Self-cleaning and surface recovery with arsine pretreatment in ex situ atomic-layer-deposition of Al2 O3 on GaAs

Article abstract of DOI:10.1063/1.3213545

Annealing native oxide covered GaAs samples in Arsine (AsH₃) prior to atomic-layer-deposition of Al₂ O₃ with trimethyaluminum (TMA) and isopropanol (IPA) results in capacitance-voltage (C-V) characteristics of the treated

Full text of DOI:10.1063/1.3213545

APPLIED PHYSICS LETTERS 95, 082106 ͑2009͒
Self-cleaning and surface recovery with arsine pretreatment
in ex situ atomic-layer-deposition of Al O on GaAs
2
3
1
,a͒
2
2
1
Cheng-Wei Cheng,
John Hennessy, Dimitri Antoniadis, and Eugene A. Fitzgerald
1
Department of Materials Science and Engineering, MIT, Cambridge, Massachusetts 02139, USA
2
Department of Electrical Engineering and Computer Science, MIT, Cambridge, Massachusetts 02139, USA
͑
Received 27 May 2009; accepted 5 August 2009; published online 26 August 2009͒
Annealing native oxide covered GaAs samples in Arsine͑AsH ͒ prior to atomic-layer-deposition of
3
Al O with trimethyaluminum ͑TMA͒ and isopropanol ͑IPA͒ results in capacitance-voltage ͑C-V͒
2
3
characteristics of the treated samples that resemble the superior C-V characteristics of p-type GaAs
grown by an in situ metal-organic chemical vapor deposition process. Both TMA and IPA show
self-cleaning effect on removing the native oxide in ex situ process, little evidence of a native oxide
was observed with high resolution transmission electron microscopy at the Al O /GaAs interface.
2
3
The discrepancy in the C-V characteristics was observed in in situ p- and n-type GaAs samples.
2009 American Institute of Physics. ͓DOI: 10.1063/1.3213545͔
©
Beyond the 22–16 nm node, an option for continuing
scaling is to use III–V channels to potentially exceed the
benefits of strained silicon technology due to higher electron
mobility. To be benefited from a III–V channel for MOSFET
application, it must be able to fabricate high quality oxide/
III–V interfaces. In situ deposition of Ga O ͑Gd O ͒ on
GaAs substrates were prepared in an AIXTRON/Thomas
Swan close-coupled showerhead cold-wall MOCVD
system. The buffer epilayers of GaAs were grown on 2 in.
heavily doped GaAs͑001͒ substrates at 650 °C with
TMG͓Ga͑CH ͒ ͔ and arsine ͑AsH3͒ ͑V/III=23͒ with
3
3
2
3
2
3
disilane ͑Si H ͒ and DMZn͓Zn͑CH ͒ ͔ as the n- and p-type
2
6
3 2
1
7
−3
GaAs and InGaAs using molecular beam epitaxy ͑MBE͒ has
dopants with doping concentration of ͑2–6͒ϫ10 cm .
For the in situ experiments, the detail procedure has been
produced interfaces with a relatively low interfacial defect
1
,2
6
density ͑D ͒, as well as the inversion-channel n-MOSFET.
it
described in previous work. In brief, we deposited in situ
In situ deposition in ultrahigh vacuum prevents direct oxida-
tion of the channel, which pins the Fermi level at the oxide/
GaAs interface. Some ex situ methods such as ALD high-k
oxide on GaAs have achieved some success due to the self-
cleaning effect of the GaAs surface. Most of these processes
use metalorganics and water as the precursors for the ALD
homoepitaxial buffer layers, followed by an annealing to
drive off excess arsenic. ALD was initiated with TMA and
IPA at a temperature of 370 °C. For the ex situ growth ex-
periments, the substrates were stored in the atmosphere for
two days between the buffer and oxide growths, no wet
cleaning was performed prior to ex situ Al O deposition.
2
3
3
–5
oxide growth. More recently, the in situ ALD Al O /GaAs
2
3
TMA/IPA and TMA/H O were chosen as the precursors for
2
in metal-organic chemical vapor deposition ͑MOCVD͒ with-
out using water as a precursor was reported. The room tem-
perature C-V characteristics for p-GaAs clearly showed a
ex situ ALD Al O deposition. The samples with Al O
2
3
2
3
grown with TMA/IPA and AsH annealing treatment were
3
performed in the MOCVD reactor with the same parameters
lower C-V frequency dispersion and D than typical results
it
as were chosen in the in situ process. Ex situ TMA/H O was
2
6
from ex situ Al O /GaAs.
2
3
applied in a Cambridge NanoTech Savannah ALD Reactor to
deposit ALD Al O . The thicknesses of all Al O layers were
TMA/H O is the most widely used precursor set for the
2
2
3
2
3
3
–5
ex situ ALD Al O on III–V substrates.
However, the
2
3
grown with thickness of 11.5 nm in all processes except the
samples for HRTEM analysis. Aluminum electrodes
MBE results previously discussed suggest that strongly oxi-
dizing the III–V channel will create a strongly pinned inter-
face. A weaker oxidizer may create a larger process window
for these interfaces. In addition, when designing an in situ
2
00 ␮m in diameter were evaporated on samples and the
C-V characteristics were measured using a HP4192A LF im-
pedance analyzer at frequencies from 1 MHz to 10 kHz at
room temperature in the dark. The x-ray photoelectron spec-
troscopy ͑XPS͒ was performed in Kratos AXIS ultraimaging
MOCVD process, the typical TMA/H O is incompatible
2
with the MOCVD system. In this work, isopropanol ͑IPA͒
was used as oxygen source for in situ and ex situ Al O on
2
3
system using a monochromatic Al K␣ source ͑1486.6 eV͒
1
both p- and n-GaAs. For comparison, TMA/H O was also
2
with argon ion gun.
applied to deposit ex situ Al O in a commercial ALD sys-
2
3
Figure 1 shows the C-V curves of the in situ p- ͑a͒ and
n-GaAs ͑c͒ samples, and ex situ p- ͑b͒ and n-GaAs͑d͒
samples, respectively. In situ p-GaAs sample shows small
frequency dispersion at flat band voltage ͑20 mV/decade͒,
but the ex situ p-GaAs one shows larger dispersion ͑153
mV/decade͒, although the complete removal of the native
oxide was observed for the ex situ oxide growth with TMA/
IPA. Interestingly, the situation of n-GaAs is totally different
from that of p-GaAs. All the C-V curves are inferior to their
p-type counterparts. A clear transition from accumulation to
depletion was observed in the C-V curve of the ex situ
tem. High resolution transmission electron microscopy ͑HR-
TEM͒ and C-V measurement were performed to evaluate the
trimethyaluminum ͑TMA͒/IPA as a good self-cleaning agent
in ex situ case. Annealing air exposed GaAs samples under
AsH at high temperature is an alternative method which has
3
7
been shown to remove the native oxide. AsH pretreatment
3
step was introduced into the ex situ process before the oxide
deposition to study its influence on interface quality.
a͒Electronic mail: chengwei@mit.edu.
0003-6951/2009/95͑8͒/082106/3/$25.00
95, 082106-1
© 2009 American Institute of Physics

