In situ studies of Al2O3 and HfO2 dielectrics on graphite

Article abstract of DOI:10.1063/1.3238560

Deposition of Al₂O₃ and HfO₂ dielectrics on graphite is studied as a route to the formation of a high- κ dielectric on graphene. Electron beam evaporation of metal Al and Hf is followed by a separate oxidation step. Reacti

Full text of DOI:10.1063/1.3238560

In situ studies of Al 2 O 3 and HfO 2 dielectrics on graphite
Adam Pirkle, Robert M. Wallace, and Luigi Colombo
Citation: Applied Physics Letters 95, 133106 (2009); doi: 10.1063/1.3238560
View online: http://dx.doi.org/10.1063/1.3238560
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/95/13?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Modifications of local structures of Gd2O3 on incorporation of SiO2
J. Appl. Phys. 110, 063527 (2011); 10.1063/1.3642083
Effects of ozone oxidation on interfacial and dielectric properties of thin HfO 2 films
J. Appl. Phys. 104, 054117 (2008); 10.1063/1.2976340
Materials and electrical characterization of molecular beam deposited CeO 2 and CeO 2 / HfO 2 bilayers on
germanium
J. Appl. Phys. 102, 024104 (2007); 10.1063/1.2756519
Low temperature ( 400 ° C ) Al 2 O 3 ultrathin gate dielectrics prepared by shadow evaporation of aluminum
followed by nitric acid oxidation
Appl. Phys. Lett. 90, 102902 (2007); 10.1063/1.2711290
Characteristics of Hf x Si y O films grown on Si 0.8 Ge 0.2 layer by electron-beam evaporation
Appl. Phys. Lett. 88, 122906 (2006); 10.1063/1.2188598
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
134.245.224.26 On: Thu, 25 Dec 2014 01:10:04

APPLIED PHYSICS LETTERS 95, 133106 ͑2009͒
In situ studies of Al O and HfO dielectrics on graphite
2
3
2
1
1,a͒
2
Adam Pirkle, Robert M. Wallace,
Department of Materials Science and Engineering, The University of Texas at Dallas, Richardson,
Texas 75080, USA
and Luigi Colombo
1
2
Texas Instruments Incorporated, Dallas, Texas 75265, USA
͑
Received 1 August 2009; accepted 8 September 2009; published online 28 September 2009͒
Deposition of Al O and HfO dielectrics on graphite is studied as a route to the formation of a
2
3
2
high-␬ dielectric on graphene. Electron beam evaporation of metal Al and Hf is followed by a
separate oxidation step. Reactive e-beam deposition of HfO by introduction of O to the deposition
2
2
chamber is also demonstrated as an alternative to the two-step metal deposition and oxidation
approach. We employ in situ x-ray photoelectron spectroscopy to study reactions between the
substrate and deposited film and ex situ atomic force microscopy to examine the dielectric film
morphology. © 2009 American Institute of Physics. ͓doi:10.1063/1.3238560͔
The experimental isolation of graphene, a single layer of
carbon atoms in a hexagonal lattice, has led to a great deal of
research due to its unique transport and physical properties.
