A catalytic alloy approach for graphene on epitaxial SiC on
silicon wafers
Author
Iacopi, Francesca, Mishra, Neeraj, Cunning, Benjamin Vaughan, Goding, Dayle, Dimitrijev,
Sima, Brock, Ryan, Dauskardt, Reinhold H, Wood, Barry, Boeckl, John
Published
2015
Journal Title
Journal of Materials Research
Version
Accepted Manuscript (AM)
DOI
https://doi.org/10.1557/jmr.2015.3
Copyright Statement
© 2015 Cambridge University Press. This is the author-manuscript version of this paper.
Reproduced in accordance with the copyright policy of the publisher. Please refer to the
journal's website for access to the definitive, published version.
Downloaded from
http://hdl.handle.net/10072/69321
Griffith Research Online
https://research-repository.griffith.edu.au
INVITED FEATURE PAPER
A catalytic alloy approach for graphene on epitaxial SiC on silicon
wafers
F. Iacopi,
a)
N. Mishra, B.V. Cunning, D. Goding, and S. Dimitrijev
Queensland Micro- and Nanotechnology Centre, Griffith University, Nathan, Queensland 4111, AustraliaAU2
R. Brock and R.H. Dauskardt
Department of Materials Science and Engineering, Stanford University, Stanford, California 94305, USA
B. Wood
Centre for Microscopy and Microanalysis, The University of Queensland, St. Lucia, Queensland 4072, Australia
J.J. Boeckl
Materials and Manufacturing Directorate, Air Force Research Laboratories, Wright-Patterson AFB, Ohio, USAAU3AU1
(Received 7 July 2014; accepted 1 January 2015)
We introduce a novel approach to the synthesis of high-quality and highly uniform few-layer
graphene on silicon wafers, based on solid source growth from epitaxial 3C-SiC films. Using
a Ni/Cu catalytic alloy, we obtain a transfer-free bilayer graphene directly on Si(100) wafers,
at temperatures potentially compatible with conventional semiconductor processing. The graphene
covers uniformly a 20 silicon wafer, with a Raman I
D
/I
G
band ratio as low as 0.5, indicative of
a low defectivity material. The sheet resistance of the graphene is as low as 25 X/square, and its
adhesion energy to the underlying substrate is substantially higher than transferred graphene. This
work opens the avenue for the true wafer-level fabrication of microdevices comprising graphene
functional layers. Specifically, we suggest that exceptional conduction qualifies this graphene as
a metal replacement for MEMS and advanced on-chip interconnects with ultimate scalability.
I. INTRODUCTION
Ever since graphene was experimentally isolated about
a decade ago,
1
the high temperature (1300–1700 °C) Si
sublimation from crystalline silicon carbide (SiC) bulk
substrates has been extensively regarded as the cleanest
and most controlled means of obtaining quality graphene
at the wafer level.
2–4
However, the transfer of this
approach to its natural pseudo-substrate, i.e., epitaxial
SiC films on silicon,
5–7
has lagged behind despite being
strongly driven by substantial cost (SiC wafers are about
100 times more expensive than silicon) and large-scale
fabrication arguments. This endeavor has proved more
challenging than expected, in particular, because of the
upper limit set by the melting temperature of silicon and
the scarce availability of a defect-free and atomically
smooth epitaxial SiC on Si(111) starting template.
8,9
In response to such challenges, we have recently
demonstrated an alternate approach to the wafer-level,
transfer-free uniform synthesis of graphene on silicon.
10
As for the sublimation process, our new methodology
relies on the use of epitaxial SiC on silicon as a solid
source of carbon. The main advantage of this approach is
that it enables a wafer-scale patterned synthesis of
graphene by prepatterning of the source SiC layer.
10,11
However, our catalyst-mediated process
10
allows for
a reduction in the optimal synthesis temperature and for
a relaxation of the strict requirements on the starting
template. In this study, we will review the alloy-mediated
catalytic synthesis and show that this process works
equally well on both SiC(100) and SiC(111) orientations
of the pseudo-substrate. We demonstrate an outstanding
graphene bilayer with remarkable adhesion to the sub-
strate and outstanding electrical conduction properties.
