Atmospheric pressure graphitization of SiC(0001) – A route
towards wafer-size graphene layers
Konstantin V. Emtsev
1
, Aaron Bostwick
2
, Karsten Horn
3
, Johannes Jobst
4
, Gary L. Kellogg
5
,
Lothar Ley
1
, Jessica L. McChesney
2
, Taisuke Ohta
5
, Sergey A. Reshanov
4
, Eli Rotenberg
2
,
Andreas K. Schmid
6
, Daniel Waldmann
4
, Heiko B. Weber
4
, Thomas Seyller
1,*
1
Lehrstuhl für Technische Physik, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
2
Advanced Light Source, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
3
Department of Molecular Physics, Fritz-Haber-Institut der Max-Planck-Gesellschaft, Berlin, Germany
4
Lehrstuhl für Angewandte Physik, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany
5
Sandia National Laboratories, Albuquerque, NM, USA
6
National Center for Electron Microscopy, Lawrence Berkeley National Laboratory, Berkeley, CA, USA
* Corresponding author: thomas.seyller@physik.uni-erlangen.de
Graphene, a single monolayer of graphite, has recently attracted considerable interest
due to its novel magneto transport properties
1-3
, high carrier mobility, and ballistic
transport up to room temperature
4
. It has technological applications as a potential
successor of silicon for the post Moore’s law era
5-7
, as single molecule gas sensor
8
, in
spintronics
9-11
, in quantum computing
12
, or as terahertz oscillators
13
. For such
applications, uniform ordered growth of graphene on an insulating substrate is
necessary. While on the one hand exfoliation of graphene from graphite yields high
quality crystals, such isolated samples with dimensions in the 10 micrometer range are
unsuitable for large-scale device production; on the other hand the vacuum
decomposition of SiC yields wafer-sized samples with small grains (30-200nm) that are
equally unsuitable. Here we show that the ex-situ graphitization of Si-terminated
SiC(0001) in an argon atmosphere of about 1 bar produces monolayer graphene films
with much larger domain sizes than previously attainable. Hall measurements confirm the
quality of the films thus obtained. For two different geometries, high electronic mobilities
were found, which reach µ = 2000 cm
2
/Vs at T=27 K. A linear decrease of the mobility
towards higher temperature is observed, which is presumably related to electron-electron
interaction. This method establishes the synthesis of graphene films on a technologically
viable basis.
The preparation of single layer graphene by the thermal decomposition of SiC is envisaged
as a viable route for the synthesis of uniform, wafer-size graphene layers for technological
applications, but the large scale structural quality is presently limited by the lack of continuity and
uniformity of the grown film
15,16
. On the Si-terminated (0001) basal plane, vacuum annealing
leads to small graphene domains typically 30-100 nm in diameter, while on the C-terminated
(
1000 ) face, larger domains (~200 nm) of multilayered, rotationally disordered graphene have
been produced
14
. The small-grain structure is due to morphological changes of the surface in
the course of high temperature annealing. Moreover, decomposition of SiC is not a self-limiting
process and, as a result, regions of different film thicknesses coexist, as shown by low-energy
electron microscopy (LEEM)
15,16
. Such inhomogeneous films do not meet the demands of large
scale device production which requires larger grains and tighter thickness control.
Homogeneous film thickness is particularly important because the electronic structure of the film
depends strongly on the number of layers. For example, while monolayer graphene is a gapless
semiconductor, a forbidden gap can be induced in bilayer graphene and tuned by an external
electrostatic potential
12,17-20
.
