(6 3×6 3)R30 periodicity. Three atomic bumps within the rings are only present in the larger
rings. Each bump is part of a diamond of four atoms (marked √ in red √
in panel (a) and panel (b)),
which in the same orientation is repeated only with the (6 3×6 3)R30◦ periodicity. In the
next subsection, it is shown that these features could be attributed to Si dangling bonds. The
periodicity of the atomic features is elucidated in panel (a)-(c) of Fig. 4.8 by a sketch of the
quasi-(6×6) rings with their dierent size and the three additional atoms (or atomic clusters).
The lower parts of panel (a) and panel
√ (b)√display ◦the same surface area as the upper parts. The
unit cells of the (6×6) and the (6 3×6 3)R30 periodicity are indicated in panel (c). With
√ √
this real space arrangement the periodicity of the LEED-patterns can be explained, which is not
possible
√ assuming
√ a (6×6) structure only. Of course, this sketch resolves only a few atoms of the
(6 3×6 3)R30 surface, its complete atomic structure is still
◦
√ unresolved.
√ This fact, however,
should not come as a surprise since
√ the √ side length of one (6 3×6 3)R30 unit cell amounts
◦
to 32 Å. Consequently, the (6 3×6 3)R30◦ unit cell contains 108 Si and 108 C atoms per
SiC bilayer or 338 atoms in a graphene layer, which can easily be calculated from the length of
the (1×1) SiC unit vectors, which amounts to 3.08 Å and from the length of the graphene unit
vectors being 2.46 Å.
√ √
4.2.2 The nature of the (6 3×6 3)R30◦ reconstruction
√ √
The very complex appearance of the (6 3×6 3)R30◦ structure is only partially √ understood
√
up to now. In the literature, several models exist that do not consider the (6 3×6 3)R30◦
structure as an inherent surface reconstruction of SiC(0001).
√ Owman
√ and Mårtensson proposed
the coexistence of a (6×6) and an incommensurate ( 2.1 × 2.1)R30◦ phase [145], which
is not in agreement with the interpretation of our LEED and STM results. It has often been
argued that the LEED√ diraction
√ image shows a Moiré pattern of graphite and the SiC substrate
((1×1) [4, 81] or ( 3× 3)R30 [149]). However, such an interpretation is in inconsistency with
◦
both band structure measurements using ARPES (see below and Ref. [49, 158160]) and the
STM images displayed in√Fig. 4.7 √ and ◦ 4.8. Furthermore, STM images show that graphene is
growing on top of the (6 3×6 3)R30 reconstruction (see Section 4.3.4 below). √ The√specic◦
properties of graphene develop only with the rst layer of graphite on top of the (6 3×6 3)R30
structure (see Section 4.3.2 and Ref. [49, 152, 153, 155, 158, 159, 161163]), while the latter
remains unperturbed during the growth of the graphene layers. A recently proposed model by
Chen et al. suggests a self-organization of surface carbon atoms in a "carbon nanomesh" with
(6×6) periodicity [151]. This model is also in contradiction to the band structure measurements.
Furthermore, it is claimed that the STM images look the same for dierent tunneling voltages.
Our STM images clearly show dierent atomic orbitals of the structure for dierent bias voltages,
which is in accordance with earlier observations [145].
√ √
In order to understand the nature of the (6 3×6 3)R30◦ reconstruction in more detail
and to suggest the structural model of a covalently bound initial carbon layer, which is the
common understanding at present, the most important experimental results are given in the
following. The Moiré pattern interpretation of the LEED data [4, 81, 149] can be tested by band
structure measurements. The bands√characteristic
√ for graphene are the σ -bands and the π -bands
(cf. Section 2.1.1). Whereas the (6 3×6 3)R30 structure already shows σ -bands [155, 159],
◦
no√π -bands
√ can ◦be observed [155, 159, 160]: Figure 4.9 (a) displays the band structure of the
(6 3×6 3)R30 reconstruction around the K-point of the graphene Brillouin zone as measured
from angular resolved photoelectron spectroscopy (ARPES) using He II excitation. For a graphene
layer, linear π -bands, which arise from the two sublattices of graphene, would be expected.
However, only two very faint delocalized and smeared out states at binding energies of around
0.1 eV to 0.5 eV and higher than 0.9 eV are visible.
The partial existence of graphene related properties is also conrmed from synchrotron based
CLPES. Figure 4.9 (b) shows the C 1s core level signal. Dierent components contributing to the
spectra were decomposed by a curve tting procedure after subtraction of a Shirley background.
The tting parameters can be found in Table 4.1. The depth position of the corresponding
species within the surface was identied by varying the incident photon energy and thus changing
the surface sensitivity. The corresponding energy in Fig. 4.9 (b) is 600 eV. Besides the SiC bulk
√ √
Figure 4.9: Unravelling the nature
√ of the
√ (6 3× 6 3)R30◦ reconstruction. (a) Inverse grayscale plot
of the band structure of the (6 3×6 3)R30 reconstruction near the K-point obtained by ARPES
◦
using He II radiation. The measurement

√ √ direction as indicated on the left is perpendicular to ΓK.
(b) C 1s spectrum of the (6 3×6 3)R30◦ reconstruction and its deconvolution measured with an
incident
√ photon
√ energy of 600 eV. These data have been published in Ref. [164]. (c) Structural model of
the (6 3×6 3)R30◦ reconstruction in side view. It consists of a covalently bound initial carbon layer
that does not yet exhibit the typical properties of graphene.
peak at 283.73 eV, two additional components called S1 at 284.99 eV and S2 at 285.60 eV can
be identied. Neither component S1 nor component S2 is located √ at
√ the position expected for
graphene, which would be at 284.7 eV. Consequently, the (6 3×6 3)R30◦ structure √ √consists◦
of two inequivalent non-graphene-like types of carbon atoms. Considering the (6 3×6 3)R30
structure as an initial carbon layer that is partially covalently bound to the substrate,
√ they can
√
be attributed to the following origin: S1 results from the carbon atoms in the (6 3×6 3)R30◦
structure bound to one Si atom of the SiC(0001) surface and to three C atoms in the sp 2 -
bonded layer, S2 is the component emitted from the remaining sp 2 -bonded carbon atoms in the
buer layer. The two components have an area ratio S1:S2 of slightly below 1:2. Note that the
interpretation of the core level spectra presented here is in accordance with Ref. [155, 159].
The experimental
√ information from LEED, STM, ARPES, and CLPES allow for a structural
√
model of the (6 3×6 3)R30◦ reconstruction, which is sketched in Fig. 4.9 (c) and in Fig. 4.10.
The interface layer represents an initial carbon layer in graphene-like honeycomb arrangement and
is partially covalently bound to the Si-terminated substrate. The bonds indicated in Fig. 4.9 (c)
should be regarded as possible bond formations since not all Si atoms can form a bond to carbon
atoms, due to the dierent lattice constants of SiC and graphene and due to the 30◦ rotation
angle of the initial carbon layer with respect to the substrate. This model is in agreement with
the observation in CLPES that slightly less than one third of the atoms of the initial carbon
layer are bonding down to the substrate. The covalent bonding breaks the hexagonal network of
π -orbitals
√ √but preserves the σ -bonds as also recently conrmed by DFT calculations based on a
(6 3×6 3)R30 unit cell [161].
◦
Now, let us have a closer look at those Si atoms that are not involved in the bonds and
therefore have a Si dangling bond. This becomes more apparent in the top view image in Fig. 4.10
that gives more insight in the registry relation between the SiC substrate and
√ the √
initial carbon
layer. Note again the complexity of this system with 338 carbon atoms per (6 3×6 3)R30◦ unit
cell being equivalent to a (13×13) graphene unit cell. The quasi-(6×6) honeycomb hexagons with
√ √
√ √
Figure 4.10: Structural model of the (6 3×6 3)R30◦ reconstruction in top view. It consists of an
initial carbon layer on√top of√the Si-terminated (1×1)-SiC substrate. The SiC and the graphene lattice
together with the (6 3×6 3)R30◦ unit cell and the three hexagons in quasi-(6×6) periodicity are
indicated. The blue shaded features along the walls of the quasi-(6×6) honeycomb are exactly the
atomic clusters observed in the STM images of Fig. 4.8 (a) and (b).
√ √
slightly varying size are highlighted in green, the true (6 3×6 3)R30◦ periodicity is highlighted
in blue. The light blue shaded areas mark the dangling-bond like features or bright spots displayed
in the STM images of Fig. 4.8 (a) and (b). They are distributed along the walls of the quasi-(6×6)
honeycomb structure and are situated exactly on (1×1) SiC-substrate grid positions. Subsequent
to the STM analysis, which is shown in Fig. 4.8 (a) and (b) (cf. Ref. [147, 156]), two DFT
studies
√ [161,√ 162] and an empirical potential calculation [165], which are all based
√ on a√unit cell
of (6 3×6 3)R30◦ size could also resolve the discrepancy between (6×6) and (6 3×6 3)R30◦
periodicty of the interface layer. Topographic [161, 162, 165] and charge density maps [161, 162]
exactly exhibit the quasi-(6×6) periodicity with the two types of hexagonal rings observed in
STM. In the calculations, the valleys of the quasi-(6×6) honeycomb are attributed to carbon
atoms covalently bound to the SiC substrate, the walls of the quasi-(6×6) structure are claimed
not to arise from Si dangling bonds but from carbon atoms that are not bonded to the SiC
substrate and therefore pushed out of the plane.
Finally, it should be emphasized again that the carbon atoms in the interface layer are not
equivalent with respect to the coupling to the SiC substrate. Accordingly, a spatially dependent
electronic inuence of the interface layer with respect to the above growing graphene layers has to
be expected. It is also one of the primary suspects of the reduced mobilities measured for epitaxial
graphene on SiC(0001). However, apart from these drawbacks, it should not be forgotten that
the interface is as well responsible for the ordered growth of the graphene layers as it imposes a
30◦ rotation with respect to the SiC substrate.
Figure 4.11: Structural model of (a) monolayer √and (b)

√ bilayer epitaxial graphene on SiC(0001) in side
view. The graphene is growing on top of the (6 3×6 3)R30◦ reconstructed interface layer.
4.3 The growth and identication of graphene layers

√ √
As elaborated in Section 4.2, the (6 3×6 3)R30◦ reconstructed interface layer serves as a pre-
cursor stage of graphitization and is therefore often called "zerolayer graphene" [160, 166]. Only
the rst graphene layer on top of this buer layer can be named monolayer graphene. Figure 4.11
shows a structural model √ of (a)
√ monolayer and (b) bilayer graphene in side view. During the
graphene growth the (6 3×6 3)R30 structure is buried below the graphene layers. However,
◦
it is shown in this section that it still does have a signicant electronic inuence on the graphene.
Concerning the graphene growth it should be considered that new graphene layers are formed due
to the desorption of Si atoms. From a simple count of the carbon atoms, it is evident that three
SiC bilayers are needed to form one graphene layer. As a new graphene layer has to be built from
the bulk √material,
√ it has◦ recently been assumed that every new layer implies the √ re-formation
√
of the (6 3×6 3)R30 reconstruction [155]. Simultaneously, the previous (6 3×6 3)R30◦
structure is released from its covalent bonding to the substrate and is transformed into a true
graphene layer. In this way, all graphene layers have the same 30◦ rotation with respect to
the substrate. As conrmed from STM and band structure measurements (see Section 4.3.2
and 4.3.4) the graphene is AB-stacked (so-called Bernal type stacking). In contrast, we will see
in Section 5.1.1 the absence of an interface layer on SiC(0001̄) causes rotational disorder in the
epitaxial graphene. In the present section a detailed structural and electronic characterization of
epitaxial graphene layers on SiC(0001) will be performed by means of CLPES, ARPES, LEED,
STM and STS, LEEM and Raman spectroscopy.
4.3.1 Core level photoelectron spectroscopy

√ √
The fact that the (6 3×6 3)R30◦ reconstruction remains intact for a dierent number of
graphene layers can clearly be demonstrated from CLPES data. Figure 4.12 shows the C 1s
core level signal measured at 600 eV incident photon energy1 . The data were acquired and
analyzed in the same way as in Fig. 4.9 (b). The tting parameters are summarized in Table 4.1.
In comparison to zerolayer graphene (Fig. 4.9 (b)) the C 1s peak for monolayer graphene in
Fig. 4.12 (a) can only accurately be tted after introducing a fourth component (G), which
arises from the graphene overlayer. This component increases in intensity for higher graphene
1 Note
that the precise calibration of the number of layers for this data set was carried out using ARPES as
shown in Section 4.3.2.
4.3. The growth and identication of graphene layers 55
Figure 4.12: C 1s spectrum for (a) mono- and (b)√ bilayer

√ epitaxial graphene together with their de-
convolution into bulk (SiC), graphene (G) and (6 3×6 3)R30◦ reconstruction (S1 and S2) related
components. The incident photon energy is 600 eV. These data have been published in Ref. [164].
Figure 4.13: Si 2p spectra for (a) dierent graphene√coverages

√ and (b) monolayer epitaxial
√ graphene
together with the deconvolution into bulk (SiC), (6 3×6 3)R30 reconstruction (6 3) and defect
◦
(def) related components. The incident photon energy is 330 eV. These data have been taken in the
framework of Ref. [164].
C 1s, 600eV SiC Graphene (G) S1 (interface) Figure

S2 (interface)
coverage EB width area EB width area EB width area
(ML) (eV) (eV) (%) (eV) (eV) (%) (eV) (%) (%)
0 283.73 0.82 42 - - - 284.98 0.89 18 4.9(b)
285.61 1.00 40
1 283.70 0.75 27 284.67 0.56 31 285.04 0.84 13 4.12(a)
285.53 0.97 29
2 283.66 0.72 16 284.56 0.53 55 285.01 0.84 9 4.12(b)
285.50 1.01 20
√
Si 2p, 330eV SiC Si to 6 3 defect Figure
coverage EB width area EB width area EB width area
(ML) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%)
1 101.57 0.62 77 102.19 1.0 13 101.00 0.95 10 4.13(b)
Table 4.1: Fit parameters for the C 1s and Si 2p core level spectra for epitaxial graphene samples with
a dierent number of layers. A linear background was subtracted from the Si 2p data and a Shirley
background from the C 1s data. The C 1s graphene related peak was tted using a Doniach-Sunjic
prole to account for the metallic behavior (asymmetry parameter 0.03, Gaussian contribution 40%), all
the other tted lines are an approximation of the Voigt function (Gaussian contribution 25% for C 1s,
70% for bulk Si 2p and 40% for the surface components of Si 2p). The values of the width are given
with respect to the full width at half maximum of the corresponding peaks.
coverage as shown for bilayer graphene in Fig. 4.12 (b). The intensity ratio of the interface related
components
√ S1 and S2 to the SiC bulk component, however, remains constant, which means that
√
the (6 3×6 3)R30◦ structure is covered by the graphene layers but is otherwise not altered.
Furthermore, the graphene peak shifts towards lower binding energy with increasing number of
layers. The explanation for this observation is coming from band structure measurements shown
in Section 4.3.2, where it is shown that epitaxial graphene is n-doped due to an electric dipole in
the interface layer. For an increasing number of layers, the corresponding electric eld is screened
and the upper layers are less doped.
The Si 2p peak measured at 330 eV incident photon energy is shown in Fig. 4.13 (a) for
varying graphene coverage. It clearly exhibits the same shape for a dierent number of layers.
Due to spin orbit splitting, the pure Si 2p peak already consists of two components, Si 2p 1/2
and Si 2p 3/2. Although, from a visual inspection, it is not easy to estimate the number of
Si 2p subcomponents, the spectra at least cannot be tted by just one Si 2p peak. Two more
components are actually necessary for an accurate tting as shown in Fig. 4.13 (b) for monolayer
graphene. Prior to curve tting with an approximation of the Voigt function, a linear background
was subtracted. Besides the SiC bulk peak, the component at higher binding energy side has
been attributed
√ √ to arise from the Si atoms of the substrate that bind to the carbon atoms of
the (6 3×6 3)R30 reconstruction [167]. As they are still in sp 3 -hybridization, only a small
◦
chemical shift with respect to the bulk peak, possibly due to strain, can be observed. The peak
at lower binding energy side has been attributed to stem from surface defects [167]. This peak is
absent for high-quality samples with large-scale homogeneity, which are prepared in an induction
furnace under an Ar atmosphere (see Ref. [10] and Section 5.2.2).
4.3.2 The band structure of epitaxial graphene

Monitoring the graphene growth layer by layer
Direct access to the band structure of epitaxial graphene is given by angle resolved photoemission
spectroscopy (ARPES). In contrast to LEED, STM and CLPES, it gives direct and precise infor-
mation about the number of graphene layers and their quality. Sharp bands are a clear indication
of a high degree of ordering and of good uniformity of the samples.
Figure 4.14 displays ARPES data from He II excitation revealing the electronic structure in the
vicinity of the K-point of the Brillouin zone of graphene for annealing temperatures of 1200 ◦ C and
1350 ◦ C [163]. The sketch in panel (a) denes the ~kk -mapping direction. The detailed dispersion
of the dierent π -bands is well resolved as shown in the 2D plots in panel (a) and compares well
to previous reports based on synchrotron data [49, 54, 168]. By counting the number of π -bands
one and two graphene layers can be distinguished. The sketch on the right side of Fig. 4.14 (a)
serves as a guide to the eye. For a quantitative evaluation, momentum and energy distribution
curves (MDC and EDC) for mono- and bilayer graphene are displayed in Fig. 4.14 (b) and (c).
Consistent with previous reports [54, 168], we nd the Dirac point shifted below the Fermi energy
by 430 meV for monolayer and 320 meV for bilayer and also the reported [54, 168] energy gap
at the Dirac energy for bilayer graphene can be observed. Using 4H- and 6H-SiC samples we
could verify that the band structure for epitaxial graphene shows no polytype dependence. The
width of the monolayer bands as determined from the MDC at 0.8 eV binding energy is around
−1
0.033 Å (full width at half maximum). For comparison the momentum resolution from the
−1
detector amounts to 0.014 Å (cf. Section 3.3).
UHV prepared samples always suer from an inhomogeneity on the micrometer scale as shown
below in Section 4.3.5. Whereas a monolayer sample is relatively easy to prepare without an
obvious bilayer contribution, the situation for bilayer graphene is already quite complicated. Both
contributions from monolayer and trilayer can be detected as highlighted by the sketch on the
right side of Fig. 4.14 (a). It should be noted, however, that for the monolayer graphene sample
it is hard to estimate the percentage of monolayer area on the surface as zerolayer graphene does
only have rather faint, delocalized and smeared out states as displayed previously in Fig. 4.9 (a).
Moreover, these states are not altered for a fully covered graphene surface as demonstrated in
Ref. [167], which shows the weak van der Waals-type
√ √ interaction between monolayer and zerolayer
epitaxial graphene [152, 153, 166]. The (6 3×6 3)R30 reconstruction of zerolayer graphene
◦
leads also to the appearance of replica bands due to diraction of the outgoing photoelectrons
as shown in Ref. [169].
Let us now have a closer look at the already mentioned n-doping of the epitaxial graphene.
The charge carrier concentration is determined by the Fermi surface and is given by Equation 2.8
of Section 2.1.1, if a circular shape of the Fermi surface is assumed. This is a reasonable
approximation as long as the Fermi energy is not too far away from the energy level of the Dirac
point. In the following, the Fermi wave vector with respect to the K-point is named kF so that the
charge carrier concentration amounts to n = kF2 /π . The resulting electron density for monolayer
graphene is about n ≈ 1 · 1013 cm−2 . The origin of this relatively high amount of n-doping is up
to now not clear. It has to be emphasized that it is independent of the polytype, of the amount of
n- or even p-doping of the SiC wafer substrate√ and √ also of the detailed preparation technique. It
however is related to the presence of the (6 3×6 3)R30 reconstruction. For epitaxial graphene
◦
on SiC(0001̄) that does not exhibit any interface layer, and for decoupled epitaxial graphene on
SiC(0001) after hydrogen intercalation (cf. Section 6.2.1), conditions around charge neutrality
Figure 4.14: Band structure of monolayer and bilayer graphene epitaxially grown on 6H-SiC(0001) near
the K-point obtained by ARPES using He II radiation. (a) Photoemission images measured in ky -direction
(as indicated on the upper left) revealing the dierent branches of the π -bands for dierent layer numbers.
The position of the Dirac energy (see lower sketch on the left) is indicated. A tentative sketch highlights
the band contributions arising from a dierent number of layers. (b) Momentum distribution curves
extracted from the images at the energies indicated by the black lines in panel (a) and noted under
the curves, integrated over 30 meV. (c) Energy distribution curves extracted at ky = 0 integrated over
0.01 Å−1 .
Figure 4.15: Band structure of epitaxial

monolayer graphene after exposure to air as
measured with ARPES (He II excitation).
Due to p-type dopants the Dirac energy is
shifted below the Fermi level only by 260 meV
and not by 430 meV as for clean epitaxial
monolayer graphene (cf. Fig. 4.14). The low
quality of the image has to be attributed to
contaminations from air.
are observed [57, 160, 170]. √ DFT calculations

√ suggest that the charge transfer takes place via the
Si dangling bonds of the (6 3×6 3)R30◦ reconstruction [152, 153]. A natural explanation for
the decrease of the displacement of the Dirac point relative to the Fermi level with an increasing
number of graphene layers comes from the fact that the top most layers are less aected by
the interface layer due to screening. In terms of applications in electronic devices, it has to be
considered that the carrier concentration changes after the samples are taken out of the UHV
chamber. ARPES measurements for monolayer graphene after exposure to air show a Dirac
energy of 260 meV below the Fermi level (cf. Fig. 4.15). Using this value, we obtain an electron
concentration of n ≈ 4.2 · 1012 cm−2 . The reduced electron concentration may be due to
compensation by p-type dopants such as adsorbed oxygen, hydrocarbons or water molecules from
the ambient environment. The ARPES data shown here serve as a rst proof of principle of
the feasibility of p-type doping of epitaxial graphene, which is an essential issue for applications
in graphene based nanoelectronics since the strong intrinsic n-doping has to be compensated.
More details and more specic means of p-type doping are presented in Section 6.1 and in the
respective publications in Ref. [51, 171].
In the literature, there is an intensive controversy about the interpretation of the monolayer
graphene band structure around the Dirac point. Obviously, the π bands above and below the
Dirac point are not following a straight line. The band is either getting a kink when crossing the
Dirac point region or an energy gap is opening at the Dirac point, with the intensity astonishingly
not dropping to zero inside the energy gap. Quite strong arguments against the existence of
a gap are coming from an interpretation in terms of many-body interaction eects such as
electron-phonon scattering, electron-hole pair creation, electron-electron interaction and electron-
plasmon scattering, which have a signicant inuence on the graphene band structure [166, 169].
Electron-plasmon scattering is claimed to be responsible for the kink directly around the Dirac
point region [166, 169, 172]. The interpretation of the band structure data in terms of an energy
band gap is argued to arise from an A-B-asymmetry of the graphene sublattices of zero- and
monolayer graphene [161, 168, 173175]. The non-zero intensity in the energy is explained with
a mid-gap state originating from the zerolayer [161]. At the moment, no denite agreement can
be found in the literature. The possible existence of a gap mainly depends on dierent ways of
interpreting the data. However, recent transport measurements did not show any evidence for a
gap opening [33]. Furthermore, STM data [162, 176178], calculations of the local density of
states (LDOS) [177] and ARPES simulations [179] show that a possible gap opening is probably
not due to A-B sublattice breaking as claimed in Ref. [168].
It should be noted that for samples that do not have a fully developed graphene layer but
consist of many small islands or ribbons the opening of a gap at the Dirac point is expected due
to connement [35, 36], and is commonly accepted to contribute to the observed band structure
in ARPES [172, 173]. Such an explanation is of course not valid any more for a fully developed
graphene layer.
In contrast to monolayer epitaxial graphene, there is no controversy about the existence of a
band gap for bilayer epitaxial graphene. As already introduced in Section 2.1.1, the band structure
of Bernal (AB) stacked bilayer graphene can be determined from a tight binding approach [48].
Provided there is an electrostatic asymmetry between the rst and second layer, an energy gap
opens. For epitaxial bilayer graphene, the n-doping from the substrate and the partial screening of
the corresponding dipole eld by the rst graphene layer indeed render the rst and second layer
inequivalent and evoke the opening of a band gap. A control of the charge carrier concentration
by surface transfer doping even allows for a precise control over the band gap size as shown in
Ref. [49] for further n-doping and in Section 6.1.2 (or Ref. [51]) for p-doping.
Figure 4.16: π -band dispersion of monolayer epitaxial graphene along ΓK for dierent incident photon
energies using (a) (right) circular polarized light and (b) linearly polarized light.
The inuence of the incident photon beam in the band structure measurements
The detailed appearance of the graphene band structure varies in the literature [49, 54, 163, 168].
An explanation can be found in dierent experimental conditions so that it is possible to get access
to dierent characteristics of graphene's band structure by changing the measurement parameters.
In the following, a short overview of the inuence of measurement direction, incident photon beam
energy and incident photon beam polarization is given. The ARPES results were acquired at the
Surface and Interface Spectroscopy beamline at the Swiss Light Source. Here, the dispersion
scans were measured parallel to the ΓK-direction at a sample temperature of 80 K.
Fig. 4.16 shows dispersion scans for a selection of dierent photon energies for (a) right
circular polarized light and (b) (horizontally) linearly polarized light. It should be noted here that
the beamline setup allows only a relative and not an absolute denition of right and left circular
polarization. From a rst inspection, it becomes clear that, in contrast to the measurements
taken perpendicular to the ΓK-direction (see earlier in this section), only one branch of the π -
bands is clearly visible in most cases. Only for right circular polarized light, the second branch
becomes also visible at low incident photon energies of around 25 eV to 30 eV. In such a case
an asymmetry of the bands can be observed. This asymmetry is not a matter of measurement
accuracy but it is intrinsic to graphene's band structure. It reects the so-called trigonal warping,
which becomes more evident in constant energy maps at binding energies higher than 1 eV as
shown below. For single dispersion scans such as those in Fig. 4.16, this asymmetry can only be
Figure 4.17: Constant energy maps for monolayer epitaxial graphene at dierent binding energies for
40.8 eV incident photon energy using (right) circular polarized light.
Figure 4.18: Inuence of the polarization of the incident photon beam on the measured intensity distribu-
tion of the graphene band structure. (a),(b) π -band dispersion of monolayer graphene along ΓK, (c),(d)
π -band dispersion of monolayer graphene perpendicular to ΓK for an incident photon energy of 30 eV
using (a),(c) left circular polarized and (b),(d) right circular polarized light. (e),(f) Constant energy
maps at dierent binding energies for an incident photon energy of 30 eV using (e) left circular polarized
and (f) right circular polarized light. Consider that left circular polarized light is dened as the opposite
direction to right circular polarized light and does not follow an absolute denition.
detected if the second π -band branch is visible and if the measurement direction is along the ΓK-
direction. Therefore, existing dispersion scans in the literature do not show this asymmetry. For
linearly polarized light, signicant changes in dependence of the incident photon energy cannot
be observed. To understand the detailed appearance of the ARPES measurements for dierent
photon energies and dierent polarization, a sophisticated modelling of the ARPES intensities
would be necessary, which is beyond the scope of this thesis. Irrespective of the polarization, the
bands are most intense for energies around 40 eV, which explains that He II excitation in the home
laboratory is ideal for the analysis of the graphene band structure, whereas He I is not suitable
despite its higher intensity. For 40 eV incident photon energy, the kink in the band structure
around the Dirac point, which was already discussed earlier in this section, is well pronounced.
Higher incident photon energies than 40 eV such as 90 eV are also suitable for a detailed inspection
of the graphene band structure [49, 54]. The images in Fig. 4.16 at 70 eV and 90 eV are of low
quality only due to experimental limitations of the detector's angular dispersion setup. Note that
the sharpness of the bands is of comparable size for all energies. The full width at half maximum
−1 −1
at 0.8 eV binding energy amounts to 0.036 Å for 40 eV incident photon energy, 0.052 Å for
−1
90 eV incident photon energy and 0.055 Å for 30 eV incident photon energy (all values are
for (right) circular polarization). For comparison, the momentum resolution from the detector is
approximately one order of magnitude lower (cf. Section 3.3).
The already mentioned trigonal warping becomes more apparent in constant energy maps of
graphene's band structure, which are displayed in Fig. 4.17 for dierent binding energies at 40.8 eV
incident photon energy using (right) circular polarized light. 40.8 eV corresponds exactly to the
energy of He II radiation. For binding energies up to around 1 eV, a circular shape is observed,
which conrms the correctness of the calculation of graphene's charge carrier concentration earlier
in this section. For higher binding energies, the pattern is not of circular shape any more but shows
trigonal warping as can be seen in the left image of Fig. 4.17. It just reects the non-linearity of
the π -bands far below the Fermi level.
The fact that under certain measurement conditions only one branch of the π -bands can be
detected is due to an anisotropy in the measured intensity of the graphene band structure. From
an inspection of the constant energy maps in Fig. 4.17, it is evident that the measurement along
the ΓK-direction leads to the extinction of the band at higher momentum for energies below
the Dirac point and of the band at lower momentum for energies above the Dirac point. The
reason for the observed angular dependence of the intensity is directly related to graphene's lattice
structure. As graphene has two inequivalent atomic sites, a photoemission experiment can be seen
as a two-source interference like Young's double slits [179]. Using a tight binding Hamiltonian
and Fermi's golden rule, Mucha et al. have simulated such constant energy maps for ARPES
experiments [179]. The angular dependence calculated from this model exactly corresponds to
the measured constant energy maps as seen in Fig. 4.17. Furthermore, the constant energy maps
directly reect the chiral nature of the electrons. The chirality is determined from the projection
of the pseudospin of the electron to the direction of its wave vector (see Section 2.1.1). Electrons
in graphene below the Dirac point have the positive chirality +1, electrons above the Dirac point
have the negative chirality -1. Since the chirality of the electrons changes its sign when crossing
the Dirac point, the intensity distribution in the constant energy maps is also ipped with respect
to energies below the Dirac point. Correspondingly, it is obvious that a band structure slice in
ΓK-direction for the constant energy maps shown in Fig. 4.17 only shows one π -band branch,
with all electrons having the same pseudospin (compare Fig. 4.16 (a) at 40 eV). Concerning this
discussion, it should be noted that a gap at the Dirac point due to an AB-asymmetry, as claimed
in Ref. [168], would be accompanied by the loss of the chirality-related anisotropy of the constant
energy maps near the Dirac point [179], which is not observed in our data.
The inuence of the polarization of the incident photon beam is elucidated in Fig. 4.18, which
compares data for left and right circular polarized light for an incident photon energy of 30 eV.
Note again that left circular polarized light is not dened in absolute but only in relative terms
as the opposite direction to right circular polarized light. The π -band dispersion along ΓK does
not show any obvious dierences for the two polarization directions as displayed in panel (a) and
panel (b), with both π -band branches exhibiting signicant intensity. However, for dispersion
scans perpendicular to ΓK, the picture is completely dierent: for left and right circular polarized
light, only one of the two branches of the π -bands is visible. The intensity of the two branches
is inverted when changing the polarization direction. Correspondingly, the constant energy maps
displayed in panel (e) and panel (f) show major changes in the angular dependence of the intensity.
The main intensity is shifted from positive to negative ky values for energies below the Dirac point
and from negative to positive ky values for energies above the Dirac point. Such an observation
has not yet been reported in the literature. The reason remains for the moment unclear and
requires a detailed modelling of the ARPES intensities in dependence of the polarization. It also
has to be considered that chirality of light and pseudo-chirality in graphene are of dierent origin.
The detailed appearance of constant energy maps requires further studies in order to analyze
and understand the inuence of the incident photon beam polarization and energy in more detail.
Comparing the constant energy maps for 40.8 eV (Fig. 4.17) and 30 eV (Fig. 4.18 (b)), the
angular dependence is strikingly dierent. The theoretical calculations in Ref. [179] do not take
such an energy dependence into account. Only for bilayer graphene, deviations are expected as
dierent phase shifts for dierent photon energies arise due to the emission of electrons from the
rst and the second graphene layer.
4.3.3 LEED ngerprints for the determination of the number of

layers
One crucial issue that deserves further attention is the accurate and at the same time easy and
practical determination of the number of graphene layers on top of SiC(0001). The methods used
in the literature at present are ARPES at a synchrotron facility (see Ref. [49, 54, 159, 168] or
Section 4.3.2) or from He II excitation in the "home laboratory" (see Ref. [163] or Section 4.3.2),
STM and STS (see Ref. [154, 156, 157, 176] or Section 4.3.4), LEEM (see Ref. [111, 163,
180] or Section 4.3.5), AES [4, 12, 181], CLPES (see Ref. [4, 137, 155] or Section 4.3.1) and
Raman spectroscopy (see Ref. [182184] or Section 4.3.6). However, these techniques have the
limitation that they either depend on a complex experimental setup (e.g. synchrotron), cannot
be applied continuously during the preparation procedure or have a limited accuracy such as
due to the inaccurate knowledge of the inelastic attenuation of the surface probing electrons in
AES or CLPES (cf. Section 3.1). After calibrating CLPES C 1s spectra with ARPES, they can
successfully be used as ngerprints for the determination of the number of layers [155]. The most
common analysis equipment in practically every UHV sample preparation chamber is LEED. In
this subsection, it is shown that ngerprints in LEED can overcome present diculties and allow
for an exact thickness determination for at least zerolayer, monolayer, and bilayer graphene on
SiC(0001), so that the number of layers can be continuously monitored during the preparation
process with high accuracy. The LEED ngerprints are calibrated by means of ARPES with He II
excitation from a laboratory based UV source as already introduced in Section 4.3.2. The results
presented here are published in Ref. [156, 163]. Note that the calibration of the number of
graphene layers from ARPES has been revised in the meantime and is slightly dierent from that
Figure 4.19: LEED intensity spectra in dependence of√the annealing

√ temperature of the graphene (10)
spot marked in green in the inset. On the right, (6 3×6 3)R30◦ LEED patterns at representative
annealing temperatures are shown.
one in the publication.

During graphitization, the LEED pattern continuously undergoes variations visible by eye as
shown in Fig. 4.19 for epitaxial graphene on 4H-SiC(0001). The inset above the I(E)-spectra
indicates the spots being discussed √in this√ context (see also Fig. 4.3 of Section 4.2.1):
√ in√the rst◦
stage (scenario at 1090 C), the (6 3×6 3)R30 reconstruction coexists with the ( 3× 3)R30
◦ ◦
phase as noted already in Section 4.1 and 4.2. Higher √temperatures

√ (approximately 1160◦ C) lead
to the disappearance of the spots related to the ( 3× 3)R30 structure, i.e. the diraction
◦
spots at the ( 31 , 31 )-position and at the ( 23 , 23 )-position. When graphene layers are growing on
√ √
top of this pure (6 3×6 3)R30◦ structure (above ≈1180◦ C), the graphene (10) spot, which
is marked in green in the inset of Fig. 4.19, shows an increasing intensity. Simultaneously, the
diuse background in the LEED patterns continuously √ increases.
√ However, as known from STM,
the surface has a similar quality as for the pure (6 3×6 3)R30 √ ◦
reconstruction. The graphene
(10) spot is positioned next to the ( 23 , 23 ) spot position with a 1/(6 3) distance (green dot in the
inset of Fig. 4.19).
√ We√have analyzed the energy dependent intensity of all clearly visible LEED
spots of the (6 3×6 3)R30 pattern in dependence of the annealing temperature. Despite
◦
the dierent scattering properties for a dierent stage of graphitization, the LEED spectra do
not show any signicant changes for all diraction spots except for the graphene (10) spot.
The corresponding spectra are shown in Fig. 4.19. Signicant changes in the peak shape and
peak position can be seen and are highlighted. As shown next, the observed changes in the
LEED
√ √ intensities in Fig. 4.19 cover the graphene growth regime from the interface layer with
( 3× 3)R30◦ domains, the pure interface layer, monolayer graphene up to around bilayer or
trilayer graphene.
The calibration of these intensities is performed via ARPES, which allows for an unambiguous
assignment of the LEED intensities to the number of graphene layers, as shown in Fig. 4.20
for epitaxial graphene on 4H-SiC(0001). Thus, specic features in the intensity spectra, which
are again highlighted in the gure and which are generated by the complex multiple scattering
Figure 4.20: LEED spot intensity spectra for dierent numbers of epitaxial graphene layers grown (at the
indicated temperatures) on 4H-SiC(0001).
√ √ As indicated in the inset, the spectra were obtained for the
green marked spot of the (6 3×6 3)R30◦ reconstruction, which is the graphene (10) spot. The rst
order diraction spot for √the SiC√substrate is indicated in the LEED pattern in yellow, the position of the
(2/3,2/3) spot of the (6 3×6 3)R30◦ reconstruction in the inset. Yellow patches indicate ngerprint
like features in the spectra that allow the unambiguous determination of the number of graphene layers.
LEED patterns at 126 eV are shown on the right, also allowing for a discrimination.
processes in LEED, allow for an accurate determination of the number of graphene layers using
LEED alone. Note that the additionally upcoming peak at around 80 eV for bilayer graphene was
not visible in the uppermost curve of Fig. 4.19. This is due to the intensity scale in Fig. 4.19, which
was chosen in a way to highlight the changes for the growth of the rst monolayer. Although the
determination of the number of graphene layers by means of ARPES would already be a rather
convenient method, it cannot be applied during preparation, and a machine with the appropriate
precision might not normally be at hand. However, a LEED data acquisition system is available
in practically every preparation chamber. Not every LEED system may permit one to record
such intensity spectra, but already a LEED pattern at 126 eV, taken at normal incidence, allows
for an approximate determination of the number of layers by comparing √ the √relative intensity
of the graphene (10) spot to that of the surrounding spots in the (6 3×6 3)R30◦ pattern
as shown in Fig. 4.20. For the pure interface layer the graphene (10) spot is weaker than its
surrounding spots, for monolayer graphene they approximately display the same intensity, and
for bilayer graphene, the graphene spot is brighter than its surrounding spots. We note that
epitaxial graphene on 4H-SiC(0001) and 6H-SiC(0001) results in the same LEED spectra [185],
a nding that corroborates that - at least with the same preparation procedure - the polytype
has no inuence on the structural and electronic properties of epitaxial graphene on SiC.
To understand the origin of the changes in the LEED intensity spectra in more detail, a direct
structural√correlation
√ by a quantitative LEED analysis would be necessary. Due to the complexity
of the (6 3×6 3)R30◦ structure with 338 possible carbon atoms, such a structural analysis is
very demanding and was beyond the scope of this thesis. Such an analysis might not resolve
the complete surface structure but might at least give some additional insight into the structural
Figure 4.21: STM image of epitaxial graphene on 4H-SiC(0001) (panel (a)) showing terraces
√ with
√ dier-
ent corrugation (panel (b)) due to a dierent number of graphene layers on top of the (6 3×6 3)R30◦
reconstruction.
development of the surface before and during the graphitization process.