vacuum. Large ͑ϳ1 cm͒ crystals of natural graphite are em-
ployed in order to facilitate XPS measurements which re-
quire a large spot size ͑ϳ0.5 cm͒. Since the individual
graphene layers of bulk graphite are weakly bonded to one
another and thus interplanar interactions will not play a sub-
stantial role in surface reactions, the use of a graphite crystal
as a model system for graphene is justifiable in the context of
this study. Sample surfaces are prepared by mechanical ex-
foliation of natural graphite using adhesive tape, similar to
the common method of obtaining graphene flakes on
1
These characteristics have encouraged studies to establish
whether graphene can become a viable candidate material
for nanoelectronic switches beyond the end of the Si-
complementary metal-oxide semiconductor roadmap. Such
devices will require the integration of a scalable high-␬ di-
electric, but a number of challenge arise in the deposition
of robust ultrathin gate dielectrics on graphene. The hydro-
phobic and chemically inert nature of the basal plane of
graphene is not conducive to direct application of standard
atomic layer deposition ͑ALD͒ processes. As a particular ex-
ample, the ALD Al O process which uses cycles of trim-
1
0
Si/SiO2. Samples are mounted on a titanium plate and
loaded in the UHV system within 5 min. For selected
samples, a “predeposition vacuum anneal” is performed at
500 °C for 30 min to remove any physisorbed species, such
2
3
as H O or organic compounds. After cooling the sample to
ethyl aluminum ͑TMA͒ and H O results in nonuniform
2
2
room temperature, the metal is deposited by electron beam
evaporation to a thickness of 1 nm as indicated by a quartz
crystal microbalance. Following deposition, samples with
metal films are oxidized in a clustered chamber in high-
deposition on graphene where growth is observed at step
2
edges with little to no deposition on the basal plane. Surface
treatment by means of O or NO have been shown to permit
3
2
3
uniform deposition, but any C–C bond scission or introduc-
tion of charged impurities at the graphene surface is expected
to be detrimental to device operation, particularly in the limit
of single-layer graphene devices. For example, an ALD pro-
cess which employs an O pretreatment to nucleate Al O
purity dry O introduced via a precision leak valve. In situ
2
XPS spectra are collected between each processing step.
Figure 1 shows XPS data for the as-deposited and oxi-
dized Al on graphite with and without a predeposition anneal
to remove physisorbed species. The graphite surface pretreat-
ment is performed at 500 °C under UHV conditions. Figure
1͑a͒ shows the XPS spectrum of the as-deposited Al located
3
2
3
deposition by TMA has been shown to introduce C–O bond-
ing, as observed in the C 1s x-ray photoelectron spectros-
2
copy ͑XPS͒ spectrum, and these bonds are likely to degrade
the performance of graphene devices. Recent efforts have
demonstrated top-gated graphene devices utilizing dielectrics
O 1s
C 1s
π−π*
C-C
Al 2p
4
–6
such as SiO , ALD Al O , and ALD HfO , including the
2
2
3
2
7
(
d)
(c)
b)
(d)
(d)
(c)
use of an evaporated Al nucleation layer. This Letter exam-
ines the detailed mechanisms of processes for deposition of
scalable Al O and HfO dielectrics on graphene.
x10
(c)
(b)
2
3
2
0
The in situ XPS work detailed here is carried out using
an ultrahigh vacuum ͑UHV͒ cluster system with analytical
and deposition modules, including XPS with monochromatic
Al K␣ x-ray source and Omicron EA 125 hemispherical
analyzer, electron beam evaporation, oxidation and annealing
Al
(
(b)
x10
(a)
(a)
(a)
5
36 532 528 290 286 282
76
72
8
,9
as described elsewhere. A UHV sample transfer system
Binding energy (eV)
−
10
͑
PՅ1ϫ10
mbar͒ allows for deposition, oxidation and
annealing with intermediate XPS analysis without breaking
FIG. 1. ͑Color online͒ O 1s, C 1s, and Al 2p spectra for graphite sample
with predeposition anneal at 500 °C to remove physisorbed interfacial hy-
droxyls ͑a͒ Al as-deposited and ͑b͒ Al oxidized in 1000 mbar O at 200 °C
for 10 min. Sample without predeposition anneal ͑c͒ Al as-deposited and ͑d͒
oxidized under the same conditions as ͑b͒.
2
a͒Author to whom correspondence should be addressed. Electronic mail:
rmwallace@utdallas.edu.
0
003-6951/2009/95͑13͒/133106/3/$25.00
95, 133106-1
34.245.224.26 On: Thu, 25 Dec 2014 01:10:04
© 2009 American Institute of Physics
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
1

133106-2
Pirkle, Wallace, and Colombo
Appl. Phys. Lett. 95, 133106 ͑2009͒
Hf 1s O 1s C 1s
Hf 4f
1
3.0 nm
13.00
(
e)
(
a)
(b)
(
d)
c)
(d)
(c)
(d)
(c)
6
00nm
600nm
0
.00
0
.0 nm
(
0
1
2
3 0
1
2
3
x (µm)
x (µm)
HfC
(b)
x10
FIG. 2. ͑Color online͒ AFM image and feature heights showing comparison
of ͑a͒ 1 nm Al O and ͑b͒ 1 nm HfO on natural graphite. Large ͑10 nm͒
clusters form after deposition of Al and oxidation at 200 °C in dry O , in
3
(b)
(a)
(b)
(a)
1
2
3
2
2
1
contrast to HfO which is substantially more uniform.