II. EXPER IMENTAL
Monocrystalline 3C-(cubic polytype) SiC films,
250 nm thick, were epitaxially grown on ,111. and
,100. oriented 20 silicon substrates in a hot wall,
horizontal low-pressure chemical-vapor deposition reac-
tor, as described by Wang et al.,
12
yielding two types of
films that will be referred to as SiC(111) and SiC(100),
respectively
13
[Fig. 1(a)]. Subsequently, either a single
layer of Ni or a double layer of Ni and Cu was sputtered
on the SiC samples, as illustrated in Fig. 1(b), and then
annealed in a Carbolite HT furnace for 75 min at 1100 °C
in a medium vacuum atmosphere (below 10
3
mbar) to
induce graphitization. As per our recent work,
10
the
annealing generates a highly intermixed layer, which is
then removed by immersion of the samples in a wet etch
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Contributing Editor: Mauricio Terrones
a)
Address all correspondence to this author.
e-mail: f.iacopi@griffith.edu.au
This paper has been selected as an Invited Feature Paper.
DOI: 10.1557/jmr.2015.3
J. Mater. Res., Vol. 30, No. 0, 2015 Ó Materials Research Society 2015 1
solution (Freckle) and sonication, to reveal a few-layer
graphene at the top surface of SiC [Fig. 1(c)]. In this
study, we have kept the Cu thickness constant to ;15
nm, while varying the Ni thickness to analyze and
compare the properties of the produced graphene.
The few-layer graphene on SiC/Si samples was char-
acterized with Raman spectroscopy with a Renishaw
inVia Raman, using a 514 nm laser. The laser spot size
was ;1 lm (50x objective), and the laser power at the
sample was about 7.5 mW. Atomic force microscopy
(AFM) was performed with a Park NX20 instrument in
a noncontact mode with a 512 512 pixel resolution
on 5 5 and 1 1 lm
2
scan areas to extract the root-
mean-square (RMS) surface roughness as well as the
power-spectral-density (PSD) of the spatial frequencies
composing the surface topography pattern. X-ray photo-
electron spectroscopy (XPS) was carried out with a Kratos
Axis ULTRA
AU4 high resolution system with monochro-
matic Al K
a
(1486.6 eV) incident radiation. Survey
(wide) scans were taken with an analyzer pass energy
of 160 eV while multiplex (narrow) high resolution scans
at 20 eV. The used aperture leads to an analysis area of
700 300 lm
2
. The base pressure in the analysis
chamber was below 1.0 10
8
Torr. Data analysis
was performed with the CasaXPS software and a Shirley
baseline with Kratos library relative sensitivity factors.
Sample foils for transmission electron microscopy
were prepared via a focused ion beam liftout technique
using a FEI Strata DB235 FIB/SEM with a Ga
1
ion
source. Prior to ion milling, the samples were protected
with a plasma sputtered gold layer of ;300 nm and then
a2lm Pt cap was deposited at 5 keV by an e-beam to
preserve the initial surface integrity. Site specific foils
were excavated from the bulk samples and thinned to
;500 nm. Subsequent Ar
1
ion milling was conducted in
a Fiscione NanoMill™ to remove Ga ion damage and
provide a high resolution transmission electron micro-
scope (HRTEM) foil. HRTEM characterization was
performed using a FEI Cs corrected Titan3 microscope
under 80 keV source illumination.
An initial analysis of the adhesive fracture energies of the
graphene/SiC/Si system was made by loading these samples
within a four point bend sandwiched configuration.
14,15
The
sandwiched structures containing the graphene on SiC were
prepared by sputtering an additional 500 nm Si “spacer”
layer, subsequently bonded with epoxy to the mirroring
silicon slab. 5 40 mm beams were diced out of the
sandwiched structures and tested. The average critical strain
energy release rates measured from such specimens yield an
indication of the adhesion energy along the debonded
pathway. Extensive detailed descriptions of this methodol-
ogy are found in the literature.
14,15
The location of the
debonding is verified through failure analysis with surface
XPS analysis on both sides of the delaminated interface.
Sheet resistance measurements were taken using four
electrical contacts in a van der Pauw configuration on the
corners of
AU51 1cm
2
samples. 150 nm of Ni were
sputtered at 100 °C through a shadow mask for forming
the contacts. A total of eight measurements were taken
for each sample by positioning the electrical probes on
the sputtered metal contacts and sweeping the DC
input current from a HP4145B parameter analyzer from
0 to 10 mA, which were then separately averaged into
“vertical” and “horizontal” sheet resistance groups. Pro-
vided the material is uniform and isotropic, the measure-
ments are expected to all converge to the same line with
only marginal variation due to probe positioning.