It should be noted that the probing area of LEED is approximately 1 mm2 , similar to ARPES
from UV excitation, CLPES and AES. Consequently, the number of layers is only an average
number of this macroscopic area. ARPES at a synchrotron light source allows a bit lower probing
areas of around 0.005 mm2 . Raman spectroscopy has a beam diameter of down to below µm
but it cannot be used in situ and the Raman data require a thourough interpretation as shown
in Section 4.3.6. For a thickness determination with spatial resolution on a micrometer scale,
LEEM is the most appropriate technique, which is shown in Section 4.3.5. To better understand
the unique properties of graphene on an atomic level, STM is the method of choice as shown in
the following section.
4.3.4 STM characterization of epitaxial graphene on SiC(0001)

Imaging the graphene unit cell
√ √
As outlined in the previous sections the graphene is growing on top of the (6 3×6 3)R30◦ re-
constructed interface layer, which remains intact as seen in LEED, CLPES and ARPES. Therefore,
it is expected that a graphene covered surface also displays a (6×6) or quasi-(6×6) honeycomb
structure as previously demonstrated for√ the interface
√ layer alone. It is a priori not clear, how and
to what extent, a distinction between (6 3×6 3)R30◦ surfaces covered with a dierent number
of layers can be made.
Figure 4.21 shows STM measurements after graphitization using a bias where the apparent
(6×6) corrugation can be observed√best.√Clearly, the development of large scale homogeneous
graphene layers on top of the (6 3×6 3)R30◦ structure seems to be prevented as will be
discussed in detail in Section 4.3.5. As illustrated in the STM image in panel (a), dierent
terraces display a dierent corrugation of the (6×6) honeycomb lattice [156]. This can be seen
more clearly in the linescans in panel (b). For reference, note that the vertical spacing of bulk SiC
bilayers amounts to 2.52 Å and that the layer spacing in (bulk) graphite is 3.35 Å. The regions
[AB], [DE] and [FG] have the same large corrugation, whereas the height of the corrugation
of the regions [BD] and [GH] amounts to approximately one half of the former value. These
eects
√ can √ be interpreted to be caused by a dierent number of graphene layers on top of the
(6 3×6 3)R30 surface reconstruction. A lower corrugation means a higher number of layers,
◦
Figure 4.22: Atomically resolved STM images of epitaxial graphene on 4H-SiC(0001) (Utip = 2.0 V)
showing (6×6) corrugations with two dierent contrasts for two dierent tip conditions √
shown√in panel
(a) and panel (b). Only for the tip condition in panel (b), graphene on top of the (6 3×6 3)R30◦
reconstruction can be resolved as can be seen from the zoom into the lower part of the STM image.
The annealing temperature was around 1300◦ C.
a higher corrugation a stronger inuence of the interface layer and thus a lower number of layers.
A quantitative interpretation seems to be dicult since it appears that electronic eects cover
topographical eects in the STM height measurements for a dierent number of graphene layers.
However, one scenario could be the interpretation of the higher corrugation in terms of monolayer
graphene and the lower corrugation in terms of bilayer graphene. It might be that at the point
B only the number of layers increases (although a substrate step cannot be excluded either). It
can at least be stated that a substrate step is located either at point B or at point D since a
direct linescan between point A and point E exhibits a topographic height change of 2.5 Å (not
shown). At the point D, one assumption could be that both the number of layers decreases
and one substrate step upwards could be detected, which would result in an overall decrease of
(3.35 - 2.5) Å = 0.85 Å. This value corresponds well to the measured dierence of around 0.6 Å.
The increase of the topographic height of 2.5 Å at the point G is again related to an increase of
the number of layers plus an optional substrate step.
In Fig. 4.22, we show that graphene layers can not only be seen indirectly by means of a
contrast change but they can also directly be identied [156]. The annealing temperature in that
case was around 1300◦ C. For two dierent tip conditions, we compare the images of the same
surface area that contrast dierently
√ √on the corresponding terraces. The rst tip condition allows
for atomic resolution of the (6 3×6 3)R30◦ reconstruction, the second one seems to allow only
for a reduced quality at a rst glance. However,
√ √ scaling down the scanning area leads to atomic
resolution of graphene on top of the (6 3×6 3)R30◦ surface reconstruction. Yet, this is only
the case for the region with the reduced contrast. The unit-cell size of about 2.5 Å corresponds to
that of graphite. Yet, since we observe only one of the two carbon atoms comprising the graphene
unit cell, we identify this surface region as (at least) bilayer graphene. The Bernal stacking of
two graphene sheets leads to the observation of such a diamond-shaped lattice [186]. The crucial
point that should be emphasized is that the visibility of the graphene layers strongly depends on
the actual tip condition, thus complicating the determination of the number of graphene layers by
means of STM. In the upper part of Fig. 4.22, the higher contrast in the STM image represents
a lower coverage with graphene, possibly monolayer graphene. As shown in the next paragraph,
the visibility of monolayer graphene depends on the bias value. However, also zerolayer graphene
cannot be excluded for the upper part of Fig. 4.22.
The visibility of monolayer epitaxial graphene in dependence of the applied bias is highlighted
Figure 4.23: Representative topographic images of the rst graphene layer grown on 4H-SiC(0001). The
images, taken on the same surface area (as can be recognized by the defect), show dierent contrast
as a function of the sample bias (I remains constant at 0.5 nA). The image size is 10 nm x 10 nm.
The sample bias is (a) -900 mV, (b) -300 mV, (c) 200 mV, and (d) 450 mV. The inset in panel (d)
is measured at 440 mV and identies monolayer graphene since all six carbon atoms in the graphene
honeycomb are resolved.
with a series of topographic STM images shown in Fig. 4.23. The sample temperature for
these images was around 6K. Although the samples have been transported through air from the
UHV system for sample preparation to the low temperature STM, the graphene quality has not
degraded. It was sucient to anneal the sample in the STM UHV-system up to 750 ◦ C in order to
desorb impurities. It is evident from Fig. 4.23 that the surface pattern changes with the applied
sample bias. The graphene honeycomb pattern can clearly be distinguished only at low energies
(panel (b), (c) and (d)), although modulated by the quasi-(6×6) periodicity of the interface
layer. At higher energies√(panel √ (a)), the graphene becomes transparent for the tip and the
typical features of the (6 3×6 3)R30◦ reconstruction can be observed through the graphene
layer. The inset in panel (d) is a high-resolution image of the graphene layer showing all six
atoms of the hexagonal cell. This again suggests that the rst graphene layer is not inuenced
by Bernal AB stacking with respect to the interface layer as claimed in Ref. [168] and discussed
in Section 4.3.2. The electronic dierence of carbon atoms in A and B position would allow to
image only three carbon atoms out of six [186, 187]. It should be noted that the identication
of the number of graphene layers by means of STM is a quite delicate task and should always
refer to more than one criterion. It has recently been shown that also for bilayer graphene all six
atoms of the graphene unit cell can be displayed under certain conditions [154]. Consequently,
the interface induced roughness and the bias dependence of the images should be considered,
too.
Surface defects inside the monolayer graphene lattice can clearly be distinguished from the
subsurface structure only at low sample bias. This can be seen from an inspection of the defect
located at the bottom
√ left
√ of the topographic images in Fig. 4.23. Such defects are known to
induce a periodic ( 3× 3)R30 pattern [178, 188], which is attributed to quantum interference
◦
generated by inter-valley scattering of the electrons. Inter-valley scattering means the elastic
0
scattering between the two inequivalent corners of the graphene Brillouin zone K and K . Such
scattering is allowed, as epitaxial graphene is intrinsically n-doped and therefore possesses states
in the Fermi surfaces (cf. Section 4.3.2), which are mainly probed by STM at low sample bias.
0 √
The length of the momentum vector between the two valleys √ K√and K ◦is 1/ 3 of the length
of the reciprocal graphene unit cell vector, which evokes a ( 3× 3)R30 superstructure in real
Figure 4.24: Topographic image of bilayer graphene showing the graphene lattice superimposed by the
quasi-(6×6) honeycomb lattice of√the √
interface layer. In the top right, a defect in the graphene structure
can be observed. It leads to a ( 3× 3)R30◦ reconstructed pattern from inter-valley scattering. The
sample bias in this image is 150 mV, the tunneling current amounts to 0.5 nA and the size of the
micrograph is 10 nm x 10 nm.
space.
Inter-valley scattering can also be observed in bilayer graphene, as can be seen √ in the
√ STM
image in Fig. 4.24, which was also taken at a sample temperature of 6K. The ( 3× 3)R30◦
superstructure can clearly be identied in the top right part of the micrograph. It is worth men-
tioning that epitaxial bilayer graphene mainly shows a graphene structure only weakly modulated
by the subsurface potential and is preserved at all applied bias values.
Intra-valley scattering, on the contrary, is only allowed for epitaxial bilayer graphene and not for
epitaxial monolayer graphene [177, 189]. Intra-valley scattering is elastic backscattering between
0
states at the Fermi surface of one K- (or K -) point only. The result is a long-range modulation
in very low bias (≈ mV) bilayer STM images of around 5 nm as reported by Rutter et al. [188]
and Brihuega et al. [177] (and not shown here). The absence of this intra-valley scattering in
monolayer graphene is a direct consequence of the (pseudo-)chirality of graphene's charge carriers
and related to the "chiral" quantum Hall eect, the Berry phase of π , Klein tunneling and the
strong suppression of weak localization in graphene (i.e. the absence of a magneto-resistance)
(cf. Section 2.1.1).
Scanning tunneling spectroscopy for mono- and bilayer epitaxial graphene

With the help of scanning tunneling spectroscopy (STS), the density of states (DOS) in graphene
can be examined on an atomic scale. In the dI /dV -spectra, the π -bands of graphene can be
accessed, integrated in k -space but with spatial resolution instead. Dierent features in the
dI /dV -spectra and their possible spatial dependence have to be interpreted with respect to
the electronic structure of graphene and of the interface layer. The UHV prepared graphene
sample was transferred ex situ to the low temperature STM/STS as mentioned earlier. The data
presented in the following have been published in Ref. [176].
An STM topographic image of a graphene monolayer is shown in Fig. 4.25 (a). As shown in
Figure 4.25: Scanning tunneling spectroscopy for monolayer graphene: (a) topographic region (image
size 5 nm x 5 nm, sample bias 100 mV, 0.5 nA, inset at 440 mV). The green hexagons indicate the
quasi-(6×6) periodicity. (b) Color scale plot of dI /dV -spectra obtained along the yellow line in (a). (c)
Selected dI /dV -spectra (red/green) obtained on the positions marked with a dot (red/green) in (a),
displayed with a vertical shift for better visualization. (d) dI /dV - spectrum at the Fermi energy. (e)-(g)
Topographic image and conductance maps (image size 100 nm x 50 nm) simultaneously measured at
-600 mV and -400 mV.
the previous section, it shows the graphene structure superimposed onto the periodicity of the
underlying buer layer. This quasi-(6×6) honeycomb shaped wall structure is indicated by the
green hexagons. In panel (c), dI /dV -spectra are shown in the respective color for the positions
marked with the red and green dot in panel (a). For monolayer graphene, a linear DOS with a
minimum at the Dirac point at around -0.4 V to -0.45 V sample bias is expected. The dI /dV -
spectra shown in panel (d) indeed do show two peaks at energies close to the Dirac point. The
median of these two peaks is located at approximately -0.46 eV and is close to the expected
Dirac point value. The maxima could be attributed to the onsets of the π -bands, which would
be separated by an energy gap of around 220 mV. Such an interpretation would be consistent
with ARPES data of Zhou et al. [168]. However, the gap cannot simply be due to an AB
asymmetry of the two atoms per unit cell caused by a Bernal-type stacking with respect to the
buer layer [176, 178, 187, 190] as all six atoms of the hexagonal honeycomb of graphene are
retrieved (inset in panel (a)). It could rather be due to the superperiodicity of the buer layer
as suggested by Mañes et al. [191]. It has to be considered, however, that - as discussed in
Section 4.3.2 - the presence of a gap in epitaxial monolayer graphene is questionable and has to
be treated with care. The interpretation of tunneling
√ spectra
√ suers from the lack of knowledge
about the inuence of the interface layer. The (6 3×6 3)R30◦ structure does indeed have a
weakly dispersing state exactly around the energy position of the Dirac point as shown previously
in Fig. 4.9 (a). Consequently, a denite statement about a possible gap cannot be drawn from
STS [154, 190].
To further emphasize the strong inuence of the buer layer, our STS measurements reveal
that the positions of the maxima discussed in the previous paragraph vary in energy according to
the periodicity of the interface structure. This modulation is visualized in Fig. 4.25 (b). It shows
in a color scale map the dI /dV -spectra measured along the yellow line marked in panel (a). The
positions of the maxima vary smoothly and uctuate with the quasi-(6×6) periodicity (green
hexagons) of the buer layer. The distance between the two peaks varies by about ±40 mV
between valley (maximum) and rim (minimum) positions. In close proximity to defects (blue
arrow in Fig. 4.25 (a)), the distance between the peaks is signicantly reduced.
The dI /dV -spectrum at the Fermi energy on the monolayer is shown in panel (d). The
Figure 4.26: Scanning tunneling spectroscopy for bilayer graphene: (a) topographic region (image size
5 nm x 3.5 nm, sample bias -400 mV, 0.5 nA). (b) Color scale plot of dI /dV -spectra obtained along the
yellow line in (a). The dashed lines indicate the positions (red) of the π -band onsets and the mid-gap
state (black). Horizontal lines correlate the three high symmetry positions in the graphene unit cell with
typical spectra displayed with a vertical shift in (c). The inset shows a sketch of the graphene bilayer
stacking. (d)-(f) Topographic image and conductance maps simultaneously measured at -280 mV and
-400 mV, respectively (image size 5 nm x 2.5 nm).
line shape shows a well pronounced local minimum at zero bias, which is also reported in the
literature for monolayer up to at least four layer graphene [154, 157, 176, 190]. It has been stated
both that the dI /dV -curve stays nite [154, 188] at the Fermi level and that it drops down to
zero at the Fermi level [190]. The origin of this gap-like feature has recently been attributed to
stem from the opening of a new electron tunneling channel due to inelastic coupling to surface
phonons [190, 192].
In order to get more information about the origin of the peaks in the dI /dV -curves shown in
Fig. 4.25 (b) and (c), energy resolved conductance maps were obtained by recording the dI /dV -
signal at the corresponding bias values of -600 mV and -400 mV. Together with a topographic
map (panel (e)), they are shown in panel (f) and (g) of Fig. 4.25. The impact of the buer layer
structure on the graphene monolayer is evident. Note that defects within the graphene lattice
together with the corresponding quantum interference patterns [188] cannot be detected here. At
-400 mV a modulating potential with a quasi-(6×6) periodicity is superimposed onto the structure
of the graphene monolayer (panel (g)). At -600 mV dark dots (points of lower conductance) can
be seen (panel (f)). The absence of electronic scattering around them excludes their origin to be
from adsorbed impurities on the graphene layer. At the rst glance, these dots do not seem to
form an ordered pattern. However, by superimposing the topographic image onto the conductance
map the dots can be correlated with the bright atom-like features in the quasi-(6×6) honeycomb
pattern of the buer layer. This suggests that dierent atoms of the graphene monolayer are
electronically distinct due to the structure of the buer layer, which does not come as a surprise
when considering the sketch of the interface layer shown previously in Fig. 4.10 of Section 4.2.2.
For bilayer graphene, the topographic image in Fig. 4.26 (a) contains much less contributions
from the interface as already shown in Section 4.3.4. Note that defects within the graphene lattice
and the corresponding quantum interference patterns [188] cannot be detected here either. The
local DOS was evaluated along the yellow line. The dI /dV -map (panel (b)) and the averaged
spectra (panel (c), see below) can be interpreted in terms of a gap around the Dirac point in
agreement with ARPES results shown in Section 4.3.2. The gap is centered around 300 mV
below the Fermi level and its size of about 180 mV is only slightly larger than the value from
ARPES (120 meV). As indicated by the dashed red lines, the peaks do not vary in energy along
the measured line. Instead, the π -band varies substantially in intensity with the atomic graphene
periodicity. Also a localized mid-gap state appears in strict correlation with the atom position in
the graphene unit cell. For the three dierent high symmetry positions in the unit cell, averaged
spectra are shown in panel (c). At the position of the B-type atoms, which corresponds to the
bright spots in the topographic image [187], only the π -band onsets are seen. On one of the dark
areas aside of a B-type atom, these states are attenuated while on the other side an additional
peak at -280 mV appears. This mid-gap state is localized on one of the hollow site positions, as
can be seen also in the conductance map (panel (e)) achieved simultaneously to the topographic
image (panel (d)). Since a mid-gap state is, in at free-standing graphene, not predicted nor
experimentally observed, we can tentatively assign it to a tip induced state. Similar to graphite,
the inuence of the tip on graphene multilayers can distort the interlayer distance [193, 194]. As
a curvature of the graphene layer is theoretically predicted to induce the formation of a mid-gap
state [195], we speculate that the tip-induced layer distortion is the origin of the observed mid-
gap state [193, 194]. The tip interaction is predicted to be stronger on the H-type atom at our
tip-sample distance, which can be estimated to be 5 Å. Accordingly, we assigned the observed
mid-gap state to the H-type position (blue curve in panel (c)). It should be noted that also the
conductance maps show practically no modulation with the periodicity of the buer layer (panels
(e) and (f)).
The observation of the gap in the STS spectra of bilayer graphene is in accordance with
the STS analysis in Ref. [154]. In contrast to our results, which are published in Ref. [176],
Ref. [154] does not report any spatial dependence of the STS spectra within one unit cell. As
already explained, the actual tip condition plays a major role and could cause these dierent
observations. Furthermore, Ref. [154] reports on a spatial modulation of the energy gap within
the quasi-(6×6) unit cell. It was concluded that the energy gap varies by about 25 meV depending
on the position within the quasi-(6×6) unit cell. This could be due to an inhomogeneous charge
distribution in the rst graphene layer [154] as the band gap in bilayer graphene arises from
dierent charge carrier concentrations in the rst and second graphene layer (cf. Section 4.3.2).
Such an observation could also be dierent for dierent samples or dierent sample preparations.
4.3.5 Homogeneity of the graphene lms

The typical terrace size in STM images is around 100 nm x 100 nm, which is suitable for a
dedicated STM analysis but is of course far away from a large scale production of epitaxial
graphene. To get more insight into the morphology of UHV prepared epitaxial graphene samples,
we show large scale STM and AFM measurements √ as √well as◦ LEEM data in the following.
Fig. 4.27 displays an AFM image of the (6 3×6 3)R30 reconstruction before the graphiti-
zation process (panel (a)) and a large area STM image after graphitization (panel (b)) [156]. The
typical
√ length
√ scale of the surface morphology appears to be determined already in the state of the
(6 3×6 3)R30 reconstruction. The AFM and STM images show a lot of small terraces where
◦
the growth conditions seem to be dierent so that, for instance, a dierent number of graphene
layers may develop or linear defects of the graphene layer are present as a consequence of sub-
strate steps. UHV prepared samples typically do not produce substantially larger homogeneous
graphene surfaces (compare STM image in Ref. [12, 156] and AFM-images in Ref. [158, 181]). It
should be noted that after the initial hydrogen etch process the surface exhibits regular terraces
with straight step edges of typically micrometer distance (cf. Fig. 4.1 (a) in Section 4.1).
STM and AFM do only show the morphology of the sample surface but not of the graphene
√ √
Figure 4.27: Panel (a) shows a typical AFM image of the (6 3×6 3)R30◦ reconstruction before the
graphitization process, panel (b) a large area STM image after graphitization. The maximum terrace
size amounts to a few hundred of nm2 .
Figure 4.28: Bright eld LEEM micrographs at 2.9 eV electron energy (eld of view 10µm and zoom
to 1µm) for a 4H-SiC(0001) sample with a mixture of mono- and bilayer graphene after (a) prolonged
annealing to 1220 ◦ C and (b) a short heating step to 1275 ◦ C. Bilayer areas (bright patches) are grown
in small seeds after the ash.
Figure 4.29: Schematic representation of the physical origin of (a) the G band, (b) the D band and (c)
the 2D (or G') band in Raman spectroscopy of graphene (for details see text). The images display the
scattering processes in the electronic band structure scheme and are adapted from Ref. [123].
layers. Since it has been shown in STM images that the graphene is growing over the steps [158],
it could in principle be that the graphene domain size is substantially larger than the typical
terrace size. Furthermore, ARPES data of monolayer graphene suggest a large area production of
graphene samples with a uniform monolayer coverage over the whole measurement area of 1 mm2 .
To clarify these issues, LEEM is the method of choice since it allows to directly image the spatial
distribution of the number of graphene layers on the micrometer scale. In the low energy regime,
the maxima and minima of the (0,0) diraction spot of graphene clearly indicate the number
of graphene layers [111, 163, 180], which can be understood from simple kinematic diraction
theory [111]. Using the appropriate electron energy, the homogeneity of epitaxial graphene can
thus easily be displayed in bright eld LEEM. Figure 4.28 (a) shows a LEEM micrograph (2.9 eV)
for a sample imaged in the transition regime between mono- and bilayer graphene obtained after
heating at 1220 ◦ C for 10 minutes [163]. Monolayer areas appear dark, bilayer regions bright
as identied in panel (c) from the corresponding intensity versus energy spectra of the reected
electrons. Panel (b) demonstrates the inhomogeneous, micro-seeded growth of the bilayer regions
after a short heating step to 1275 ◦ C for 10 seconds. We explicitly show exactly the same surface
area before and after heating directly in front of the LEEM optics. It has to be emphasized that
further heating does not lead to a fully closed graphene bilayer. Instead, the third layer starts to
grow. It therefore has to be concluded that the morphology of graphene is comparable to the
one of the sample surface [163].
LEEM turns out to be ideal for analyzing the homogeneity of the graphene growth on the
micrometer scale. It is the method of choice to control the homogeneity of modied growth
techniques for epitaxial graphene. This has been shown for atmospheric pressure graphitization
(Ref. [10] and Section 5.2.2 and 6.2.1), for instance, which preserves the initial morphology of
hydrogen etched SiC surfaces and yields graphene terraces of several tens of micrometer in length
and a few micrometer in width.
4.3.6 Raman Spectroscopy

Before discussing the results of Raman spectroscopy in detail, it is instructive to have a closer look
at the physical background of the Raman features in graphene. Raman spectroscopy is known to
be a powerful tool to determine the electronic properties of carbon-based materials, and there have
been several reports about Raman measurements on graphene layers that were micromechanically
exfoliated from highly oriented pyrolytic graphite (HOPG) on SiO2 substrates [123, 196200]. As
the unit cell of graphene contains two carbon atoms, there are six phonon dispersion bands,
three of which are acoustic branches and the other three optic phonon branches [123]. However,
only a few phonon modes are Raman active and are involved in the creation of the features in
Raman spectroscopy. The most prominent peaks in the Raman spectra are the so-called G band
at around 1580 cm−1 , the 2D band at around 2700 cm−1 using a laser wavelength of 488 nm
and, in the case of disorder, the D band positioned at about half of the frequency of the 2D
band (around 1350 cm−1 using a laser wavelength of 488 nm) [123]. The physical origin of
these bands is shown in Fig. 4.29, which displays the scattering processes in the electronic band
structure scheme. The starting point is the excitation of an electron by a photon of energy Elaser .
Electrons with appropriate binding energy are only available in the π -bands around the K- (or
0
K -) point. After excitation, the electron is inelastically scattered by a phonon of energy Ephonon .
In case of the G band, a phonon, which is located at the Γ-point of the phonon dispersion, is
scattering with the electron so that the electron momentum remains unchanged (Fig. 4.29 (a)).
Afterwards, the electron-hole pair is annihilated accompanied by the emission of a photon of
energy Elaser − Ephonon . In fact, the G-band in graphene is the only band arising from such a
normal rst order Raman scattering process. The type of vibration of the graphene lattice is
dened by the phonon dispersion. The G band is actually due to a doubly degenerate phonon
mode and represents an in-plane vibration of the graphene honeycomb lattice as displayed in
Fig. 4.29 (a). The 2D and D bands are stemming from a second-order process, involving two
graphene phonons near the K-point for the 2D band or one graphene phonon and one defect
in the case of the D band. The notation 2D is actually misleading as no defect is involved.
Therefore, the 2D band is sometimes called G' band [123]. Note that both the D and the 2D
band frequency are dependent on the laser energy. The frequencies shift upwards with increasing
laser energy, however with dierent slopes for D and 2D band. As shown in Fig. 4.29 (b) and
(c), the double-resonance process in case of the D and 2D band begins with an electron of wave-
vector ~k around the K-point, which is excited by a photon of energy Elaser . This electron is then
scattered by a phonon of wavevector ~q and energy Ephonon or by a defect of wavevector ~q to a
0
point belonging to a circle around the K -point. It now has the wavevector ~k + ~q . This type of
scattering is intervalley scattering. In the next step, the electron is scattered back to a ~k -state
and emits a photon by recombining with a hole at a ~k -state. In the case of the D band, the two
scattering events consist of one elastic scattering process by defects and one inelastic scattering
event by emitting or absorbing a phonon. For the 2D band, both processes are inelastic scattering
events and two phonons are involved. In principle, more Raman features than the three bands
discussed here can be observed. However, in most cases, their intensity is very small.
Ferrari et al. have demonstrated that the shape of the 2D Raman peak may serve as the
ngerprint to distinguish mono-, bi-, and few-layer graphene [197]. For monolayer graphene, the
2D peak can be tted to a single Lorentzian, whereas the multiple bands in few-layer graphene
require tting to more Lorentzians. For bilayer graphene in particular, the two π -bands allow for
four dierent double resonance scattering processes. Although Raman data of few-layer epitaxial
graphene were recently reported [182, 183, 201], a similar, clear procedure to dierentiate between
single layer, bilayer, and multilayer graphene on SiC has been lacking so far. In the following,
a possible Raman ngerprint to distinguish the layer thickness is shown by correlating Raman
spectroscopy experiments with ARPES data on mono-, bi- and multilayer epitaxial graphene.
In addition, Raman data of epitaxial graphene akes transferred onto SiO2 are shown in order
to investigate the inuence of the substrate material or buer layer on the Raman spectrum.
The Raman data presented in the next two subsections have been published in Ref. [184]. Note
that the ARPES and LEED calibration of the number of graphene layers has been revised in the
meantime and is slightly dierent from that one in the publication. However, the main conclusions
of the Raman data analysis remain unchanged.
Raman spectroscopy for epitaxial graphene on SiC(0001)

For a series of SiC(0001) samples, the number of layers was determined in situ by ARPES
and/or LEED (cf. Sections 4.3.2 and 4.3.3 as well as Ref. [163]) before they were moved to
the Raman apparatus. For an additional series of samples, the graphene layer thicknesses were
determined from an optical technique. This series includes graphene samples on both SiC(0001)
and SiC(0001̄). Raman spectra were measured under ambient conditions using an Ar+ laser with
a wavelength of 488 nm and a spot size of around one micrometer. Because the Raman signal of
graphene is much smaller than that of SiC, as will be discussed later, it is necessary to use high
power levels and long acquisition times. It was veried that no signicant spectral changes were
induced by the extensive laser exposure during Raman scans with long acquisition times.
The key Raman results are displayed in Fig. 4.30. Spectra are plotted for a monolayer (1 ML),
a 1.5 ML, and a bilayer (2 ML) graphene SiC(0001)-sample, together with reference spectra
measured on clean, hydrogen etched 4H-SiC and 6H-SiC substrates as well as on bulk HOPG and
mechanically exfoliated monolayer and bilayer graphene from HOPG, which were transferred onto
an oxidized Si substrate. The well-isolated G and 2D peaks for HOPG and exfoliated graphene
layers on SiO2 near 1580 and 2700 cm−1 can clearly be identied. The defect D peak, which
should be near 1350 cm−1 , is absent for graphene layers exfoliated from HOPG, attesting to the
high crystalline quality of graphene layers exfoliated from HOPG. Note also that the 2D peaks of
the monolayer and bilayer appear dierent. For the bilayer, a shoulder structure is clearly visible.
For the epitaxial graphene layers on SiC(0001), it is apparently dicult to distinguish the G and
D peaks in the spectra because their intensities are much smaller than those of the surrounding
peaks from the SiC substrate. This can be concluded from the reference spectra, which were
recorded on pure SiC samples immediately after hydrogen etching. Only the 2D peaks are well
isolated around ≈ 2700-2750 cm−1 and are out of the range of the SiC peaks. The 2D peak is
slightly blue-shifted in comparison to graphene exfoliated from HOPG. The 2D double resonance
process is sensitive to strain and the mentioned blue-shift has been attributed to compressive
strain [183], building up during the cool-down procedure after the graphene preparation since
graphene expands with decreasing temperature unlike any other material. A further discussion of
the 2D peak position is following in the next subsection.
Higher resolution measurements on epitaxial graphene as well as on the SiC reference samples
using a grating with more lines are depicted in Fig. 4.31. In these spectra, the D and G peaks
can be discerned more easily. Boxes mark the approximate location of the D and G features. But
even now it is clearly dicult to obtain quantitative information about the G and D peaks in order
to identify the layer thickness from the raw data as plotted in Fig. 4.31 (a) and (b). Figure 4.31
(c) shows dierential spectra obtained by subtracting the Raman spectrum of the SiC reference
sample from the spectra of the graphitized samples. This procedure has recently been proposed
by our group [184] and in Ref. [183]. Now, a small D peak is visible near ≈ 1360 cm−1 . Also the
G peak, located in between the SiC peaks at ≈ 1500 and ≈ 1620 cm−1 , is brought out in these
dierential spectra. The G peak position (P(G)) does not vary with the number of layers (P(G)
≈ 1591 cm−1 ). However, it is blue-shifted as compared to micromechanically cleaved graphene
(P(G) ≈ 1587 cm−1 ) and graphite (see Fig. 4.30). A possible explanation in terms of charge
carrier concentration is given later in this section. The full width at half-maximum (FWHM) of
the G peak (FWHM(G)) is also not inuenced by the number of layers.
Figure 4.30: Raman spectra of HOPG, mono-, and bilayer graphene exfoliated from HOPG and trans-
ferred onto SiO2 , mono- up to bilayer graphene grown on SiC (monolayer (ML) and 1.5 ML: 6H-
SiC(0001); 2 ML: 4H-SiC(0001)) and reference data on clean 4H-SiC and 6H-SiC substrate pieces after
hydrogen etching. All data were obtained using a wavelength of 488 nm. The scale for the data of the
graphene samples on SiC and clean SiC samples are normalized with respect to the highest SiC peak at
≈ 1510 cm−1 . The data of HOPG and the graphene akes from HOPG are scaled down arbitrarily to t
the image.
Figure 4.31: (a),(b) Raman spectra around the D and G peaks of (a) 1 ML (black) and 1.5 ML graphene
(red) together with the 6H-SiC(0001) reference (green) and of (b) 2 ML graphene (blue) with the 4H-
SiC(0001) reference (green). The dotted boxes mark the region of the D and G peaks, which are partly
masked by Raman contributions from the SiC substrate. (c) Dierential spectra between the Raman data
recorded on the graphene samples and on the clean, hydrogen etched SiC reference pieces. (d) Raman
spectra (raw data) near the 2D peak for epitaxial graphene samples with 1, 1.5, and 2 layers. The data
are normalized to the peak height. (e),(f) Curve tting (yellow lines) of the spectra around the 2D peak,
for 1 ML graphene (lled circles in (e)) with a single Lorentzian and for 2 ML graphene (lled circles in
(f)) with four Lorentzian curves. The lower panel shows the 2 ML spectrum (lled circles) and a best
t to the data using a single Lorentzian curve.
Figure 4.32: The ratio between the intensity of the strongest SiC substrate Raman peak obtained on
graphitized samples and on a hydrogen etched reference sample as a function of the graphene layer
thickness. Experimental data points are for 1 ML, 1.5 ML and 2 ML samples (mainly discussed in the
manuscript) and a 2.3 ML thick graphene sample. The error bars mark the maximum and minimum
values from a set of 4 or 5 measurements taken at dierent positions on the sample. The red curve
corresponds to an exponential attenuation t.
In order to nd signatures in the Raman spectra for a determination of the number of layers,
let us now have a closer inspection of the 2D peak. As mentioned, the number of layers for
graphene samples exfoliated from HOPG can be successfully assigned by analyzing the shape of
the 2D peak [197]. Exfoliated monolayer graphene for instance exhibits a single and sharp 2D
peak (FWHM < 30 cm−1 ). For bilayer and trilayer graphene, the 2D feature becomes broader
(50 cm−1 for the bilayer in Fig. 4.30 for instance), and asymmetric and shoulder structures appear.
In the following, it is demonstrated that in the case of epitaxial graphene it is not straightforward
as with graphene akes exfoliated from HOPG to determine the number of layers by simple visual
inspection of the shape of the 2D peak in the raw data. Afterwards, a ngerprint feature will
be isolated that allows to facilitate the identication of the number of epitaxial graphene layers
by means of Raman spectroscopy. For our epitaxially grown graphene, the 2D spectra for 1 ML,
1.5 ML and 2 ML graphene are shown in Figure 4.31 (d). The peak position, P(2D), increases
monotonously from 2721 to 2760 cm−1 with increasing layer number. The peak widths FWHM
(2D) are substantially larger than that for exfoliated layers and rise from ≈ 46 to 64 and 74 cm−1
when comparing a monolayer sample with 1.5 ML and 2 ML graphene, respectively. Despite
the broadening, the 2D peak of monolayer epitaxial graphene can still be tted with a single
Lorentzian (Figure 4.31 (e)). In bilayer graphene grown on SiC, however, a visual inspection of
the 2D peak in Fig. 4.31 (d) shows no clear asymmetry nor a shoulder structure. Even if so,
the peak can only be poorly tted by a single Lorentzian as demonstrated in Fig. 4.31 (f). As
in the case of exfoliated bilayers, a decomposition into four Lorentzians is necessary to obtain
better agreement between the t and the experimental data. Even for up to 8 stacked layers of
graphene, the shape of the 2D peak does not develop a shoulder structure but rather resembles
closely a single peak in contrast to exfoliated layers from HOPG. Our observation [184] is in
agreement with other Raman data for epitaxial graphene reported at the same time [182, 201].
In order to nd Raman ngerprints for the determination of the number of layers, let us have
a closer look at the 2D peak taking a larger database into account. Besides ARPES and LEED
intensities [163], we estimate the thickness of epitaxial lms using a simple model based on Beer's
Figure 4.33: P(2D) and FWHM(2D) of graphene grown on SiC samples plotted against the number of
layers. The dierent symbols refer to the dierent methods used to identify the layer thickness. Stars
represent data obtained on 1 ML, 1.5 ML, and 2 ML graphene, which were grown on SiC(0001) and
whose layer thickness was determined via in-situ ARPES. Filled diamonds are the data, where the number
of layers was assigned by LEED measurements. The number of layers of the other samples (lled squares
and open circles) was estimated by calculating the attenuation of the laser light. The open circles mark
samples grown on SiC(0001̄). All other samples (lled symbols) were fabricated on SiC(0001). The color
of the data points refers to dierent polytypes of SiC: black for 6H-SiC and red for 4H-SiC. The dashed
line in the lower panel is a least-squares linear t (as a function of the inverse thickness) through the
data of graphene samples grown on the Si-face of SiC (stars, squares, and diamonds).
Law [182, 202]. The model assumes that light passing through a lm is attenuated exponentially
according to exp(−t/λ), where t is the lm thickness and λ the penetration depth of the laser
light. The SiC peak at ≈ 1510 cm−1 (see Fig. 4.31), which has the highest intensity, was used for
these purposes. In Fig. 4.32 we plot this peak intensity as a function of the number of graphene
layers for graphene samples, whose layer thicknesses were predetermined from ARPES. These
intensities were normalized to the intensity of this peak recorded on the bare SiC substrate, IRef .
The data points are tted to ISample /IRef = exp(−2t0 N/λ), where t0 is the stacking distance of
the graphene layers and N the number of layers. The factor of 2 is added to take into account
the reection geometry. The tting parameter amounts to 2t0 /λ ≈ 0.22, and thus λ ≈ 9.3t0 .
From the graphite stacking distance t0 = 0.335 nm, we deduce that λ is about 3.1 nm. This
t serves to determine the number of layers on samples, on which no LEED or ARPES data
were available. For thick graphene (more than four layers), the estimated thickness will exhibit
a larger error since we extrapolate from data recorded on few layer graphene only. Applying
this procedure to the thickest graphene samples used in our studies, we estimated layer numbers
between 6 and 8 (the uppermost three red squares in the lower panel of Fig. 4.33). CLPES was
carried out on these thicker layers as an additional check. The emission intensity of the core level
also follows a similar attenuation model (cf. Section 3.1.3), and the CLPES analysis led to an
estimated 6 - 9 layers.
The P(2D) and FWHM(2D) values for a large number of samples are plotted against the
number of layers in Fig. 4.33. The stars represent the 1 ML, 1.5 ML, and 2 ML data discussed
above. Note that the monolayer data has been measured on a not fully developed graphene layer
(around 0.8 ML), a fact that is amplied by the choice of the inverse number of layers as unit
of the x -axis. The number of layers for the remaining data points is determined either using
LEED intensities [163] or an optical attenuation model based on Beer's Law [182, 202]. The
tendency that P(2D) increases as the number of layers increases is not unambiguous. There is a
relatively large spread of P(2D) even for samples with the same number of layers. Hence, P(2D)
is not suitable to precisely identify the number of graphene layers a sample has. However, for
graphene on SiC(0001), the FWHM(2D) value exhibits a clear (negative) linear relationship with
the inverse number of layers: FWHM(2D)= (-41(1/N) + 92) [cm−1 ]. Note that for samples
prepared on SiC(0001̄), the FWHM(2D) does not vary with layer thickness. This suggests that
the electronic properties of graphene layers prepared on SiC(0001̄) are quite dierent from those
of samples fabricated on SiC(0001) (cf. Section 5.1.1 and Ref. [4, 153, 203]).
In the past paragraphs, it became obvious that the interpretation of Raman spectra of epitaxial
graphene on SiC turns out to be more challenging in comparison to graphene akes exfoliated
from HOPG. Before moving on, a brief recapitulation of the existing results may be helpful. For
the isolation of the D and G peak of epitaxial graphene, reference spectra of clean SiC samples
have to be subtracted in order to eliminate the Raman peaks of the SiC substrate. The disorder
related D peak is small and accounts for the high crystalline quality of the graphene. However,
the D peak is even absent for graphene exfoliated from HOPG. A slight blue-shift with respect to
graphene exfoliated from HOPG can be detected for the G peak and the 2D peak, which both will
be discussed in more detail in the next subsection. The 2D peak is much broader compared to
graphene exfoliated from HOPG and shows the peak shape of a single peak even for multilayers.
With some eorts, it is possible to analyze the peak shape (single peak or four peak contributions)
to distinguish monolayer from bilayer graphene, provided a high resolution spectrum is recorded.
However, it is the width of the 2D peak that stands out as a suitable ngerprint to assign the
number of layers in the case of graphene on SiC(0001).
Figure 4.34: Optical micrographs of graphene transferred from SiC onto Si-substrates with a thermally
grown SiO2 -layer. The contrast of the image was enhanced to improve the visibility of the akes. The
red arrows indicate the visible blueish objects, on which further Raman measurements were carried out to
identify them as graphene or glue residues. The yellow objects are Au markers deposited by conventional
e-beam lithography and thermal evaporation. They help in relocating akes. The spacing between the
markers is 80 µm.
Raman spectroscopy for epitaxial graphene transferred to SiO2