2
x10
2
3
(a)
at 72.6 eV as a reference. When the Al is deposited on the
annealed surface and oxidized at 200 °C using oxygen, the
Al does not fully oxidize as shown in Fig. 1͑b͒ where the
metallic Al 2p ͑Al ͒ peak is still present. However, when the
graphite surface is not annealed, a small amount of oxygen is
536532528 286 282 22 18 14
Binding energy (eV)
FIG. 3. ͑Color online͒ 1 nm Hf deposited on ͑a͒ exfoliated graphite with
0
−8
deposition chamber walls at 25 °C and P=1ϫ10 mbar, ͑b͒ annealed
͑500 °C/30 min͒ graphite with chamber walls cooled by liquid N and P
2
−
10
=
4ϫ10
mbar, ͑c͒ Hf oxidized in 1000 mbar O at 25 °C, ͑d͒ HfO
2 2
−6
detected and attributed to physisorbed hydroxyl species ͓Fig.
deposited by reactive e-beam deposition with 1ϫ10 mbar partial pressure
of O . A difference spectrum between the two graphite+HfO C 1s peaks is
1
͑c͔͒. Subsequent Al deposition and oxidation results in fully
2
2
oxidized Al, as indicated by the absence of the Al 2p metal
shown in ͑e͒; the peak for sample ͑c͒ is broader than ͑d͒ by 0.13 eV.
0
͑
Al ͒ peak in Fig. 1͑d͒. Previous studies of Al oxidation in-
dicate that the Al O thickness on Al saturates at as little as
–2 nm at low temperature ͑TՅ300 °C͒, suggesting that our
Al metal thickness is somewhat larger than the limiting oxide
thickness. Calculation of Al thickness using the attenuation
of the XPS C 1s peak after deposition is in good agreement
Ϯ0.15 nm͒ with the 1 nm thickness indicated by the quartz
microbalance. However, the calculation assumes a uniform
layer of Al on the graphite surface. Examination of the sur-
face morphology by atomic force microscopy ͑AFM͒ shows
that Al forms nonuniform clusters on the graphite surface,
and that these clusters have heights as large as 10 nm, as
shown in Fig. 2. The formation of the clusters is consistent
with previous studies of evaporated Al on highly oriented
pyrolytic graphite surfaces. This clustering effect explains
the incomplete oxidation of the deposited Al, as the cluster
radii are larger than the saturated thickness of Al O .
2
3
1
denoted “2,” and a broad Hf 1s feature denoted “3.” Decon-
volution of the Hf 4f spectrum is carried out using an asym-
metric Doniach–Sunjic doublet to fit the metal state, as well
as symmetric Gaussian–Lorentzian doublets to fit oxide
states, observed at higher binding energy relative to the metal
line if present. In Fig. 3͑a͒, the substrate was exfoliated
͑nonannealed͒ natural graphite. After deposition of 1 nm Hf,
a HfC peak is observed in addition to substantial oxide states
in both the Hf 4f and O 1s regions. A Hf 4f doublet shifted
+1.02 eV relative to the metallic Hf peak is attributed to Hf
11
͑
1
6
suboxides, and a second doublet at +3.05 eV is due to HfO ,
2
1
7
consistent with previously reported peak assignments. Oxi-
dation of the deposited Hf is attributed to residual phys-
isorbed water on the sample surface. Samples are mounted
on “4” Ti plates, and liberation of water from the unannealed
sample holder plate by radiative heating from the e-beam
crucible is another potential source of oxidants and hydroxyl
species. The chamber pressure was observed to be 1
ϫ10 mbar during this deposition. The sample shown in
Fig. 3͑b͒ received a predeposition anneal and the deposition
chamber walls were cooled to 77 K by means of an internal
1
2
2
3
The C 1s XPS line shape for graphite exhibits a charac-
teristic asymmetry due to screening of the core hole pro-
1
3
duced by photoemission. No additional bonding states or
changes in asymmetry are observed in the C 1s line after
deposition or oxidation of Al, indicating that there is no de-
tectable covalent bonding with the substrate at low tempera-
ture ͑TՅ300 °C͒. This study is conducted in this low tem-
perature range since physisorbed species such as hydroxyls
catalyze formation of Al–C bonds at temperatures greater
than 400 °C. In the case of single layer graphene, Al–C
bonding will disrupt the structure of the graphene lattice, and
likely induce substantial mobility degradation. Al–C bonds
appear at a binding energy of 281.5 eV in the C 1s spectrum
and at 73.6 eV in the Al 2p spectrum, and these states are
−
8
LN shroud. Cooling the chamber walls reduced the pressure
2
−
10
during deposition to 4ϫ10
mbar, and no HfC bonds were
observed within the detection limits of XPS. The Hf 4f peak
is well described by a metallic doublet state, and only a very
small residual O 1s signal is present.