III. RESULTS AND DISCUSSION
A. Mechanisms for catalytic alloy graphitization
and physical analysis
Figure 2 shows an overview of the I
D
/I
G
band ratios,
indicative of the defect density of graphene layers,
16
calculated from the Raman spectra of graphene prepared
with different compositions of metal catalysts on SiC
(100) and SiC(111), as measured after the removal of the
reacted metal layer by wet etching. The error bar
represents the variation obtained by measuring five
different sites over the prepared samples. Note that data
extracted from samples produced over more than 10
experimental runs fell within those error bars. Confirming
and extending the scope of our recent findings for
graphene on SiC(111),
10
these data show that the use
of a Ni/Cu mixture improves dramatically the quality of
the obtained graphene on SiC(100), as compared to when
only Ni is used. In fact, the I
D
/I
G
ratio of graphene on SiC
(100) decreases from ;2 for 5 nm Ni catalyst alone
to ;0.8 when the Cu catalyst layer is added. Note that the
graphene on SiC(100) seems to be consistently of better
quality than that grown on SiC(111) under the same
conditions. Figure 3 shows an example of Raman
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
FIG. 1. Schematic of the process steps for the preparation of graphene on epitaxial 3C-SiC on silicon. The starting substrate can be either a ,100. or
,111. oriented substrate (a), on which a single- or double-thin catalyst layer of Ni alone or Ni and Cu is deposited (b). An anneal at 1100 °C leads to
the formation of a few-layer graphene covered with a metallic highly intermixed layer (c) that is subsequently removed by wet etching. (color online)
F. Iacopi et al.: A catalytic alloy approach for graphene on epitaxial SiC on silicon wafers
J. Mater. Res., Vol. 30, No. 0, 20152
spectrum of graphene grown on SiC(100) with 8 nm Ni
and 15 nm Cu. The intensity of the 2D band around
2700 cm
1
is slightly higher than that of the G band
(;1580 cm
1
), which indicates monolayer to bilayer
graphene.
16
The spectrum in Fig. 3 includes Raman shifts
down to 750 cm
1
to reveal the TO peak of the
underlying SiC(100) at 795 cm
1
. Note from Fig. 2 that
the final graphene quality, as seen from the Raman
spectrum, appears rather insensitive to the further increase
of Ni thickness when the Ni/Cu mixture is used.
We have also compared the surface morphology of
graphene samples prepared with Ni only (8 nm) and the
Ni/Cu approach with AFM. As shown in Table I, the use
of the alloy halves the surface RMS roughness of the final
graphene sample. This is also clearly shown by the power
spectral distributions in Fig. 4, indicating that the use of
Ni/Cu reduces dramatically the large-range (spatial fre-
quencies below 5 lm
1
) as well as the smallest-range
roughness (frequencies larger than 20 lm
1
). This means
that asperities larger than 200 nm and smaller than 50 nm
are considerably reduced with the use of Ni/Cu.
In our recent work,
10,17
we had presented a preliminary
elucidation of the mechanisms that allow the Ni/Cu alloy
to induce a better quality and more controlled graphiti-
zation as compared to Ni alone. The use of Ni only had
been in fact already reported in an earlier literature on
SiC wafers and then abandoned as yielding graphene on
the metal surface rather than on SiC, and of inferior
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
FIG. 2. Comparison of the defe ctivity of different graphenes on
SiC preparations as estimated from the ratio of the intensities of the
D over G Raman band s. T he thickness of the Cu layer, when
present, was kept constant. The error bar is calculated from the
measurements variation over five different sites. The use of a Ni/Cu
alloy leads to a remarkable decrease in defectivity. Graphene on
SiC(100) shows consistently a slightly lower defect density than
that on SiC(111). (color online)
FIG. 3. Example of Raman spectrum (graphene D, G, and 2D bands
region) from graphene on SiC(100) prepared with ;8 nm Ni and
;15 nm Cu. Additionally to the low D over G band intensity,
indicative of a good quality graphene, note that the 2D over G
intensity ratio higher than 1 is indicative of a few-layer graphene.
The Raman shift region for 3C-SiC is also shown.
FIG. 4. Comparison of the spatial frequency spectra defining the
surface topography of different graphenes on SiC preparations (power
spectral densities), related to the AFM measurements in Table I. Note
that the graphene prepared with Ni only has substantially more large
(frequencies below 5 lm
1
) as well as very small surface asperities as
compared to the Ni/Cu preparations. (color online)
TABLE I. XXX AU6.