We now return to the particularities in the G peak and 2D peak position (blue-shift) as well as the
width and shape of the 2D peak for graphene on SiC as compared to graphene exfoliated from
HOPG. The following issues may be relevant: epitaxial graphene layers may have dierent
√ √ defects
and curvature and the as grown graphene layers likely interact strongly with the (6 3×6 3)R30◦
buer layer reconstruction. To shed some light on the role of this buer layer for the Raman
features, we compare, for the rst time, Raman data of epitaxial graphene on the SiC substrate
and epitaxial graphene, which was transferred from SiC onto a Si substrate with a 300 nm thick
SiO2 top layer [184].
For the transfer process, the conventional adhesive tape technique was used. As starting
material, we used a bilayer sample and a sample with a thickness varying from one to three
layers. Both samples were grown on a 4H-SiC(0001) substrate. The yield and the properties
of the akes did not depend on the specic starting material. After attaching several times a
small piece of adhesive tape onto the graphitized substrate, graphene akes were transferred by
gently pressing the tape onto the oxidized Si-substrate and subsequently rubbing it with a soft
tip of plastic tweezers. The adhesive tape, that is commonly used for dicing wafers as well as
mechanical exfoliation of graphene akes from HOPG, leaves little residue. However, it turned
out that it does not oer sucient adhesion for our epitaxial graphene samples and hence a tape
with a larger glueing force was needed. In most cases, the transferred graphene akes were around
1 µm2 or less in size, and thus hardly visible by eye using an optical microscope. Even if they were
visible, it was hard to distinguish the graphene akes from glue residues. For example, Fig. 4.34
shows optical micrographs made on an oxidized Si-substrate, on which epitaxial graphene akes
were transferred. The substrate was cleaned with acetone, isopropanol and deionized water before
optical inspection. Four blueish objects are highlighted with red boxes and red arrows. However,
only two of them were identied as graphene akes based on Raman measurements.
Figure 4.35: (a-c) AFM images of representative graphene akes transferred from SiC onto SiO2 . All
scale bars are 500 nm. (d) A representative Raman spectrum recorded on a monolayer graphene ake.
(e) Raman spectra around the 2D peak for transferred mono-, bi-, and multilayer akes. (f) P(2D) as
a function of I(G)/I(2D) of the graphene akes. Black circles correspond to monolayer samples, red
squares to bilayer, and blue triangles to multilayer samples. The inset depicts a histogram of I(D)/I(G)
of the graphene akes. (g) FWHM(2D) and FWHM(G) as a function of I(G)/I(2D). The FWHM(2D)
increases but the FWHM(G) decreases slightly with the number of layers. The dashed lines are guides
to the eye.
It's due to the initial morphology of the graphene layers on SiC [111, 156, 163, 180] that
the transferred graphene akes on SiO2 were only about 1 µm2 or less in size. With the help
of atomic force microscopy (AFM) two dierent types of graphene akes can be distinguished
(Figure 4.35 (a)-(c)). Some akes show a at surface over their entire area similar to normal
graphene akes exfoliated from HOPG (Figure 4.35 (a)), but more often akes are not at as
illustrated in Fig. 4.35 (b) and (c). This suggests that, even after the transfer onto SiO2 , the
akes might maintain some of their initial morphology imposed by the ne terrace structure of
the SiC substrate formed after the annealing process (cf. Section 4.3.5). Note that the graphene
is known to grow over the SiC substrate steps in a carpet-like manner [158]. Although the
akes were not at, the Raman data for dierent points on each ake were nearly the same.
Figure 4.35 (d) depicts a representative Raman spectrum obtained on a graphene monolayer
exfoliated from SiC onto SiO2 . Raman data around the 2D peak are displayed in Fig. 4.35 (e) for
a monolayer, bilayer, and a multilayer (more than 2) sample. Finally, the position of the 2D peak
as well as the FWHM of the G and 2D features are shown in parts (f) and (g) of Figure 4.35
for a large collection of akes exfoliated from epitaxial graphene. The shape of the 2D peak
changes upon the transfer of the akes and allows to determine the layer thickness for each ake
by analyzing the shape of the 2D peak together with the ratio of the G and 2D peak strength,
I(G)/I(2D) [198]. Accordingly, the akes are classied into three groups: monolayer (black),
bilayer (red) and multilayer (more than 2, blue) [197]. For all akes, the D peak intensity is
small as shown in the inset of Fig. 4.35 (f) (around half of the akes have an I(D)/I(G) ratio of
less than 0.05), which indicates high crystalline quality of the initial epitaxial graphene on SiC
because the disorder would certainly not be reduced after the graphene akes are transferred.
After transferring epitaxial graphene onto SiO2 , the G peak is red-shifted back to ≈ 1587 ±
2 cm−1 , which is the position for exfoliated graphene from HOPG. Also very similar, the exact
location of the G peak does not have any correlation to the number of layers (not shown) and
also its peak width is more or less constant (Figure 4.35 (g)). Hence, the G peak blue-shift for
graphene on SiC can be attributed to the presence of the SiC substrate and buer layer. A possible
cause for this blue-shift may be the already mentioned compressive strain, which builds up during
the cool down procedure [183] or charge doping from the substrate [199, 204207]. Assuming
that charge neutrality exists at ≈ 1583 cm−1 (the minimum value of the measured P(G) for the
akes), the G peak blue-shift of ≈ 8 cm−1 from 1583 to 1591 cm−1 (the value for graphene on
SiC) corresponds to n ≈ 5 · 1012 cm−2 according to the literature [205, 207]. This carrier density
is consistent with the carrier density estimated from ARPES data for epitaxial graphene samples
exposed to air (n ≈ 4.2 · 1012 cm−2 , see Fig. 4.15 of Section 4.3.2). This good agreement between
the Raman and ARPES data conrms our interpretation of the G peak in terms of charge carrier
concentration. Additional evidence is given in Section 6.1.2 and Ref. [51] for epitaxial graphene
samples with dierent carrier concentration levels due to a dierent amount of adsorbed p-type
dopants. Furthermore, there is no signicant red-shift for the 2D peak after the graphene has
been transferred to SiO2 (cf. Fig. 4.35 (f)). This suggests that the SiC substrate does not cause
phonon stiening, which would inuence the 2D peak. Strain related eects should therefore not
inuence the shift of the G peak position. However, it is rather astonishing that the 2D peak
position does not change after transferring the graphene to SiO2 . It has to be concluded that
either the possible compressive strain remains within the graphene layer or the position of the 2D
peak is inuenced by other factors. In view of the large variation of P(2D), a denite conclusion
can not yet be drawn.
After the discussion of the positions of the G and the 2D peak, let us now come to the 2D
peak shape. It has been reported that the 2D peak for epitaxial graphene on both SiC(0001)
(see earlier in this subsection and Ref. [182, 184]) and SiC(0001̄) [182, 201] does not exhibit a
clear asymmetry or shoulder structure. For SiC(0001̄), this was attributed to the "turbostratic"
stacking, which decouples the layers electronically [4, 170, 201]. We have grown up to 8 graphene
layers on our SiC(0001) samples and still obtained a 2D Raman line with the shape of a single
peak. Such thick layers are most likely AB stacked. Upon transferring to the SiO2 substrate, the
2D peak becomes more asymmetric as seen in Fig. 4.35 (e). The peak now largely resembles
that of exfoliated multilayer graphene from HOPG. Hence, the anomalous shape of the 2D
peak of epitaxial graphene on SiC cannot be assigned solely to the stacking sequence, at least for
graphene on the Si-terminated surface of SiC(0001), because the stacking should not √ change
√ after
being transferred onto SiO2 . Instead, presumably either the strain from the (6 3×6 3)R30◦
reconstruction or charge doping aect the double resonance process.
Although the shape of the 2D peak changes after the graphene akes are transferred, the
peak width does not change, as shown in Fig. 4.35 (g). The FWHM(2D) value for the monolayer
graphene akes is ≈ 46 ± 4 cm−1 , which is similar to the value of graphene on SiC. Thus,
apparently the broadening of the 2D peak of graphene on SiC does not originate from a substrate
eect but is due to the graphene structure itself. One possible explanation for the broadening of
the 2D peak is a nonuniform number of layers: if a monolayer graphene sample on SiC consisted
of a small fraction of bi- or multilayer graphene, the 2D peak would be broadened [200], whereas
the eect would be smeared out in the ARPES measurement on the large scale of ≈ 1 mm2 .
Indeed, recent progress in the production of epitaxial graphene by means of atmospheric pressure
graphitization yields a higher degree of homogeneity accompanied by a sharper 2D peak with a
FWHM of 37 cm−1 [10]. Another factor to be considered in the discussion of the 2D peak width is
Figure 4.36: (a) 5×5 µm2 Rayleigh image of a monolayer of graphene grown on SiC. Regions with yellow
and black colors indicate a high and a low optical reection signal, respectively. The dierence between
the signals for yellow and black regions is approximately 4 %. The scale bar is 2 µm. The bottom is a
sketch of the graphene morphology. It follows the terrace structure of the SiC substrate. (b) The top
panel depicts a cross section at the dashed blue line of the optical reection map of panel (a). The
bottom panel displays the FWHM(2D) in the Raman spectrum at various locations. The dashed lines
mark the position of terrace edges as seen from the optical reection signal.
intrinsic disorder in epitaxial graphene, notably the curvature introduced near the numerous small
steps formed during the UHV graphitization process at the surface. It may also be responsible
for the 2D peak broadening as discussed in the next paragraph.
Figure 4.36 (a) displays a 5×5 µm2 Rayleigh map for a monolayer of graphene fabricated
on SiC. This map illustrates the typical morphology of such samples [199]. The spatial map
resembles AFM images of UHV prepared epitaxial graphene samples (cf. Section 4.3.5). Raman
spectra were recorded along the blue dashed line in 500 nm steps with a laser spot size of around
400 nm. Figure 4.36 (b) shows a cross section of the Rayleigh signal of panel (a). In the
bottom panel, the width of the 2D peak, i.e. FWHM(2D), is plotted, as measured along that
line at the same positions. The FWHM(2D) values undulate around 44 - 48 cm−1 . These are
widths characteristic for monolayer graphene. However, near the edges of terraces, the FWHM
values are larger by 2-3 cm−1 than on the terraces themselves. These increased values still
seem too small to attribute them to a bilayer region. Note that also P(2D) is constant around
2723 cm−1 . Since the laser spot size is much larger than the spatial inhomogeneity (< 100 nm)
reported in Section 4.3.5 and Ref. [10, 111, 163, 180], it remains dicult to draw a denite
conclusion from these measurements. Another plausible explanation for the broadened 2D peak
could be the following: since it is expected that at the edges of the terraces the graphene sheet
is bent [158], the broadening of the 2D peak may be explained by the curvature at the edge and
the accompanied local displacement of carbon atoms in the honeycomb lattice to conform to the
terrace structure. However, this observed variation in the FWHM(2D) at the terrace edges is not
sucient to explain the larger dierence in the full FWHM(2D) of about 10-20 cm−1 between
graphene grown on SiC and graphene exfoliated from HOPG. Since the larger broadening for SiC
based graphene does not disappear upon transferring the graphene to an oxidized Si substrate,
it can be proposed that structural defects such as vacancies on the terraces are responsible for
this large dierence in the FWHM.
In this subsection, the rst successful transfer of epitaxial graphene from the SiC substrate
to a SiO2 wafer has been demonstrated accompanied by a dedicated Raman spectroscopy analy-
sis [184]. The transferred akes are about 1 µm2 in size and follow the initial ne terrace
structure of UHV grown epitaxial graphene. The comparison of Raman data before and after
the transfer process allows to clarify the origin of the characteristics of the Raman features of
epitaxial graphene on SiC, which have been shown in the previous subsection. The G peak blue-
shift of epitaxial graphene on SiC in comparison to exfoliated graphene from HOPG is reverted
after transferring of the graphene to SiO2 . From the interpretation of the G feature in terms
of charge carrier concentration, it can be concluded that the n-doping of epitaxial graphene on
SiC is not present any more for the transferred akes. The 2D peak shape changes after the
transfer and resembles the 2D peak shape of exfoliated graphene from HOPG. However, the 2D
peak position and the 2D peak width remain unchanged after the transfer and are intrinsic to
the graphene structure.
4.4 Conclusion
In this chapter, an extensive analysis of the growth as well as the structural and electronic
properties of UHV grown epitaxial graphene on SiC(0001), the Si-terminated side of the basal
plane surfaces of SiC, has been given. Epitaxial graphene on SiC(0001) has the most promising
prospects for a well-ordered and homogeneous growth of a precisely controlled number of graphene
layers on large SiC wafers. The growth of graphene on SiC crystals is based on a thermal
decomposition at high temperatures of typically more than 1200 ◦ C. Owing to the higher vapor
pressure of Si in comparison to C, the Si atoms sublimate and √ leave
√ graphene layers behind. The
graphene growth is governed by the formation of the (6 3×6 3)R30◦ reconstruction, which
acts as a precursor phase of graphitization. This structure has been discussed in detail in the rst
part of this chapter, whereas the growth of graphene layers has been elaborated in the second
part. √ √
For a long time, the (6 3×6 3)R30◦ interface layer has been controversially discussed in the
literature. As shown above, the nature of this structure is now claried to a large extent. It con-
sists of an initial carbon layer in a graphene-like honeycomb arrangement with approximately 30%
percent of the carbon atoms bonding down to the Si atoms of the substrate as conrmed from
a core level analysis. The remaining two thirds of the carbon atoms remain in sp 2 -hybridization.
As a consequence of the covalent bonds to the substrate, the π -bands with the typical linear
dispersion at the K-point of the graphene Brillouin zone cannot develop in this layer, whereas the
σ -bands
√ are √ still present. These results are also in agreement with theoretical work. Therefore,
the (6 3×6 3)R30◦ reconstruction is often called buer layer or zerolayer graphene. It has to
be emphasized that this structure indeed has its designated periodicity as conrmed from LEED
and low bias STM images [147, 156], although for most √ tunneling
√ conditions an apparent (6×6)
honeycomb structure can be observed. The true (6 3×6 3)R30 periodicity is actually formed
◦
by a quasi-(6×6) honeycomb lattice consisting of smaller and larger rings. The walls of these
rings are probably formed by Si dangling bonds of the SiC substrate. Furthermore, the role of
(5×5) domains, which have also been controversially discussed in the past years, has been clari-
ed and their amount depends on the detailed UHV preparation procedure√ [156].√A minimization
of the number of (5×5) domains and thus a larger homogeneity of the (6 3×6 3)R30◦ recon-
struction can be obtained by performing a Si-deposition
√ √ at 800 ◦ C in UHV conditions. Despite
the relatively good understanding of the (6 3×6 3)R30◦ reconstruction, the detailed atomic
4.4. Conclusion 87
√ √
structure is not yet resolved in contrast to the (3×3) or ( 3× 3)R30◦ structure [142, 144].
Further investigations such as a quantitative LEED analysis should therefore be envisioned for
the future. However, the complexity of this structure comprising 338 atoms in a carbon layer
has been impressively demonstrated in this chapter and imposes severe diculties in theoretical
calculations.
From core level spectroscopy, band structure measurements, LEED and STM/STS, a coher-
ent model for the graphene growth could be developed in this chapter: during graphitization,
the interface layer is buried below the graphene layer and remains√intact. √ Putting it more pre-
cisely, every new graphene layer implies the re-formation of the (6 3×6 3)R30 reconstruction
◦
and the release of the former buer layer into a true graphene layer. Owing to this mecha-
nism,
√ the√graphene layers are all rotated by exactly 30◦ with respect to the SiC substrate. The
(6 3×6 3)R30◦ reconstruction therefore has a vital importance in the understanding of the
graphene growth characteristics on SiC(0001). As we will see in the next chapter, on SiC(0001̄)
no interface
√ layer
√ is present, which leads to a completely dierent growth mechanism. Although
the (6 3×6 3)R30 structure guarantees a well ordered growth of epitaxial graphene, some
◦
disadvantages are accompanied by its presence: rst, due to an electric dipole in the interface
layer, epitaxial graphene is intrinsically n-doped in the 1013 cm−2 regime, as can be deduced most
impressively from band structure measurements [163]. Due to screening of the electric eld of
the interface dipol, the displacement of the Dirac point relative to the Fermi level decreases with
an increasing number of√ layers.√Second, the epitaxial graphene (notably monolayer) suers from
the complexity of the (6 3×6 3)R30◦ reconstruction underneath, with its dangling bonds seen
in STM and STS [156, 176], or the possible introduction of scattering centers in graphene. In
particular, a spatial dependent electronic inuence of the interface layer with respect to monolayer
graphene could be observed in STS. Furthermore, the inuence of the interface layer is reduced
for bilayer graphene
√ in√comparison to monolayer graphene as conrmed from conductance maps
in STS. The (6 3×6 3)R30◦ reconstruction of the SiC-graphene interface is most probably re-
sponsible for the reduced mobilities measured so far in epitaxial graphene on SiC(0001). Possible
solutions to these problems will be oered later in Chapter 6.
The most direct way to monitor the graphene growth layer by layer is given by ARPES band
structure measurements [163], and the origin of all the spectacular √ physics
√ of graphene can be
condensed in one single image: the rst layer on top of the (6 3×6 3)R30 reconstruction
◦
exhibits the linear π -bands of monolayer graphene at the K-point of the graphene Brillouin zone,
with the crossing point of the π -bands (Dirac point) being displaced from the Fermi level by around
430 meV due to the already mentioned n-doping. Bilayer epitaxial graphene shows a parabolic
dispersion and has a band gap of around 100 meV arising from the gradient of the interface
dipole over the rst and second layer. The band gap can also be detected on a microscopic
level in STS [176]. This intrinsic band gap makes bilayer graphene extremely appealing for
applications in electronic devices. For monolayer graphene, on the contrary, there is still a lively
discussion about the existence of a band gap, and our data do not allow for a denite answer
to this problem. Additionally, it has to be considered that the detailed appearance of the band
structure depends on the incident photon beam energy and incident photon beam polarization
in the ARPES measurements. As shown above, constant energy maps have a shape like the
crescent of the moon and directly reect the chiral nature of the charge carriers. The intensity
distribution in these constant energy maps has been shown to dier for linear, left and right
circular polarization.
By correlating the band structure measurements for zero-, mono- and bilayer graphene with
LEED intensity spectra of the graphene (10) diraction spot, it was demonstrated for the rst time
that LEED alone can be used to determine the number of graphene layers on SiC(0001) [156, 163].
Since in practically every UHV system a LEED optics is mounted, this nding is of extremely
practical use for a continuous monitoring of the graphene preparation process. For the future, it
should be envisioned to correlate these spectra with quantitative √ LEED√ calculations.
In STM measurements, the graphene together with the (6 3×6 3)R30◦ structure can be
examined with atomic resolution [156, 176]. A higher number of graphene layers is reected by
a lower corrugation of the quasi-(6×6) honeycomb lattice of the interface layer, which remains
visible despite the graphene layers lying above. Provided the interpretation of the STM data is
taken with care, zero-, mono- and bilayer graphene can be clearly distinguished. In particular,
dierent tip conditions and dierent bias values have to be considered. In contrast to bilayer
graphene, monolayer graphene becomes transparent for high bias values. At low bias values, all
six atoms of the graphene honeycomb can be resolved only for monolayer graphene. Furthermore,
it has been shown in this chapter that STM allows to get spectacular insight into quantum
interference patterns of mono- and bilayer graphene due to intervalley scattering between the K-
0
and K -point.
Vibrational modes of graphene can be probed by Raman spectroscopy. Although the interpre-
tation of the Raman data is more complicated in comparison to graphene exfoliated from HOPG,
important information can be gained [184]. In particular, the disorder related D peak is small
and accounts for the high crystalline quality of the graphene. The width of the 2D peak stands
out as a suitable ngerprint to assign the number of layers in the case of graphene on SiC(0001),
since it exhibits a (negative) linear relationship with the inverse number of layers. Furthermore,
the rst successful transfer of epitaxial graphene from the SiC substrate to a SiO2 wafer has been
demonstrated yielding ake sizes of about 1 µm2 . A comparison of the Raman spectra before and
after the transfer process leads to an interpretation of the G feature mainly in terms of charge
carrier concentration. It can be concluded that the n-doping of epitaxial graphene on SiC is not
present any more for the transferred akes.
The discussion of the properties of epitaxial graphene on SiC(0001) in this chapter has shown
that SiC is a perfectly suited template for graphene growth on a semiconducting substrate with
wafer scale size. Despite the presence of the substrate, the graphene exhibits all of the spectacular
properties that are known from freestanding graphene akes exfoliated from HOPG. Of course,
the inuence of the substrate has to be treated and analyzed with care, it however does not impose
principal problems. With the knowledge accumulated in this chapter, excellent preconditions are
given for a manipulation of the graphene and the graphene-SiC interface with the goal to reduce
the inuence of the substrate on the above growing graphene layers. This task will be faced
in Chapter 6. Before, a dedicated analysis of dierent growth conditions will be given in the
next chapter. In particular, the problem of the low homogeneity of UHV grown graphene has
to be addressed since the typical terrace sizes are only of some hundreds of nm2 , as has been
demonstrated above.
Chapter 5
Variation of the graphene growth

conditions
In Chapter 4 the basic properties of epitaxial graphene on SiC(0001) have been discussed. The
Si-terminated side of the basal plane surfaces of SiC has the most promising prospects for a
well-ordered and homogeneous growth of a precisely controlled number of epitaxial graphene
layers. For the epitaxial growth of graphene on SiC, also the inuence of dierent types of
the substrate should be investigated, which is elaborated in Section 5.1. The most common
alternative to SiC(0001) is the corresponding carbon terminated side of the basal plane surfaces
of SiC. It has proven to be an appropriate template for graphene growth. In fact the amazing
properties of epitaxial graphene on SiC have rst been demonstrated on SiC(0001̄) [4, 12]. It
has to be considered, however, that the two faces SiC(0001) and SiC(0001̄) exhibit completely
dierent graphitization characteristics. In Section 5.1.1 more details about epitaxial graphene on
SiC(0001̄) are discussed. The role of the SiC polytype (4H, 6H, 3C) in the graphitization process
is subsequently examined in Section 5.1.2. The graphene growth is in principle not limited to
the basal plane surfaces of SiC but could rather be extended from slight o-axis samples to
SiC(112̄0), SiC(11̄02) and SiC(1̄102̄), which is briey discussed in Section 5.1.3. The variations
of the growth conditions do not only include the variation of the substrate but also the variation of
the graphene growth process itself. It turned out that the basic structural and electronic properties
of epitaxial graphene are to a large extent independent of the detailed graphitization procedure so
that Chapter 4 focussed on the widely used UHV preparation of epitaxial graphene on SiC. Since
the standard UHV growth does not yield a large scale homogeneity as shown in Section 4.3.5,
the goal of modied growth techniques is to improve the morphology of graphene. Section 5.2.1
discusses the epitaxial graphene growth by molecular beam epitaxy using carbon evaporation
in UHV conditions at reduced graphitization temperatures, and Section 5.2.2 investigates the
recently developed atmospheric pressure graphitization technique [10]. The latter yields large
terrace graphene surfaces and has promising prospects towards graphene based nanoelectronics.
90 Chapter 5. Variation of the graphene growth conditions
Figure 5.1: Phase diagram of the SiC(0001̄) surface with LEED images of (a) the (2×2)Si , (b) the (3×3)
and (c) the (2×2)C reconstruction phase as well as (d) a graphitic surface. Arrows with encircled process
numbers indicate dierent preparation steps. The image is taken from Ref. [9, 137].
5.1 Variation of the SiC substrate

5.1.1 Epitaxial graphene on SiC(0001̄)
The phase diagram of SiC(0001̄)
The phase diagram of SiC(0001̄) has a similar complexity as the one on SiC(0001) shown in
Fig. 4.1 of Section 4.1. Also similar preparation procedures were applied although the anneal-
ing temperatures are generally higher on SiC(0001̄). The sequence of surface reconstructions,
however, is completely dierent [9, 208] and is shown in √Fig. √
5.1. After insertion into the UHV
chamber, the hydrogen etched samples show an oxidic ( 3× 3)R30◦ structure, similar to the
one observed for the Si-face [209]. The Si rich side of the phase diagram starts with a (2×2)
phase with a LEED pattern as shown in Fig. 5.1 (a), which is denoted (2×2)Si in order to distin-
guish it from a more carbon rich (2×2)C phase (see below). The (2×2)Si phase develops upon
annealing in Si ux at about 1150 ◦ C (indicated as process "A" in Fig. 5.1). This procedure is
also used in order to remove surface oxides from the hydrogen etched samples. Further anneal-
ing at 1050 ◦ C (process "1") results in a (3×3) reconstruction, cf. Figure 5.1 (b). The (3×3)
reconstruction can also be observed after directly annealing the hydrogen etched sample up to
1000 ◦ C (process "B"). Heating of the (3×3) structure (at 1075 ◦ C, process "2") leads to the
carbon rich (2×2)C phase (LEED pattern in Fig. 5.1 (c)), which in most cases, can only be
prepared in coexistence with the (3×3) reconstruction [208, 210]. We note that the (2×2)C is
the only phase on SiC(0001̄), for which the atomic structure of the surface (a Si adatom model)
has been resolved [211]. Further heating at temperatures of at least 1150 ◦ C (process "3") leads
to a graphitic phase with (1×1) spots of the SiC substrate and ring-like diraction features from
the graphite layer as shown in Fig. 5.1 (d). However, before observing a pure graphitic (1×1)
structure, graphene already coexists with the (3×3) and (2×2)C phase [137, 155, 212, 213].
It should be emphasized once again that epitaxial graphene grown on the carbon face of the
basal plane surfaces of SiC exhibits completely dierent growth characteristics than graphene on
the silicon face. Most of the transport measurements have so far been conducted on SiC(0001̄) [4,
12], whereas the structural and electronic properties have more extensively been analyzed on
SiC(0001). The following paragraphs provide deeper insight into the initial stages of graphitization
of SiC(0001̄) as investigated by LEED, CLPES and STM. These results have been published in
Ref. [137].
5.1. Variation of the SiC substrate 91
Figure 5.2: LEED pattern at 75 eV showing a coexistence of the (3×3) and the (2×2)C phase with a
graphitic (1×1)graphitic structure on SiC(0001̄). The graphene does not show a distinct spot but rather
a ring-like structure as shown in the inset.
Low energy electron diraction

Figure 5.2 displays a LEED pattern of the rst stage of the graphene growth. It shows sharp
and intense spots that correspond to both the (3×3) and the (2×2)C periodicity, but also quite
weak graphitic diraction rings that are located close to the ( 32 , 23 )-position of the SiC substrate
as shown in the enlarged section in the gure. The position of these graphitic intensities suggests
that the graphene is rotated by 30◦ with respect to the SiC-substrate as it is the case for
SiC(0001). However, the fact that the LEED pattern does not show a single spot, but rather a
ring, indicates that graphene patches are existing in domains of dierent orientations with respect
to the substrate.
Core level photoemission spectroscopy

The essential dierences between the graphitization processes on both polarities of the basal
plane surfaces of SiC√already√ arise in these initial stages of the graphitization process. In case
of SiC(0001), the (6 3×6 3)R30◦ reconstruction corresponds to a covalently bound graphene
layer so that it can act as an electronically inactive buer layer as outlined above. Such a
precursor phase of graphitization is missing for SiC(0001̄). The existence of rotated domains
can already be seen as an indication for the absence of a strongly bonded rst graphene layer.
Spectra of the C 1s core level, in particular, allow for a detailed analysis of dierently bound
carbon atoms as shown in Fig. 5.3 for 0.3, 0.6 and 0.8 ML graphene with the thickness evaluated
by consideration of the layer dependent electron attenuation (cf. Section 3.1.3). In contrast to
epitaxial graphene on SiC(0001), where three components can already be observed for the buer
layer reconstruction, only two components can be resolved on SiC(0001̄). This also holds for
higher graphene coverages [155], which are not shown here. The SiC bulk component appears at
282.8 eV binding energy and the graphene related component at 284.6 eV. The binding energy
of the latter slightly decreases for higher coverages thus approaching that of graphite (284.4 eV).
Note that the data has been obtained in the home laboratory using Mg Kα radiation. However,
the resolution is sucient to conclude that there is no covalent bonding of graphene to the SiC
substrate upon graphitization. For graphene on SiC(0001), the covalent bonding gives rise to two
surface related components in the buer layer, which remain unperturbed during the graphene
growth (cf. Section 4.2.2). There, as further graphene layers are starting to grow from the
Figure 5.3: C 1s core level spectra for the initial stages of graphitization of SiC(0001̄) with dierent
graphene layer thicknesses as determined from layer dependent electron attenuation. The graphene is
growing without the formation of an interface layer.
Figure 5.4: STM micrographs for the initial stages of graphitization on SiC(0001̄). (a) Large area
image, which shows the inhomogeneous growth characteristics. (b) Overview of the dierent surface
reconstructions: the pure (3×3) phase and the (2×2) and (3×3) phases covered with graphene showing
a Moiré type superlattice. (c) Enlarged scan of the graphene covered (3×3) phase (panels (a)-(c):
Utip = 2.0 V). (d) Enlarged scan of the graphene covered (2×2) phase (Utip = 1.0 V).
bottom, the orientation of the graphene layers will always be the same as their growth direction
is imprinted by the buer layer. Due to the weak coupling of graphene in case of SiC(0001̄), the
graphene can grow in rotated domains, which results in a turbostratic growth mode [214, 215].
Scanning tunneling microscopy

It should be noted that density functional theory calculations [152, 153] as well as studies with
inverse photoemission [216] and X-ray reectivity [170, 215] have claimed a strong interaction
between the rst graphene layer and the SiC(0001̄)-substrate. Deeper insight on a microscopic
level can be given by STM. It can directly be analyzed how and on what surface reconstruction(s)
the graphene growth takes place. We investigate an early graphene stage, where still considerable
contributions from the (2×2)C and, in particular, from the (3×3) reconstructions are present.
STM data, where patches of both phases can be seen, are presented in Fig. 5.4. Figure 5.4 (a)
shows a large scale STM micrograph highlighting the fact that, as for SiC(0001), UHV grown
epitaxial graphene on SiC(0001̄) does not exhibit a large homogeneity. The LEED pattern corre-
sponding to the STM images displays an even weaker graphitic ring than shown in Fig. 5.2 and
Figure 5.5: Atomically resolved graphene layer covering the (3×3) reconstruction acquired at a tip bias
of +1.0 V. The graphene layer is rotated by 18◦ with respect to the substrate.
mainly consists of the (3×3) reconstruction (not shown). Correspondingly, most of the area in
Fig. 5.4 (a) is covered by the (3×3) structure. Panel (b) shows a smaller area of this surface with
higher resolution as indicated by the zoom sketch between panels (a) and (b). Here, the left side
is covered with the (3×3) phase in particular. On the right side in panel (b), however, the two
surface reconstructions (2×2)C and (3×3) can be observed with dierent superlattices on top as
pointed out by the enlargements in panel (c) and (d). Panel (c) displays the (3×3) reconstruction
with a superlattice periodicity of 4.2 nm and panel (d) shows the (2×2)C reconstruction with a
supperlattice periodicity of 3.6 nm. The period of these superlattices are related to the rotation
of the graphene-like overlayers with respect to the SiC substrate and can be seen as Moiré pat-
terns [137, 212, 213]. Such superlattices have already been observed in earlier STM studies on
the (2×2)C phase [208, 210, 211] but could not directly be linked to the growth of graphene. For
monolayer graphene on SiC(0001) it is known that its visibility depends on the bias voltage (see
Section 4.3.4 and Ref. [154, 178, 188, 190]). The STM micrographs in Fig. 5.4 (a)-(c) have been
taken at high bias. For these values the graphene layer is transparent for tunneling and only the
bare surface reconstructions, which have earlier been analyzed extensively [208, 210, 211, 217],
can be observed. At the chosen tunnelling bias of 2.0 V (tip voltage), the (3×3) structure dis-
plays a regular array of hexagonal holes. For low tunnelling bias, however, a single protrusion
per unit cell can be resolved which is shown in Fig. 5.5. Furthermore, the graphene layer can be
seen as a honeycomb pattern and the graphene unit cell can indeed be resolved. The fact that
every graphene atom of the unit cell can be imaged is another indication for monolayer graphene
(see Section 4.3.4 and Ref. [154, 176, 178, 186, 188, 190]). The rotation of the graphene unit
cell with respect to the SiC substrate amounts to 18◦ 1 . These observations conrm the picture
of a weak interaction of the rst graphene layer on SiC(0001̄). Furthermore, Hiebel et al. have
concluded from STM data that the substrate-overlayer coupling is a bit stronger for (2×2)C
phase [212, 218] but still much smaller than
√ the √covalent bonding of the rst graphitic layer on
the Si face, which is represented by the (6 3×6 3)R30◦ reconstruction.
For a higher graphene coverage, the SiC surface reconstructions cannot be resolved any more.
1 Note
that this orientation angle is not close to the main orientation angle of 30 ◦ C as seen in LEED. However,
the observation of rotational disorder in LEED do not exclude such an orientation angle.
Instead, dierent kinds of Moiré patterns can be observed which corroborates the proposal of
a turbostratic growth mode on SiC(0001̄) [214]. The rotational disorder leads to an electronic
decoupling of the graphene layers thus preserving some unique properties of monolayer graphene
even in thicker lms [4, 12, 56, 170]. Recent photoemission experiments have revealed the ideal
graphene monolayer band structure for a multilayer sample, with the Dirac point being located
at the Fermi level so that charge neutrality is obtained [57]. Despite its complicated graphene
growth characteristics, SiC(0001̄) represents an attractive support for potential graphene based
or carbon electronics. Very recently, even the quantum Hall eect was measured for epitaxial
graphene samples on SiC(0001̄) [39].
5.1.2 Inuence of the polytype

A polytype dependence of the growth of epitaxial graphene on SiC surfaces could in principle be
expected as a dierent stacking in the near surface layers could inuence the growth dynamics.
The most widely used polytypes are 4H and 6H. As already mentioned at the appropriate positions
in Chapter 4, the structural and electronic properties of graphene do not seem to be inuenced
by the choice of the polytype. In particular, the band structure and the atomic structure as
measured from STM look identical. Direct evidence for the same structural arrangement on the
surface is given by LEED intensity vs. energy
√ √spectra. As shown in Ref. [185] the graphene (10)
spot (which is the only spot in the (6 3×6 3)R30◦ diraction pattern that shows variations
upon graphitization) exhibits identical spectra for the 4H- and 6H-polytype. This can only be due
to the same atomic arrangement at the surface. In terms of homogeneity, it could be expected
that the polytype 6H is of advantage as the stacking sequence ("S3") provides three SiC bilayers
in one stacking direction instead of two for the polytype 4H. It should be noted again that
carbon atoms of three SiC bilayers are needed to form one graphene layer. However, samples of
dierent polytype do not show any signicant dierences in their morphology. This observation
is reasonable for samples prepared in UHV under a Si ux as it is known that such a preparation
procedure yields an S3-termination at the surface [144]. As 4H-SiC-samples prepared without
Si ux do not show any signicant degradation in homogeneity, it has to be assumed that the
stacking arrangement in the near surface layers is of minor inuence in the graphitization process.
The polytype 3C is of major interest for graphene based nanoelectronics, as 3C-SiC is com-
monly grown on Si-wafers with a thickness of a few tens of µm so that an implementation in
existing Si device technology seems to be easily possible. Until now, only a few reports exist on
the graphene growth using 3C-SiC(111) on a Si substrate [219, 220]. Whereas Suemitsu et al. for
instance succeeded in growing epitaxial graphene on 3C-SiC(111) on a Si(110) substrate [219],
Ouerghi et al. were using a Si(111) substrate [220], which exhibits a larger lattice mismatch
between Si and SiC. The analysis of these graphene lms is unfortunately rather limited and does
not allow for a denite statement about their quality. The surface reconstructions are the same
for 3C-, 4H- and 6H-SiC samples so that a comparable graphitization process is expected. The
main problem, however, is the fact that the silicon of the Si substrate starts to evaporate or
even melt at those annealing temperatures that are necessary for conventional UHV graphitiza-
tion. Even if the substrate remains intact, extremely Si-rich conditions seem to be present at
the 3C-SiC surface. This is exactly the problem that we also suered from in our experimental
eorts on 3C-SiC(111)/Si(111). We actually could observe a coexistence between the Si-rich
surface
√ √reconstructions (such as the (3×3) structure or transition phases between (3×3) and
( 3× 3)R30◦ reconstruction) and already graphitized regions. It seems that Si substrate atoms
could diuse to the 3C-SiC surface via defects, micropipes or along areas that are peeled o the
Figure 5.6: (a) Schematic orientation of non-basal plane surfaces within the SiC structure. (b) Atomic
model of 4H-SiC in side-view perspective along the (112̄0) direction, with bonds within the (112̄0) plane
indicated by the black line. The orientation of the (1̄102̄) plane is denoted by the red line. Large (green)
atoms represent Si, small (black) atoms C. The image is adapted from Ref. [105].
surface. Consequently, a controlled graphene growth seems to be impossible. On our samples,

the graphene developed only at 3C-SiC regions that have already peeled o the Si substrate. The
more appropriate way to analyze epitaxial graphene on 3C-SiC is to grow the 3C-SiC on a 4H- or
6H-SiC substrate, which has been performed in Ref. [221]. However, a detailed analysis of such
graphene lms is not yet available.
5.1.3 Non-basal plane SiC surfaces

The focus of the epitaxial growth of graphene on SiC surfaces is currently only on the basal
plane surfaces SiC(0001) and SiC(0001̄). Naturally, this is the most reasonable way to start the
investigation of the graphene growth on SiC as the structural and electronic properties of the
corresponding surface reconstructions are quite well understood. Also basal plane SiC wafers are
easily available in large quantities. Although the graphene growth on non-basal plane surfaces
has never been investigated, there is no principal restriction to study basal plane surfaces only.
Such wafers are produced by cutting the grown SiC bulk crystal perpendicular to the (0001)
direction. In a similar way, the bulk crystal can be cut in other main symmetry directions as
schematically illustrated in Fig. 5.6 (a) for SiC(11̄00), SiC(112̄0) and SiC(1̄102̄). In panel (b)
of Fig. 5.6, the (1̄102̄) direction is highlighted in a structural model of 4H-SiC. The non-basal
plane surfaces directly exhibit the long period of the SiC bulk polytype stacking in the direction of
the (0001)-axis. The resulting reduced surface symmetry makes these directions quite promising
for the investigation of a reduced dimensionality of SiC surfaces, which then could be imposed
onto the epitaxially grown graphene layers. Instead of large scale two-dimensional graphene as
grown on the basal plane surfaces, the goal would be to study graphene with a reduced scale
in one dimension, i.e. epitaxial graphene nanoribbons, where the development of gaps can be
expected [35, 36].
Figure 5.7: LEED patterns at 95 eV for dierent surface reconstructions of the 4H-SiC(11̄02) surface
at dierent annealing temperatures including the graphitization process: (a) the Si-rich (2×1), (b) the
c(2×2) and (c) the carbon rich (1×1) structure for dierent annealing temperatures.
O-axis SiC(0001) surfaces

The rst step towards a reduced dimensionality would be to study vicinal SiC surfaces, i.e. wafers
that are cut out of the SiC bulk crystal by a few degrees o the basal plane axis. Whereas
every on-axis wafer has a slight misalignment of some tenths of a degree, which results in the
stepped terrace structure after hydrogen etching (cf. Section 3.2), o-axis wafers are cut o the
(0001)-axis by around four to eight degrees. As a consequence, the step density on such surfaces
is much higher, and therefore the terrace width much smaller. However, it has recently been
shown that the graphene is growing over the steps [158, 167, 222] so that a direct access to
graphene nanoribbons is not possible. It should also be considered that the graphene thickness
control is much more dicult on such samples as the Si out diusion is enhanced due to the
higher step density [167]. More studies on these surfaces could however be promising. Controlled
step bunching could lead to suciently large steps so that the graphene cannot grow over the
steps any more. Furthermore, the etching of the graphene by metal nanoparticles [223] along the
terrace edges for nanoribbon development deserves detailed investigation. The stepped substrate
morphology generated during hydrogen etching could induce a controlled etching direction, which
is probably much more dicult for large homogeneous terraces on on-axis SiC wafers.
The 4H-SiC(11̄02) surface