Figure 3͑c͒ shows the post-oxidation behavior of metal-
lic Hf with no detectable HfC. The sample was oxidized in
1
4
1000 mbar dry O at 25 °C. The Hf 4f peak indicates that
2
1
5
below the limit of detection for the samples discussed here.
this sample is fully oxidized and does not include a detect-
able hydroxide component. However, the O 1s spectrum has
an additional state at 531.8 eV, which does not correspond to
any state in the Hf 4f spectrum. In addition to the approach
of metal deposition and subsequent oxidation, HfO was de-
posited by means of reactive e-beam deposition in which dry
In contrast to Al, upon deposition of Hf on exfoliated
natural graphite by e-beam evaporation at 25 °C substrate
temperature, carbide ͑HfC͒ bonding is observed by XPS.
This behavior is shown in Fig. 3͑a͒. The Hf 1s and O 1s
features are close in binding energy and not well resolved
from one another. Deconvolution reveals O 1s peaks corre-
sponding to hafnium oxide, denoted “1,” hafnium hydroxide,
2
O was introduced during the evaporation of Hf at a partial
pressure of 1ϫ10 mbar, as shown in Fig. 3͑d͒. This ap-
2
−
6
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
1
34.245.224.26 On: Thu, 25 Dec 2014 01:10:04

133106-3
Pirkle, Wallace, and Colombo
Appl. Phys. Lett. 95, 133106 ͑2009͒
proach has been shown previously to produce high-quality
sition annealing and low residual chamber pressure during
deposition are shown to be essential to the deposition of
carbide-free Hf films. Reactive e-beam deposition of Hf ap-
pears to be a viable method for deposition of uniform,
carbide-free films with sufficient uniformity that scalability
beyond 3 nm thickness is feasible.
1
8
HfO films on Si. The deposition chamber walls were
2
cooled to 77 K using the LN shroud and the chamber base
2
pressure prior to the introduction of O2 was ϳ1
ϫ10− mbar. Under the above chamber conditions, no car-
10
bide bonds are detectable after deposition of HfO , as shown
2
in Fig. 3͑d͒. The Hf 4f and O 1s peaks indicate that the film
is fully oxidized with a small additional component in each
peak due to hafnium hydroxide. Comparison of C 1s peaks
reveals that the sample upon which metallic Hf was depos-
ited and subsequently oxidized has a C 1s line which is
broader than that of the reactive e-beam sample by 0.13 eV,
and the aforementioned O 1s state at 531.8 eV can therefore
be attributed to C–O bonding. AFM images of the graphite
The authors acknowledge useful discussions with Pro-
fessor J. Kim and Mr. B. Lee, and TEM support by Mr. C.
Floresca and Professor M. J. Kim. This work is sponsored by
the Nanoelectronic Research Initiative ͑NRI SWAN Center
Contract No. 2006-NE-1464͒.
1
A. H. C. Neto, F. Guinea, N. M. R. Peres, K. S. Novoselov, and A. K.
surface with HfO ͓Fig. 2͑b͔͒ do not exhibit clustering as
Geim, Rev. Mod. Phys. 81, 109 ͑2009͒.