We have devised a method of preparing graphene on SiC which results in a significantly
improved film quality. Consider the data in Fig. 1, where we compare samples prepared by
vacuum annealing with samples produced by
ex-situ annealing under argon atmosphere. Panels
(a)-(c) show the morphology of the SiC (0001) surface before and after the formation of a
graphene monolayer by annealing in ultra high vacuum (UHV) as determined by atomic force
microscopy (AFM) and LEEM. The initial SiC(0001) surface in Fig.1(a), obtained after hydrogen
etching, is characterized by wide, highly uniform, atomically flat terraces. The step direction and
terrace width (on the order of 300-700 nm) are determined by the incidental misorientation of the
substrate surface with respect to the crystallographic (0001) plane. The step height is 15 Å
which corresponds to the dimension of the 6H-SiC unit cell in the direction perpendicular to the
surface (c-axis). On defect-free areas of the sample, the terraces typically extend undisturbed
over 50 µm in length. The morphology of the surface covered with a monolayer of graphene
prepared by vacuum annealing is shown in Fig. 1(b). The surface obviously undergoes
significant modifications; it is now covered with small pits up to 10 nm in depth, and the original
steps are hardly discernible any longer. This indicates that graphene growth is accompanied by
substantial changes in the morphology of the substrate itself, leading to a considerable
roughening. As a consequence of this roughening, the graphene layer acquires an
inhomogeneous thickness distribution as can be seen in the LEEM image shown in Fig.1(c). The
irregularly-shaped graphene islands are at most a few hundred nm in size, in agreement with x-
ray diffraction
14
. Moreover, monolayer graphene areas coexist with graphene bilayer islands as
well as with uncovered regions of the (6√3×6√3) buffer layer
21
.
In stark contrast to the low quality resulting from vacuum graphitization (Fig. 1(b)), films
grown under 900 mbar of argon have a greatly improved surface morphology, as demonstrated
by the AFM image in Fig. 1(d). Step bunching is manifested by the formation of macro-terraces
with a width that increases from about 0.5 µm on the original surface (Fig. 1(a)) to about 3 µm.
Correspondingly, the macro-steps which are running in the same crystallographic direction as
the original steps reach a height of about 15 nm. Parallel to the steps, uninterrupted
macroterraces more than 50 µm long have been observed.
The thickness distribution of the graphene film grown
ex-situ under an argon atmosphere is
determined by LEEM as shown in figs. 1(e,f). Series of spatially-resolved LEEM I-V spectra
taken along a vertical and a horizontal line in fig. 1(f) are shown in figs. 1(g,h). The layer
thickness is easily determined from the number of minima in the individual spectra; the LEEM
image taken at a particular energy shows stripes that follow in width and orientation the
macroterraces with a contrast that is determined by the graphene layer thickness.
15,16
Hence, we
can unambiguously conclude that except for narrow stripes at the edges, the large atomically flat
macro-terraces are homogeneously covered with a graphene monolayer. The domain size of
monolayer graphene is significantly larger than that of the vacuum annealed samples as a
comparison between figs. 1(c) and 1(f) shows. In fact, the domain size appears to be limited by
the length and width of the SiC terraces only. Narrower, darker regions at the downward edges
of the terraces correspond to bilayer and in some cases trilayer graphene (see region 3 in fig.
1(f)). In the AFM image these regions (see fig. 1(i)) appear as small depressions of around 4 Å
and 8 Å amplitude located at the very edge of the macrostep. This indicates that the nucleation
of new graphene layers starts at step edges of the substrate surface. We also note that the
laterally averaged graphene thickness determined by LEEM is in perfect agreement with the
average layer thickness value of 1.2 ML obtained by x-ray photoelectron spectroscopy (XPS).
The graphene layers grown under an argon atmosphere exhibit high structural and
electronic quality as demonstrated by the LEED and photoelectron spectroscopy data in Figure
2, taken from an Ar-grown film with a thickness of 1.2 ML. The LEED pattern demonstrates that
the graphene layer is well ordered and aligned with respect to the substrate, such that the basal
plane unit vectors of graphene and SiC subtend an angle of 30 degrees. The C1s core level
spectrum shows the characteristic signals of the SiC substrate, the (6√3×6√3) interface layer
and the graphene monolayer, respectively, in excellent agreement with previous work
21,22
. The
angle-resolved photoelectron spectroscopy (ARPES) measurement reveals the characteristic
band structure of monolayer graphene
20,23,24
. Note that, as for vacuum grown layers
20,23,24
, the
Dirac point (
E
D
) is shifted below the Fermi level (E
F
) due to electron doping from the substrate.
Therefore, while our epitaxial growth process results in a dramatic improvement in surface
morphology all other important properties such as crystalline order, electronic structure, and
charge carrier density remain unaltered as compared to vacuum grown layers.