The SiC(1̄102̄) and the SiC(11̄02) surface, which is the isomorphic opposite of the (1̄102̄) orien-
tation, are situated diagonally in the bulk unit cell as depicted in Fig. 5.6 (b) for the 4H polytype.
The stacking sequence induces an anisotropic bulk truncated surface structure, which contains
alternating stripes of dierent bond conguration, namely basal plane type, and cubic type sur-
face atom conguration [105, 224]. On these intrinsically anisotropic surfaces, the growth of
graphene with reduced dimensionality could be promising.
The 4H-SiC(11̄02) surface has recently been investigated intensively [104, 105, 224, 225]. Its
phase diagram exhibits dierent surface reconstructions similar to the basal plane surfaces of SiC
(cf. Section 4.1 and 5.1.1). For a Si rich (2×1) reconstruction, obtained only after Si-evaporation
at 800 ◦ C onto the hydrogen etched 4H-SiC(11̄02) crystal, a one-dimensional structure was dis-
covered, both atomically and electronically, with atomic Si chains at the outermost surface,
which host a one-dimensional surface state [104]. The corresponding LEED pattern is shown in
Figure 5.8: Raman spectrum of the graphitized (1×1) structure of the 4H-SiC(11̄02) surface after
annealing to 1150 ◦ C. The huge D peak and the appearance of the (D+G)- and the G*-feature is a sign
for high disorder and a large amount of defects on the surface.
Fig. 5.7 (a). For a c(2×2) phase, which develops after annealing the (2×1) structure to around
1000 ◦ C (see panel (b) in Fig. 5.7), the surface is terminated by alternating rows of Si adatoms
and carbon dimers, again displaying a strongly anisotropic structure [225, 226].
A further increase of the annealing temperature of 50 ◦ C with respect to the c(2×2) structure
leads to the development of a carbon rich (1×1) structure as depicted in the LEED pattern
in Fig. 5.7 (c). After another heating step at 1080 ◦ C, the (1×1) LEED pattern already gets
signicantly weaker, at 1100 ◦ C the LEED pattern has practically disappeared. The nature of the
(1×1) structure, however, does not change as the LEED I/V curves of the diraction spots do
not change (not shown). However, a graphitic diraction spot or graphitic rings as for the basal
plane surfaces of SiC cannot be detected. Obviously, the graphene growth seems not to take
place in an ordered way. Also ARPES measurements do not show any development of π -bands
near the Fermi level (not shown). In comparison to SiC(0001), the graphene seems to start to
grow at lower temperatures and with a higher growth rate.
Fig. 5.8 shows the Raman spectrum of the same sample after annealing to 1150 ◦ C. Reference
data of a hydrogen etched reference sample have been subtracted in order to eliminate SiC related
peaks. The strong intensity of the D peak in comparison to the G peak is a clear indication for
the strong disorder in the graphene and corroborates the LEED observations. Furthermore, the
(D+G) and G* peak can only be observed for low quality graphene samples. The G peak value
of 1602 cm−1 could indicate strong n-doping. Also strain should not be neglected as the 2D
peak position at 2754 cm−1 indicates compressive strain (compare Section 4.3.6). Similar to
epitaxial graphene on SiC(0001), the 2D peak has a single peak like appearance. The 2D peak
is however much broader with a full width at half maximum of around 100 cm−1 , which would
not be expected for a high quality graphene sample.
The role of the (1×1) structure during graphitization remains so far unclear. Figure 5.9
displays the C 1s spectra for the (1×1) structure at 1050 ◦ C, 1100 ◦ C and 1150 ◦ C. Note that
the measurements have been performed with a Mg Kα source. Whereas the C 1s peak for the
Figure 5.9: The C 1s spectrum as measured with CLPES for the carbon rich (1×1) structure of the
4H-SiC(11̄02) surface for dierent annealing temperatures including graphitization.
c(2×2) structure is sharp and consists of practically one component only [105, 224, 225], the
(1×1) structure already exhibits a clear shoulder. From synchrotron measurements, it has been
attributed to be a graphitic component [105, 224, 225]. With higher annealing temperature, this
component is getting more intense as can be seen from the Mg Kα based spectra in Fig. 5.9.
Although at 1150 ◦ C not more than two graphene layers should be present on the surface, it is
quite astonishing that the LEED pattern at this temperature does not show any clear diraction
spots any more. The need for further studies of the initial stages of the graphitization of the
SiC(11̄02) surface and the role of the (1×1) structure is quite evident.
For the isomorphic opposite of the (11̄02) surface orientation, i.e. the 4H-SiC(1̄102̄) surface,
the present knowledge of the surface phase diagram is much more limited. The preparation of
dierent surface phases seems to be more dicult. Preliminary experiments show also several
ordered phases, namely a (1×2), a c(2×2) phase and a (1×1) carbon rich surface structure [130].
The (1×2) phase appears in particular interesting, since for this phase the long periodicity on
the surface is doubled, instead of the short one as for the (2×1) structure of SiC(11̄02). This
promises a further separation of the adatom rows, and thus an even higher degree of anisotropy
of this phase. It unfortunately is not present at the surface when the graphitization temperatures
are reached. Similar to the SiC(11̄02) surface, an ordered graphene growth could not be obtained
in the framework of my thesis. The LEED spots of the (1×1) carbon rich surface reconstruction
again disappear upon graphitization, also no π -band system could be observed near the Fermi
level (not shown).
From a more general point of view, the graphitization characteristics of SiC(11̄0n) surfaces,
i.e. surfaces having an arbitrary angle to the basal plane direction, have a signicant impact
on the understanding of the graphitization of the basal plane surfaces of SiC. The homogeneity
of graphene on SiC(0001) is so far limited by the size of the terraces generated during the
initial hydrogen etching [10] (cf. Section 5.2.2). At the terrace steps, the next graphene layer
starts to grow while this process is governed by the graphitization characteristics of SiC(11̄0n)
surfaces. Two recent investigations of the epitaxial graphene step edges by Raman spectroscopy,
transmission electron microscopy and LEEM are in accordance with our results of the SiC(11̄02)
surface [227, 228]: the graphene growth starts at lower temperatures and with a higher growth
rate in comparison to SiC(0001); furthermore, the defect density is much higher. A better control
of the graphene growth on SiC(11̄0n) surfaces is crucial for a further optimization of the graphene
growth conditions on SiC(0001).
The SiC(112̄0) surface

The SiC(112̄0) surface that was examined in the framework of my thesis showed similar graphi-
tization characteristics as the SiC(11̄02) and SiC(1̄102̄) surface. An ordered graphene growth
could not be observed. The SiC(112̄0) surface is of particular interest as it directly exhibits the
long period of the SiC bulk polytype stacking in direction of the (0001) axis. Quite astonish-
ingly, only a (1×1) surface reconstruction can be observed although various kinds of Si-deposition
and annealing cycles have been performed for hydrogen etched 4H-SiC(112̄0) samples. In some
preparation stages, very faint superstructure spots could be observed, which, however, do not
present a clear surface reconstruction. This behavior corresponds well to reports in the litera-
ture [229]. While the surface remains in a (1×1) periodicity throughout the dierent annealing
and Si deposition processes, the spot intensities change at certain temperatures. A quantitative
inspection of the LEED spots reveals the presence of three unique phases [229]. To understand
the nature of these (1×1) reconstructions, a quantitative analysis of their structure by means
of LEED calculations remains to be performed. Upon graphitization, the LEED pattern is again
getting weaker without the formation of any ordered graphene layers.
Until now, the reasons for the reported graphitization characteristics of non-basal plane SiC
surfaces remain unclear. Evidently, more experiments are needed to draw denite conclusions. It
could well be that slightly modied preparation conditions may lead to more promising results.
Another possible approach to shed light into non-basal plane surfaces would be a systematic
surface study from o-axis samples to non-basal plane surfaces by cutting wafers under a large
variety of dierent angles out of a SiC bulk crystal. Such an investigation is of importance for
the graphene growth on the basal plane surfaces as the graphene starts to grow at the terrace
edges where (11̄0n) surfaces are present.
5.2 Variation of the growth technique

The variations of the growth conditions do not only include the variation of the substrate but
also the variation of the detailed procedure of the graphene growth. The basic structural and
electronic properties of epitaxial graphene are mainly independent of the detailed graphitization
procedure so that the discussion in Chapter 4 focussed on the widely used UHV preparation
of epitaxial graphene on SiC. The graphene growth in UHV conditions, however, requires a
rather complex equipment and a lengthy preparation procedure with dierent Si evaporation and
annealing steps. Furthermore, for a large scale production, this technique is not suitable as the
size of a homogeneous graphene domain is much too small as shown previously in Section 4.3.5.
For dierent preparation techniques, the thermodynamics and growth kinetics could be inuenced.
In this section, two alternative methods are presented. First, the graphene growth induced by
carbon evaporation is elaborated. It requires reduced temperatures in comparison to the UHV
preparation technique and should be regarded as a proof of principle for such an alternative
preparation process. Second, the graphene growth in an Ar atmosphere is discussed. It has
recently been shown to yield large terrace graphene surfaces [10].
5.2.1 Carbon evaporation in UHV conditions

Phase diagrams for SiC surfaces show more carbon rich surface reconstructions for higher an-
nealing temperatures (cf. Section 4.1 and 5.1.1). As silicon has the higher vapor pressure in
5.2. Variation of the growth technique 101
Figure
√ √5.10: (a) LEED pattern of the (3×3) phase on SiC(0001) and (b) of the carbon source induced
( 3× 3)R30◦ structure. (c) LEED intensity spectra√of the√ (1/3,1/3) spot of the carbon induced (solid
line) and the conventionally prepared (dashed line) ( 3× 3)R30 phase. (d) LEED pattern of carbon
◦
deposition assisted monolayer graphene.
comparison to carbon, its evaporation leads to a carbon rich surface, a fact which is convention-
ally used for graphene growth. By the deposition of silicon, it is possible to move back to silicon
rich surface reconstructions. Similarly, although never investigated, it should be possible, to de-
velop carbon rich surface structures by oering carbon at lower annealing temperatures. In this
section, we use this idea for an alternative approach of epitaxial graphene growth on SiC(0001)
and SiC(0001̄), which is based on carbon evaporation under UHV conditions. In particular, we
succeeded in getting monolayer graphene on SiC(0001) at signicantly lower temperatures in
comparison to the existing UHV based growth technique so that the initial surface morphology
of the SiC crystals is not degraded as it is the case for conventionally UHV grown epitaxial
graphene. Besides the improved surface morphology, this technique has promising prospects for
the epitaxial growth of graphene on 3C-SiC/Si(111) as the Si substrate starts to evaporate at
the conventionally used graphitization temperatures of around 1200 ◦ C. The results presented in
this section have been published in Ref. [230].
After introducing the hydrogen etched samples into the UHV chamber, the samples were
cleaned in a Si ux of around 1 monolayer (ML)/min at 800 ◦ C for SiC(0001) and 1150 ◦ C for
SiC(0001̄) (cf. Section 4.1 and 5.1.1). The evaporation of carbon is performed using a commercial
carbon source (MBE Komponenten GmbH) with a ux of around 1ML/min evaporated from highly
oriented pyrolytic graphite (HOPG) (cf. Section 3.3).
The starting point for the carbon deposition on SiC(0001) is the Si enriched (3×3) structure
as shown in Fig. 5.10 (a). It develops during annealing the SiC crystal
√ √at 800 ◦ C and simulta-
◦
neous Si deposition (≈ 1 ML/min ux) (cf. Section 4.1). The ( 3× 3)R30 reconstruction
normally
√ evolves
√ during subsequent annealing of the (3×3) phase to about 950 ◦ C. Note that
this ( 3× 3)R30◦ phase contains 1/3 ML√of Si √ on top◦ of the SiC substrate (cf. Fig. 4.2 (c)
and (d) of Section 4.1). Alternatively, a ( 3× 3)R30 structure also develops after 10 min-
utes of carbon evaporation on the (3×3) reconstruction at room temperature and subsequent
annealing
√ √ to about 550 ◦ C (see LEED pattern in Fig. 5.10 (b)). This carbon source induced
( 3× 3)R30◦ phase can also be generated by evaporating carbon during simultaneous anneal-
ing at around 400 ◦ C for 30 minutes.
√ √The quality, however, was worse as concluded from LEED.
Since on SiC(0001) an oxidic ( 3× 3)R30◦ surface √ reconstruction
√ can also develop [209], we
used LEED intensity spectra to determine which ( 3× 3)R30 phase evolves during carbon
◦
evaporation.
√ The√ (1/3,◦ 1/3) spot spectra are displayed in Fig. 5.10 (c) for the carbon source
induced ( 3× 3)R30 reconstruction (solid line) and the one prepared conventionally by an-
Figure 5.11: Epitaxial monolayer graphene on SiC grown by carbon deposition and simultaneous annealing
to about 950 ◦ C: (a) LEED pattern at 140 eV (right) with (10) spot of the SiC substrate indicated, and a
close-up view (left) at 126 eV for the region around the rst order graphene diraction spot (center spot in
the schematic inset). Below, intensity spectra of the graphene spot are shown for carbon assisted growth
(solid line) and for conventional growth by annealing in UHV (dashed line). (b) π -band dispersion
perpendicular to the ΓK-direction as obtained from ARPES (He II radiation). (c) AFM micrograph
showing the morphology of the carbon source induced monolayer graphene sample.
nealing alone (dashed line). They show the same characteristics, which claries that the same
atomic arrangement is present at the surface.
√ √
The successful preparation of the ( 3× 3)R30◦ reconstruction by carbon assistance at a
much lower temperature than by annealing alone serves as a proof of principle that by carbon
deposition it should also be possible to obtain epitaxial graphene at lower temperatures. Indeed,
this can be achieved by annealing with carbon to about 950 ◦ C for 30 minutes as compared
to 1200 ◦ C used conventionally
√ √ (cf. Chapter 4). Note again that at 950 ◦ C, without carbon
evaporation, the ( 3× 3)R30◦ structure develops (cf. Section 4.1). The corresponding LEED
pattern
√ of
√ carbon◦ source induced epitaxial graphene is displayed in Fig. 5.10 (d). It shows the
(6 3×6 3)R30 periodicity of the interface layer, which is located between the SiC substrate
and the graphene. As demonstrated in Section 4.3.3, LEED patterns exhibit this periodicity up to
several graphene layers. Yet, a variation with the graphene thickness can be√ seen√in the intensity
spectra of the rst order graphene spot in relation to the surrounding (6 3×6 3)R30◦ spots.
The intensity spectrum of the graphene spot can be used as a ngerprint for the determination
of the number of graphene layers as elaborated in Section 4.3.3 and Ref. [156, 163]. The
corresponding spectrum of the carbon source induced graphene is displayed in Fig. 5.11 (a)
and shows exactly the characteristics of the spectrum from conventionally prepared epitaxial
monolayer graphene shown underneath. Direct evidence is coming from ARPES measurements
using He II radiation. The carbon deposition assisted graphene displays well resolved linearly
dispersing π -bands with no apparent zero- or bilayer contributions as shown in Fig. 5.11 (b). The
Dirac point is shifted below the Fermi energy level by about 400 meV as typical for epitaxial
monolayer graphene on SiC (cf. Section 4.3.2). This n-doping is known to be independent from
the preparation procedure as well as from
√ the√polytype and doping level of the SiC substrate, but
it is related to the presence of the (6 3×6 3)R30 interface reconstruction (cf. Section 6.2.1
◦
and Ref. [160]).

The modication of the growth dynamics of graphene inuences the morphology of the sur-
face. Figure 5.11 (c) shows an AFM image of the monolayer graphene sample prepared by carbon
Figure 5.12: LEED patterns and close-up views of graphene of approximately one monolayer thickness
on 4H-SiC(0001̄) prepared from two dierent procedures in UHV: (a) assisted by carbon evaporation
showing diraction rings from graphene with a predominant orientation of around 0◦ with respect to the
SiC-substrate (10) spot, and (b) prepared by conventional annealing showing rings with a predominant
orientation of 30◦ with respect to the substrate spot.
deposition. In contrast to monolayer graphene conventionally prepared in UHV, the regular array
of atomically at terraces generated by the initial hydrogen etching procedure is preserved to a
large extent with the carbon source aided method. Possibly, due to the reduced annealing tem-
peratures, a temporary surface roughening can be avoided. Alternatively, also the necessity of Si
diusion to leave the graphene layer system might be reduced. However, it has to be noted that
the observed morphology does not necessarily have to be identical with the homogeneity of the
graphene layer. It should be noted that a French group has been working in parallel on the same
approach [231] as presented here and in Ref. [230]. However, their thickness determination has
only been performed via CLPES so that a statement about the electronic quality is not possible.
Similar to our results, a spatially resolved analysis of the graphene homogeneity was not presented
either.
Epitaxial graphene can also be prepared at reduced temperatures on the SiC(0001̄) surface
by additional carbon deposition. After cleaning the hydrogen etched 4H-SiC(0001̄) crystal at
1150 ◦ C in a Si ux of approximately 1 ML/min, the Si rich (2×2)Si structure develops (cf. Sec-
tion 5.1.1). The starting point for carbon evaporation is the (3×3) structure, which is generated
after subsequent heating at 1050 ◦ C. Annealing the crystal at exactly this temperature under
a simultaneous carbon ux of around 1 ML/min leads to the growth of epitaxial graphene as
conrmed by the LEED pattern in Fig. 5.12 (a). A comparison to the LEED and CLPES analysis
shown in Section 5.1.1 lets us assume to have approximately one monolayer of graphene [137].
The graphene is growing immediately on top of the (3×3) structure without the presence of an
interface layer (cf. Section 5.1.1). It should be recalled that in the LEED pattern rings instead of
spots develop at the rst order diraction angle of graphene, which is an indication for rotational
disorder. Graphene patches are existing in domains of dierent orientations with respect to the
substrate. However, in case of carbon source assisted growth the predominant orientation is
around 0◦ as seen from the magnication in Fig. 5.12 (a). For comparison, Fig. 5.12 (b) displays
the LEED pattern for epitaxial graphene of approximately one monolayer, prepared in UHV by
heating to 1200 ◦ C alone. Here, the predominant graphene orientation is 30◦ with respect to the
substrate (see magnied inset), which is also reported in the literature [155, 212]. The graphene
shown in Fig. 5.12 (b) is mainly growing on top of a carbon rich (2×2)C reconstruction, which
coexists with the (3×3) structure in this temperature regime and does not have any inuence on
the predominant orientation of graphene [137, 155, 212]. Interestingly, our preferential orienta-
tion of around 0◦ coincides with results found for furnace grown epitaxial graphene on SiC(0001̄)
under low vacuum conditions [170]. For epitaxial graphene on SiC(0001̄), no interface layer is
present that could impose a certain rotation angle to the graphene lattice as it is the case for
SiC(0001). Consequently, it may be subtle dierences in the preparation procedure that result
in dierent orientations of the graphene structure with respect to the substrate. Incidentally,
the already mentioned French group reported a predominant rotation of 30◦ with respect to the
substrate [231] and not 0◦ as it is in our case [230].
The method of carbon deposition on SiC crystals opens up new pathways in the investigation
of SiC surface phase diagrams and for low temperature procedures in graphene growth. This tech-
nique, which resembles classical molecular beam epitaxy, has in particular promising prospects for
the epitaxial growth of graphene on 3C-SiC/Si since the Si substrate does not start to evaporate,
which is the case for the elevated temperatures in conventional graphitization (cf. Section 5.1.2).
This could extremely facilitate the implementation in existing Si device technology. However,
we note that the direct evaporation of carbon on top of a Si substrate, which would be the
more appropriate and more appealing way, was not successful in our experiments. The carbon
evaporation onto the clean (7×7) reconstruction of the Si(111) surface followed by subsequent
annealing steps did not form any ordered graphene pattern [129].
5.2.2 Graphitization of SiC(0001) in an argon atmosphere

A signicant improvement in the morphology of graphene surfaces on SiC(0001) requires a better
understanding of the graphene growth mechanism. By means of STM and LEEM, the forma-
tion process of epitaxial graphene in UHV conditions has recently been investigated in more
detail [232, 233]. Hannon et al. suggest from LEEM measurements that the surface √ √roughening
of hydrogen
√ etched samples already occurs in the transformation from the ( 3× 3)R30◦ to
√
the (6 3×6 3)R30◦ reconstruction [232]. The Si atoms are assumed to evaporate mainly at
the bottom of step edges, where the interface layer starts to form. This is accompanied by a
retraction of the steps so that
√ interface
√ layer islands are formed. Due to gaps or discontinuities
in the interface layer, the ( 3× 3)R30◦ steps continue to retract onto the next uphill terrace
resulting in the formation of pits and canyons. In an STM study, Hupalo et al. also come to the
conclusion that the Si evaporation starts at the step edges with the graphene morphology being
determined by dierent retraction speeds of the steps as well as a dierent mobility of carbon
atoms along and perpendicular to steps [233]. It is stated that graphene formation out of samples
with single bilayer steps avoids roughening in comparison to hydrogen etched samples with larger
step heights due to step bunching.
Both Hannon et al. and Hupalo et al. claim that fast high-temperature annealing results
in more homogeneous graphene samples than prolonged heating at slightly lower temperatures,
which gives a hint to a solution to the problem of the low quality of UHV prepared epitaxial
graphene samples. Obviously, the restructuring of the surface has to be completed before the
interface layer and the graphene layers are growing. The mobility of silicon and carbon atoms has
Figure 5.13: Homogeneity of monolayer epitaxial graphene on SiC(0001) for (a) UHV grown epitaxial
graphene on an initially hydrogen etched crystal and (b) epitaxial graphene grown in an induction furnace
in Ar at atmospheric pressure using a non-hydrogen etched CMP polished sample. LEEM data for
samples from atmospheric pressure graphitization with slightly better morphology are shown in Fig. 6.16
of Section 6.2.1.
to be increased, which is fullled by higher annealing temperatures. In order to avoid complete

graphitization of the surface, the evaporation of the Si atoms has to be suppressed. Tromp
and Hannon have shown that this goal can be accomplished by means of an external ux of Si
using the gas disilane [234]. Due to the external Si background pressure of 10−6 mbar, the Si
atoms that evaporate from the surface are reected back to the surface. As a consequence, the
phase transformation temperatures can be shifted up by several hundred degrees. The result is a
strongly increased morphology of the graphene.
In a similar approach, Emtsev et al. and Virojanadara et al. are using Ar as a background
gas [10, 79]. The Ar pressure is at around 1000 mbar. The temperature in the radio frequency
(RF) induction furnace is at around 1500 ◦ C to 2000 ◦ C. This graphitization procedure is techno-
logically well adapted as it does not need any UHV equipment. In addition to the step bunching in
the hydrogen etching process, further step bunching takes place upon graphitization [10]. How-
ever, no pit formation is observed when the terraces retract upon Si out diusion. The interface
layer and graphene is actually starting to grow at the upper edge of a step in contrast to the UHV
preparation technique [10]. Eventually, homogeneous terraces with a width of a few micrometers
and a length of several tens of micrometers are observed, both for the interface layer and mono-
layer graphene [10]. This is a dramatic improvement of the graphene homogeneity in comparison
to the UHV preparation technique with terraces of around 100 nm x 100 nm. Transport meas-
urements indicated higher mobilities of around 2000 cm2 /Vs in comparison to 700 cm2 /Vs for
UHV grown epitaxial graphene [10]. However, the mobility values are still lower in comparison
to graphene akes exfoliated from HOPG or epitaxial graphene on SiC(0001̄). Possible reasons
are the negative electronic inuence of the interface layer (cf. Chapter 4 and Section 6.2). Nev-
ertheless, several groups recently succeeded in measuring the quantum Hall eect on epitaxial
graphene on SiC(0001) [33, 40, 41].
Figure 5.13 demonstrates the tremendously improved surface morphology in our graphene
samples after preparation in an induction furnace under an Ar atmosphere. Whereas panel (a)
of Fig. 5.13 shows an AFM micrograph of epitaxial monolayer graphene obtained by the UHV
preparation technique, panel (b) displays the corresponding surface morphology after atmospheric
pressure graphitization. Note that in this case, an initially hydrogen etched sample was used for
the UHV preparation, whereas an initially CMP polished sample was chosen for the atmospheric
pressure graphitization. This kind of polishing is provided by the companies that sell SiC wafers
("NovaSiC" in this case) and is known to yield atomically at surfaces (cf. Section 3.2), which is a
satisfying starting condition for graphene preparation. Besides some passivation eect, hydrogen
etching generates a more regular array of terraces. At the same time step bunching takes place so
that the graphene terraces are eventually broader and the step height between terraces is much
larger in comparison to graphitized CMP samples [10, 79]. It can therefore be stated that none
of the two starting materials is a priori better, it rather depends on the experimental goal.
In the meantime, we succeeded in slightly improving the furnace process parameters. Cor-
responding LEEM images for initially CMP polished samples can be found in Fig. 6.16 of Sec-
tion 6.2.1. The terrace size of one domain is in the range of some µm2 or some tens of µm2 .
The data shown in that chapter are taken on hydrogen treated samples2 , which will be shown
to produce quasi-freestanding epitaxial graphene without altering the surface morphology. The
combination of atmospheric pressure graphitization and hydrogen intercalation seems to be the
future method of choice for the preparation of epitaxial graphene on SiC(0001).
5.3 Conclusion
Whereas in the previous chapter, only UHV prepared epitaxial graphene on SiC(0001), the Si-
terminated side of the basal plane surfaces of SiC, has been discussed, the main goal of this
chapter was to present dierent ways to grow epitaxial graphene on SiC crystals. Although many
groups focus on one certain recipe for the graphene growth, it is instrumental for future advances
in this research eld to vary the growth conditions in order to fully exploit the potential of this new
material system. The variations of the growth conditions manifests in two main directions: rst,
the variation of the type of SiC substrate, and second, the variations of the growth parameters.
Epitaxial graphene on SiC(0001) has the most promising prospects for a well-ordered and ho-
mogeneous growth of a precisely controlled number of graphene layers. However, the spectacular
physics of epitaxial graphene has rst been realized on SiC(0001̄), the C-terminated counterpart,
and deserves particular attention, too. This chapter provided detailed insight into the initial stages
of UHV grown epitaxial graphene on SiC(0001̄) [137]. Since the phase diagram of SiC(0001̄) is
already dierent from that of SiC(0001), major dierences in the graphitization process could be
expected. The most striking dierence to SiC(0001), actually, is the lack of an interface layer.
The graphene is growing directly on top of the (3×3) and (2×2)C structures, which are intrinsic
surface reconstructions of SiC(0001̄). Due to a much weaker interaction between substrate and
graphene, azimuthal disorder is observed from the beginning of the graphene growth. Stacks of
several graphene layers exhibit a turbostratic growth mode so that the dierent layers have a mu-
tual rotation with respect to each other [57, 170]. As a consequence, the layers are electronically
decoupled, and even multilayer graphene shows the typical properties of monolayer graphene.
Owing to this fact, the graphene thickness control, which is much more dicult in compari-
son to epitaxial graphene on SiC(0001), is not that important. Multilayer epitaxial graphene
2 "Hydrogen treatment" means hydrogen intercalation below epitaxial graphene as discussed in Section 6.2.1
and should not be mixed with "hydrogen etching".
5.3. Conclusion 107
on SiC(0001̄) often consists of several tens of layers. However, it has to be considered that
an external electric eld is screened within just a couple of near-surface layers thus limiting the
potential for electronic applications. Nevertheless, for certain applications and for fundamental
studies, epitaxial graphene on SiC(0001̄) will probably play a major role in the future.
It is interesting to note that, in the present literature, the focus is on the basal plane surfaces
of SiC only. Some publications can be found discussing the graphene growth on samples with a
slight miscut of a few degrees with respect to the (0001) direction (o-axis samples). However, a
systematic study of the graphitization behavior of non-basal plane surfaces is lacking so far. In this
chapter, a short introduction to the UHV-graphitization characteristics of SiC(11̄02), SiC(1̄102̄)
and SiC(112̄0) has been provided. Since the non-basal plane surfaces directly exhibit the long
period of the SiC bulk polytype stacking in the direction of the (0001)-axis, they are promising
candidates for the investigation of graphene with a reduced dimensionality, i.e. epitaxial graphene
nanoribbons. Although the corresponding experiments presented above unfortunately did not
show any ordered graphene growth on a large scale, further investigations could be promising. It
might be that dierent preparation parameters give better results. Alternatively, a microscopic
analysis with STM, for instance, might identify small ordered graphene patches, which then
could be examined in detail. Further research is highly justied, since an understanding of the
graphitization of non-basal plane surfaces could also optimize the graphitization procedures of the
basal plane surfaces. The basal plane surfaces always have a slight miscut in the range of some
tenths of a degree so that the surface has a stepped structure. It has to be considered that the
graphitization characteristics of the basal plane surfaces is governed by the graphitization behavior
of these steps, which are represented by non-basal plane surfaces. Evidently, an understanding of
the graphitization of those surfaces is an essential precondition for improved graphene preparation
procedures on the basal plane surfaces. Apart from this issue, non-basal plane surfaces of SiC
could be important for the synthesis of graphene nanoribbons. Methods to produce graphene
nanoribbons in a reliable way, and on a large scale, are one of the key issues in graphene research
for the next years. The reason is quite obvious, since the band structure of graphene does not
exhibit any band gap. Graphene nanoribbons, on the contrary, can have a band gap, with its size
being dependent on the ribbon width. O-axis samples, in particular, exhibit small terrace widths
that could represent a template for nanoribbons. Since the graphene is growing over terrace
steps, the etching of the graphene by metal nanoparticles along the terrace edges is one possible
way to produce epitaxial graphene nanoribbons on a wafer scale.
Besides the surface orientation, another possible variation of the substrate would consist of
changing the polytype. It has been shown above that structural or electronic dierences cannot
be detected for epitaxial graphene on the polytype 4H in comparison to the polytype 6H [185].
Furthermore, signicant changes of the homogeneity of the graphene surfaces could not be
detected. Note also that the doping of the substrate (weak or strong n-doping, or semi-insulating)
does not have any inuence on the epitaxially grown graphene. Another widely used polytype
is the polytype 3C. Since 3C-SiC is commonly grown on Si wafers, a reliable graphene growth
process has not been presented due to extremely Si-rich growth conditions. However, epitaxial
graphene on 3C-SiC/Si is very appealing as it would, in principle, give access to graphene on Si
wafers and thus tremendously facilitate the implementation of graphene electronics into existing
device technology.
Although the basic structural and electronic properties can perfectly be analyzed with UHV
grown epitaxial graphene on SiC(0001), as discussed in the previous chapter, this graphitization
procedure is not suitable for a homogeneous large scale production of graphene. In UHV condi-
tions, the restructuring of the surface upon annealing is not completed when the interface layer
and the graphene layers are starting to grow. As a consequence, the surface breaks into many
small subterraces with a dierent number of layers. Homogeneous graphene terraces typically
have the size of some hundreds of nm2 . By adopting the recently proposed atmospheric pressure
graphitization technique of Ref. [10], we could solve the homogeneity problems of our UHV pre-
pared samples and now reach a homogeneous graphene coverage on terraces with a size of some
tens of µm2 . The principle of this technique is not dicult to understand. Since the samples
are annealed in an Ar atmosphere, the graphitization temperatures are higher by some hundred
degrees Celsius in comparison to UHV graphitization. The increased mobility of the silicon and
carbon atoms of the surface allows a restructuring of the surface before the graphitization sets
in. Another advantage of this method is the use of technologically well adapted CVD equip-
ment instead of the relatively complicated UHV setup. However, relatively high temperatures of
around 1600 ◦ C are needed, which naturally impose strong boundary conditions to the experi-
mental setup and sample holder system. More generally speaking, the lower the temperature of
a fabrication process is, the less problematic it is. For instance, high temperatures seem to be
counterproductive for the graphene formation on 3C-SiC/Si as mentioned above. Therefore, in
this chapter, a low temperature graphene growth process has been presented, which consists of
carbon evaporation in UHV conditions on the SiC crystals [230], similar to classical molecular
beam epitaxy. For SiC(0001), the conventional UHV graphitization temperature of 1200 ◦ C for
monolayer graphene is reduced by around 250 ◦ C, for SiC(0001̄) graphene is obtained at tem-
peratures of around 100 ◦ C below that of conventional graphitization. Furthermore, the terraces
obtained after initial hydrogen etching of the SiC substrate are preserved, in contrast to con-
ventional UHV preparation. Apart from these promising results in terms of graphitization of SiC
surfaces, this new approach opens up new pathways in the investigation of SiC surface phase
diagrams, since it adds an additional degree of freedom besides the parameters temperature and
amount of deposited Si.
We have learned in this chapter that the graphitization of SiC surfaces can be quite diverse.
Depending on the purpose, dierent preparation procedures have been presented. For fundamen-
tal surface science studies, the UHV preparation is still suitable to some extent. However, for
applications in electronic devices, large scale homogeneous samples prepared from atmospheric
pressure graphitization, are indispensable. With respect to the choice of substrate material, both
faces of the basal plane surfaces are of importance, which, however, address dierent elds of
applications and research. Finally, non-basal plane surfaces could be one promising approach
for alternative graphitization procedures and might play an important role in reliable production
methods for epitaxial graphene nanoribbons.
Chapter 6
Tuning the material properties of
epitaxial graphene on SiC(0001)
While epitaxial graphene on SiC(0001̄) allows only for a poor thickness control and the graphene
grows with rotational disorder (cf. Section 5.1.1), on SiC(0001) a dened number of epitaxially
ordered graphene layers can be grown√(cf. Chapter
√ 4 and Section 5.2). However, for SiC(0001)
the graphene is growing on top of a (6 3×6 3)R30 reconstructed interface layer. This interface
◦
or buer layer is constituted of carbon atoms, which are arranged in a graphene-like honeycomb
structure. As we learned earlier in Section 4.2.2, about 30% of these carbon atoms are bound to
the Si atoms of the SiC(0001) surface, which prevents π -bands with the typical linear dispersion
of graphene to develop in this layer. As a consequence, the interface layer is electronically inactive
in terms of the characteristic graphene properties so that it is called zerolayer graphene. The next
carbon layer, growing on top of the interface layer, then has well developed π -bands. However,
the interface causes an intrinsic electron doping of epitaxial graphene (n ≈ 1013 cm−2 ). It is also
attributed to the inuence of this covalent interlayer bonding that a strongly reduced mobility is
found in epitaxial graphene on SiC(0001) as compared to exfoliated graphene akes. The reason
probably is the introduction of scattering centers into the graphene layer. So, for a practical
application of epitaxial graphene on SiC(0001), it would be desirable to counteract the intrinsic
doping and to reduce the inuence of the interface bonding.
In this chapter, several solutions to these problems are provided. Reduction of the intrinsic
charge carrier density by surface transfer doping can be achieved by the deposition of an elec-
tron acceptor. Whereas atomic layers of Bi and Sb deposited in UHV conditions can partially
reverse the intrinsic n-doping as will be shown in Section 6.1.1, only the deposition of the strong
electron acceptor molecule F4-TCNQ results in charge neutrality, which is preserved in air and is
temperature resistant up to 200 ◦ C (cf. Section 6.1.1).
Surface transfer hole doping, however, only compensates the intrinsic n-doping, and, beyond
that, does not solve the actual problem. To overcome the inuence of the interface layer, a
manipulation on the atomic level would therefore be necessary. The elimination of the covalent
bonding at the interface in order to decouple the epitaxial graphene layers from the SiC substrate
would require to break and saturate the respective bonds. In Section 6.2.1, we demonstrate that
hydrogen intercalation can induce the desired decoupling, so that the outstanding properties of
graphene can be made accessible in quasi-free standing epitaxial graphene layers on large-scale
SiC(0001) wafers suitable for a practical technological application. Besides hydrogen, also the
element gold can be applied for intercalation, which will be briey discussed in Section 6.2.2.
110 Chapter 6. Tuning the material properties of epitaxial graphene on SiC(0001)
Figure 6.1: Position of the Dirac point and Fermi level of monolayer epitaxial graphene on SiC(0001) as
a function of doping. Panel (a) stands for an n-doped monolayer, panel (b) for an as grown monolayer,
panel (c) for a charge neutral monolayer, and panel (d) visualizes truly p-doped monolayer graphene.
6.1 Hole doping of epitaxial graphene

A prerequisite for the development of graphene based electronics is the reliable control of the type
and density of the charge carriers by external (gate) and internal (doping) means. While gating
has been successfully demonstrated for graphene akes [1, 3, 235] and epitaxial graphene on
silicon carbide [181, 236, 237], the development of reliable chemical doping methods, hole doping
in particular, turns out to be a real challenge. In the following two subsections, it is shown that
atomic doping with bismuth or antimony and molecular doping with the strong electron acceptor
tetrauoro-tetracyanoquinodimethane (F4-TCNQ) present eective means of p-type doping. As
a consequence, the intrinsic n-doping of epitaxial graphene on SiC(0001) can be reversed. The
challenge here is to interact with the system just enough to add or remove electrons but not too
much so as to modify or even collapse the electronic structure. Therefore, it is not an option to
replace atoms within the graphene layer, as is common practice when doping silicon.
For graphene the doping is usually realized by adsorbing atoms and/or molecules on its surface,
i.e. surface transfer doping [238240]. For n-type (p-type) doping the electrons have to be released
into (extracted out of) the graphene layer. A schematic sketch that visualizes the band structure
of monolayer epitaxial graphene on SiC(0001) for dierent doping levels is displayed in Fig. 6.1.
Directly after the growth, the graphene is n-doped due to charge transfer from the substrate
(cf. panel (b) of Fig. 6.1). Even more n-type doping can be induced very eectively by alkali
atoms that easily release their valence electron [49, 169]. The Dirac point is shifted further into
the occupied states away from the Fermi level as illustrated in Fig. 6.1 (a). However, aside from
the fact that epitaxial graphene on SiC is already n-doped, alkali atoms are very reactive and
their suitability in electronic devices is more than questionable.
P-type doping for epitaxial graphene is essential to exploit the ambipolar properties of graphene
at conditions around charge neutrality, where EF ≈ ED . This situation is demonstrated in
Fig. 6.1 (c), whereas panel (d) exemplies the true p-type regime in order to round up the picture.
The successful implementation of p-type doping is quite challenging. Many of the elements with a
high electronegativity - e.g. nitrogen, oxygen, or uorine - form strong dimer bonds. They would
not likely form a stable adlayer on the graphene surface. Therefore, dierent molecules such as
NO2 or H2 O have been used to induce p-type doping in graphene [241243]. However, NO2 for
instance is a very reactive chemical and therefore not suitable for use in an electronic material.
Despite these diculties, we have already seen in Fig. 4.15 of Section 4.3.2 that p-type doping
of epitaxial graphene can simply be accomplished by exposing an UHV prepared sample to air.
Adsorbed oxygen, hydrocarbons or water molecules may induce this p-doping, which proceeds
in an uncontrolled way and serves only as a proof of principle. A more promising and reliable
6.1. Hole doping of epitaxial graphene 111
Figure 6.2: Experimental band structure of epitaxial graphene doped with Bi atoms. (a) Clean graphene
layer, (b)-(d) increasing amounts of bismuth atoms have been deposited. The momentum scans are
taken with UV excited ARPES perpendicular to the ΓK-direction of the graphene Brillouin zone.
approach is given by organic molecules [244246]. Many of these molecules possess good thermal
stability, have limited volatility after adsorption, and can be easily applied via wet chemistry. An
eective p-type dopant is the strong electron acceptor F4-TCNQ. It is of great technological
relevance as it plays an important role in optimizing the performance in organic light emitting
diodes [247, 248]. It only would be incompatible with high temperature processes. Details
about the hole doping process for F4-TCNQ on epitaxial graphene are given in Section 6.1.2
and Ref. [51]. However, in the now following section, we start with a viable alternative, being
presented by the heavier elements, which are not as reactive as, e.g., oxygen or uorine. Although
not obvious, because their electron anity is somewhat lower than that of atomic carbon, Bi as
well as Sb turn out to be able to extract electrons out of the graphene sheet. This work is
published in Ref. [171].
6.1.1 Atomic hole doping by means of Bi and Sb