5 5mm
2
RMS
(nm 60.15)
1 1mm
2
RMS
(nm 60.15)
Graphene/SiC(100) {Ni only} 28 37
Graphene/SiC(100) {Ni/Cu} 15 12
Graphene/SiC(111) {Ni/Cu} 11 9
Surface roughness comparison of different graphenes on SiC prepara-
tions. Note that the graphene prepared with Ni/Cu indicates a 50%
decrease in RMS as compared to the use of Ni alone as a catalyst, and
that the graphene roughness does not seem dependent on the orientation
of the starting substrate. Note that the typical RMS roughness of the
initial SiC surface is 2–3nm.
9
F. Iacopi et al.: A catalytic alloy approach for graphene on epitaxial SiC on silicon wafers
J. Mater. Res., Vol. 30, No. 0, 2015 3
quality compared to the sublimation approach.
18,19
The
catalytic action of Ni is essential because it reacts with
SiC at relatively low temperature to form stable NiSi
x
,
20
and therefore allows for the release of the atomic
C needed for the synthesis of graphene. However, the
control of a uniform reaction of a thin Ni film over a large
SiC surface is not trivial, as nickel silicidation appears
dominated by strong local driving forces leading to
typical clustering.
20,21
This is confirmed by the high
nonuniformity of graphene observed using only Ni in the
past and confirmed with Raman bands in Fig. 2 and by
the substantial topography of about 30 nm RMS shown
by our surface roughness measurements in Table I.
The addition of Cu has a twofold beneficial action.
Firstly, Cu acts like an efficient medium to dilute and
helps to distribute as uniformly as possible the Ni over
the SiC surface. Note in this regard that the thermal
process for graphitization takes place very close to the
melting temperature of Cu. Secondly, the Cu is known as
a very efficient catalyst for graphitization, since free C
has an extremely low saturation in melt Cu (only a few
ppm, lower than Ni
22
), as calculated from a simple binary
C–Cu phase diagram. This allows for a fast precipitation
and graphitization of C released through the Ni silicida-
tion reaction, leading to an extraordinary improvement in
the uniformity and quality of the graphene obtained
through Ni/Cu.
The second important advance of this study is the
demonstration that the Ni/Cu alloy graphitization
achieves quality graphene on both SiC(111) and SiC
(100) surface orientations, with a consistently better
quality obtained from SiC(100), as inferred from the
I
D
/I
G
trend in Fig. 2. This is in contrast with the Si
sublimation process from SiC on silicon that is typically
more successful on the SiC(111) surfaces,
23–26
which are
believed to be more suitable templates for graphene
because of their hexagonal atomic arrangement. We
suggest here that our graphene synthesis conditions are
close to those of a liquid phase epitaxy,
27
a process
whose benefits are largely known to nanotechnology,
28,29
with a larger adatom surface mobility that can somewhat
relax the requirements on the starting template as
compared to a more conventional epitaxial process like
in the case of the sublimation. The reason why the SiC
(100) template leads in our case to an even better
graphene quality than SiC(111) could be related to the
combination of a considerably higher tensile stress and
higher surface defectivity of the latter epitaxial SiC as
discussed in our previous work.
13
SiC(111) layers indeed
show a higher density of stacking faults appearing at the
surface that, combined to a residual tensile SiC stress in
the range of 700 MPa to 1 GPa, could provide a strong
driving force around distinct preferential diffusion paths
of the Ni atoms into the SiC layer and thus a less uniform
silicidation reaction.
We have selected the sample series showing the best
(lowest) Raman I
D
/I
G
ratio in Fig. 2, prepared on SiC
(100) with 8 nm Ni and our standard Cu thickness (full
spectrum in Fig. 3) for a more detailed physical and
chemical analysis. XPS analysis shows a near-surface
composition of 50% C, 39% Si, and about 9% O, as
shown in the survey spectrum in Fig. 5(a). Other
elements found in just trace amounts are F (0.7%),
N (0.7%), and Ni (0.15%). The F and N traces are
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
FIG. 5. (a) XPS surface survey spectrum and (b) high resolution spectrum of the C1s region of a graphene on SiC(100) prepared with Ni/Cu,
corresponding to the Raman spectrum in Fig. 3. The C1s spectrum reveals a peak centered at 284.5 eV for the graphene, one 282.5 eV
corresponding to the underlying SiC, plus lower intensity peaks at higher eV, attributed to C–O–Si, C–O, and C5O bonds. (color online)
F. Iacopi et al.: A catalytic alloy approach for graphene on epitaxial SiC on silicon wafers
J. Mater. Res., Vol. 30, No. 0, 20154