B. Lee, S. Park, H. Kim, K. Cho, E. M. Vogel, M. J. Kim, R. M. Wallace,
and J. Kim, Appl. Phys. Lett. 92, 203102 ͑2008͒.
D. B. Farmer and R. G. Gordon, Nano Lett. 6, 699 ͑2006͒.
M. C. Lemme, T. J. Echtermeyer, M. Baus, and H. Kurz, IEEE Electron
Device Lett. 28, 282 ͑2007͒.
Y. Lin, K. A. Jenkins, A. Valdes-Garcia, J. P. Small, D. B. Farmer, and P.
Avouris, Nano Lett. 9, 422 ͑2009͒.
I. Meric, M. Y. Han, A. F. Young, B. Ozyilmaz, P. Kim, and K. L. Shep-
ard, Nat. Nanotechnol. 3, 654 ͑2008͒.
S. Kim, J. Nah, I. Jo, D. Shahrjerdi, L. Colombo, Z. Yao, E. Tutuc, and S.
K. Banerjee, Appl. Phys. Lett. 94, 062107 ͑2009͒.
R. M. Wallace, ECS Trans. 16, 255 ͑2008͒.
A. Pirkle, Y. J. Chabal, L. Colombo, and R. M. Wallace, ECS Trans. 19,
2
2
observed for Al O samples, indicating that HfO may be a
2
3
2
better prospective material with regard to dielectric thickness
scaling. AFM images collected on multiple samples ͑not
shown͒ exhibit comparable surface roughness ͑Ͻ0.3 nm͒,
3
4
5
and TEM images over a magnification range from up to 3
5
ϫ10 ͑not shown͒ confirm that the HfO /graphite interface is
2
6
uniform for a 5 nm HfO film deposited by the e-beam
2
method.
7
Metal deposition on graphite/graphene followed by oxi-
dation may be a viable route to the formation of high-k
Al O and HfO dielectrics for top-gated graphene devices.
8
9
2
3
2
We have shown that the initial surface condition particularly
with regard to physisorbed hydroxyl species impacts the
composition of the oxidized film by the incorporation of hy-
droxides as well as catalytic carbide bond formation. Obser-
vation of cluster formation indicates that further understand-
ing of surface mobility and deposition kinetics will be
215 ͑2009͒.
1
0
K. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V.
Dubonos, and I. V. Grigorieva, and A. A. Firsov, Science 306, 666 ͑2004͒.
L. P. H. Jeurgens, W. G. Sloof, F. D. Tichelaar, and E. J. Mittemeijer,
J. Appl. Phys. 92, 1649 ͑2002͒.
Q. Ma and R. A. Rosenberg, Surf. Sci. 391, L1224 ͑1997͒.
F. Sette, G. K. Wertheim, Y. Ma, G. Meigs, S. Modesti, and C. T. Chen,
Phys. Rev. B 41, 9766 ͑1990͒.
11
12
13
required in order to obtain scalable Al O dielectrics with
2
3
1
1
1
4
5
6
thickness less than 5–10 nm. This two step process where
oxidized Al is followed by ALD Al O has been used else-
B. Maruyama, F. Ohuchi, and L. J. Rabenberg, J. Mater. Sci. Lett. 9, 864
͑1990͒.
2
3
M. Alexander, G. Thompson, and G. Beamson, Surf. Interface Anal. 29,
468 ͑2000͒.
S. Doniach and M. Sunjic, J. Phys. C: Solid State Phys. 3, 285 ͑1970͒.
C. Morant, L. Galan, and J. Sanz, Surf. Interface Anal. 16, 304 ͑1990͒.
H. Harris, K. Choi, N. Mehta, A. Chandolu, N. Biswas, G. Kipshidze, S.
Nikishin, S. Gangopadhyay, and H. Temkin, Appl. Phys. Lett. 81, 1065
͑2002͒.
where to demonstrate a low leakage dielectric that preserves
7
the high mobility of graphene FETs. In the case of Hf depo-
sition and oxidation, observation of carbide bonding after Hf
deposition indicates the presence of covalent bonding which
is likely to degrade graphene device performance by disrupt-
ing the graphene electronic structure. We find that predepo-
17
18
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
34.245.224.26 On: Thu, 25 Dec 2014 01:10:04
1