Bi and Sb were deposited on a room temperature sample √ using
√ a commercial electron beam
evaporator, which was calibrated with the help of the ( 3× 3)R30 reconstruction that is
◦
formed for 1/3 monolayer coverage of both Bi and Sb on Ag(111). All measurements were
conducted at room temperature. Fig. 6.2 shows the experimental band structure of epitaxial
graphene doped with successively higher amounts of Bi atoms. Whereas panel (a) shows the
clean graphene layer, panel (b)-(d) display the evolution of the band structure when successively
higher amounts of bismuth atoms per graphene unit cell (u.c.) (as indicated in each panel) are
deposited on the graphene layer. As the bismuth coverage increases, the Dirac point clearly
shifts towards the Fermi level. Otherwise, the band structure remains unaltered by the bismuth
adatoms, i.e. the linear dispersion is preserved. Only at high bismuth coverage, the line width of
the bands increases noticeably. This is probably related to the fact that Bi atoms do not form
an ordered structure but rather tend to form clusters on the surface, which leads to broadening
of the photoemission features. With increasing Bi coverage, the number of free charge carriers
decreases as a successively smaller cross section of the conduction band intersects the Fermi level.
A more detailed analysis of the band widths and the doping behavior is given in Ref. [171].
Very similar results for the p-type doping behavior have been obtained for antimony atoms
deposited on the graphene layer. Antimony is also located in group V of the periodic table
just above bismuth, so that a very similar doping behavior is expected. The experimental band
structure is not shown here, but looks very much like the data obtained for bismuth on graphene
except that it takes a higher antimony coverage to reach the same doping level. In both cases,
Bi and Sb, it is not possible to reach charge neutrality.
6.1.2 Molecular hole doping by means of F4-TCNQ

The deposition of Bi and Sb only partially overcomes the intrinsic n-doping of epitaxial graphene.
The goal is to reach charge neutrality and the samples should be stable in ambient conditions.
In this section, a promising solution to this problem is demonstrated, namely the non-covalent
functionalization of graphene with the strong electron acceptor molecule F4-TCNQ, which has
an electron anity of Eea = 5.24 eV. Thoroughly outgassed F4-TCNQ molecules (7,7,8,8-
Tetracyano-2,3,5,6-tetrauoroquinodimethane, Sigma Aldrich, 97% purity) were deposited on the
graphene/SiC substrate by thermal evaporation from a resistively-heated crucible. Subsequent
ARPES measurements and Raman spectroscopy prove that, due to surface transfer doping, in-
deed charge neutrality (EF = ED ) can be achieved for mono- and bilayer graphene. Epitaxial
bilayer graphene with its band gap at the Dirac point, in particular, is turned from a conducting
system into a truly semiconducting layer. Furthermore, the additional dipole generated at the
graphene surface by the molecular overlayer allows for a controllable engineering of the bilayer
band gap magnitude, which can be increased up to double of its initial value. This could make
the implementation of eld eect devices with a suitable on/o ratio feasible. Further techno-
logical relevance is given by the observation that the F4-TCNQ/graphene complex is stable when
exposed to air and the doping is preserved up to 200 ◦ C. At the end of this section, it is demon-
strated that the doping can even be achieved by application friendly wet chemistry. The results
presented here have been published in Ref. [51]. For a deeper discussion and some experimental
details the reader should refer to this publication.
F4-TCNQ on monolayer graphene

The doping level of the graphene layers is again monitored via ARPES measurements of the π -
band dispersion around the K-point of the graphene Brillouin zone. The thickness of the deposited
molecular layers was estimated from CLPES spectra calibrated through a comparison to spectra for
a well characterized surface phase of TCNQ (7,7,8,8-tetracyanoquinodimethane, Sigma Aldrich,
98% purity) on Cu(100) measured under identical conditions [249]. The starting point is the
as-grown monolayer graphene sample with its band structure shown in Fig. 6.3 (a). The Fermi
level EF is located about 0.42 eV above the Dirac point ED , corresponding to a charge carrier
concentration of n ≈ 1 · 1013 cm−2 (cf. Section 4.3.2). For increasing amounts of deposited
F4-TCNQ, EF moves back towards ED as illustrated in Fig. 6.3 (b)-(d). Meanwhile, the bands
remain sharp, which indicates that the integrity of the graphene layer is preserved. Evidently,
deposition of F4-TCNQ activates electron transfer from graphene towards the molecule thus
neutralizing the excess doping induced by the substrate. When depositing a 0.8 nm thick layer of
molecules, charge neutrality is reached, i.e. EF = ED . For a nominal thickness of the molecular
lm above 0.8 nm, no additional shift of the Fermi energy is observed as seen in Fig. 6.3 (e),
which indicates that the charge transfer saturates.
The two-dimensional π -band dispersion with energy vs. momentum plots and constant energy
maps is shown in Fig. 6.4 for clean monolayer epitaxial graphene and for two dierent deposition
stages of F4-TCNQ. These ARPES measurements were acquired at the Surface and Interface
Figure 6.3: Dispersion of the π -bands measured with UV excited ARPES through the K-point of the
graphene Brillouin zone for (a) an as-grown graphene monolayer on SiC(0001) and (b)-(e) for graphene
on SiC samples covered with an increasing amount of F4-TCNQ molecules. The momentum scans are
taken perpendicular to the ΓK-direction in reciprocal space. The Fermi level EF shifts progressively
towards the Dirac point (ED , dotted black line) with increasing thickness of the deposited F4-TCNQ
lm. Charge neutrality (EF = ED ) is reached for a nominal thickness of the molecular layer of 0.8 nm
(d). When depositing additional molecules, the Fermi level does not shift any further. Hence, the doping
eect saturates for higher coverage (e).
Spectroscopy beamline at the Swiss Light Source using (right) circular polarized light with a
photon energy of 30 eV. An in-situ calibration of the molecular thickness was not possible but the
coverage can be estimated from the results shown in Fig. 6.3. The dispersion scans in Fig. 6.4 (a)
were measured parallel to the ΓK-direction at a sample temperature of 80 K. They have the same
appearance as shown in Section 4.3.2. The slight asymmetry in the π -bands is expected in this
measurement direction and just reects the trigonal warping of graphene's band structure shown
previously. The Dirac point shifts from around 400 meV to 150 meV for the rst deposition stage
and close to the Fermi level for the second deposition stage. The constant energy maps shown in
panel (b)-(e) have exactly the appearance expected for circular polarized light (cf. Section 4.3.2).
Their shape and appearance, in particular their intensity distribution, does not change upon
molecule deposition. The position of the Dirac point can clearly be followed, too. The constant
energy maps above and below the Dirac point are inverted with respect to each other due to the
chiral nature of the charge carriers (see Section 4.3.2). The Fermi surfaces shown in panel (e) give
direct access to the charge concentration (cf. Equation 2.8 of Section 2.1.1). Assuming a circular
Fermi surface we get n ≈ 7.3 · 1012 cm−2 for clean monolayer graphene, n ≈ 9 · 1011 cm−2 for
the rst deposition stage and n ≈ 1.5 · 1011 cm−2 for the second deposition stage.
Characterization of the charge transfer complex

The location of the charge transfer process within the F4-TCNQ molecule can be elucidated
with the help of CLPES. N 1s and F 1s core level emission spectra for dierent amounts of
deposited F4-TCNQ are displayed in Fig. 6.5. For the N 1s spectra of panel (a), a line shape
analysis after the subtraction of a Shirley background reveals two main components centered at
binding energies (BE) of 398.3 and 399.6 eV, respectively. This indicates that dierent N species
exist in the deposited molecular lm. In agreement with the literature [250, 251], the peak at
398.3 eV is assigned to the anionic species N−1 while the 399.6 eV component is attributed to the
114
Figure 6.4: (a) π -band dispersion for clean monolayer graphene and monolayer graphene after two F4-TCNQ deposition levels. The measurement
direction is parallel to the ΓK-direction. (b)-(e) Constant energy maps for dierent binding energies for clean monolayer epitaxial graphene and for
graphene after two F4-TCNQ deposition levels. For all data, (right) circular polarized light with a photon energy of 30 eV was used.
Chapter 6. Tuning the material properties of epitaxial graphene on SiC(0001)
Figure 6.5: (a,b) CLPES spectra of the N 1s (a) and F 1s (b) core level emission regions from sub-
monolayer (bottom spectrum) to multilayer (top spectrum) amounts of F4-TCNQ deposited on epitaxial
monolayer graphene. Three dierent components are tted into the N 1s region by a line shape analysis
corresponding to N−1 , N0 species and a shake-up related feature. The blue dashed line indicates the
exact energy position of the N−1 component as it shifts with molecular layer thickness. (c) Schematic
structure of a F4-TCNQ layer deposited on top of a graphene layer grown on SiC. The charges induced
in the graphene layer due to the interface dipole and the molecular charge transfer are indicated.
neutral N0 species. The additional broad component at 401.7 eV likely originates from shake-up
processes considering its energy location and the relative intensity of ≈ 20% as compared to
the main peak [252]. The F 1s spectra in Fig. 6.5 (b) are in contrast dominated by a single
component. Only at low coverages, a weak asymmetry develops. The appearance of the N−1
anion species indicates that the electron transfer takes place through the C≡N groups of the
molecules while the uorine atoms are largely inactive. A similar mechanism with electronically
active cyano groups has been found for F4-TCNQ on other surfaces [252254]. However, in
the present case, not all C≡N groups are involved in the charge transfer process. While for low
coverages the N−1 species dominate (71%), for molecular lm thicknesses from 0.4 nm to 0.8 nm
about 45% of the C≡N groups are uncharged (N0 ) as determined from the peak areas of the
tted components. This indicates that, when the lms are densely packed, most of the molecules
are standing upright as sketched in Fig. 6.5 (c) (apparently, in dilute layers not all molecules
are arranged perpendicular to the surface). The energy position of the dierent core level peaks
shifts with increasing molecular coverage as indicated by the blue dashed line in Fig. 6.5 (a). For
0.8 nm lm thickness this shift is exactly the same as the shift of the π -bands with respect to
the Fermi energy EF (i.e. 0.4 eV for saturation) in agreement with our working hypothesis of a
strong electronic coupling between the F4-TCNQ molecule and the graphene surface. At lm
thicknesses larger than 0.8 nm, the shift of both the N−1 peak and the band structure saturates.
Only the N0 peak continues to grow, indicating the formation of a charge neutral second layer
of molecules. The saturation eect at 0.8 nm lm thickness also supports the model of a dense
layer of upright standing molecules since the size of an F4-TCNQ molecule along its axis is indeed
about 0.8 nm.
A comparison of the experimental band shifts when using the non-uorinated version of the
F4-TCNQ molecule, i.e. TCNQ, shows that the charge transfer is greatly enhanced when the
Figure 6.6: Dispersion of the π -bands measured via UV excited ARPES through the K-point of the
graphene Brillouin zone for (a) pristine epitaxial monolayer graphene on SiC(0001) and (b)-(d) after
increasing deposition of TCNQ, the non-uorinated version of F4-TCNQ. The Fermi level (EF ) shifts
progressively towards the Dirac point (ED , black dotted line) for increasing thickness up to a value of
ED − EF = - 0.25 eV.
F species are present, even though they are not directly involved in the charge transfer process.
With TCNQ, which has a much smaller electron anity than F4-TCNQ (i.e., 2.8 eV for TCNQ
compared to 5.24 eV for F4-TCNQ), the Fermi energy remains at least 0.25 eV above the Dirac
point (see Fig. 6.6). The maximum shift of the band structure measured upon TCNQ deposition
is obtained for a 0.4 nm thick deposited molecular layer (Fig. 6.6 (d)), and no additional shift is
observed for higher amounts of deposited molecules.
Raman spectroscopy analysis

The inuence of the F4-TCNQ coverage on graphene can also be studied with Raman spec-
troscopy. Raman spectra were measured in the same setup as in Section 4.3.6, i.e. under ambient
conditions using an Argon ion laser with a wavelength of 488 nm at a power level of 12 mW
and a laser spot size of ≈ 1 µm in diameter. Figure 6.7 (a) compares Raman spectra for an
as-grown epitaxial layer of graphene (bottom trace) and for a sample that has been covered with
a 1.5 nm thick F4-TCNQ layer (top trace). Peaks related to the SiC substrate are marked by
arrows. The 2D peak of graphene is highlighted with gray shading. So is the G peak, which again
is hidden in the strong contributions of the SiC substrate (cf. Section 4.3.6 and Ref. [184]). The
Raman spectrum for graphene, covered with F4-TCNQ, reveals numerous additional features
that are marked by stars. By illuminating a sample that is covered with F4-TCNQ molecules
with the laser light, it is possible to gradually remove the deposited molecules through evapora-
tion. Features associated with the SiC substrate and graphene hardly change, while the peaks
attributed to the F4-TCNQ molecules decrease in amplitude. Laser heating can therefore be
used to trim the molecule coverage and hence tune the charge carrier concentration in graphene.
As in Section 4.3.6 the large contributions of the SiC substrate are eliminated by analyzing dif-
ferential spectra, obtained by subtracting the Raman data of the untreated but clean SiC from
the spectrum of the F4-TCNQ-modied graphene layer on top of SiC. The evolution of the G
peak upon successive laser illumination, i.e. for varying F4-TCNQ layer thickness, is illustrated in
Fig. 6.7 (b). The spectra can be decomposed into three peaks. Two molecular peaks at ≈ 1602
Figure 6.7: (a) Raman spectrum of pristine (bottom curve) and F4-TCNQ-modied (top curve) epitaxial
graphene. Molecular peaks are marked with stars and SiC peaks with arrows. The G peak and 2D peak
regions of graphene are shaded. (b) In order to bring out contributions of graphene and the F4-TCNQ
molecule to the Raman spectrum, dierential spectra are plotted. They are obtained by subtracting
the Raman spectrum of the clean SiC substrate (after hydrogen etching). The top trace displays the
spectrum obtained on an epitaxial graphene sample with a 1.5 nm thick layer of F4-TCNQ molecules on
top. The other dierential spectra were obtained after the same sample was exposed to an increasing
amount of laser light. Curves are labelled with the thickness of the molecular lm estimated from the
strength of the molecular peaks. The gray dashed lines are Lorentzians to t the molecular peaks. The
blue solid line is the extracted graphene contribution to the Raman spectrum (G peak).
and ≈ 1637 cm−1 decrease with the laser exposure. The molecular coverage before laser exposure
was calibrated with CLPES (top curve in panel (b)). The other molecular coverages marked in
Fig. 4.30 (b) are calculated from the relative intensity of the molecular peaks. The intensity of the
remaining peak, which we attribute to the G phonons of graphene, is approximately constant and
not inuenced by laser exposure. The peak position shifts however from ≈ 1583 to ≈ 1591 cm−1 .
As discussed in Section 4.3.6, the G peak position in our epitaxial graphene samples correlates
with the carrier concentration. The G peak position of the F4-TCNQ saturated sample (1583.3
± 0.9 cm−1 ) is nearly the same as for charge neutral graphene akes (cf. Section 4.3.6 and
Ref. [205, 207]), which is consistent with the ARPES data. As the molecules are successively
removed, the G peak is blue-shifted and nally reaches 1591 cm−1 , the value for clean monolayer
graphene on SiC exposed to air, which corresponds to a charge concentration of ≈ 5 · 1012 cm−2
or a band shift of EF − ED ≈ 0.3 eV (cf. Fig. 4.15 of Section 4.3.2 and Section 4.3.6). We note
again that this value is less than measured by ARPES (EF − ED = 0.4 eV) due to the additional
doping under ambient conditions. Finally, it should be remarked again that the interpretation of
our G peak data implies that the G peak position seems to be hardly inuenced from strain and
is rather dominated by changes in the charge carrier concentration (see also Section 4.3.6).
F4-TCNQ on bilayer graphene

For as-grown bilayer graphene on SiC, the intrinsic band shift is only about 0.3 eV. In addition,
the electric dipole present at the graphene/SiC interface imposes an electrostatic asymmetry in
the bilayer stack which causes a band gap to open of roughly 0.1 eV as shown previously in
Section 4.3.2. Figure 6.8 (a) displays the band structure of pristine bilayer graphene. In the
gure, simulated bands based on tight-binding calculations are superimposed to the dispersion
plot. This facilitates an analytical evaluation of the Dirac energy position and the size of the band
gap. The calculations are based on a symmetric bilayer Hamiltonian as described by McCann and
Fal'ko (cf. Section 2.1.1 and Ref. [48]). We note that, due to the inevitable inhomogeneity of
UHV prepared graphene samples and the nite size of the beam spot, the ARPES data contain
contributions of lm areas with dierent thickness. This can be seen by a comparison with data
from a sample prepared at a slightly lower temperature in Fig. 6.8 (f). Here, the contribution
from monolayer patches is notably stronger and obstructs a clear view on the bilayer bands.
Therefore, the sample annealed at higher temperatures was used for panel (a), where the bilayer
bands are well isolated, although trilayer contributions are clearly present. The sketch in panel (g)
identies the band contributions stemming from dierent graphene thicknesses. Similar to the
monolayer case, F4-TCNQ deposition onto this sample causes a progressive shift of the bilayer
bands, i.e. a reduction of the intrinsic n-type doping. This is illustrated in the dispersion plots
in Fig. 6.8 (b)-(e). Concurrent with the drop of EF − ED , the size of the band gap increases as
seen from the bands tted with the tight binding simulations. The evolution of the characteristic
energies of these tted bands with the amount of deposited molecules is plotted in Fig. 6.8 (h).
Shown are the energy gap Eg , the mid gap or Dirac energy ED , the energy at the bottom of the
lowest conduction band Econd and the top of the uppermost valence band Eval . These energies
are marked in panel (c). The band gap Eg increases from 116 meV for a clean as-grown bilayer
to 275 meV when a 1.5 nm thick layer of F4-TCNQ molecules has been deposited. It has been
veried that no further charge transfer occurs for higher amounts of deposited molecules. The
Fermi energy moves into the band gap for a molecular layer thickness of 0.4 nm. Hence, the
bilayer is turned from a conducting system into a truly semiconducting layer. The increase of
the band gap indicates that the molecular deposition increases the on-site Coulomb potential
Figure 6.8: UV excited ARPES band structure plots measured perpendicular to the ΓK-direction for an
epitaxial graphene bilayer on SiC(0001), (a) sample without F4-TCNQ coverage and (b)-(e) samples with
increasing F4-TCNQ deposition. Bands calculated within a tight binding model are superimposed to the
experimental data. (f) ARPES data showing the band structure of an epitaxial graphene bilayer prepared
at a lower annealing temperature. Contributions from monolayer domains are evident. (g) Schematic
band structure of mono-, bi-, and trilayer epitaxial graphene. (h) Evolution of the energy gap Eg , the
gap midpoint or Dirac point ED , the minimum of the lowest conduction band Econd and the maximum
of the uppermost valence band Eval as a function of molecular coverage. The denition of these energies
is indicated in panel (c).
Figure 6.9: Comparison of the band structure, measured parallel to the ΓK-direction, for an incomplete
bilayer sample before (a) and after (b) F4-TCNQ deposition, and (c) for a monolayer of graphene after
F4-TCNQ deposition. For these measurements, (right) circular polarized light with a photon energy of
40 eV was used. The insets of panels (a) and (b) depict schematically the contributions of the monolayer
patches (blue lines) and the bilayer patches (red lines) to the ARPES band diagram.
dierence between both layers. From the tight binding calculations, we get an increase in the
on-site Coulomb interaction from 0.12 eV for a clean bilayer to 0.29 eV for a bilayer covered with
a 1.5 nm thick molecular lm1 . This increase can be attributed to an increased electrostatic eld
due to the additional dipole developing at the graphene/F4-TCNQ interface. The tunability of the
band gap due to the deposition of a strong electron acceptor corresponds to a chemical gating
technique. For graphene akes, it has been demonstrated that a dual-gate bilayer eld-eect
transistor allows to tune the band gap in a similar way as presented in this work [53].
Synchrotron ARPES data taken at the Surface and Interface Spectroscopy beamline at the
Swiss Light Source conrm that hole doping with F4-TCNQ requires higher coverages on bilayer
than on monolayer samples by comparing data obtained on an as-grown and a molecule covered
sample with an average graphene thickness between one and two layers. These data are depicted
in Fig. 6.9 (a) and (b). As indicated in the schematic band diagram in the insets, the band
originating from the monolayer regions (blue lines) is shifted more than the bands originating
from the bilayer parts (red lines) of the sample. The molecule deposition also appears to inuence
the intensity distribution within an individual band. For example, near the minimum of the upper
band, the intensity increases signicantly when molecules are deposited as also sketched in the
inset. Note, that this intensity enhancement cannot be seen in pure monolayer data as shown in
Fig. 6.9 (c).
Technological relevance
An important aspect of the F4-TCNQ/graphene system is the robustness of its preparation: the
Raman experiments after transport through ambient environment already demonstrated that the
charge transfer complex is stable in air. On a monolayer sample covered with a multilayer of
F4-TCNQ molecules, the band structure was measured with ARPES before and after several
hours of air exposure. This experiment revealed no change in the band structure. CLPES
1 The dimer coupling γ1 varies from 0.40 eV for the clean bilayer to 0.52 eV for a bilayer with molecular lm
coverage of 1.5 nm. The next-nearest neighbour coupling γ3 remains xed at 0.12 eV. The Fermi velocity vF
equals 1.07 x 106 m/s.
Figure 6.10: (a) Shift of the Fermi level EF with respect to the Dirac energy ED vs. annealing temperature.
The shift was determined from ARPES data recorded after each annealing step in UHV of the graphene
sample with an initial F4-TCNQ lm thickness of 1.5 nm. (b) π -band dispersion after F4-TCNQ wet
chemical application. Charge neutrality is achieved, as indicated by EF = ED .
measurements also conrmed the inert nature of the graphene substrate. The experiment with
laser light exposure suggests that the F4-TCNQ layer is sensitive to temperature. The volatility
of F4-TCNQ was probed by stepwise annealing in UHV a sample with an F4-TCNQ molecular
lm on top. The sample was annealed for a time t = 1 min at successively higher temperatures
between 25 ◦ C to 230 ◦ C. After each annealing step, the shift of the Fermi level EF with respect
to the Dirac energy ED was determined from ARPES spectra recorded at room temperature.
As the annealing temperature increased, the dierence between the Dirac energy and the Fermi
energy increased back to the value of a pristine graphene layer. This increase is direct evidence for
molecular desorption from the graphene surface. As is evident from Fig. 6.10 (a), desorption of
the molecules is initiated at temperatures around 75 ◦ C and completed at 230 ◦ C. Since thermal
desorption is amplied by UHV conditions, we anticipate that even higher temperatures are needed
under atmospheric pressure to remove the molecular layer.
Finally, we show that the F4-TCNQ layer can also be applied by immersing the sample in a
chemical F4-TCNQ solution. In order to apply the molecular layer on graphene via wet chemistry,
F4-TCNQ (Sigma Aldrich, 97% purity) was rst dissolved in dimethyl sulfoxide (DMSO) until
saturation. Then the sample was left immersed in the solution for 12 hours. ARPES spectra taken
immediately after introduction into UHV show a considerable background due to contamination
by residual chemicals from the solution as displayed in Fig. 6.10 (b). Nevertheless, it is well
apparent that a condition of charge neutrality is achieved (i.e., EF = ED ).
√ √
Figure 6.11: (a)-(d) Side view models for (a) the (6 3×6 3)R30◦ reconstruction of SiC(0001) (zero-
layer graphene) and (b) epitaxial monolayer graphene. After hydrogen intercalation, (c) the zerolayer
and (d) monolayer graphene are decoupled from the substrate. (e) Schematic setup for the hydrogen
intercalation process.
6.2 Decoupling epitaxial graphene from the SiC sub-

strate
Surface transfer hole doping of epitaxial graphene on SiC(0001), as presented in the previous
section, is a suitable and technologically well adapted approach to overcome the intrinsic n-
doping√of the√graphene layers. However, the actual problem, which is given by the presence of
the (6 3×6 3)R30◦ reconstructed interface layer between SiC and graphene, is not solved. A
solution to this problem is demonstrated in Section 6.2.1 by the method of hydrogen intercalation.
Hydrogen decouples the epitaxial graphene from the substrate so that the outstanding properties
of graphene can be made accessible in quasi-free standing epitaxial graphene layers on large-scale
SiC(0001) wafers suitable for a practical technological application. Decoupling is also possible
with Au, as will briey be introduced in Section 6.2.2. In that case, however, the role of the
intercalated Au is not a priori clear and more dicult to analyze.
6.2.1 Hydrogen intercalation below epitaxial graphene

For a better understanding of the potential of this new technique, which is proposed in this
section and published in Ref. [160, 255], Fig. 6.11 (a)-(d) explains √
the main
√ principle of hydrogen
intercalation below epitaxial graphene. Panel (a) shows the (6 3×6 3)R30◦ reconstructed
interface layer (zerolayer graphene) as a partially covalently bound initial carbon layer. The rst
carbon layer on top of this buer layer is displayed in panel (b) of this gure and exhibits the
typical properties of monolayer graphene. Upon hydrogen treatment of zerolayer and monolayer
graphene, the hydrogen migrates below the interface layer, breaks the covalent bonds between
carbon and silicon, and subsequently saturates the Si atoms of the Si-terminated SiC substrate.
This passivation decouples the graphene from the substrate and allows to get exclusive access
to the exceptional properties of quasi-freestanding mono- and bilayer epitaxial graphene in an
elegant, eective, and processing friendly method.
Although the principle of hydrogen intercalation is not dicult to understand, the question
arises how to realize the experimental goal. Recently, several studies investigated the interaction
between hydrogen and graphene [37, 256258], mainly in terms of graphene hydrogenation and
6.2. Decoupling epitaxial graphene from the SiC substrate 123
√ √
Figure 6.12: (a) and (b) LEED
√ patterns
√ at 126 eV for (a) the (6 3×6 3)R30◦ reconstruction (zerolayer
graphene) and (b) the (6 3×6 3)R30 reconstruction after hydrogen intercalation. (c) and (d) LEED
◦
patterns at 126 eV for (c) around monolayer epitaxial graphene and (d) monolayer graphene after hydrogen
intercalation. In all LEED patterns, the rst order diraction spots are indicated for SiC and graphene.
the formation of graphane, which is an insulating graphene's derivative [37]. Whereas these
experiments using atomic hydrogen at room temperature did not result in any intercalation
below graphene, we achieved success by adopting the hydrogen etching [124, 125] and passiva-
tion [99, 259, 260] process for SiC surfaces. Our graphene samples were transported through
air from the graphitization setup (UHV or atmospheric pressure graphitization) to a chemical
vapor deposition reactor for the hydrogen treatment. The corresponding experimental setup is
shown in Fig. 6.11 (e). The samples were annealed for 10 min at temperatures between 600 ◦ C
and 1000 ◦ C in ultra-pure molecular hydrogen at atmospheric pressures. The best temperature
for zerolayer graphene turned out to be around 600 ◦ C, for monolayer graphene around 750 ◦ C2 .
Before the gas inlet, the hydrogen was cleaned in a palladium diusion cell. A more detailed
experimental description is given in Section 3.3.
Geometrical decoupling from LEED

Let us now have a detailed look at the experimental data that conrm the decoupling of the
epitaxial graphene samples
√ from
√ the◦ substrate. Figure 6.12 (a) and (b) display LEED patterns
at 126 eV for the (6 3×6 3)R30 reconstructed interface layer (i.e. zerolayer) before and
after hydrogen treatment, respectively. For the pristine zerolayer√graphene
√ sample, the LEED
pattern shows a very weak graphene diraction spot but intense (6 3×6 3)R30 superstructure
◦
spots (panel (a)), corresponding to the pronounced atomic displacements in this layer due to
the covalent bonding to the SiC substrate (cf. Section 4.2). After hydrogen treatment, the
superstructure spots are strongly suppressed as depicted in panel (b), indicating much smaller
atomic displacements in the reconstructed layers, which in turn suggests the absence or weakening
of the interlayer bonding. The same observation can be made for a sample with a graphene
coverage of slightly above one monolayer as can√ be √seen in panel (c) and (d) of Figure 6.12. In
that case, after hydrogen intercalation, the (6 3×6 3)R30◦ superstructure spots are practically
not visible any more. These results are already a clear indication of a geometrical decoupling
of the interface layer from the substrate. Interestingly, for the zerolayer sample in particular,
2 Note that the temperature is measured with a thermocouple at the backside of the graphite susceptor so that
the real temperature of the samples might be higher.
Figure 6.13: Dispersion of the π -bands measured with UV excited ARPES perpendicular to the ΓK-
direction of the graphene Brillouin zone for (a) an as-grown graphene zerolayer (ZL) on SiC(0001),
(b) after hydrogen treatment and (c)-(e) subsequent annealing steps; (f) for an as-grown monolayer (ML),
(g) after hydrogen treatment and (h)-(j) subsequent annealing steps.
√ √
the (6 3×6 3)R30◦ superstructure spots have not completely vanished. Although a denite
explanation cannot be given at the moment, √one possible
√ reason could be partial hydrogenation.
However, it should be considered √ the (6◦ 3×6 3)R30 diraction spots do not necessarily
√ that
◦
indicate the existence of a (6 3×6 3)R30 reconstruction in the graphene layer but can arise
alone from the combined periodicity of the graphene and substrate layers (double diraction).
Electronic decoupling from ARPES

Apart from this structural aspect, the hydrogen treatment has a dramatic eect on the electronic
structure of the samples. This is shown in Fig. 6.13 by ARPES measurements of the valence band
structure around the K-point of the graphene Brillouin zone. As already noted in Section 4.2.2
for a pristine zerolayer, no π -bands are observed (cf. panel (a)). Only two very faint delocalized
and smeared out states at binding energies of around 0.1 eV to 0.5 eV and higher than 0.9 eV are
visible. After hydrogen treatment, quite dierently, the linearly dispersing π -bands of monolayer
graphene appear, see panel (b). In addition, while as-grown monolayer graphene is n-doped, as
mentioned before and shown in panel (f) so that its Fermi level EF is located about 420 meV
above the Dirac point, ED , this eect is absent after hydrogen treatment of the zerolayer sample.
Even more, the sample is slightly p-doped so that EF is shifted below ED by ≈ 100 meV. The
appearance of the graphene type π -bands and the absence of n-doping corroborates our working
hypothesis that the covalently bound carbon layer is decoupled from the substrate. Apparently,
the hydrogen atoms migrate under this layer, break the bonds between C and Si and bind to the Si
atoms as previously sketched in Fig. 6.11 (a) and (c). Correspondingly, the buer layer is lifted and
displays the electronic properties of a quasi-free standing graphene monolayer. It should be noted
that the hydrogen treated sample was outgassed at 400 ◦ C before acquiring the spectrum shown
in Fig. 6.13 (b) in order to reduce the background. This process has no eect on the structure and
position of the π -bands. Yet, after heating the sample up to 700 ◦ C (panel (c)) the slight p-doping
practically vanishes and conditions around charge neutrality (EF = ED ) are retrieved, so that we
tentatively attribute the p-doping eect to the presence of chemisorbed species on the graphene
surface and the subsequent downshift of the band structure to their desorption [184]. From
recent synchrotron data, measured at the Surface and Interface Spectroscopy beamline at the
Swiss Light Source, we get a hole carrier concentration of around 1-3 · 1011 cm−2 , corresponding
to EF − ED ≈ 40-70 meV [261]. At temperatures above 700 ◦ C the π -bands progressively weaken
as indicated in panel (d). Since Si-H bonds are known to break at temperatures just above
700 ◦ C [262], this eect can be correlated to progressive hydrogen desorption. Around 900 ◦ C,
the hydrogen has completely desorbed and the zerolayer structure is re-established as seen from
the absence of π -bands (Fig. 6.13 (e)) and also from the LEED pattern which is similar again to
the one shown in Fig. 6.12 (a).
Consistent results were observed for monolayer epitaxial graphene. The corresponding band
structure measured by ARPES before hydrogen treatment and after hydrogen treatment plus
subsequent outgassing at 400 ◦ C is shown in Fig. 6.13 (f) and (g). The intrinsic doping of the
pristine monolayer vanishes completely and the monolayer turns into bilayer graphene as previously
sketched in Fig. 6.11 (b) and (d). Again, the hydrogen treated sample shows a slight p-doping,
which disappears after annealing to 700 ◦ C (panel (h)). From recent synchrotron data, measured
at the the Swiss Light Source, together with tight binding ts, we get a hole carrier concentration
in the range of 8 · 1011 cm−2 to 1.8 · 1012 cm−2 , corresponding to EF − ED ≈ 25 - 35 meV [261].3
For temperatures higher than 700 ◦ C, the intensity of the bilayer π -bands decreases while the
monolayer bands reappear (Fig. 6.13 (i)). The hydrogen progressively desorbs until at 1000 ◦ C
the original monolayer band structure is completely recovered (panel (j)).
Structural decoupling from CLPES

The structural arrangements and atomic bonding congurations can be precisely determined by
monitoring the core level spectra of hydrogen intercalated samples as prepared and in the course
of progressive hydrogen desorption by annealing. Since the resulting picture is very similar for
zerolayer and monolayer graphene, only the monolayer results will be discussed. Si 2p and C 1s
core level spectra were measured for the hydrogen-treated monolayer sample annealed at dierent
temperatures, as depicted in Fig. 6.14. Dierent components contributing to the spectra were
decomposed by a curve tting procedure. The depth position of the corresponding species within
the surface was identied by varying the incident photon energy and thus changing the surface
sensitivity. The energies shown in Fig. 6.14 (a) and (c) are 450 eV and 140 eV for the C 1s and
Si 2p spectra, respectively; the energies shown in Fig. 6.14 (b) and (d) are 600 eV and 330 eV for
the C 1s and Si 2p spectra, respectively, corresponding to a lower surface sensitivity in comparison
to panel (a) and (c). The experimental data points are displayed in black dots. The gray solid
line is the envelope of the tted components. The tting parameters are summarized in Table 6.1
and 6.2. Beside the C 1s graphene related peak, which was realized through a Doniach-Sunjic
prole to account for the metallic behavior, all the other tted lines are an approximation of the
Voigt function (i.e., obtained as a convolution of Lorentzian and Gaussian line shapes). Prior to
3 According to these data, the band gap is practically closed (≈ 30 meV in size) since n-doping from the
substrate is not present. In other words, the on-site Coulomb potential dierence between the lower and upper
layer in pristine epitaxial bilayer graphene seems to be absent in quasi-freestanding epitaxial bilayer graphene.
126
Figure 6.14: (a),(b) C 1s and (c),(d) Si 2p spectra for a hydrogen treated monolayer graphene sample (bottom spectra) and the same sample annealed
at increasing temperatures. The experimental data are displayed in black dots. Dierent components, accordingly labeled in the spectra, are tted into
the C 1s and Si 2p regions by a line shape analysis. To distinguish between surface and bulk related peaks, dierent incident photon energies have been
used (450 eV and 600 eV for C 1s, 140 eV and 330 eV for Si 2p). The gray solid line is the envelope of the tted components. The tting parameters
are summarized in Table 6.1 and Table 6.2. Each spectrum is normalized to the corresponding integrated spectrum at 600 ◦ C.
C 1s SiC Graphene Graphene S1 (interface) S2 (interface)
450 eV (lifted BL) (ML)
T EB FWHM Area EB FWHM Area EB FWHM Area EB FWHM Area EB FWHM Area
( ◦ C) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%)
600 283.03 0.53 14 284.61 0.29 86 - - - - - - - - -
730 283.18 0.60 17 284.70 0.29 70 284.93 0.34 6 284.89 0.80 2 285.30 1.10 4
800 283.61 0.73 17 284.85 0.28 40 285.00 0.42 23 285.10 0.80 6.5 285.70 1.10 13
830 283.84 0.77 20 284.85 0.28 15 285.00 0.38 26 285.14 0.80 13 285.74 1.10 26
1060 284.00 0.74 25 - - - 285.00 0.42 27 285.29 0.80 16.5 285.84 1.00 31
C 1s SiC Graphene Graphene S1 (interface) S2 (interface)

600 eV (lifted BL) (ML)
T EB FWHM Area EB FWHM Area EB FWHM Area EB FWHM Area EB FWHM Area
( ◦ C) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%)
6.2. Decoupling epitaxial graphene from the SiC substrate
600 283.03 0.56 24 284.60 0.37 76 - - - - - - - - -

730 283.20 0.61 26 284.71 0.35 64 285.00 0.34 4 284.85 0.80 2 285.45 1.20 4
800 283.68 0.79 27 284.88 0.33 36 285.09 0.47 19 285.19 1.05 6 285.84 1.00 11
830 283.84 0.77 27 284.87 0.30 13 285.07 0.42 25 285.00 0.95 12 285.77 1.00 23
1060 283.96 0.72 32 - - - 285.00 0.49 24 285.33 0.80 15 285.84 1.00 29
Table 6.1: Fit parameters for the C 1s level spectra for hydrogen treated monolayer epitaxial graphene after outgassing to the specied temperature
values. The incident photon energy was 450 eV and 600 eV, respectively. Before curve tting, a Shirley background was subtracted from the raw data.
The C 1s graphene related peak was tted using a Doniach-Sunjic prole to account for the metallic behavior (asymmetry parameter 0.04, Gaussian
contribution 56%), all the other tted lines are an approximation of the Voigt function (Gaussian contribution 30% for C 1s).
127
128
√
Si 2p SiC Si-H Si to 6 3 defect
140 eV
T EB FWHM Area EB FWHM Area EB FWHM Area EB FWHM Area
( ◦ C) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%)
600 100.65 0.45 63 100.94 0.56 37 - - - - - -
730 100.72 0.50 49 100.98 0.55 37 101.56 0.8 11 100.62 0.80 3
800 101.61 0.65 55 101.08 0.51 33 102.02 1.05 18 100.90 0.80 4
830 101.64 0.62 58 101.12 0.52 19 102.17 1.00 18 101.00 0.80 5
1060 101.65 0.59 69 - - - 102.22 1.00 22 100.98 0.89 9
√
Si 2p SiC Si-H Si to 6 3 defect
330 eV
T EB FWHM Area EB FWHM Area EB FWHM Area EB FWHM Area
( ◦ C) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%) (eV) (eV) (%)
600 100.67 0.45 67 100.93 0.54 33 - - - - - -
730 100.79 0.50 55 100.98 0.55 35 101.52 0.9 8 100.62 0.75 2
800 101.54 0.64 57 101.09 0.46 30 102.92 1.15 10 100.95 0.75 3
830 101.61 0.55 67 101.17 0.44 17 102.08 1.17 11 101.00 0.80 5
1060 101.65 0.53 82 - - - 102.14 1.18 11 101.08 1.05 7
Table 6.2: Fit parameters for the Si 2p core level spectra for hydrogen treated monolayer epitaxial graphene after outgassing to the specied temperature
values. The incident photon energy was 140 eV and 330 eV, respectively. Before curve tting, a linear background was subtracted from the raw data.
All tted lines are an approximation of the Voigt function (Gaussian contribution 70% for bulk Si 2p and 40% for the surface components of Si 2p).
curve tting, a linear background was subtracted from the Si 2p data and a Shirley background
from the C 1s data. The bottom curves in Fig. 6.14 (a) and (b) show the C 1s spectrum
measured after outgassing the hydrogen-treated sample at a temperature around 600 ◦ C. The
dominant peak at 284.6 eV is the graphene related component (red line) while the broader, less
intense peak at 283.0 eV is the SiC (bulk) related component (dark blue line). Clearly, for a
higher incident photon energy, the bulk component exhibits a higher √ intensity.
√ The most striking
feature is the complete absence of components related to the (6 3×6 3)R30◦ reconstruction
(cf. Section 4.2.2), which clearly identies the surface as a quasi-free standing epitaxial graphene
layer (bilayer in this case). No Si-C covalent bonds are present at the interface. The possible
presence of hydrocarbon contamination, which might cause the slight p-type doping observed in
the ARPES data of hydrogen-treated samples (see above), is hard to detect since hydrocarbons
also have a binding energy of 284.6 eV. At annealing temperatures higher than 700 ◦ C, the
hydrogen starts to desorb, as indicated by the appearance of the interface components S1 and
S2, marked with light blue lines. As for the graphene component, their intensity is slightly lower
for higher
√ incident
√ photon energy. The components S1 and S2 result from the carbon atoms in
the (6 3×6 3)R30◦ structure (cf. Section 4.2.2): S1 results from the carbon atoms in the buer
layer bound to one Si atom of the SiC(0001) surface and to three C atoms in the sp 2 -bonded layer;
S2 is the component emitted from the remaining sp 2 -bonded carbon atoms in the buer layer.
The two components have an area ratio S1:S2 of ≈ 1:2. The hydrogen desorption implicates the
appearance of a second graphene related peak (black line), representing those patches, where the
hydrogen has left. For an annealing temperature of 830 ◦ C, this monolayer component becomes
more signicant than the one resulting from the quasi-free bilayer patches. After annealing at
1000 ◦ C, the hydrogen is completely desorbed: the C 1s spectrum acquires the shape typically
obtained for epitaxial monolayer graphene. The dierence in binding energy location between
the monolayer and decoupled bilayer components is about 0.4 eV, which perfectly agrees with
the shift of the Fermi level measured via ARPES. The total shift of the SiC component is 1 eV,
which conrms that on the SiC surface hydrogen bonds are present, which cause a respective
band bending or modify the band bending induced by the interface.
Further evidence of the existence of Si-H bonds is brought by the Si 2p data. In contrast
to C 1s data, the Si 2p spectra of SiC are known to show much more similar appearance for
dierent surface reconstructions and surface sensitivities (cf. Section 4.3.1 and Ref. [99, 262]).
The hydrogen intercalation and subsequent desorption, however, cause major changes in the Si 2p
spectra. The Si 2p spectrum obtained after initial outgassing (bottom curves in Fig. 6.14 (c)
and (d)) consists of two spin-orbit split doublets. The binding energies are given with respect
to the Si 2p3/2 component. According to the surface sensitivity variations and in agreement
with Ref. [262], the dominant peak at 100.6 eV (dark blue line) can be assigned to the bulk
component and the one at 100.9 eV (red line) to Si-H bonds. After annealing at 730 ◦ C, the Si 2p
spectrum can be accurately tted only by introducing two additional surface related components:
the√one √ at higher binding energy (light blue line) is attributed to the Si atoms bonded to the
(6 3×6 3)R30◦ reconstructed overlayer, the small one at lower binding energy (green line)
to surface defects (cf. Section 4.3.1). These components increase in intensity for increasing
annealing temperatures while the Si-H component gradually vanishes and completely disappears
after annealing at 1000 ◦ C. The total shift observed for the Si 2p bulk component amounts to
1 eV in agreement with the C 1s bulk peak.
Monitoring the decoupling by LEEM