082106-2
Cheng et al.
Appl. Phys. Lett. 95, 082106 ͑2009͒
FIG. 1. ͑Color online͒ C-V characteristics of in situ Al O on ͑a͒ p-GaAs and ͑c͒ n-GaAs; ex situ Al O on ͑b͒ p-GaAs and ͑d͒ n-GaAs. All the Al O were
2
3
2
3
2
3
grown with TMA/IPA, except the inset of ͑b͒ ͑for inserted figures in ͑a͒ and ͑d͒, AsH annealing treatment was performed before the Al O growth for ex situ
3
2
3
process. For inserted figure in ͑b͒, the Al O were grown ex situ with TMA/H O͒.
2
3
2
n-GaAs sample, but this was not observed in that of the in
situ n-GaAs sample. The C-V curves of the in situ n-GaAs
sample shows a huge frequency dispersion in accumulation
region and this suggests that a very high defect density exists
at the interface.
removal of arsenic oxide and the arsenic, which are believed
4
,5
as one source to cause Fermi level pinning. Both TMA and
IPA could remove the arsenic oxide and arsenic directly by
forming volatile products, tertiary arsines ͓As͑CH ͒ ͔, and
3
3
arsenic alkoxides ͕As͓OCH͑CH ͒ ͔ ͖ ͑Refs. 8–10͒ and this
3
2 3
Figures 2͑a͒–2͑c͒ show the HRTEM image of the native
doubles the cleaning effect.
oxide/GaAs substrate, an ex situ Al O /GaAs, and Al O /Si
2
3
2
3
Heating GaAs substrates above 590 °C could also des-
orb the native oxide completely, but the surface becomes
rough due to the reaction between the substrate and the na-
interface using TMA/IPA as the precursors, respectively. The
H-passivated Si substrate was used as a reference. The thick-
ness of Al O grown on the Si and native oxide covered
2
3
tive oxide. Annealing in AsH is required to prevent the de-
3
GaAs are almost the same and sharp Al O /GaAs interfaces
2
3
composition of the GaAs and keep the surface smooth during
native oxide desorption. This could undo the oxidation pro-
cess and recover the surface when the ex situ procedure was
performed. The resemblance of the C-V characteristics ͓the
insets of Figs. 1͑a͒ and 1͑d͔͒ with that of in situ sample
were observed. The C-V characteristics of ex situ ALD
Al O grown with TMA/H O is shown in the inset of Fig.
2
3
2
1͑b͒. The incomplete removal of residual native oxide may
be the reason for the lower accumulation capacitance and
larger frequency dispersion for this sample. This implies the
optimal condition for self-cleaning effect should be carefully
controlled because only TMA is the agent to perform the
͓
Figs. 1͑a͒ and 1͑c͔͒ clearly shows the surface recovery. This
suggests the ex situ process that introduces the thermal clean-
ing process with AsH can imitate the in situ process.
3
8
cleaning effect. The self-cleaning effect is attributed to the
The bonding states of the Ga–O bonds were recently
shown to correlate with the C-V characteristics. Recently, the
3
+
removal of the Ga oxidation state at the interface was
shown to result in the reduction of frequency dispersion in
11
n-type sample. Figure 3͑a͒ shows the Ga 2p3/2 spectrum
from clean GaAs surface. This fitted single peak ͑GaAs͒ pro-
vides the reference to the following spectrum fitting. In Fig.
3
͑b͒, the Ga 2p3/2 spectrum from the in situ Al O /p-GaAs
2
3
interface is fitted carefully and two peaks are revealed. The
major peak is the Ga 2p3/2 in GaAs with the binding energy
of about 1117.1 eV and the smaller one is with the binding
energy of 1117.7 eV. The existence of the small peak could
also be contributed from plasmon loss and inelastic scatter-
ing of the photoelectrons. But, the plasmon loss usually pro-
duces a distinct and rather sharp hump 20–25 eV above the
FIG. 2. HRTEM image of the interfaces between ͑a͒ native oxide/GaAs, ͑b͒
ex situ Al O /GaAs, and ͑c͒ ex situ Al O /Si. The Al O was grown with
2
3
2
3
2
3
TMA and IPA. The scale are the same in all figures.