What is the reason for the observed improvement of the surface morphology of the Ar-
annealed samples compared to the samples annealed in UHV? From the data in Fig. 1. it is
clear that the surface undergoes considerable morphological changes at the temperature where
graphitization occurs. The large roughness of the UHV annealed samples suggests that the
surface is far from equilibrium, such that a transformation to a smooth morphology cannot be
achieved under these conditions. The key factor in achieving an improved growth is the
significantly higher annealing temperature of 1650°C that is required for graphene formation
under argon at a pressure of 900 mbar as compared to 1280°C in UHV. Graphene formation is
the result of Si evaporation from the substrate. For a given temperature, the presence of a high
pressure of argon leads to a reduced Si evaporation rate because the silicon atoms desorbing
from the surface have a finite probability of being reflected back to the surface by collision with
Ar atoms, as originally pointed out by Langmuir
25,26
. The significantly higher growth temperature
thus attained results in an enhancement of surface diffusion such that the restructuring of the
surface that lowers the surface free energy (by step bunching, for example) is completed before
graphene is formed. Ultimately, this leads to the dramatically improved surface morphology that
we observe here. The macrostep structure is also responsible for the tighter thickness control.
As shown above, a new graphene layer starts to grow from the step edges; hence having fewer
steps along well defined crystallographic directions reduces the nucleation density of multilayer
graphene.
In order to evaluate the electronic quality of our graphene layers we determined the carrier
mobility of monolayer epitaxial graphene on SiC(0001) using Hall effect measurements. Two
different geometries were investigated, both patterned with electron beam lithography: square
graphene films (100 µm × 100 µm) with contact pads at the four corners for van der Pauw
measurements as well as Hall bars (2 µm × 30 µm) placed on macroterraces. No significant
difference in electron mobility was observed between the two geometries indicating that step
edges play a minor role. Mobilities of 930 cm
2
/Vs and 2000 cm
2
/Vs were measured at 300 and
27 K, respectively. At the same time the electron density remained basically constant
increasing only by 3%. Experiments performed in other groups reported on
maximum values of 1200 cm
2
/Vs, but on many-layer graphene
6,27
.
213
cm101
−
×≈n
Figure 3 shows the temperature dependence of the electron mobility measured in van der
Pauw geometry. The linear
µ(T) dependence is unexpected. Scattering at acoustic phonons of
graphene would result in
T
-4
behavior at low temperatures
28
. A theoretical treatment of the effect
of static impurities in graphene predicts 1/T dependence of the mobility
29
. Candidates for such
impurities are certainly dangling bonds below the graphene layer. Also adsorbates might play a
certain role. The linear dependence of the scattering rates rather fits to the case of electron-
electron interaction in a 2D electron gas
30
. Clearly more work is required to understand the
temperature dependence of the mobility in epitaxial graphene.
In conclusion, we have shown that the growth of epitaxial graphene on SiC(0001) in an Ar
atmosphere close to atmospheric pressure provides morphologically superior graphene layers in
comparison to vacuum graphitization. Extensive step bunching taking place during processing
yields arrays of parallel terraces up to 3 µm wide and more than 50 µm long. The terraces are
essentially completely and homogeneously covered with a monolayer of graphene. At present,
downward step edges, where the initiation of second and third layer graphene growth is
detected, are prohibiting an even larger extend of the graphene domains. Because the substrate
step direction and step width are determined by the magnitude and azimuthal orientation of the
surface misorientation with respect to major crystallographic directions, a proper choice of these
parameters controls terrace width and length and hence the ultimate uninterrupted lateral extent
of the graphene layer. An improved substrate quality in terms of crystallographic orientation is
therefore expected to lead to further improvements. In comparison to the UHV treatment, the
technique presented here is much closer to standard preparation conditions in semiconductor
manufacture, permitting the use of standard CVD (chemical vapor deposition) equipment for the
fabrication of graphene layers. All necessary processing steps, i.e. hydrogen etching and
graphene synthesis, can be carried out in a single reactor. Electrical measurements confirm the
picture of improved film quality: mobilities around 1000 cm
2
/Vs at room temperature, which
increases linearly up to 2000 cm
2
/Vs at 27 K.