The eect of hydrogen intercalation on the graphene structure can be analyzed with spatial reso-
lution using LEEM, which can identify the number of graphene layers on SiC by the number of dips
in the electron reectivity spectra between 0 and 8 eV [111] as previously shown in Section 4.3.5.
In Fig. 6.15, LEEM micrographs are shown for an electron energy of 5.1 eV measured in the same
area of the sample with intercalated hydrogen (panel (a)) and subsequent hydrogen desorption
due to annealing (panel (c),(e) and (g)). At this energy, regions of dierent graphene thickness
can be distinguished by the reected intensity. The electron reectivity spectra for the dierent
surface domains A, B and C, as labeled in panel (a), are plotted in panel (b). The number of dips
in the spectra identies region A, B and C as bi-, tri-, and four layer graphene. The advantage
of the low homogeneity of the sample is that the eect of hydrogen on patches with dierent
thickness can be observed. As we have already seen in the discussion of the band structure and
core level data, the hydrogen starts to desorb at temperatures between 700 ◦ C and 800 ◦ C. This
process of hydrogen desorption can be followed in the LEEM images of panel (c) and (e), and in
the corresponding reectivity spectra of panel (d) and (f) with the annealing temperatures being
800 ◦ C and 850 ◦ C, respectively. Whereas the spatial distribution of the dierent domains remains
unchanged, their LEEM intensity changes, and the reectivity spectra indicate the beginning of
a transformation of (n+1)-layer thick areas into (n)-layer thick areas (n=1,2,3). The most
prominent variations can thereby be observed in the low energy regime. After full desorption
of the hydrogen through an annealing step at 900 ◦ C, this transformation is completed, as can
be seen in panel (g) and (h). Note also that, for the same graphene thickness, the spectra in
panel (b) present dips located at slightly lower energies than those in panel (h). Apparently, this
is a characteristic feature of the hydrogen treated samples. Furthermore, note that the region
labeled D in Fig. 6.15 displays the same intensity before, during and after desorption of the
hydrogen (and a at reectivity spectrum (not shown)) and is attributed to surface defects, e.g.
from residual polishing damage.
Powerful combination of atmospheric pressure graphitization and hydrogen interca-

lation
The powerful combination of hydrogen intercalation and atmospheric pressure graphitization
(cf. Section 5.2.2) leads to the growth of a dened number of quasi-freestanding epitaxial
graphene layers with a homogeneity in the µm2 range and could be an essential step for-
ward towards a future implementation of graphene-based nanoelectronics. Figure 6.16 (a)
displays
√ √a bright eld LEEM image measured on quasi-freestanding monolayer graphene. A
(6 3×6 3)R30◦ zerolayer phase was prepared by Ar atmosphere RF heating on a CMP polished
4H-SiC(0001) sample, and subsequently treated in hydrogen in order to achieve the intercala-
tion/decoupling. The surface shows a state of the art homogeneity with terrace sizes of some tens
of µm2 . An inspection of dierent surface areas, which are marked in the LEEM image, clearly
identies all regions to be quasi-freestanding monolayer graphene since the intensity curves exhibit
one minimum only (panel (b)). As noted before for bilayers and above, also here the minimum
in the intensity spectrum is shifted by around 1 eV to lower energies with respect to conventional
epitaxial monolayer graphene (cf. Fig. 6.15 (h)). In our quasi-freestanding graphene samples,
there is practically no bilayer contribution at the step edges, which is in contrast to conventional
monolayer epitaxial graphene grown under atmospheric pressure conditions [10]. Only the thin
black lines forming the terrace edges turned out to be (decoupled) bilayer graphene as conrmed
at a dierent position on the sample (not shown). Possible reasons for this improved quality
Figure 6.15: 4×4 µm2 LEEM micrographs recorded with an electron energy of 5.1 eV for the same area
of a hydrogen-treated graphene sample after (a) outgassing at 400 ◦ C and (c),(e),(g) after progressive
hydrogen desorption due to annealing at 800 ◦ C, 850 ◦ C and 900 ◦ C, respectively. Representative regions
are labeled A, B, C, and D. The electron reectivity spectra obtained for the regions A, B, and C are
plotted in panels (b), (d), (f) and (h), respectively, labeled with the number of graphene monolayers (ML).
The region D is attributed to surface defects.
could either be dierent growth characteristics for zerolayer and monolayer samples or dierent
growth characteristics for CMP polished and hydrogen etched samples.
Although practically the whole surface consists of quasi-freestanding epitaxial graphene, a
slight contrast dierence between dierent domains can be detected in the LEEM image at
5.0 eV in panel (a), which is quite obviously not due to a dierent number of graphene layers.
The intensity curves are nearly identical except for slight deviations at around 5 eV. At a dierent
sample position, we investigated the lighter and darker gray areas in more detail by micro-LEED
to reveal the nature of the dierent domains. The corresponding diraction patterns measured
at 111 eV are shown in the insets of panel (a) and exhibit a threefold symmetry of the SiC-
substrate spots. However, the lighter and darker gray regions dier by a rotation of the substrate
diraction spots of 60◦ with respect to each other, which just reects the two dierent possible
surface terminations of 4H-SiC(0001) (cf. Section 2.2.2). Such observations are only possible in
micro-LEED and not in conventional LEED [97, 98]. For precise normal incidence of the electron
beam, no contrast dierence should be detected for these domains so that the observed slight
deviations have to be attributed to a tiny misalignment of the incident electron beam.
Further evidence for the extraordinary homogeneous graphene coverage of the surface shown
in Figure 6.16 (a) is given from a core level analysis of the dierent domains by using micro-
CLPES. The corresponding C 1s spectra for the same regions as for the LEEM intensity curves of
Figure 6.16 (b) are displayed in panel (c) of the same gure. Quite evidently, quasi-freestanding
monolayer
√ √ graphene is present for all measured surface regions, interface components of the
(6 3×6 3)R30◦ reconstruction cannot be detected in the spectra.
After hydrogen desorption at an annealing temperature of around 900 ◦ C, we expect monolayer
graphene to have completely transformed back to zerolayer graphene. Fig. 6.16 (d) displays the
corresponding LEEM image at 2.0 eV of the same surface area as before. The morphology of
the sample has apparently not changed. The intensity spectra shown in panel (e) for dierent
surface regions indeed do not have any pronounced minimum and are of zerolayer character [111].
Again, slight contrast changes can be detected in the LEEM image, which, however, are not as
obvious as in the hydrogen intercalated sample. The C 1s core level data from micro-CLPES
on a dierent position of the sample give complementary evidence of the back transformation
after annealing.
√ As √ shown◦ in Fig. 6.16 (f), the spectrum exhibits the shape of zerolayer graphene
with the (6 3×6 3)R30 interface components S1 and S2 in accordance with Figure 4.9 (b) of
Section 4.2.2.
Discussion
The question arises how the hydrogen migrates both below the interface layer and even through
several graphene layers. A direct penetration of the hydrogen through graphene is not possible
since graphene is impermeable to gases [46]. The energy barrier for such a direct passage is in
the range of several eV as predicted from theory [263265] and is too high, even if the graphene
samples are annealed to some hundred degrees Celsius. Therefore, it is not astonishing that, until
now, there has been no experimental evidence for hydrogen penetration through graphene [37].
In particular, several experiments were dedicated to passivate the SiC substrate and decouple the
graphene layers [256258]. However, these experimental eorts have not been successful in that
respect. Only the passivation of substrate Si dangling bonds has been reported in Ref. [257].
However, in our experiments, the hydrogen obviously nds a way through the epitaxial
graphene in order to generate quasi-freestanding graphene [160, 255]. In contrast to the already
mentioned experiments, we use molecular hydrogen at atmospheric pressures and our graphene
6.2. Decoupling epitaxial graphene from the SiC substrate
Figure 6.16: (a),(b) LEEM micrographs together with electron reectivity spectra of the indicated surface areas showing quasi-freestanding monolayer
graphene on a large scale on 4H-SiC(0001). Hereby, the methods of atmospheric pressure graphitization and hydrogen intercalation have been successfully
combined. The micro-LEED patterns for surface domains with light and dark gray contrast are shown in the insets of panel (a) and correspond to SiC
substrate domains with a mutual rotation of 60◦ . (c) Micro-CLPES C 1s core level spectra of quasi-freestanding monolayer graphene taken from the
surface regions indicated in panel (a). (d),(e) Corresponding LEEM micrographs and electron reectivity spectra of the same surface area after thermal
hydrogen desorption. (f) Micro-CLPES C 1s core level spectrum of zerolayer graphene taken after hydrogen desorption.
133
samples were annealed up to 1000 ◦ C, which might facilitate a reactive passage of the hydrogen
through the graphene lattice. Another possibility is that the hydrogen intercalation starts at grain
boundaries or defects on the surface. However, the hydrogen incorporation does not show any
signicant dierences between samples with high and low morphological quality. For samples with
a low homogeneity, the hydrogen desorption upon annealing was shown in Fig. 6.15. However,
a migration pathway of the hydrogen could not be deduced. For the high quality samples of
Fig. 6.16, the hydrogen desorption in our LEEM experiments unfortunately proceeded too fast to
follow the transformation back to the interface layer.
Another unsolved issue is the transformation from molecular to atomic hydrogen, which ob-
viously has to take place during the hydrogen treatment of the epitaxial graphene samples. A
recent theoretical study [265], that was motivated by our experimental results, suggested that the
hydrogen molecule breaks after intercalation, and one hydrogen atom binds to a silicon dangling
bond of the substrate. Afterwards, the second hydrogen atom breaks the covalent bonds between
zerolayer graphene and the substrate and then passivates the corresponding Si atom. However,
it has to be noted that it still seems to be too early for denite conclusions about the detailed
sequence of this process.
It should be pointed out that the hydrogen intercalated samples are extremely stable in
ambient atmosphere, at least for several months. Also, no measurable eects were caused by
extended exposure to ultraviolet or soft X-ray illumination. The stability up to 700 ◦ C is remarkable
and technologically appealing. Furthermore, the hydrogen passivation and desorption can be
repeated several times without notable changes in the sample quality.
Yet, for our samples, there is still room for improvement of the reproducibility of high quality
samples. In some cases, slightly dierent doping levels can be observed that coexist on dierent
domains on the surface. It can furthermore be expected that the hydrogen not only passivates the
SiC substrate but partially becomes active as it could bind to the carbon atoms in graphene [257,
258, 263], similar to the case of graphane formation reported for free-standing graphene akes
exfoliated from HOPG [37].
The technique of hydrogen intercalation presented in this section opens up the possibility to
produce quasi-free standing epitaxial graphene on large SiC wafers. The hydrogen passivates the
underlying SiC substrate similar to the case of bare SiC surfaces [99, 259, 260]. The intercalated
hydrogen is sustained in ambient conditions and stable up to 700 ◦ C. For future experiments,
the deposition of n-dopants such as potassium or electron donor molecules could give further
insight into the band structure of hydrogen intercalated graphene samples. The possible existence
and tunability of a band gap in decoupled bilayer graphene should be investigated in particular.
Furthermore, the important question as to if the transport properties are improved in comparison
to conventional epitaxial graphene arises. A better control of the process parameters will probably
soon result in even higher quality samples. Triggered by our results [160, 255], research eorts in
this eld have recently been intensied and led to comparable ndings [266]. By all means, the
hydrogen intercalation process is technologically well adapted and represents a highly promising
route towards epitaxial graphene based nanoelectronics.
Figure 6.17: (a)-(c) Band structure perpendicular to the ΓK direction of the graphene Brillouin zone
for (a) zerolayer graphene, (b) decoupled p-doped monolayer after Au intercalation and (c) decoupled
n-doped monolayer after Au intercalation. The momentum scans are taken by means of UV excited
ARPES. (d) Model sketch for Au intercalation below epitaxial graphene. The images are adapted from
Ref. [267].
6.2.2 Gold intercalation below epitaxial graphene

4
The starting
√ √point of◦ this work, which is published in Ref. [267] , is again the preparation of
the (6 3×6 3)R30 interface layer, i.e. zerolayer graphene, as displayed in Fig. 6.17 (a). On
top of the zerolayer, Au atoms are deposited at room temperature from a commercial Knudsen
cell. Post annealing up to around 800 ◦ C for around ve minutes decouples the zerolayer and
the linear dispersion typical for graphene appears (cf. Fig. 6.17 (b) and (c)). The Au obviously
intercalated below the interface layer and turns this layer into monolayer graphene. Whereas
conventional monolayer graphene is n-doped with the Fermi level shifted away from the Dirac
point by about 420 meV (cf. Section 4.3.2), the graphene obtained after Au deposition and
post-annealing is either n- or p-doped, depending on the gold coverage (about one third or one
monolayer, respectively). The Dirac point for the decoupled p-doped monolayer is about 100 meV
above the Fermi level. For the decoupled n-doped monolayer, the π -bands cross at -800 meV.
The charge carrier densities deduced from the size of the Fermi surface amount to -1 · 1013 cm−2
for conventional monolayer graphene, +7 · 1011 cm−2 for decoupled p-doped monolayer graphene
and -5 · 1013 cm−2 for decoupled n-doped monolayer graphene.
The main dierence between decoupled p-doped and n-doped monolayer graphene manifests
itself in the electronic quality and in the degree of decoupling from the substrate. One striking
feature in the band structure of the decoupled p-doped monolayer graphene is the sharpness of
the bands. An analysis of the full width at half maximum of the bands by tting momentum
distribution curves shows that decoupled p-doped monolayer graphene has much lower electron-
electron interactions and therefore a larger carrier lifetime in comparison to conventional mono-
layer graphene and the decoupled n-doped phase. Dierences can also be detected in LEED. For
the p-doped monolayer phase,√the graphene
√ diraction spot shows a much higher intensity and
stronger suppression of the (6 3×6 3)R30◦ satellite peaks compared to the decoupled n-doped
phase, indicating a much better geometrical decoupling.
Of course, the question arises what the detailed nature of these two dierent phases is. Core
4A detailed analysis of the gold intercalation process is performed in this reference, and the reader should refer
to this publication for a deeper understanding. The most important results are summarized in this section.
level photoelectron spectroscopy of the Au 4f core levels show both contributions from Au-Si
bonds and from Au-Au bonds. For a decoupling of the graphene, the Si-C bonds in the interface
layer have to be broken and be replaced by Au-Si bonds. From the core level peak intensities
about one Au atom per graphene unit cell is intercalated for the decoupled n-doped phase. This
is consistent with the observation that every third carbon atom in zerolayer graphene forms a C-Si
bond. For the formation of decoupled p-doped graphene, about one monolayer of gold is necessary.
Here, Au clusters on top of the graphene layer can be observed in AFM and STM. In contrast to
hydrogen intercalation below epitaxial graphene, a clear structural model cannot be given at the
moment. A model sketch of the present understanding is given in Fig. 6.17 (d). In particular, it is
not clear to what extent the gold is intercalating and to what extent the intercalated gold forms
bonds to the SiC substrate. The doping behavior for dierent Au coverages is nicely explained by
the theoretical work of Giovannetti et al. [268], who predicted p-type doping for graphene on a
Au substrate. Reducing the Au-graphene distance to dAuG < 3.2 Å, however, will lead to n-type
doping. The larger amount of intercalated Au as well as the observed LEED pattern in the case of
the decoupled p-doped graphene are strong indications for an increase of the graphene-substrate
distance and are consistent with the observed doping behavior.
The high quality of the decoupled p-doped graphene mentioned above is also reected in
Raman spectroscopy measurements (cf. Ref. [267]). Whereas the p-doped phase is stable in air,
decoupled n-doped graphene degrades fast so that no Raman measurements are possible. In
comparison to the conventional monolayer, a smaller D peak has been recorded for decoupled
p-doped graphene. Furthermore, the 2D peak shifts to lower values of around 2685 cm−1 . For
comparison, conventional monolayer graphene shows a 2D peak at around 2700 - 2750 cm−1 as
discussed previously in Section 4.3.6. This compressive strain is released during the Au intercala-
tion process and even turned into slight tensile strain when considering the 2D peak position of
strain free graphene of around 2700 cm−1 .
So far, the intercalation of epitaxial graphene below zerolayer graphene has been elaborated.
The question arises if gold can also be intercalated below an epitaxial monolayer graphene in
order to create decoupled bilayer graphene. However, in our experiments we could not make such
an observation. As reported in Ref. [171], a p-doped phase with sharp bands developed, quite
similar to the p-doped phase after Au intercalation below zerolayer graphene. The preparation
parameters were also the same. Although a denite explanation cannot be given at the moment,
it could well be that, for a monolayer sample, the gold atoms intercalate only below monolayer
graphene and not below the interface layer.
At present, it seems more instructive to concentrate the research on the p-doped phase
obtained after decoupling the interface layer. The low electron-electron interaction and the low
defect density of the quasi free-standing graphene sample in the case of the p-doped phase
should increase the phase coherence time of the graphene charge carriers as well as their mobility.
Furthermore, the implementation of a strong spin-orbit coupling through the presence of heavy
atoms in the vicinity of the graphene monolayer increases the importance of graphene for the
eld of spintronics [269271]. These aspects should, in particular, be claried by transport
measurements. Despite the high complexity of the intercalation process of gold, more exciting
results could well be expected.
6.3. Conclusion 137
Figure 6.18: Schematic representation of dierent approaches to reduce the inuence of the interface
layer present between graphene and SiC.
6.3 Conclusion
In contrast to the previous two chapters, which were discussing the properties of epitaxial graphene
on SiC and dierent methods of epitaxial graphene synthesis, the present chapter gave direct
insight into dierent ways of tuning and manipulating the structural and electronic properties of
epitaxial √
graphene
√ on SiC(0001). Hereby, the main motivation was to circumvent the drawbacks
of the (6 3×6 3)R30 reconstructed interface layer that is located between the graphene layers
◦
and the SiC substrate. This buer layer is named zerolayer graphene since it represents an
initial carbon layer with about 30% of the carbon atoms forming covalent bonds to the Si-
terminated substrate, which prevent the development of π -bands with the typical linear dispersion
of graphene. The problematic of the interface layer is twofold: rst, it causes n-doping in the
1013 cm−2 regime so that the ambipolar properties of graphene cannot be exploited in electronic
devices. Second, the interface layer is one of the primary suspects of reduced mobilities in epitaxial
graphene, probably due to the introduction of scattering centers.
The dierent approaches to solve these problems are summarized in Fig. 6.18. The most direct
way has already been introduced in Section 4.3.6 and consists of removing the SiC substrate by
transferring the epitaxial graphene from SiC to SiO2 . The transfer is accomplished via mechanical
exfoliation by using an adhesive tape, similar to the production of graphene akes from HOPG. Of
course, this technique is not suitable for large scale implementation. It however serves as a proof
of principle to distinguish those physical eects that are due to the presence of the SiC substrate
from those that are intrinsic to the epitaxially grown graphene. In particular, the n-doping typical
for epitaxial graphene on SiC(0001) is completely reversed after transferring to SiO2 .
A technologically well adapted and much more elegant way of reducing the intrinsic charge
carrier concentration can be implemented by the technique of surface transfer doping. The
dopant materials need to have a high electronegativity to extract the electrons out of the epitaxial
graphene. In Chapter 4, it has already been demonstrated that unintentional doping from oxygen,
hydrocarbons or water from the ambient environment induces p-type doping. In this chapter, more
controllable and dened means of p-type doping have been presented, which are schematically
sketched in Fig. 6.18. Whereas atomic layers of Bi and Sb, deposited in UHV conditions, can
only partially reverse the intrinsic n-doping [171], the molecule F4-TCNQ results in conditions
around charge neutrality [51]. The latter approach seems to be extremely promising since the
corresponding samples are stable in air and temperature resistant up to 200 ◦ C. Furthermore,
the doping can even be achieved by application friendly wet chemistry. In the future, particular
attention should be drawn to electrical transport measurements of F4-TCNQ doped epitaxial
bilayer graphene. The size of the intrinsic band gap can be controlled via the amount of deposited
molecules. At the same time, the graphene can be rendered truly semiconducting and could lead
to eld eect devices with a suitable on/o ratio.
However, the ultimate solution to the problems of the interface layer cannot be given by
surface transfer doping since only the negative consequences of the presence of this layer are
cured, but the actual problem remains open. Therefore, a manipulation of the interface layer on
the atomic level would be necessary. In this chapter, it has been demonstrated that the elegant,
eective, and processing friendly technique of hydrogen intercalation fullls this purpose in an
impressive way [160, 255]. The hydrogen breaks the covalent bonds between the Si-terminated
substrate and the interface layer, and subsequently saturates the Si bonds to generate quasi-
freestanding epitaxial graphene on large-scale SiC(0001) wafers. This procedure has been shown
to work for at least up to trilayer graphene. The exclusive access to the outstanding properties
of graphene is extremely appealing for applications in graphene based nanoelectronics.
Intercalation below epitaxial graphene is not only possible for hydrogen but also for the
element gold [267]. Gold decouples the graphene from the substrate in a similar way, although
the detailed nature of the intercalated samples is much less understood. Depending on the
amount of intercalated gold, two dierently doped phases exist. Since the element gold is a
metal, its role in terms of transport and application in electronic devices is not a priori clear. At
rst sight, this might be of disadvantage; however, the strong spin-orbit interaction of the heavy
element gold might induce spin-splitting, and therefore increase the importance of graphene for
the eld of spintronics.
The decoupling of the interface layer from the SiC(0001) substrate by gold or hydrogen in-
tercalation opens up the possibility for a large variety of similar approaches addressing dierent
physical or technological questions. Other elements than gold or hydrogen can be envisioned
for intercalation below epitaxial graphene. A non-metallic element such as germanium already
shows promising results [112, 272] and could be highly interesting for technological applica-
tions. Elements such as iron, nickel or cobalt could aim at magnetic phenomena. The elements
calcium or magnesium, which are of importance in superconducting graphite intercalation com-
pounds [273, 274], could address superconductivity in graphene and graphene based materials.
In the future, it is most probably the combination of decoupling and doping techniques that
will be essential for applications in graphene based nanoelectronic devices. It could be envisioned
that epitaxial graphene grown on a large scale on SiC(0001) wafers via atmospheric pressure
graphitization is rst decoupled via hydrogen intercalation. In a next step, the graphene can be
doped with n-type or p-type dopants. Since molecules typically are sensitive to light irradiation
(F4-TCNQ, e.g., can be completely removed from the graphene surface), irradiation with an
optical fringe would create a periodic doping variation and thus give access to p-n-junctions.
Furthermore, to exploit the benets of a band gap, either epitaxial bilayer graphene can be taken
or epitaxial graphene nanoribbons. The latter could be produced from lithographical patterning or
from a controlled physical structuring of the graphene lattice such as etching with nanoparticles.
Apart from preparation issues as well as the structural and electronic characterization, it should
be emphasized at this point that the investigation of the electric transport properties has to be
intensied for epitaxial graphene on SiC(0001), since the knowledge is still behind the one for
graphene exfoliated from graphite.
6.3. Conclusion 139
In addition to the non-covalent functionalization, which has been discussed in this chapter, a
covalent functionalization of epitaxial graphene on SiC, for instance with oxygen (graphene oxide)
or hydrogen (graphane) has not yet been investigated in detail and deserves further attention.
Another promising research pathway could be the connection of graphene electronics with the
electronics of single molecules. Since the conductivity of organic molecules relies on π -electrons,
graphene-molecular interconnects may serve as a building block of future electronic devices.
The possible upcoming research directions that have been mentioned here, are of course only a
small fraction of all ideas that are currently circulating in this very active research area. Above all,
it should not be forgotten that experiments concerning the non-electronic properties of graphene
have just been started. Many research topics are still at the beginning and promising results will
have to be expected in the future.
Chapter 7
Summary
The aim of the present thesis is to provide detailed insight into the growth of epitaxial graphene
layers on SiC surfaces, the precise characterization of their structural and electronic properties,
and, eventually, the manipulation of their qualities according to technological relevance. The
material graphene can be seen as a single atomic plane pulled out of bulk graphite. In 2004, the
discovery of simple ways to produce graphene has led to the realization of its potential for both
fundamental physics and applications in nanoelectronics. This has caused an enormous increase
of graphene related research over the past six years and resulted in the denition of the research
eld of relativistic condensed matter physics. The graphitization of SiC crystals turns out to be
one of the most promising ways for a large scale production of graphene and can be implemented
in existing device technology. The main focus in my thesis is lying on the investigation of epitaxial
graphene on SiC(0001), the Si-terminated side of the basal plane surfaces of SiC, which has the
most promising prospects for a well-ordered and homogeneous growth of a precisely controlled
number of graphene layers on large SiC wafers. A few considerations are also devoted to epitaxial
graphene on the C-terminated counterpart, SiC(0001̄), and on non-basal plane surfaces. Special
attention has to be drawn to the initial stages of graphitization and the graphene-SiC interface
formation since it diers for dierent SiC surfaces and is the basis for a well ordered and controlled
growth on SiC(0001). Apart from this precise knowledge about the basic properties of epitaxial
graphene on various kinds of SiC surfaces, the main results of this thesis concern the successful
manipulation of epitaxial graphene on SiC(0001) in terms of technological relevance. By means of
surface transfer hole doping and the decoupling of the graphene from the substrate via hydrogen
intercalation below the interface layer, the graphene properties on SiC(0001) could be improved
tremendously with respect to applications in future electronic devices. In combination with the
method of atmospheric pressure graphitization, which is shown to be the method of choice to
obtain homogeneous large area graphene, the technique of hydrogen intercalation sets a new
standard in the production of high quality epitaxial graphene on SiC.
In the framework of my thesis as a multidisciplinary surface science work, a large variety of
surface science related techniques has been applied to study the epitaxial graphene layers on SiC
wafers. Low energy electron diraction (LEED), core level photoemission spectroscopy (CLPES)
and scanning tunneling microscopy (STM) were mainly used for structural investigations, atomic
force microscopy (AFM) and low energy electron microscopy (LEEM) for the analysis of the
morphology on a large scale. Angle resolved photoemission spectroscopy (ARPES) and scanning
tunneling spectroscopy (STS) were the methods of choice for band structure studies. Finally,
Raman spectroscopy allowed for an insight into vibrational properties of the graphene.
The growth of graphene on SiC crystals is based on a thermal decomposition at temperatures
142 Chapter 7. Summary
of typically more than 1200 ◦ C. Owing to the higher vapor pressure of Si in comparison to C, the
Si atoms sublimate and leave graphene layers behind. This process is mainly governed by the
annealing temperature and only to a minor degree by the annealing time. The graphitization of
the SiC(0001) surface was extensively investigated for samples prepared under ultra-high vacuum
(UHV) conditions. However, most of the ndings also hold for dierent preparation procedures.
In addition, there seems to be no inuence of the SiC-polytype with respect to the properties
of the epitaxial graphene. Before the growth of well isolated graphene layers on SiC(0001),
an interface layer is formed. This precursor stage of graphitization is sometimes also called
"zerolayer graphene" or buer layer. In the past years there have been many controversial
interpretations of this surface
√ structure.
√ In this thesis, it has been shown that the true periodicity
of the interface layer is (6 3×6 3)R30 as conrmed from both LEED and STM. Furthermore,
◦
the role of (5×5) domains has been claried and depends on the detailed UHV preparation
procedure:
√ √ a minimization of the number of (5×5) domains and thus a larger homogeneity of the
(6 3×6 3)R30 reconstruction can be obtained by performing a Si-deposition at 800 ◦ C before
◦
graphitization.
√ √ From band structure measurements and a core level analysis the model of the
(6 3×6 3)R30◦ structure as a partially covalently bound initial carbon layer has been proposed.
About one third of the carbon atoms form covalent bonds to the Si atoms of the SiC substrate,
the remaining two thirds of the carbon atoms remain in sp 2 -hybridization.
√ √ During graphitization,
every new graphene layer implies the re-formation of the (6 3×6 3)R30◦ reconstruction and
the release of the former buer layer into a true graphene layer. The structure of the shifted
interface layer remains unchanged. Owing to this mechanism, the graphene layers are all rotated
by exactly 30◦ with respect to the SiC substrate.
Whereas zerolayer graphene does not yet show the typical graphene π -bands but only the
graphene σ -bands, the rst graphene layer on top of the interface layer exhibits the linear disper-
sion of the π -bands at the K-point of the graphene Brillouin zone and therefore is the rst "true"
graphene layer. Due to an electric dipole in the interface layer, monolayer epitaxial graphene is
intrinsically n-doped in the 1013 cm−2 regime. In contrast to monolayer epitaxial graphene, bilayer
epitaxial graphene shows a parabolic dispersion. Since the dipole in the interface layer induces
an electric eld, a eld gradient develops over the rst and second layer. This gradient causes
the opening of a band gap of around 120 meV in bilayer graphene. Furthermore, by correlating
the band structure measurements at the K-point with LEED intensity spectra of the graphene
(10) diraction spot for dierent graphene thicknesses, it was shown for the rst time that LEED
alone can be used to determine the number of graphene layers on SiC(0001). Since in practically
every UHV system a LEED optics is mounted, this nding is of extremely practical use for a
continuous monitoring of the graphene preparation process. √ √
In STM measurements the graphene together with the (6 3×6 3)R30◦ structure were ex-
amined on an atomic level in real space. Considering dierent tip conditions and dierent bias
values, zero-, mono- and bilayer graphene can be clearly distinguished. In contrast to bilayer
graphene, monolayer √ graphene
√ samples still exhibit a large electronic inuence and a high cor-
rugation of the (6 3×6 3)R30◦ reconstruction. Furthermore, monolayer graphene is visible
only for low bias values and shows all six atoms of the graphene unit cell (in contrast to bilayer
graphene).
Vibrational modes of graphene can be probed by Raman spectroscopy. On the samples that
have been investigated, the disorder related D peak is small and accounts for the high crystalline
quality of the epitaxial graphene. A suitable ngerprint to assign the number of layers in the case
of UHV prepared graphene on SiC(0001) is given by the width of the 2D peak since it exhibits a
(negative) linear relationship with the inverse number of layers. The position of the G peak can
143
be correlated to a large extent to the charge carrier concentration. Furthermore, the successful
transfer of epitaxial graphene from SiC to SiO2 has been reported for the rst time in the scope
of my thesis, which allows to distinguish Raman features due to the presence of the SiC substrate
from those being intrinsic to the graphene structure itself.
The growth of epitaxial graphene on SiC(0001̄) occurs in a completely dierent way than
on SiC(0001). Besides the observation that the thickness control is much more dicult in
comparison to SiC(0001), the initial graphene layer develops in coexistence with the intrinsic
surface reconstructions (3×3) and (2×2)C without the presence of an interface layer as conrmed
from CLPES and STM. This weak interaction between the substrate and the graphene causes
rotational disorder in the graphene layers, which can be detected in form of diraction rings instead
of diraction spots in LEED. The most important consequence of this turbostratic growth mode
is the electronic decoupling of the layers so that even multilayer graphene exhibits the typical
properties of monolayer graphene.
There is no principal restriction to study epitaxial graphene on the basal plane surfaces of SiC,
but the focus could rather be extended to non-basal plane surfaces such as SiC(11̄02), SiC(1̄102̄)
and SiC(112̄0), which, for the rst time, was briey discussed in this thesis. The reduced surface
symmetry in these directions provokes a reduced dimensionality of the SiC surfaces, which could
be imposed onto the epitaxially grown graphene layers. However, the present results do not show
any ordered growth of graphene so that high quality graphene samples do not seem to be feasible
at the moment on those surfaces.
The search for graphene samples with a large homogeneity comes to its limits in the conven-
tional UHV preparation technique since the homogeneity of the graphene is only in the range of
a few hundreds of nm2 . It has recently been shown in the literature that the graphitization of
SiC(0001) samples in an Ar atmosphere at elevated temperatures of around 1600 ◦ C to 2000 ◦ C
allows for a higher mobility of the atoms at the surface, which results in a tremendously increased
homogeneity. In the framework of my thesis, a homogeneous monolayer coverage of at least ten
µm2 could be demonstrated by adopting this technique. In addition, a completely new prepa-
ration approach was proposed, which consists of the epitaxial growth of monolayer graphene on
SiC(0001) at a reduced temperature of around 950 ◦ C by the assistance of carbon deposition in
UHV, resembling classical molecular beam epitaxy. The original, well ordered terrace structure
of SiC(0001) is preserved in contrast to the conventional graphitization method in UHV.
Although highly ordered graphene can be grown on a large scale with a well dened num-
ber√of layers
√ on ◦SiC(0001), it suers from two disadvantages arising from the presence of the
(6 3×6 3)R30 reconstructed interface layer. The rst one is the electron doping in the
1013 cm−2 range, which makes applications in ambipolar eld eect devices cumbersome. The
second one is the strong electronic inuence of the interface layer, which is one of the primary
suspects for the lower electron mobilities in comparison to graphene exfoliated from graphite.
Reduction of the intrinsic charge carrier density by surface transfer doping can be achieved by
the deposition of an electron acceptor. Electrons are extracted out of the graphene layer without
deteriorating the integrity of the band structure of graphene, which has been shown for the rst
time in the scope of this thesis and with two dierent approaches. Atomic hole doping with Bi
and Sb can partially reverse the intrinsic n-doping for monolayer graphene. Charge neutrality for
mono- and bilayer graphene can be reached by a non-covalent functionalization of the graphene
surface with the technologically well adapted electron acceptor molecule F4-TCNQ. Epitaxial
bilayer graphene, in particular, is turned from a conducting system into a truly semiconducting
layer. Furthermore, the additional dipole, generated at the graphene surface by the molecular
overlayer, allows for a controllable engineering of the bilayer band gap magnitude, which can be
144 Chapter 7. Summary
increased up to double of its initial value. This could make the implementation of eld eect
devices with a suitable on/o ratio feasible.
However, surface transfer hole doping only compensates the intrinsic n-doping, and, beyond
that, does not reduce the electronic inuence of the interface layer. Apart from the proof of
principle of transferring epitaxial graphene from SiC to SiO2 , a technologically more promising
approach requires a manipulation of the interface layer on an atomic level. In the framework of
this thesis, it has been successfully demonstrated for the rst time that the elegant, eective, and
processing friendly technique of hydrogen intercalation fullls this purpose in an impressive way
by decoupling the epitaxial graphene from the substrate. The hydrogen breaks the covalent bonds
between the Si-terminated substrate and the interface layer, and subsequently saturates the Si
bonds so that the outstanding properties of graphene can be made accessible in quasi-free standing
epitaxial graphene layers. Moreover, the powerful combination of hydrogen intercalation and
atmospheric pressure graphitization leads to the growth of a dened number of quasi-freestanding
epitaxial graphene layers on large-scale SiC(0001) wafers and could be an essential step forward
towards a future implementation of graphene-based nanoelectronics. Besides hydrogen, also the
element gold can be intercalated to decouple the epitaxial graphene from the substrate. However,
the role of the intercalated Au is not yet entirely clear, it may alter the electronic transport
properties but could also lead to new physical phenomena.
Although the rst promising characterizations of graphene around six years ago have been
regarded with some skepticism with respect to a sustainable impact on the research landscape,
the new material graphene, and epitaxial graphene on SiC in particular, can in the meantime
be regarded as rising stars in material science. The results of this thesis have given a deeper
understanding of epitaxial graphene on SiC and are a valid basis to trigger new developments in
this eld. Evidently, there is still plenty of room in "carbon atland".
Chapter 8
Zusammenfassung
Das Ziel der vorliegenden Dissertationsschrift ist es, einen detaillierten Einblick in das Wachstum
und die Charakterisierung epitaktischer Graphen-Lagen auf SiC-Oberächen zu geben, sowie die
gezielte Beeinussung ihrer strukturellen und elektronischen Eigenschaften in Hinblick auf techno-
logische Anforderungen zu diskutieren. Das Material Graphen kann als eine einzelne, aus Graphit
herausgelöste Kohlenstoage angesehen werden. Mit der Entdeckung im Jahr 2004, Graphen
auf eine einfache Art und Weise herzustellen, wurde auch das entsprechende Potential sowohl für
die Grundlagenphysik als auch für Anwendungen in der Nanoelektronik erkannt. Dies hat über
die letzten sechs Jahre zu einem starken Anstieg der Forschungsaktivitäten an dem Materialsys-
tem Graphen und der Denition eines Forschungsgebiets der relativistischen Physik kondensierter
Materie geführt. Die Graphitisierung von SiC-Kristallen ist dabei eine der vielversprechensten Me-
thoden für eine groÿtechnische Herstellung von Graphen und kann in die bereits existierende Tech-
nologie elektronischer Bauelemente integriert werden. Der Schwerpunkt meiner Doktorarbeit liegt
auf der Untersuchung epitaktischen Graphens auf SiC(0001), der Si-terminierten Seite der Basal-
Oberächen von SiC. Diese Oberäche erscheint am aussichtsreichsten für ein geordnetes und
homogenes Wachstum einer exakt denierten Anzahl von Graphen-Lagen auf groÿen SiC-Wafern.
Jedoch wird auch der Fall epitaktischen Graphens auf der C-terminierten Gegenseite, SiC(0001̄),
sowie auf nicht-basalen Oberächen beleuchtet. Besonderes Augenmerk muss bei allen Ober-
ächenrichtungen auf das Anfangsstadium der Graphitisierung und die Ausbildung der Graphen-
SiC-Grenzäche gerichtet werden, da diese für verschiedene SiC-Oberächen unterschiedlich ist
und die Grundlage für ein geordnetes und kontrolliertes Wachstum auf SiC(0001) darstellt.
Abgesehen von den genauen Kenntnissen der grundlegenden Charakteristika von epitaktischem
Graphen auf verschiedenen SiC-Oberächen, liegen die Hauptergebnisse dieser Doktorarbeit in
der hinsichtlich technologischer Anforderungen erfolgreichen Manipulierung der Eigenschaften von
epitaktischem Graphen auf der SiC(0001)-Oberäche. Mittels Oberächen-Transferdotierung und
dem Entkoppeln des Graphens vom Substrat durch Wassersto-Interkalation unter die Grenzä-
chenlage können die Eigenschaften von Graphen auf SiC(0001) in Hinblick auf mögliche Anwen-
dungen in zukünftigen elektronischen Bauelementen entscheidend verbessert werden. In Verbin-
dung mit der Graphitisierung unter Atmosphärendruck, der Methode der Wahl, um homogene
und groÿächige Graphen-Lagen zu erhalten, setzt das Wassersto-Interkalations-Verfahren neue
Maÿstäbe in der Herstellung von hochqualitativem epitaktischen Graphen auf SiC.
Im Rahmen meiner Dissertationsschrift als eine multidisziplinär angelegte Arbeit in den Ober-
ächenwissenschaften wurde eine Vielzahl an Untersuchungsmethoden verwendet. Während die
Beugung langsamer Elektronen (low energy electron diraction (LEED)), die Rumpfniveau-Photo-
elektronenspektroskopie (core level photoemission spectroscopy (CLPES)) und die Rastertunnel-
146 Chapter 8. Zusammenfassung
mikroskopie (scanning tunneling microscopy (STM)) hauptsächlich für strukturelle Untersuchun-

gen benutzt wurden, dienten die Rasterkraftmikroskopie (atomic force microscopy (AFM)) und
die Mikroskopie mit niederenergetischen Elektronen (low energy electron microscopy (LEEM)) der
groÿächigen Untersuchung der Morphologie. Winkelaufgelöste Photoemission (angle resolved
photoemission spectroscopy (ARPES)) und Rastertunnelspektroskopie waren die Methoden der
Wahl für Untersuchungen der Bandstruktur. Einblick in die phononischen Eigenschaften von
Graphen konnte mit Hilfe der Raman-Spektroskopie gewonnen werden.
Die Herstellung von Graphen auf SiC-Kristallen beruht auf der thermischen Zersetzung des SiC
bei Temperaturen von typischerweise mehr als 1200 ◦ C. Aufgrund des höheren Dampfdrucks von
Si im Vergleich zu C verdampfen die Si-Atome, und der verbleibende Kohlensto verbindet sich zu
einer oder mehreren Graphen-Lagen. Dieser Prozess wird im Wesentlichen durch die Heiztempera-
tur und nur in geringem Maÿe durch die Heizdauer bestimmt. Die Graphitisierung der SiC(0001)-
Oberäche wurde ausgiebig für Proben untersucht, die unter Ultrahochvakuum (UHV)-Bedin-
gungen präpariert wurden. Jedoch gelten die meisten Erkenntnisse auch für davon abweichende
Herstellungsmethoden. Weiterhin scheint der SiC-Polytyp keinen Einuss auf die Eigenschaften
des epitaktischen Graphens zu haben. In Hinblick auf das Graphen-Wachstum auf SiC(0001) muss
beachtet werden, dass sich zunächst eine Graphen-SiC-Grenzächenlage ausbildet. Diese Vorstufe
der Graphitisierung wird manchmal auch als "nullte Graphen-Lage" oder als Puerlage (buffer
layer) bezeichnet. In den vergangenen Jahren wurden viele, teils kontroverse Interpretationen die-
ser Oberächenstruktur publiziert. In der vorliegenden Doktorarbeit wurde sowohl durch LEED-
als auch √ durch√ STM-Messungen gezeigt, dass die tatsächliche Periodizität der Grenzächen-
lage (6 3×6 3)R30◦ beträgt. Weiterhin wurde die Rolle der (5×5)-Domänen geklärt, die von
den Details der UHV-Herstellungsmethode √ abhängt:
√ eine Minimierung der (5×5)-Domänen und
damit eine gröÿere Homogenität der (6 3×6 3)R30◦ -Rekonstruktion kann durch einen Auf-
dampfschritt von √ Si √
bei 800 ◦ C vor der Graphitisierung erreicht werden. Die Struktur der überaus
komplexen (6 3×6 3)R30◦ -Rekonstruktion konnte mit Hilfe von Bandstruktur- und Rumpf-
niveau-Untersuchungen insoweit geklärt werden, dass es sich um eine teilweise kovalent gebunde-
ne erste Kohlenstoage handelt. Etwa ein Drittel der C-Atome bilden dabei kovalente Bindun-
gen zu den Si-Atomen des SiC-Substrats aus, die restlichen zwei Drittel der C-Atome bleiben
sp√2
-hybridisiert.
√ Das Wachstum einer jeden neuen Graphen-Lage beinhaltet die Neubildung der
(6 3×6 3)R30 -Rekonstruktion und die Umwandlung der vormaligen Grenzächenlage in eine
◦
echte Graphen-Lage. Die Struktur der verschobenen Grenzäche bleibt unverändert, sodass alle
Graphen-Lagen in gleicher Weise um genau 30◦ bezüglich des SiC-Substrats gedreht sind.
Während die nullte Lage Graphen noch nicht die typischen π -Bänder sondern nur σ -Bänder
aufweist, zeigt die erste Kohlenstoage, die oberhalb der Grenzächenlage liegt, die lineare Dis-
persion der π -Bänder am K-Punkt der Brillouin-Zone und stellt damit die erste "echte" Graphen-
Lage dar. Aufgrund eines elektrischen Dipols in der Grenzächenlage weist die Monolage eine
n-Dotierung im Bereich von 1013 cm−2 auf. Im Gegensatz zur Monolage besitzt epitaktisches
Bilagen-Graphen eine parabolische Dispersion. Da der Dipol in der Grenzäche ein elektrostati-
sches Feld induziert, entsteht ein Gradient dieses Feldes über die erste und zweite Lage hinweg.
Dieser Gradient führt zur Bildung einer Bandlücke von ca. 120 meV in Bilagen-Graphen. Wei-
terhin wurde durch Korrelation der Bandstruktur-Messungen am K-Punkt mit den entsprechen-
den LEED-Intensitätskurven des Graphen-(10)-Beugungsreexes für verschieden dicke Graphen-
Schichten zum ersten Mal gezeigt, dass LEED für sich genommen bereits ausreicht, um die
Anzahl der Graphen-Lagen auf SiC(0001) zu bestimmen. Da in praktisch jedem UHV-System
eine LEED-Optik zur Verfügung steht, ist diese Erkenntnis von sehr groÿem praktischen Wert in
Hinblick auf eine kontinuierliche Überprüfung des Graphen-Herstellungsprozesses.
147
√ √
Ein direkter Einblick in die atomare Struktur des Graphengitters und der (6 3×6 3)R30◦ -
Struktur konnte durch STM-Messungen gewonnen werden. Unter Verwendung unterschiedlicher
Tunnelspannungen und Spitzenbedingungen können die nullte Lage Graphen sowie Mono- und
Bilagen-Graphen deutlich voneinander unterschieden werden. Im Gegensatz zu Bilagen-Graphen
weist Monolagen-Graphen
√ √ immer noch einen groÿen elektronischen Einuss und eine hohe Kor-
rugation der (6 3×6 3)R30 -Rekonstruktion auf. Darüber hinaus ist Monolagen-Graphen nur
◦
bei niedrigen Tunnelspannungen aufzulösen und zeigt dann (im Gegensatz zur Bilage) alle sechs
Atome der Graphen-Einheitszelle.
Um die phononischen Eigenschaften von Graphen zu untersuchen, ist Raman-Spektroskopie
die Methode der Wahl. Die D-Mode, die nur bei Defekten auftritt, ist auf den untersuchten Pro-
ben nur schwach ausgeprägt, was ein Zeichen für die hohe kristalline Qualität des epitaktischen
Graphens ist. Ein Kriterium zur Bestimmung der Anzahl der Lagen für UHV-präpariertes Graphen
auf SiC(0001) wurde in der Breite der 2D-Mode gewonnen, da diese einen (negativ-)linearen Zu-
sammenhang mit der inversen Anzahl an Lagen aufweist. Die Position der G-Mode wiederum ist
im Wesentlichen mit der Ladungsträgerkonzentration korreliert. Weiterhin wurde im Rahmen mei-
ner Doktorarbeit erstmalig der erfolgreiche Transfer epitaktischen Graphens von SiC auf ein SiO2 -
Substrat beschrieben. Dadurch können Charakteristika in den Raman-Spektren, die vom SiC-
Substrat herrühren, von denjenigen unterschieden werden, die intrinsisch zur Graphen-Struktur
gehören.
Auf SiC(0001̄) verhält sich das Wachstum von epitaktischem Graphen völlig anders als auf
SiC(0001). Abgesehen davon, dass die Kontrolle der Dicke im Vergleich zu SiC(0001) erheb-
lich schwieriger ist, entwickelt sich die erste Graphen-Lage in Koexistenz mit den intrinsischen
Oberächenrekonstruktionen (3×3) und (2×2)C ohne Ausbildung einer Grenzächenlage, was
durch CLPES und STM bestätigt wurde. Diese schwache Wechselwirkung zwischen Substrat
und Graphen bewirkt Rotations-Unordnung in den Graphen-Lagen, die sich in LEED-Bildern in
Form von Beugungsringen an Stelle von Beugungsreexen manifestiert. Dabei ist die wichtigste
Folge dieses turbostratischen Wachstums die elektronische Entkopplung der Lagen, sodass sogar
Multilagen-Graphen die typischen Eigenschaften von Monolagen-Graphen aufweist.
Prinzipiell sollte es keine Beschränkung des Graphen-Wachstums auf die Basal-Oberächen
von SiC geben, der Fokus könnte vielmehr auf nicht-basale Oberächen wie z.B. SiC(11̄02),
SiC(1̄102̄) und SiC(112̄0) erweitert werden. Dieses Thema wurde erstmals in dieser Disser-
tationsschrift behandelt. Die reduzierte Oberächensymmetrie in diesen Richtungen ruft eine
reduzierte Dimensionalität der SiC-Oberächen hervor, welche auf die epitaktisch gewachsenen
Graphen-Lagen übertragen werden könnte. Die bisherigen Ergebnisse zeigen jedoch kein geord-
netes Graphen-Wachstum, sodass hochqualitative Graphen-Proben auf diesen Oberächen zur
Zeit nicht machbar erscheinen.
Die Suche nach Graphen-Proben mit groÿer Homogenität gelangt bei der konventionellen
UHV-Präparationsmethode sehr schnell an ihre Grenzen, da die Homogenität des Graphens
hier nur im Bereich einiger hundert nm2 liegt. Vor Kurzem wurde gezeigt, dass die Graphi-
tisierung von SiC(0001)-Proben in Argon bei Atmosphärendruck und erhöhten Temperaturen von
ca. 1600 ◦ C bis 2000 ◦ C eine höhere Mobilität der Atome an der Oberäche ermöglicht, welche
zu einer grundlegend besseren Homogenität führt. Im Rahmen meiner Doktorarbeit konnte mit
dieser Methode eine homogene Monolagen-Bedeckung von mind. zehn µm2 demonstriert werden.
Darüber hinaus wurde eine komplett neue Methode eingeführt, die darin besteht, eine Monolage
epitaktischen Graphens auf SiC(0001) bei einer reduzierten Temperatur von 950 ◦ C mittels Auf-
dampfen von Kohlensto im UHV in Analogie zur klassischen Molekularstrahlepitaxie wachsen zu
lassen. Die ursprünglich geordnete Terrassenstruktur von SiC(0001) bleibt dabei im Gegensatz
148 Chapter 8. Zusammenfassung
zur konventionellen UHV-Graphitisierung erhalten.

Obwohl eine denierte Anzahl wohl geordneter Graphen-Lagen groÿächig auf SiC(0001)
hergestellt √
werden√kann, gibt es dennoch zwei entscheidende Nachteile, die von der Anwesen-
heit der (6 3×6 3)R30◦ rekonstruierten Grenzächenlage herrühren. Zum einen ist Graphen
n-dotiert im Bereich von 1013 cm−2 , was Anwendungen in ambipolaren Feldeekttransistoren
schwierig macht. Zum anderen unterliegen die Graphen-Lagen einem starken elektronischen
Einuss der Grenzächenlage. Dieser ist die wahrscheinlichste Ursache dafür, dass die bislang
beobachteten Elektronen-Beweglichkeiten in epitaktischem Graphen niedriger sind als in Graphen,
das von Graphit abgelöst wurde. Die Verringerung der intrinsischen Ladungsträgerkonzentration
mittels Oberächen-Transferdotierung kann durch das Aufbringen eines starken Elektronenak-
zeptors bewirkt werden. Die Elektronen werden dabei der Graphen-Lage entzogen, ohne deren
Bandstruktur negativ zu beeinträchtigen. Dies wurde im Rahmen dieser Arbeit zum ersten Mal,
und mit zwei verschiedenen Herangehensweisen, demonstriert. Eine Loch-Dotierung durch Atome
wie z.B. Bi und Sb kann die intrinsische n-Dotierung von Monolagen-Graphen teilweise reduzieren.
Vollständige Ladungsneutralität für Mono- und Bilagen-Graphen kann hingegen durch die nicht-
kovalente Funktionalisierung der Graphen-Oberäche mit dem für technologische Zwecke gut
geeigneten Elektronenakzeptor-Molekül F4-TCNQ erreicht werden. Insbesondere wird epitakti-
sches Bilagen-Graphen von einem leitenden in ein halbleitendes System umgewandelt. Zudem
erlaubt der durch die Moleküllage zusätzlich aufgebrachte Dipol an der Graphen-Oberäche eine
Kontrolle über die Bilagen-Bandlücke, deren Gröÿe bis auf das Doppelte ihres ursprünglichen
Wertes erhöht werden kann. Die Herstellung von Feldeekt-Bauelementen mit einem ausreichend
groÿen On/O-Verhältnis könnte dadurch ermöglicht werden.
Oberächen-Transferdotierung kompensiert allerdings nur die intrinsische n-Dotierung und
verringert darüber hinaus nicht den elektronischen Einuss der Grenzächenlage. Abgesehen von
der prinzipiell vorhandenen Möglichkeit, epitaktisches Graphen von SiC auf ein SiO2 -Substrat zu
transferieren, beinhaltet eine vielversprechendere Herangehensweise die Manipulation der Grenz-
ächenlage auf atomarer Ebene. Im Rahmen dieser Doktorarbeit wurde zum ersten Mal gezeigt,
dass die elegante und in technologischer Hinsicht bestens geeignete Methode der Wassersto-
Interkalation diesen Zweck auf eindrucksvolle Weise erfüllt, indem das epitaktische Graphen vom
Substrat entkoppelt wird. Der Wassersto bricht die kovalenten Bindungen zwischen dem Si-
terminierten Substrat und der Grenzächenlage auf und sättigt anschlieÿend die Si-Bindungen
ab, sodass die auÿergewöhnlichen Eigenschaften von Graphen in Form von quasi-freistehendem
epitaktischen Graphen zugänglich werden. Weiterhin führt die leistungsfähige Kombination von
Wassersto-Interkalation und Graphitisierung unter Atmosphärendruck zum groÿächigen Wachs-
tum einer wohl denierten Anzahl an quasi-freistehenden epitaktischen Graphen-Lagen auf groÿen
SiC(0001)-Wafern. Dies könnte ein entscheidender Schritt hin zu einer zukünftigen Realisierung
Graphen-basierter Nanoelektronik sein. Neben dem Element Wassersto kann auch das Element
Gold interkaliert werden, um epitaktisches Graphen vom Substrat zu entkoppeln. Allerdings ist
die Rolle des interkalierten Golds noch nicht vollständig klar, es könnte sowohl die elektronischen
Transporteigenschaften beeinträchtigen als auch zu neuen physikalischen Phänomenen führen.
Obwohl die ersten aussichtsreichen Untersuchungen von Graphen vor sechs Jahren mit einem
gewissen Grad an Skepsis in Hinblick auf einen nachhaltigen Einuss auf die Forschungslandschaft
betrachtet wurden, können das neue Material Graphen, und epitaktisches Graphen auf SiC im
Speziellen, als rising stars in den Materialwissenschaften betrachtet werden. Die Ergebnisse dieser
Doktorarbeit liefern dabei ein tieferes Verständnis über epitaktisches Graphen auf SiC und sind
eine geeignete Grundlage dafür, neue Entwicklungen anzustoÿen. Es gibt in der Tat noch viel
Raum für neue Entdeckungen auf dem Gebiet des zweidimensionalen Kohlenstos.
Bibliography
[1] K.S. Novoselov, A.K. Geim, S.V. Morozov, D. Jiang, Y. Zhang, S.V. Dubonos, I.V.
Grigorieva, A.A. Firsov, Science 306, 666 (2004).
[2] K. S. Novoselov, A.K. Geim, S.V. Morosov, D. Jiang, M.I. Katsnelson, I.V. Grig-
orieva, S.V. Dubonos, and A.A. Firsov, Nature 438, 197 (2005).
[3] Y. Zhang, Y.-W. Tan, H.L. Stormer, and P. Kim, Nature 438, 201 (2005).
[4] C. Berger, Z. Song, X. Li, X. Wu, N. Brown, C. Naud, D. Mayou, T. Li, J. Hass,
A. N. Marchenkov, E. H. Conrad, P. N. First, and W. A. de Heer, Science 312,
1191 (2006).
[5] Nobel Lectures, Chemistry 1996-2000, Editor Ingmar Grenthe, World Scientic
Publishing Co., Singapore (2003).
[6] P.R. Wallace, Phys. Rev. 71, 622 (1947).
[7] G.W. Semeno, Phys. Rev. Lett. 53, 2449 (1984).
[8] 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, J. Phys. Chem. B 108, 19912 (2004).
[9] U. Starke, in: Silicon Carbide, Recent Major Advances (eds.: W.J. Choyke, H.
Matsunami and G. Pensl), p. 281-316, Atomic structure of SiC surfaces, (Springer,
2004).
[10] K.V. Emtsev, A. Bostwick, K. Horn, J. Jobst, G.L. Kellogg, L. Ley, J.L. McChesney,
T. Ohta, S.A. Reshanov, E. Rotenberg, A.K. Schmid, D. Waldmann, H.B. Weber,
and T. Seyller, Nature Mater. 8, 203 (2009).
[11] A.K. Geim and K.S. Novoselov, Nature Mater. 6, 183 (2007).
[12] W.A. de Heer, C. Berger, X. Wu, P.N. First, E.H. Conrad, X. Li, T. Li, M. Sprinkle,
J. Hass, M.L. Sadowski, M. Potemski and G. Martinez, Solid State Commun. 143,
92 (2007).
[13] L. Jiao, L. Zhang, X. Wang, G. Diankov, and H. Dai, Nature 458, 877 (2009).
[14] D.V. Kosynkin, A.L. Higginbotham, A. Sinitskii, J.R. Lomeda, A. Dimiev, B.K.
Price, and J.M. Tour, Nature 458, 872 (2009).
[15] J.C. Meyer, A.K. Geim, M.I. Katsnelson, K.S. Novoselov, T.J. Booth, and S. Roth,
Nature 446 60-63 (2007).
150 Bibliography
[16] R.E. Peierls, Ann. I. H. Poincaré 5, 177 (1935).
[17] L.D. Landau, Phys. Z. Sowjetunion 11, 26 (1937).
[18] A. Barth, and W. Marx, arXiv:0808.3320v3 (2008).
[19] R. Saito, G. Dresselhaus, and M. S. Dresselhaus, Physical Properties of Carbon

Nanotubes, Imperial, London (1998).
[20] S. Reich, J. Maultzsch, C. Thomsen, P. Ordejón, Phys. Rev. B 66, 035412 (2002).
[21] A.H. Castro Neto, F. Guinea, N.M.R. Peres, K.S. Novoselov, and A.K. Geim, Rev.
Mod. Phys. 81, 109 (2009).
[22] S.V. Morozov, K.S. Novoselov, M.I. Katsnelson, F. Schedin, L.A. Ponomarenko,
D. Jiang, and A. K. Geim, Phys. Rev. Lett. 97, 016801 (2006).
[23] M.I. Katsnelson, K.S. Novoselov and A.K. Geim, Nature Phys. 2, 620 (2006).
[24] A.F. Young, P. Kim, Nature Phys. 5, 222 (2009).
[25] N. Stander, B. Huard, D. Goldhaber-Gordon, Phys. Rev. Lett. 102, 026807 (2009).
[26] M.I. Katsnelson, K.S. Novoselov, Solid State Commun. 143, 3 (2007).
[27] N.W. Ashcroft, and D.N. Mermin, Festkörperphysik, (Oldenbourg Wis-

senschaftsverlag Gmbh, 2007).
[28] X. Du, I. Skachko, A. Barker, and E.Y. Andrei, Nature Nanotech. 3, 491 (2008).
[29] K.I. Bolotin, K.J. Sikes, J. Hone, H.L. Stormer, and P. Kim, Phys. Rev. Lett. 101,
096802 (2008).
[30] A. Itoh, H. Matsunami, CRC Critical Rev. Solid state and Mater. Sci. 22, 111
(1997).
[31] D.L. Rode, Phys. Rev. B 3, 3287 (1971).
[32] F. Schedin, A.K. Geim, S.V. Morozov, E.W. Hill, P. Blake, M.I. Katsnelson, and
K.S. Novoselov, Nature Mater. 6, 652 (2007).
[33] J. Jobst, D. Waldmann, F. Speck, R. Hirner, D.K. Maude, T. Seyller, and H.B.
Weber, Phys. Rev. B 81, 195434 (2010).
[34] 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, Phys. Rev.
Lett. 101, 267601 (2008).
[35] Y.-W. Son, M.L. Cohen, and S.G. Louie, Phys. Rev. Lett. 97, 216803 (2006).
[36] M.Y. Han, B. Ozyilmaz, Y. Zhang, and P. Kim, Phys. Rev. Lett. 98, 206805
(2007).
Bibliography 151
[37] D.C. Elias, R.R. Nair, T.M.G. Mohiuddin, S.V. Morozov, P. Blake, M.P. Halsall,
A.C. Ferrari, D.W. Boukhvalov, M.I. Katsnelson, A.K. Geim, K.S. Novoselov, Sci-
ence 323, 610 (2009).
[38] K.S. Novoselov, Z. Jiang, Y. Zhang, S.V. Morozov, H.L. Stormer, U. Zeitler, J.C.
Maan, G.S. Boebinger, P. Kim, A. K. Geim, Science 315, 1379 (2007).
[39] X. Wu, Y. Hu, M. Ruan, N.K. Madiomanana, J. Hankinson, M. Sprinkle, C. Berger,

and W.A. de Heer, Appl. Phys. Lett. 95, 223108 (2009).
[40] T. Shen, J.J. Gu, M. Xu, Y.Q. Wu, M.L. Bolen, M.A. Capano, L.W. Engel, and
P.D. Ye, Appl. Phys. Lett. 95, 172105 (2009).
[41] A. Tzalenchuk, S. Lara-Avila, A. Kalaboukhov, S. Paolillo, M. Syväjarvi, R. Yaki-

mova, O. Kazakova, T.J.B.M. Janssen, V. Fal'ko, and S. Kubatkin, Nature Nan-
otech. 5, 186 (2010).
[42] K.I. Bolotin, F. Ghahari, M.D. Shulman, H.L. Stormer, and P. Kim, Nature 462,
196 (2009).
[43] X. Du, I. Skachko, F. Duerr, A. Luican, and E.Y. Andrei, Nature 462, 192 (2009).
[44] C. Lee, X. Wei, J.W. Kysar, J. Hone, Science 321, 385 (2008).
[45] A.A. Balandin, S. Ghosh, W. Bao, I. Calizo, D. Teweldebrhan, F. Miao, C.N. Lau,
Nano Lett. 8, 902 (2008).
[46] J.S. Bunch, S.S. Verbridge, J.S. Alden, A.M. van der Zande, J.M. Parpia, H.G.
Craighead and P.L. McEuen, Nano Lett. 8, 2458 (2008).
[47] B. Partoens, and F.M. Peeters, Phys. Rev. B 74, 075404 (2006).
[48] E. McCann, and V.I. Fal'ko, Phys. Rev. Lett. 96, 086805 (2006).
[49] T. Ohta, A. Bostwick, T. Seyller, K. Horn, E. Rotenberg, Science 313, 951 (2006).
[50] K.S. Novoselov, E. McCann, S.V. Morozov, V.I. Fal'ko, M.I. Katsnelson, U. Zeitler,
D. Jiang, F. Schedin and A.K. Geim, Nature Phys. 2, 177 (2006).
[51] C. Coletti, C. Riedl, D.S. Lee, B. Krauss, K. von Klitzing, J.H. Smet, and U.
Starke, Phys. Rev. B 81, 235401 (2010). See also the accompanying "Viewpoint"
by A. Fedorov, Physics 3, 46 (2010) and the press release 167 in 2010 of the
Max-Planck-Society (9th July 2010).
[52] E.V. Castro, K.S. Novoselov, S.V. Morozov, N.M.R. Peres, J.M.B. Lopes dos San-
tos, J. Nilsson, F. Guinea, A.K. Geim, and A.H. Castro Neto, Phys. Rev. Lett. 99,
216802 (2007).
[53] Y. Zhang, T.-T. Tang, C. Girit, Z. Hao, M.C. Martin, A. Zettl, M.F. Crommie,
Y.R. Shen, and F. Wang, Nature 459, 820 (2009).
[54] T. Ohta, A. Bostwick, J.L. McChesney, T. Seyller, K. Horn, E. Rotenberg, Phys.

Rev. Lett. 98, 206802 (2007).
152 Bibliography
[55] S. Latil, and L. Henrard, Phys. Rev. Lett. 97, 036803 (2006).
[56] S. Shallcross, S. Sharma, and O. Pankratov, Phys. Rev. Lett. 101, 056803 (2008).
[57] M. Sprinkle, D. Siegel, Y. Hu, J. Hicks, A. Tejeda, A. Taleb-Ibrahimi, P. Le Fèvre,

F. Bertran, S. Vizzini, H. Enriquez, S. Chiang, P. Soukiassian, C. Berger, W. A. de
Heer, A. Lanzara, and E. H. Conrad, Phys. Rev. Lett. 103, 226803 (2009).
[58] A.K. Geim, Science 324, 1530 (2009).
[59] Y. Hernandez, V. Nicolosi, M. Lotya, F. Blighe, Z. Sun, S. De, I.T. McGovern, B.

Holland, M. Byrne, Y. Gunko, J. Boland, P. Niraj, G. Duesberg, S. Krishnamurti,
R. Goodhue, J. Hutchison, V. Scardaci, A.C. Ferrari, and J.N. Coleman, Nature
Nanotech. 3, 563 (2008).
[60] D.A. Dikin, S. Stankovich, E.J. Zimney, R.D. Piner, G.H.B. Dommett, G. Evme-
nenko, S.T. Nguyen, and R.S. Ruo, Nature 448, 457 (2007).
[61] S. Stankovich, D.A. Dikin, G.H.B. Dommett, K.M. Kohlhaas, E.J. Zimney, E.A.
Stach, R.D. Piner, S.T. Nguyen, and R.S. Ruo, Nature 442, 282 (2006).
[62] C. Gómez-Navarro, R.T. Weitz, A.M. Bittner, M. Scolari, A. Mews, M. Burghard,

and K. Kern, Nano Lett. 7, 3499 (2007).
[63] H.P. Boehm, A. Clauss, G. Fischer, U. Hofmann, Z. Naturforschg. 17g, 150 (1962).
[64] C. Schafhaeutl, Phil. Mag. 16, 570 (1840).
[65] H. Zi-Pu, D.F. Ogletree, M.A. Van Hove, and G.A. Somorjai, Surf. Sci. 180, 433
(1987).
[66] M. Sasaki, Y. Yamada, Y. Ogiwara, S. Yagyu, and S. Yamamoto, Phys. Rev. B 61,
155653 (2000).
[67] A.B. Preobrajenski, M.L. Ng, A.S. Vinogradov, and N. Mårtensson, Phys. Rev. B
78, 073401 (2008).
[68] N.A. Kholin, E.V. Rut'kov and A.Y. Tontegode, Surf. Sci. 139, 155 (1984).
[69] I. Pletikosi¢, M. Kralj, P. Pervan, R. Brako, J. Coraux, A.T. N'Diaye, C. Busse,

and T. Michely, Phys. Rev. Lett. 102, 056808 (2009).
[70] R. Rosei, M. De Crescenzi, F. Sette, C. Quaresima, A. Savoia, and P. Perfetti,

Phys. Rev. B 28, 1161 (1983).
[71] A.M. Shikin, G.V. Prudnikova, V.K. Adamchuk, F. Moresco, and K.-H. Rieder,
Phys. Rev. B 62, 13202 (2000).
[72] A. Varykhalov, J. Sanchez-Barriga, A.M. Shikin, C. Biswas, E. Vescovo, A. Rybkin,

D. Marchenko, and O. Rader, Phys. Rev. Lett. 101, 157601 (2008).
[73] P.W. Sutter, J.-I. Flege, and E. Sutter, Nature Mater. 7, 406 (2008).
Bibliography 153
[74] A. Nagashima, K. Nuka, K. Sato, H. Itoh, T. Ichinokawa, C. Oshima and S. Otani,

Surf. Sci. 291, 93 (1993).
[75] D. Eom, D. Prezzi, K.T. Rim, H. Zhou, M. Lefenfeld, S. Xiao, C. Nuckoll, M.S.
Hybertsen, T.F. Heinz, and G.W. Flynn, Nano Lett. 9, 2844 (2009).
[76] X. Li, W. Cai, J. An, S. Kim, J. Nah, D. Yang, R. Piner, A. Velamakanni, I. Jung,
E. Tutuc, K. Banerjee, L. Colombo, and R.S. Ruo, Science 324, 1312 (2009).
[77] A. Reina, X. Jia, J. Ho, D. Nezich, H. Son, V. Bulovic, M.S. Dresselhaus, and J.
Kong, Nano Lett. 9, 30 (2009).
[78] K.S. Kim, Y. Zhao, H. Jang, S.Y. Lee, J.M. Kim, K.S. Kim, J.-H. Ahn, P. Kim,
J.-Y. Choi, and B.H. Hong, Nature 457, 706 (2009).
[79] C. Virojanadara, M. Syväjarvi, R. Yakimova, L.I. Johansson, A. A. Zakharov, and

T. Balasubramanian, Phys. Rev. B 78, 245403 (2008).
[80] D.V. Badami, Nature 193, 569 (1962).
[81] A.J. van Bommel, J.E. Crombeen and A. van Tooren, Surf. Sci. 48, 463 (1975).
[82] E.G. Acheson, Amer. Pat. 568323 (1896).
[83] H. Moissan, Comptes Rendus de l'Académie des Sciences 140, 405 (1905).
[84] J.J. Berzelius, Annalen d. Physik 77, 169 (1824).
[85] P.J. Guichelaar, in: Carbide, nitride, and boride materials synthesis and processing
(ed.: A.W. Weimer), p. 115, Acheson process, (Springer, 1997).
[86] E.G. Acheson, Chem. News 68, 179 (1893).
[87] H.J. Round, Electrical World 19, 309 (1907).
[88] J.A. Lely, Berichte der deutschen Keramischen Gesellschaft e.V. 32, 229 (1955).
[89] Y.M. Tairov, and V.F. Tsvetkov, J. Cryst. Growth 43, 209 (1978).
[90] H. Matsunami, S. Nishino, and H. Ono, IEEE Trans. Electron Devices ED-28,
1235, (1981).
[91] N. Kuroda, K. Shibahara, W.S. Yoo, S. Nishino, and H. Matsunami, Extended

Abstracts of the 19th Conference on Solid State Devices and Materials, 227 (1987).
[92] U. Starke, J. Bernhardt, J. Schardt and K. Heinz, Surf. Rev. Lett. 6, 1129 (1999).
[93] G. Pensl, and R. Helbig. Silicon Carbide (SiC) Recent Results in Physics and
in Technology, Band 30 von Festkörperprobleme/Advances in Solid State Physics,
(Vieweg Verlag, 1990).
[94] L.S. Ramsdell, Amer. Mineralogist 32, 64 (1947).
[95] J. Bernhardt, PhD Thesis, University of Erlangen-Nürnberg (2001).

154 Bibliography
[96] G. Ertl, and J. Küppers: Low Energy Electrons and Surface Chemistry (VCH, 1985).
[97] D.A. Siegel, S.Y. Zhou, F. El Gabaly, A.K. Schmid, K.F. Mc Carty, and A. Lanzara,
Phys. Rev. B 80, 241407(R) (2009).
[98] J.B. Hannon, R.M. Tromp, N.V. Medhekar, and V.B. Shenoy, Phys. Rev. Lett.
103, 256101 (2009).
[99] T. Seyller, J. Phys.: Condens. Matter 16, S1755 (2004).
[100] E.A. Wood, J. Appl. Phys. 35, 1306 (1964).
[101] M. Sabisch, P. Kruger, and J. Pollmann, Phys. Rev. B 55, 10561 (1997).
[102] F. Bechstedt, and J. Furthmüller, J. Phys.: Condens. Matter 16, S1721 (2004).
[103] T. Seyller, R. Graupner, N. Sieber, K.V. Emtsev, L. Ley, A. Tadich, J.D. Riley, and
R. C. G. Leckey, Phys. Rev. B 71, 245333 (2005).
[104] C. Virojanadara, M. Hetzel, C. Riedl, L.I. Johansson, W. J. Choyke and U. Starke,

Surf. Sci. 602, 3506 (2008).
[105] U. Starke, Phys. Status Solidi B 246, 1569 (2009).
[106] P. Hofmann (www.philiphofmann.net/surec/node11.html), after A. Zangwill:

Physics at surfaces, (Cambridge University Press, 1996).
[107] M. Henzler, and W. Göpel, Oberächenphysik des Festkörpers, (Teubner, 1994).
[108] K. Heinz, LEED and DLEED, Rep. Prog. Phys. 58, 637 (1995).
[109] M.A. van Hove, W.H. Weinberg, and C.-M. Chan, Low Energy Electron Diraction.
Experiment, Theory and Surface Structure Determination, (Springer, 1986).
[110] E. Bauer, Rep. Prog. Phys. 57, 895 (1994).
[111] H. Hibino, H. Kageshima, F. Maeda, M. Nagase, Y. Kobayashi, and H. Yamaguchi,

Phys. Rev. B 77, 075413 (2008).
[112] K.V. Emtsev, ..., and U. Starke, to be published (2010).
[113] G. Binnig, C.F. Quate and C. Gerber, Phys. Rev. Lett. 56, 930 (1986).
[114] N. Sasaki and M. Tsukada, in: Advances in Scanning Probe Microscopy (eds.: T.
Sakurai, and Y. Watanabe), p. 1 (Springer, 1999).
[115] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel, Phys. Rev. Lett. 49, 57 (1982).
[116] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel, Phys. Rev. Lett. 50, 120 (1983).
[117] C.J. Chen: Introduction to Scanning Tunneling Microscopy, (Oxford University

Press, 1993).
Bibliography 155
[118] S. Hüfner: Photoelectron Spectroscopy: Principles and Applications, 3rd edition

(Springer, 2003).
[119] H. Lüth: Solid Surfaces, Interfaces and Thin Films, 4th Edition (Springer, 2001).
[120] R.C. Fang, and L. Ley, Phys. Rev. B 40, 3818 (1989).
[121] F.J. Himpsel, F.R. McFeely, A. Taleb-Ibrahimi, J.A. Yarmo, G. Hollinger, Phys.
Rev. B 38, 6084 (1988).
[122] M. Hundhausen, R. Püsche, J. Roehrl, and L. Ley, Phys. Status Solidi B 245, 1356
(2008).
[123] L.M. Malard, M.A. Pimenta, G. Dresselhaus, and M.S. Dresselhaus, Physics Re-
ports 473, 51 (2009).
[124] S. Soubatch, S.E. Saddow, S.P. Rao, W.Y. Lee, M. Konuma and U. Starke, Mater.
Sci. Forum 483-485, 761 (2005).
[125] C.L. Frewin, C. Coletti, C. Riedl, U. Starke and S.E. Saddow, Mater. Sci. Forum
615-617, 589 (2009).
[126] C. Coletti, M.J. Jaroszeski, A. Pallaoro, A.M. Ho, S. Iannotta, S.E. Saddow, IEEE
EMBS Proceedings 2007, 5849 (2007).
[127] W. Kern, RCA Review 31, 187 (1970).
[128] U. Starke, C. Bram, P.-R. Steiner, W. Hartner, L. Hammer, K. Heinz, K. Müller,

Appl. Surf. Sci. 89, 175 (1997).
[129] A. Al-Temimy, Master thesis, University of Stuttgart (2009).
[130] M. Hetzel, Diplomarbeit, Institut für Allgemeine Physik der Technischen Universität
Wien (2009).
[131] K. Besocke, Surf. Sci. 181, 145 (1987).
[132] I. Horcas, R. Fernández, J.M. Gómez-Rodríguez, J. Colchero, J. Gómez-Herrero,

and A.M. Baro, Rev. Sci. Instrum. 78, 013705 (2007).
[133] P. Wahl, PhD thesis, University of Konstanz (2005).
[134] R. Nyholm, J.N. Andersen, U. Johansson, B.N. Jensen and I. Lindau, Nucl. Instr.
and Meth. in Phys. Res. A 467-468, 520 (2001).
[135] L. Patthey, T. Schmidt, U. Flechsig, C. Quitmann, M. Shi, R. Betemps, M.

Botkine, R. Abela, in PSI Scientic Report 1999/Volume VII, Surface/Interface,
Spectroscopy Beamline, (Paul Scherrer Institut, 2002).
[136] R. Kaplan, Surf. Sci. 215, 111 (1989).
[137] U. Starke, and C. Riedl, J. Phys.: Condens. Matter 21, 134016 (2009).
156 Bibliography
[138] K. Heinz, J. Bernhardt, J. Schardt and U. Starke, J. Phys.: Condens. Matter 16,
S1705 (2004).
[139] K. Heinz, U. Starke, J. Bernhardt and J. Schardt, Appl. Surf. Sci. 162-163, 9
(2000).
[140] U. Starke, Mater. Sci. Forum 353-356, 205 (2001).
[141] U. Starke, Mater. Res. Soc. Proc. 742, 35 (2003).
[142] U. Starke, J. Schardt, J. Bernhardt, M. Franke, K. Reuter, H. Wedler, K. Heinz,

J. Furthmüller, P. Käckell and F. Bechstedt, Phys. Rev. Lett. 80, 758 (1998).
[143] J. Schardt, J. Bernhardt, U. Starke and K. Heinz, Phys. Rev. B 62, 10335 (2000).
[144] U. Starke, J. Schardt, J. Bernhardt, M. Franke and K. Heinz, Phys. Rev. Lett. 82,
2107 (1999).
[145] P. Mårtensson, F. Owman and L.I. Johansson, Phys. Status Solidi B, 202, 501
(1997).
[146] U. Starke, J. Schardt and M. Franke, Appl. Phys. A 65, 587 (1997).
[147] U. Starke, M. Franke, J. Bernhardt, J. Schardt, K. Reuter and K. Heinz, Mater.

Sci. Forum 264-268, 321 (1998).
[148] I. Forbeaux, J.-M. Themlin and J.-M. Debever, Phys. Rev. B 58, 16396 (1998).
[149] L. Simon, J.L. Bischo and L. Kubler, Phys. Rev. B 60, 11653 (1999).
[150] A. Charrier, A. Coati, T. Argunova, F. Thibaudau, Y. Garreau, R. Pinchaux, I.

Forbeaux, J.-M. Debever, M. Sauvage-Simkin and J.-M. Themlin, J. Appl. Phys.
92, 2479 (2002).
[151] W. Chen, H. Xu, L. Liu, X. Gao, D. Qi, G. Peng, S.C. Tan, Y. Feng, K.P. Loh and
A.T.S. Wee, Surf. Sci. 596, 176 (2005).
[152] A. Mattausch and O. Pankratov, Phys. Rev. Lett. 99, 076802 (2007).
[153] F. Varchon, R. Feng, J. Hass, X. Li, B. Ngoc Nguyen, C. Naud, P. Mallet, J.-Y.
Veuillen, C. Berger, E.H. Conrad, and L. Magaud, Phys. Rev. Lett. 99, 126805
(2007).
[154] P. Lauer, K.V. Emtsev, R. Graupner, T. Seyller, L. Ley, S.A. Reshanov, and H.B.
Weber, Phys. Rev. B 77, 155426 (2008).
[155] K.V. Emtsev, F. Speck, T. Seyller, L. Ley, and J.D. Riley, Phys. Rev. B 77, 155303
(2008).
[156] C. Riedl, U. Starke, J. Bernhardt, M. Franke, and K. Heinz, Phys. Rev. B 76,
245406 (2007).
[157] G. M. Rutter, N. P. Guisinger, J. N. Crain, E. A. A. Jarvis, M. D. Stiles, T. Li, P.

N. First, J. A. Stroscio, Phys. Rev. B 76, 235416 (2007).
Bibliography 157
[158] T. Seyller, K.V. Emtsev, K. Gao, F. Speck, L. Ley, A. Tadich, L. Broekman, J.D.
Riley, R.C.G. Leckey, O. Rader, A. Varykhalov, and A. M. Shikin, Surf. Sci. 600,
3906 (2006).
[159] K.V. Emtsev, T. Seyller, F. Speck, L. Ley, P. Stojanov, J.D. Riley, R.G.C. Leckey,
Mater. Sci. Forum 556-557, 525 (2007).
[160] C. Riedl, C. Coletti, T. Iwasaki, A.A. Zakharov, and U. Starke, Phys. Rev. Lett.
103, 246804 (2009). See also the press release 167 in 2010 of the Max-Planck-
Society (9th July 2010).
[161] S. Kim, J. Ihm, H.J. Choi, and Y.-W. Son, Phys. Rev. Lett. 100, 176802 (2008).
[162] F. Varchon, P. Mallet, J.-Y. Veuillen, and L. Magaud, Phys. Rev. B 77, 235412
(2008).
[163] C. Riedl, A.A. Zakharov, and U. Starke, Appl. Phys. Lett. 93, 033106 (2008).
[164] C. Riedl, C. Virojanadara, and U. Starke, MAX-lab activity report 2007 (eds.: U.
Johansson, A. Nyberg, R. Nyholm, H. Ullman), National Laboratory Lund Sweden
(2008).
[165] E. Lampin, C. Priester, C. Krzeminski, and L. Magaud, J. Appl. Phys. 107, 103514
(2010).
[166] A. Bostwick, T. Ohta, J.L. McChesney, K.V Emtsev, T. Seyller, K. Horn, and E.
Rotenberg, New J. Phys. 9, 385 (2007).
[167] K.V. Emtsev, Dissertation, University Erlangen-Nürnberg (2009).
[168] S. Y. Zhou, G.-H. Gweon, A. V. Fedorov, P. N. First, W. A. de Heer, D.-H. Lee,

F. Guinea, A. H. Castro Neto, and A. Lanzara, Nature Mater. 6, 770 (2007).
[169] A. Bostwick, T. Ohta, T. Seyller, K. Horn, and E. Rotenberg, Nature Phys. 3, 36

(2007).
[170] J. Hass, W.A. de Heer, and E.H. Conrad, J. Phys.: Condens. Matter 20, 323202
(2008).
[171] I. Gierz, C. Riedl, U. Starke, C.R. Ast and K. Kern, Nano Lett. 8, 4603 (2008).
[172] E. Rotenberg, A. Bostwick, T. Ohta, J.L. McChesney, T. Seyller, and K. Horn,

Nature Mater. 7, 258 (2008).
[173] S.Y. Zhou, D.A. Siegel, A.V. Fedorov, F. El Gabaly, A.K. Schmid, A.H. Castro
Neto, D.-H. Lee, and A. Lanzara, Nature Mater. 7, 259 (2008).
[174] S.Y. Zhou, D.A. Siegel, A.V. Fedorov, and A. Lanzara, Physica E 40, 2642 (2008).
[175] C.-H. Park, F. Giustino, C.D. Spataru, M.L. Cohen, and S.G. Louie, Nano Lett. 9,
4234 (2009).
158 Bibliography
[176] L. Vitali, C. Riedl, R. Ohmann, I. Brihuega, U. Starke and K. Kern, Surf. Sci. 602,
L127 (2008).
[177] I. Brihuega, P. Mallet, C. Bena, S. Bose, C. Michaelis, L. Vitali, F. Varchon, L.

Magaud, K. Kern, and J. Y. Veuillen, Phys. Rev. Lett. 101, 206802 (2008).
[178] P. Mallet, F. Varchon, C. Naud, L. Magaud, C. Berger, and J.-Y. Veuillen, Phys.
Rev. B 76, 041403(R) (2007).
[179] M. Mucha-Kruczy«ski, O. Tsyplyatyev, A. Grishin, E. McCann, V.I. Fal'ko, A.

Bostwick, and E. Rotenberg, Phys. Rev. B 77, 195403 (2008).
[180] T. Ohta, F. El Gabaly, A. Bostwick, J.L. McChesney, K.V. Emtsev, A.K. Schmid,
T. Seyller, K. Horn, and E. Rotenberg, New J. Phys. 10, 023034 (2008).
[181] G. Gu, S. Nie, R.M. Feenstra, R.P. Devaty, W.J. Choyke, W.K. Chan and M.G.
Kane, Appl. Phys. Lett. 90, 253507 (2007).
[182] Z.H. Ni, W. Chen, X.F. Fan, J.L. Kuo, T. Yu, A.T.S. Wee, and Z.X. Shen, Phys.
Rev. B 77, 115416 (2008).
[183] J. Röhrl, M. Hundhausen, K.V. Emtsev, T. Seyller, R. Graupner, and L. Ley, Appl.
Phys. Lett. 92, 201918 (2008).
[184] D.S. Lee, C. Riedl, B. Krauss, K. v. Klitzing, U. Starke and J.H. Smet, Nano Lett.
8, 4320 (2008).
[185] C. Riedl, and U. Starke, Mater. Sci. Forum 615-617, 219 (2009).
[186] E. Stolyarova, K.T. Rim, S. Ryu, J. Maultzsch, P. Kim, L.E. Brus, T.F. Heinz,
M.S. Hybertsen and G.W. Flynn, Proc. Natl. Acad. Sci. USA 104, 9209 (2007).
[187] D. Tománek, S.G. Louie, H.J. Mamin, D.W. Abraham, R.E. Thomson, E. Ganz,
and J. Clarke, Phys. Rev. B 35, 7790 (1987).
[188] G.M. Rutter, J.N. Crain, N.P. Guisinger, T. Li, P.N. First, and J.A. Stroscio, Science
317, 219 (2007).
[189] C. Bena, Phys. Rev. Lett. 100, 076601 (2008).
[190] W.V. Brar, Y. Zhang, Y. Yayon, T. Ohta, J.L. McChesney, A. Bostwick, E. Roten-
berg, K. Horn, and M.F. Crommie, Appl. Phys. Lett. 91, 122102 (2007).
[191] J.L. Mañes, F. Guinea, and M.A.H. Vozmediano, Phys. Rev. B 75, 155424 (2007).
[192] Y. Zhang, V.W. Brar, F. Wang, C. Girit, Y. Yayon, M. Panlasigui, A. Zettl, and
M.F. Crommie, Nature Phys. 4, 627 (2008).
[193] J.M. Soler, A.M. Baro, N. Garcia, and H. Rohrer, Phys. Rev. Lett. 57, 444 (1986).
[194] S. Ciraci, and I.P. Batra, Phys. Rev. B 36, 6194 (1987).
[195] F. Guinea, M.I. Katsnelson, and M.A.H. Vozmediano, Phys. Rev. B 77, 075422
(2008).
Bibliography 159
[196] L.M. Malard, J. Nilsson, D.C. Elias, J.C. Brant, F. Plentz, E.S. Alves, A.H. Castro
Neto, and M.A. Pimenta, Phys. Rev. B 76, 201401 (2007).
[197] A.C. Ferrari, J.C. Meyer, V. Scardaci, C. Casiraghi, M. Lazzeri, F. Mauri, S. Pis-
canec, D. Jiang, K.S. Novoselov, S. Roth, and A.K. Geim, Phys. Rev. Lett. 97,
187401 (2006).
[198] A. Gupta, G. Chen, P. Joshi, S. Tadigadapa, and P.C. Eklund, Nano Lett. 6, 2667
(2006).
[199] C. Casiraghi, S. Pisana, K.S. Novoselov, A.K. Geim, and A.C. Ferrari, Appl. Phys.
Lett. 91, 233108 (2007).
[200] D. Graf, F. Molitor, K. Ensslin, C. Stampfer, A. Jungen, C. Hierold, and L. Wirtz,

Nano Lett. 7, 238 (2007).
[201] C. Faugeras, A. Nerrière, M. Potemski, A. Mahmood, E. Dujardin, C. Berger, and

W.A. de Heer, Appl. Phys. Lett. 92, 011914 (2008).
[202] T.W. Sharf, and I.L. Singer, Tribol. Lett. 14, 137 (2003).
[203] J. Hass, F. Varchon, J.E. Millán-Otoya, M. Sprinkle, N. Sharma, W.A. de Heer,

C. Berger, P.N. First, L. Magaud, and E.H. Conrad, Phys. Rev. Lett. 100, 125504
(2008).
[204] S. Pisana, M. Lazzeri, C. Casiraghi, K.S. Novoselov, A.K. Geim, A.C. Ferrari, and
F. Mauri, Nature Mater. 6, 198 (2007).
[205] J. Yan, Y.B. Zhang, P. Kim, and A. Pinczuk, Phys. Rev. Lett. 98, 166802 (2007).
[206] C. Stampfer, F. Molitor, D. Graf, K. Ensslin, A. Jungen, C. Hierold, and L. Wirtz,

Appl. Phys. Lett. 91, 241907 (2007).
[207] A. Das, S. Pisana, B. Chakraborty, S. Piscanec, S.K. Saha, U.V. Waghmare, K.S.
Novoselov, H.R. Krishnamurthy, A.K. Geim, A.C. Ferrari, and A.K. Sood, Nature
Nanotech. 3, 210-215 (2008).
[208] J. Bernhardt, M. Nerding, U. Starke, and K. Heinz, Mater. Sci. Eng. B 61-62, 207
(1999).
[209] J. Bernhardt, J. Schardt, U. Starke and K. Heinz, Appl. Phys. Lett. 74, 1084
(1999).
[210] J. Bernhardt, A. Seubert, M. Nerding, U. Starke and K. Heinz, Mater. Sci. Forum
338-342, 345 (2000).
[211] A. Seubert, J. Bernhardt, M. Nerding, U. Starke and K. Heinz, Surf. Sci. 454-456,
45 (2000).
[212] F. Hiebel, P. Mallet, F. Varchon, L. Magaud, and J.-Y. Veuillen, Phys. Rev. B 78,
153412 (2008).
160 Bibliography
[213] F. Hiebel, P. Mallet, L. Magaud, and J.-Y. Veuillen, Phys Rev. B 80, 235429
(2009).
[214] F. Varchon, P. Mallet, L. Magaud, and J.-Y. Veuillen Phys. Rev. B 77, 165415
(2008).
[215] J. Hass, R. Feng, J.E. Millán-Otoya, X. Li, M. Sprinkle, P.N. First, W.A. de Heer,
E.H. Conrad, and C. Berger, Phys. Rev. B 75, 214109 (2007).
[216] I. Forbeaux, J.-M. Themlin, and J.-M. Debever, Surf. Sci. 442, 9 (1999).
[217] H.M. Hoster, M.A. Kulakov, and B. Bullemer, Surf. Sci. 382, L658 (1997).
[218] L. Magaud, F. Hiebel, F. Varchon, P. Mallet, and J.-Y. Veuillen, Phys. Rev. B 79,
161405(R) (2009).
[219] M. Suemitsu, Y. Miyamoto, H. Handa, and A. Konnoe, e-J. Surf. Sci. Nanotech.
7, 311 (2009).
[220] A. Ouerghi, M. Portail, A. Kahouli, L. Travers, T. Chassagne, and M. Zielinski,
Mater. Sci. Forum 645-648, 585 (2010).
[221] N. Camara, J.R. Huntzinger, A. Tiberj, G. Rius, B. Jouault, F. Perez-Murano, N.

Mestres, P. Godignon, and J. Camassel, Mater. Sci. Forum 615-617, 203 (2009).
[222] C. Virojanadara, R. Yakimova, J.R. Osiecki, M. Syväjarvi, R.I.G. Uhrberg, L.I.

Johansson, and A.A. Zakharov, Surf. Sci. 603, L87 (2009).
[223] L.C. Campos, V.R. Manfrinato, J.D. Sanchez-Yamagishi, J. Kong, and P. Jarillo-
Herrero, Nano Lett. 9, 2600 (2009).
[224] C. Virojanadara, M. Hetzel and U. Starke, Appl. Phys. Lett. 92, 061902 (2008).
[225] C. Virojanadara, M. Hetzel, L.I. Johansson, W. J. Choyke, U. Starke, Surf. Sci.

602, 525 (2008).
[226] B. Baumeier, P. Krüger, and J. Pollmann, Phys. Rev. B 78, 245318 (2008).
[227] J. Robinson, X. Weng, K. Trumbull, R. Cavalero, M. Wetherington, E. Frantz, M.

LaBella, Z. Hughes, M. Fanton, and D. Snyder, ACS Nano 4, 153 (2010).
[228] S. Tanaka, K. Morita, and H. Hibino, Phys. Rev. B 81, 041406(R) (2010).
[229] W.Y. Lee, S. Soubatch, and U. Starke, Mater. Sci. Forum 527-529, 673 (2006).
[230] A. Al-Temimy, C. Riedl, and U. Starke, Appl. Phys. Lett. 95, 231907 (2009).
[231] E. Moreau, F.J. Ferrer, D. Vignaud, S. Godey, and X. Wallart, Phys. Status Solidi
A 207, 300 (2010).
[232] J.B. Hannon, and R.M. Tromp, Phys. Rev. B 77, 241404(R) (2008).
[233] M. Hupalo, E.H. Conrad, and M.C. Tringides, Phys. Rev. B 80, 041401(R) (2009).
Bibliography 161
[234] R.M. Tromp, and J.B. Hannon, Phys. Rev. Lett. 102, 106104 (2009).
[235] J.B. Oostinga, H.B. Heersche, X. Liu, A.F. Morpurgo, and L.M.K. Vandersypen,
Nature Mater. 7, 151 (2008).
[236] J. Kedzierski, P.-L. Hsu, P. Healey, P.W. Wyatt, C.L. Keast, M. Sprinkle, C. Berger,
and W.A. de Heer, IEEE Trans. Electron Devices 55, 2078 (2008).
[237] Y.-M. Lin, C. Dimitrakopoulos, K.A. Jenkins, D.B. Farmer, H.-Y. Chiu, A. Grill, P.
Avouris, Science 327, 662 (2010).
[238] V. Chakrapani, J.C. Angus, A.B. Anderson, S.D. Wolter, B.R. Stoner, and G.U.
Sumanasekera, Science 318, 1424 (2007).
[239] J. Ristein, Science 313, 1057 (2006).
[240] S.J. Sque, R. Jones, and P.R. Briddon, Phys. Status Solidi A 204, 3078 (2007).
[241] S.Y. Zhou, D.A. Siegel, A.V. Fedorov, and A. Lanzara, Phys. Rev. Lett. 101,
086402 (2008).
[242] E.H. Hwang, S. Adam, and S. Das Sarma, Phys. Rev. B 76, 195421 (2007).
[243] T.O. Wehling, K.S. Novoselov, S.V. Morozov, E.E. Vdovin, M.I. Katsnelson, A.K.
Geim, and A.I. Lichtenstein, Nano Lett. 8, 173 (2008).
[244] W. Chen, D. Qi, X. Gao, and A.T.S. Wee, Prog. Surf. Sci. 84, 279 (2009).
[245] E. Bekyarova, M.E. Itkis, P. Ramesh, C. Berger, M. Sprinkle, W.A. de Heer, and
R.C. Haddon, J. Am. Chem. Soc. 131, 1336 (2009).
[246] Y.H. Lu, W. Chen, Y.P. Feng, and P.M. He, J. Phys. Chem. B 113, 2, (2009).
[247] J. Blochwitz, M. Pfeier, T. Fritz, and K. Leo, Appl. Phys. Lett. 73, 729 (1998).
[248] X. Zhou, M. Pfeier, J. Blochwitz, A. Werner, A. Nollau, T. Fritz, and K. Leo,

Appl. Phys. Lett. 78, 410 (2001).
[249] T.-C. Tseng, C. Urban, Y. Wang, R. Otero, S.L. Tait, M. Alcamí, D. Écija, M.
Trelka, J.M. Gallego, N. Lin, M. Konuma, U. Starke, A. Nefedov, A. Langner,
C. Wöll, M.Á. Herranz, F. Martín, N. Martín, K. Kern, and R. Miranda, Nature
Chem., doi:10.1038/nchem.591 (2010).
[250] W. Chen, S. Chen, D.C. Qui, X.Y. Gao, and A.T.S. Wee, J. Am. Chem. Soc. 129,
10418 (2007).
[251] S.K. Wells, J. Giergel, T.A. Land, J.M. Lindquist, and J.C. Hemminger, Surf. Sci.
257, 129 (1991).
[252] J.M. Lindquist, and J.C. Hemminger, J. Phys. Chem. 92, 1394 (1998).
[253] L. Romaner, G. Heimel, J.-L. Bredas, A. Gerlach, F. Schreiber, R.L. Johnson, J.

Zegenhagen, S. Duhm, N. Koch, and E. Zojer, Phys. Rev. Lett. 99, 256801 (2007).
162 Bibliography
[254] D. Qi, W. Chen, X. Gao, L. Wang, S. Chen, K.P. Loh, and A.T.S. Wee, J. Am.
Chem. Soc. 129 8084 (2007).
[255] C. Riedl, C. Coletti, T. Iwasaki, and U. Starke, Mater. Sci. Forum 645-648, 623
(2010).
[256] A. Bostwick, J.L. McChesney, K.V. Emtsev, T. Seyller, K. Horn, S.D. Kevan, and
E. Rotenberg, Phys. Rev. Lett. 103, 056404 (2009).
[257] N.P. Guisinger, G.M. Rutter, J.N. Crain, P.N. First, and J.A. Stroscio, Nano Lett.
9, 1462 (2009).
[258] R. Balog, B. Jørgensen, J. Wells, Erik Lægsgaard, P. Hofmann, F. Besenbacher,
and L. Hornekær, J. Am. Chem. Soc. 131, 8744 (2009).
[259] C. Coletti, C.L. Frewin, A.M. Ho, and S.E. Saddow, Electrochem. Solid-State
Lett. 11, H285 (2008).
[260] H. Tsuchida, I. Kamata, and K. Izumi, J. Appl. Phys. 85, 3569 (1999).
[261] unpublished.
[262] N. Sieber, T. Seyller, L. Ley, D. James, J.D. Riley, R.C.G. Leckey, and M. Polcik,
Phys. Rev. B 67, 205304 (2003).
[263] B. Lee, S. Han, and Y.-S. Kim, Phys. Rev. B 81, 075432 (2010).
[264] A. Ito, H. Nakamura, A. Takayama, J. Phys. Soc. Japan 77, 114602 (2008).
[265] J. Soltys, J. Piechota, M. Lopuszynski, and S. Krukowski, arXiv:1002.4717v2
(2010).
[266] F. Speck, M. Ostler, J. Röhrl, J. Jobst, D. Waldmann, M. Hundhausen, L. Ley,
H.B. Weber, and T. Seyller, Mater. Sci. Forum 645-648, 629 (2010).
[267] I. Gierz, T. Suzuki, R.T. Weitz, D.S. Lee, B. Krauss, C. Riedl, U. Starke, H. Höchst,
J.H. Smet, C.R. Ast, and K. Kern, Phys. Rev. B 81, 235408 (2010).
[268] G. Giovanetti, P.A. Khomyakov, G. Brocks, V.M. Karpan, J. van der Brink, and
P.J. Kelly, Phys. Rev. Lett. 101, 026803 (2008).
[269] C.L. Kane, and E.J. Mele, Phys. Rev. Lett. 95, 226801 (2005).
[270] H. Min, J.E. Hill, N.A. Sinitsyn, B.R. Sahu, L. Kleinman, and A.H. MacDonald,
Phys. Rev. B 74, 165310 (2006).
[271] E.I. Rashba, Phys. Rev. B 79, 161409(R) (2009).
[272] L. Kubler, K. Aït-Mansour, M. Diani, D. Dentel, J.-L. Bischo, and M. Derivaz,
Phys. Rev. B 72, 115319 (2005).
[273] T.E. Weller, M. Ellerby, S.S. Saxena, R.P. Smith, and N.T. Skipper, Nature Phys.
1, 39 (2005).
[274] J.S. Kim, L. Boeri, J.R. O'Brien, F.S. Razavi, and R.K. Kremer, Phys. Rev. Lett.
99, 027001 (2007).
Publications
Peer-reviewed journals
1. C. Riedl, U. Starke, J. Bernhardt, M. Franke and K. Heinz, Phys. Rev. B 76, 245406
(2007), Structural properties of the graphene-SiC(0001) interface as a key for the prepara-
tion of homogeneous large-terrace graphene surfaces.
2. L. Vitali, C. Riedl, R. Ohmann, I. Brihuega, U. Starke and K. Kern, Surf. Sci. 602,
L127-L130 (2008), Spatial modulation of the Dirac-gap in epitaxial graphene.
3. C. Riedl, A.A. Zakharov, and U. Starke, Appl. Phys. Lett. 93, 033106 (2008), Precise in
situ thickness analysis of epitaxial graphene layers on SiC(0001) using low-energy electron
diraction and angle resolved ultraviolet photoelectron spectroscopy.
4. C. Virojanadara, M. Hetzel, C. Riedl, L.I. Johansson, W.J. Choyke and U. Starke, Surf. Sci.
602, 3506-3509 (2008), Silicon adatom chains and one-dimensionally conned electrons
on 4H-SiC(11̄ 02): The (2x1) reconstruction.
5. D.S. Lee, C. Riedl, B. Krauss, K. von Klitzing, U. Starke and J.H. Smet, Nano Lett. 8,
4320-4325 (2008), Raman spectra of epitaxial graphene on SiC and of epitaxial graphene
transferred to SiO2 .
6. I. Gierz, C. Riedl, U. Starke, C.R. Ast and K. Kern, Nano Lett. 8, 4603-4607 (2008),
Atomic hole doping in graphene.
7. U. Starke and C. Riedl, J. Phys.: CM 21, 134016 (2009), Epitaxial graphene on SiC(0001)
and SiC(0001̄ ): from surface reconstructions to carbon electronics.
8. C. Riedl, C. Coletti, T. Iwasaki, A.A. Zakharov, and U. Starke, Phys. Rev. Lett. 103,
246804 (2009), Quasi-free standing epitaxial graphene on SiC obtained by hydrogen in-
tercalation. See also the press release 167 in 2010 of the Max-Planck-Society (9th July
2010).
9. A. Al-Temimy, C. Riedl, and U. Starke, Appl. Phys. Lett. 95, 231907 (2009), Low
temperature growth of epitaxial graphene on SiC induced by carbon evaporation.
10. C. Coletti, C. Riedl, D.S. Lee, B. Krauss, K. von Klitzing, J.H. Smet, and U. Starke, Phys.
Rev. B 81, 235401 (2010), Editor's suggestion, Charge neutrality and band-gap tuning of
epitaxial graphene on SiC by molecular doping. See also the accompanying "Viewpoint"
by A. Fedorov, Physics 3, 46 (2010) and the press release 167 in 2010 of the Max-Planck-
Society (9th July 2010).
164 Publications
11. I. Gierz, T. Suzuki, R.T. Weitz, D.S. Lee, B. Krauss, C. Riedl, U. Starke, H. Höchst,
J.H. Smet, C.R. Ast, and K. Kern, Phys. Rev. B 81, 235408 (2010), Editor's suggestion,
Electronic decoupling of an epitaxial graphene monolayer by gold intercalation.
12. C. Riedl, C. Coletti, and U. Starke, to appear in J. Phys. D: Appl. Phys. (2010),
Structural and electronic properties of epitaxial graphene on SiC(0001): a review of growth,
characterization, transfer doping and hydrogen intercalation.
Conference proceedings
1. C. Riedl, J. Bernhardt, K. Heinz and U. Starke, Mat. Sci. Forum 600-603, 563 (2009),
Evolution and structure of graphene layers on SiC(0001).
2. C. Riedl and U. Starke, Mat. Sci. Forum 615-617, 219 (2009), Structural and electronic
properties of epitaxial graphene on SiC(0001).
3. C.L. Frewin, C. Coletti, C. Riedl, U. Starke and S.E. Saddow, Mat. Sci. Forum 615-617,
589 (2009), A comprehensive study of hydrogen etching on the major SiC polytypes and
crystal orientations.
4. A. Al-Temimy, C. Riedl, and U. Starke, Mat. Sci. Forum 645-648, 593 (2010), Growth
and characterization of epitaxial graphene on SiC induced by carbon evaporation.
5. C. Riedl, C. Coletti, T. Iwasaki, and U. Starke, invited, Mat. Sci. Forum 645-648, 623
(2010), Hydrogen intercalation below epitaxial graphene on SiC(0001).
Annual reports
1. C. Riedl and U. Starke, J. Bernhardt, M. Franke and K. Heinz, Annual report 2007,
Max-Planck-Institut für Festkörperforschung, Stuttgart (2008), Evolution and structure of
graphene layers on SiC(0001).
2. C. Riedl, C. Virojanadara, and U. Starke, MAX-lab activity report 2007 (eds.: U. Jo-
hansson, A. Nyberg, R. Nyholm, H. Ullman), National Laboratory Lund Sweden (2008),
Electronic and structural properties of graphene layers on SiC(0001).
3. C. Riedl, A.A. Zakharov, and U. Starke, MAX-lab activity report 2007 (eds.: U. Johansson,
A. Nyberg, R. Nyholm, H. Ullman), National Laboratory Lund Sweden (2008), LEEM
studies of epitaxial graphene growth on SiC(0001).
4. C. Riedl and U. Starke, Annual report 2008, Max-Planck-Institut für Festkörperforschung,
Stuttgart (2009), Structural and electronic properties of epitaxial graphene on SiC(0001).
5. C. Riedl, C. Coletti, T. Iwasaki, and U. Starke, Annual report 2009, Max-Planck-Institut
für Festkörperforschung, Stuttgart (2010), Quasi-free standing epitaxial graphene on SiC
obtained by hydrogen intercalation.
6. C. Riedl, C. Coletti, T. Iwasaki, and U. Starke, MAX-lab activity report 2009, National
Laboratory Lund Sweden, submitted (2010), Quasi-free standing epitaxial graphene on
SiC(0001) obtained by hydrogen intercalation.
Publications 165
7. C. Coletti, C. Riedl, K.V. Emtsev, T. Iwasaki, A. Al-Temimy, S. Forti, A.A. Zakharov,

and U. Starke, MAX-lab activity report 2009, National Laboratory Lund Sweden, submitted
(2010), LEEM studies of quasi-free standing epitaxial graphene on SiC(0001) obtained by
hydrogen intercalation.
166 Publications
Danksagung
An dieser Stelle möchte ich mich bei allen bedanken, die zum Gelingen dieser Arbeit beige-
tragen haben. Im Einzelnen geht dieser Dank an:
• meinen Doktorvater PD Dr. Ulrich Starke für die Möglichkeit der Promotion in seiner Ar-
beitsgruppe und den damit verbundenen Einstieg in ein sehr aktuelles und durchaus aufre-
gendes Themengebiet. Ihm schulde ich groÿen Dank für die sehr direkte und unkomplizierte
Betreuung, für seine Förderung und für die Möglichkeiten, die er mir im Rahmen der Arbeit
erönet hat.
• diejenigen Organisationen, die die Finanzierung dieser Doktorarbeit ermöglicht haben, d.h.
die Max-Planck-Gesellschaft und das Max-Planck-Institut für Festkörperforschung, sowie in
Hinblick auf die Strahlzeiten am MAX-lab Synchrotron in Lund/Schweden und an der Swiss
Light Source in Villigen/Schweiz die Europäische Union mit ihren Rahmenprogrammen FP6
und FP7 zur Unterstützung der Forschungslandschaft in Europa.
• Prof. Dr. Heiko Weber und Prof. Dr. Oleg Pankratov für Ihre Bereitschaft die Aufgaben im
Prüfungskollegium wahrzunehmen, sowie an Prof. Dr. Andreas Magerl für die Übernahme
des Zweitgutachtens und insbesondere an Prof. Dr. Karsten Horn (Fritz-Haber-Institut der
Max-Planck-Gesellschaft) für die nicht selbstverständliche Übernahme des Drittgutachtens
meiner Arbeit.
• Dr. Ulrich Wedig für sein groÿes Interesse, mit dem er den Fortgang meiner Arbeit im
Rahmen des "PhD Advisory Committee" begleitet hat.
• Dr. Camilla Coletti für das gemeinsame Überleben der Strahlzeiten sowie ihre Beiträge
und groÿe Unterstützung bei den Arbeiten zur Graphen-Dotierung mit F4-TCNQ und zur
Wassersto-Interkalation, die zu den Hauptergebnissen meiner Arbeit zählen.
• Ameer Al-Temimy, der im Rahmen seiner Master-Arbeit einen wesentlichen Beitrag zu
Kapitel 5 meiner Arbeit geliefert hat.
• alle Mitglieder der Grenzächenanalytik-Gruppe für ein äuÿerst angenehmes Arbeitsklima
sowie für alle fachlichen und auÿerfachlichen Diskussionen und Aktivitäten. Dies gilt ins-
besondere in Hinblick auf die Beiträge zu den Ergebnissen meiner Arbeit (soweit noch nicht
genannt) für Dr. Konstantin Emtsev, Dr. Takayuki Iwasaki, Dr. Mitsuharu Konuma und
Stiven Forti, sowie in Hinblick auf die technische Unterstützung für Artur Küster, Axel
Köhler sowie für meinen Zimmerkollegen Tolga Acartürk. Nicht unerwähnt bleiben sollen
Masrrah Al-Ahmad sowie die ehemaligen Gruppenmitglieder Dr. Chariya Virojanadara und
Martin Hetzel.
• Dr. Dong Su Lee, Benjamin Krauss, Dr. Jurgen Smet und Prof. Dr. Klaus von Klitzing
für die sehr angenehme Zusammenarbeit auf dem Gebiet der Raman-Spektroskopie und
Transportmessungen.
• Isabella Gierz, Dr. Christian Ast und Prof. Dr. Klaus Kern für die Unterstützung bei den Pho-
toemissionsmessungen im Labor 7B15 und die sehr erfolgreiche Zusammenarbeit bezüglich
der Dotierung von Graphen mit Bi und Sb sowie der Interkalation von Au.
168 Danksagung
• Dr. Lucia Vitali, Robin Ohmann, Dr. Ivan Brihuega und Prof. Dr. Klaus Kern für die
Durchführung von Tieftemperatur-STM- und STS-Messungen.
• alle Nutzer des Labors in 7B15, die trotz nicht immer einfach zu organisierender "Messzeiten"
zu einem angenehmen Laboralltag beigetragen haben.
• Hansjörg Ruppender von der Firma Omnivac für die fortwährende und unkomplizierte Un-
terstützung bei der Wartung und Weiterentwicklung unserer UHV-Anlagen.
• alle für die Infrastruktur der Max-Planck-Institute in Stuttgart zuständigen Mitarbeiter, die
zu in der Tat exzellenten Forschungsbedingungen beitragen und einem die Arbeit in vielerlei
Hinsicht sehr erleichtern.
• Dr. Alexei Zakharov für die LEEM-Messungen und Dr. Karina Schulte für den Beamline-
Support an der Beamline I311 des MAX-lab-Synchrotrons in Lund/Schweden.
• Dr. Luc Patthey, Dr. Xiaoyu Cui and Fritz Dubi von der SIS-Beamline an der Swiss Light
Source in Villigen/Schweiz.
• Dr. Christopher Frewin, Christopher Locke sowie Prof. Dr. Stephen Saddow von der Uni-
versity of South Florida, USA, v.a. für das Wassersto-Ätzen unserer Proben.
• alle Mitglieder des Lehrstuhls für Festkörperphysik (von Prof. Dr. Thomas Fauster) in Erlan-
gen, dem ich während meiner Doktorandenzeit als "Rydberg-Atom" erhalten geblieben bin.
Dies gilt vor allem für Rosemarie Feldner, Christine Staudt, (Doktor-Opa) Prof. Dr. Klaus
Heinz, Dr. Lutz Hammer, Prof. Dr. Stefan Müller sowie für meine "Weggefährten" Carsten
Tröppner, Matthias Gubo und insbesondere Dr. Andreas Dobler.
• meine Freunde und meine Studienkollegen für eine sehr abwechslungsreiche und schöne
Studien- und Doktorandenzeit mit vielen unvergesslichen Erlebnissen.
• an meine Familie, besonders an meine Eltern, die mich stets in jeglicher Hinsicht unter-
stützen.
• und in ganz besonderer Weise an meine Freundin Verena für ihre Unterstützung, Geduld
und Nachsicht bei all meinen fachlichen und auÿerfachlichen Aktivitäten.
Curriculum Vitae
PERSÖNLICHE DATEN
Name Christian Helmut Riedl

Geburtsdatum 14.06.1980
Geburtsort Nürnberg
Anschrift Robert-Leicht-Str. 135b
70569 Stuttgart
Familienstand ledig
SCHULAUSBILDUNG
1986 - 1990 Grundschule, Nürnberg

1990 - 1999 Sigmund-Schuckert-Gymnasium, Nürnberg
1999 Allgemeine Hochschulreife
ZIVILDIENST
09/1999 - 08/2000 Sana-Klinik Nürnberg GmbH
HOCHSCHULAUSBILDUNG
10/2000 - 03/2006 Studium der Physik (Diplom) an der Friedrich-Alexander-Universität

Erlangen-Nürnberg
Auslandssemester an der Université des Sciences et Technologies

de Lille, Frankreich
Diplomarbeit am Lehrstuhl für Festkörperphysik:

"Segregation an der CoAl(111)-Oberäche eine Untersuchung
auf ab-initio-Basis"
03/2006 Abschluss: Dipl.-Phys.
PROMOTION
seit 07/2006 Wissenschaftlicher Mitarbeiter in der Servicegruppe Grenzächen-

analytik am Max-Planck-Institut für Festkörperforschung, Stuttgart
07/2010 Promotion zum Dr. rer. nat.:

"Epitaktisches Graphen auf Siliziumkarbid-Oberächen:
Wachstum, Charakterisierung, Dotierung und Wassersto-Interkalation"

Dissertação - ChristianRiedl

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Dissertação - ChristianRiedl

Hochgeladen von

Copyright:

Verfügbare Formate

Epitaxial Graphene

on Silicon Carbide Surfaces:

( Epitaktisches Graphen auf Siliziumkarbid-Oberächen:

Der Naturwissenschaftlichen Fakultät

Tag der mündlichen Prüfung: 09.07.2010

Erstberichterstatter: Priv.-Doz. Dr. Ulrich Starke

Zweitberichterstatter: Prof. Dr. Andreas Magerl

Drittberichterstatter: Prof. Dr. Karsten Horn

4 Characterization of epitaxial graphene on SiC(0001) 43

5 Variation of the graphene growth conditions 89

6 Tuning the material properties of epitaxial graphene on SiC(0001) 109

2.1 Introduction to graphene

The electronic structure of monolayer graphene

E± (~q ) ≈ ± h̄vF |~q | , (2.5)

"Relativistic condensed matter physics"

where ~σ is the two-dimensional Pauli matrix. The corresponding wave function is

Graphene - a playground for new applications

Bilayer graphene and graphene stacks

γ12 U2 v32 √ 1/2

2.1.2 Synthesis of graphene

2.2 Silicon carbide as a template for epitaxial graphene

2.2.1 SiC crystal structure

Si GaAs 4H-SiC 6H-SiC 3C-SiC

2.2.2 SiC surfaces

3.1 Investigation methods

3.1.1 Low energy electron diraction (LEED) and microscopy

I (~k, ~k0 ) = |F (~k, ~k0 )|2 · |S(~k − ~k0 )|2 . (3.2)

in surface structure determinations. However, even without performing a structure analysis a

3.1.2 Scanning probe microscopy

Atomic force microscopy

Scanning tunneling microscopy (STM) and spectroscopy (STS)

the tunneling current amounts to

I (d, U) ∝ eUρsample (z = 0, EF )e −2κd ∝ eUρsample (z = d, EF ) . (3.5)

dI /dV ∝ ρsample (EF − eU) . (3.8)

3.1.3 Photoelectron spectroscopy

Core level photoelectron spectroscopy (CLPES)

EB (k)K .T . = −k . (3.11)

EB (k) = −k − ∆relax . (3.12)

∆EB = −∆k − ∆(∆relax ) . (3.13)

Angle resolved photoemission spectroscopy (ARPES)

~kik = 1/h̄ 2mEkin sin(θ) = 0.512 · Ekin [eV ] · sin(θ) [Å−1 ] .

3.1.4 Raman spectroscopy

3.3 Experimental setup and sample preparation

carbon lm [129].

UHV system for photoelectron spectroscopy

UHV system for STS at 6K

Atmospheric pressure graphitization system

Hydrogenation of graphene samples

Beamline I311 of the MAX radiation synchrotron laboratory

4.1 The phase diagram of SiC(0001)

Figure 4.1: Phase diagram of the SiC(0001) surface:

structure with Si adatoms. A further annealing step √ without

Figure 4.2: (a) Top and (b) side

mesh. Only the brightest spots of the LEED pattern √ in panel

The role of (5×5) domains

struction of 4H-SiC(0001) for three dier-

Figure 4.8: (a),(b) Atomically

using He II radiation. The measurement

Figure 4.11: Structural model of (a) monolayer √and (b)

4.3 The growth and identication of graphene layers

4.3.1 Core level photoelectron spectroscopy

Figure 4.12: C 1s spectrum for (a) mono- and (b)√ bilayer

Figure 4.13: Si 2p spectra for (a) dierent graphene√coverages

C 1s, 600eV SiC Graphene (G) S1 (interface) Figure

( Epitaktisches Graphen auf Siliziumkarbid-Oberächen:

3.1.1 Low energy electron diraction (LEED) and microscopy

EB (k)K .T . = −k . (3.11)

EB (k) = −k − ∆relax . (3.12)

∆EB = −∆k − ∆(∆relax ) . (3.13)

carbon lm [129].

struction of 4H-SiC(0001) for three dier-

4.3 The growth and identication of graphene layers

Figure 4.13: Si 2p spectra for (a) dierent graphene√coverages

4.3.3 LEED ngerprints for the determination of the number of

4.3.5 Homogeneity of the graphene lms

Low energy electron diraction

5.1.2 Inuence of the polytype

O-axis SiC(0001) surfaces

to be increased, which is fullled by higher annealing temperatures. In order to avoid complete

[7] G.W. Semeno, Phys. Rev. Lett. 53, 2449 (1984).