082106-3
Cheng et al.
Appl. Phys. Lett. 95, 082106 ͑2009͒
above the midgap for p- and n-type GaAs samples can be
accessible in the frequency region 10 kHz to 1 MHz at room
temperature, the comparisons can sufficiently be made.
In summary, interfacial self-cleaning and surface recov-
ery in ex situ ALD of Al O are demonstrated. The concept
2
3
of using the TMA/IPA as precursors to remove the native
oxide was confirmed by both HRTEM and the C-V measure-
ment, which shows small dispersion and high capacitance.
Ex situ Al O /n-GaAs sample shows better C-V characteris-
2
3
tics than that of in situ case. Moreover, annealing native
oxide covered GaAs samples in AsH is shown to undo the
3
oxidation process incurred when the ex situ procedure is per-
formed. The C-V characteristics of the AsH treated sample
3
resemble that of the sample under in situ process. The ex situ
process introduces the thermal cleaning process with AsH
3
can imitate the in situ process. No Ga+ oxidation state was
3
detected at the in situ Al O /GaAs interface. The D is
2
3
it
FIG. 3. ͑Color online͒ ͑a͒ Ga 2p3 XPS spectrum for clean GaAs surface
/2
11
2
around 5ϫ10 ͑1/cm eV͒ for both in situ and AsH an-
͑
b͒ Ga 2p3/2 XPS spectrum at in situ Al O /p-GaAs interface. The oxide
3
2
3
above the interface was thinned by the argon ion gun with etching rate of
nealed p-GaAs samples in the measureable trap energy
range.
0
2
.021 nm/min in XPS system and the residual thickness was estimated as
.4 nm. The Shirley background subtraction was included in all XPS fittings.
͑
c͒ The summary of D distribution of all samples in the bandgap. ͑Closed
it
This work is supported by the MARCO MSD Focus
Center. The authors thank Professor Akinwande for help in
electrical measurements at MIT. The authors also thank Pro-
fessor Guy Brammertz at IMEC, Dr. Shu-Wei Huang at MIT,
and Dr. Mao-Lin Huang at NTHU in Taiwan for helpful dis-
cussions.
and open symbols: in and ex situ process, respectively͒. Al O was depos-
2
3
ited with TMA/IPA unless specified. The open square represents the sample
which AsH annealing treatment was performed.
3
1
2
binding energy of the parent peak and the inelastic scatter-
ing of the photoelectron would contribute to the background
signal and determined by the Shirley model. The binding
energies of the Ga and Ga states were measured as 0.55
and 1.2 eV above the bulk peak, indicating Ga state could
be the possible peak located at 1117.7 eV and no Ga state
was seen in the spectrum. It was shown that the oxide growth
on n- and p-type GaAs substrates is identical when exposed
to identical environmental and chemical conditions. The
huge frequency dispersion in in situ n-type sample may come
from other sources like the surface kinks, which would gen-
erate a high density of donor defect states in the bandgap
1
3
1
+
1
+3
M. Passlack, M. Hong, J. P. Mannaerts, R. L. Opila, S. N. G. Chu, N.
Moriya, F. Ren, and J. R. Kwo, IEEE Trans. Electron Devices 44, 214
11
+1
2͑
1997͒.
+
3
F. Ren, J. M. Kuo, M. Hong, W. S. Hobson, J. R. Lothian, J. Lin, H. S.
Tasi, J. P. Mannaerts, J. Kwo, S. N. G. Chu, Y. K. Chen, and A. Y. Cho,
IEEE Electron Device Lett. 19, 309 ͑1998͒.
M. M. Frank, G. D. Wilk, D. Starodub, T. Gustafsson, E. Garfunkel, Y. J.
Chabal, J. Grazul, and D. A. Muller, Appl. Phys. Lett. 86, 152904 ͑2005͒.
M. L. Huang, Y. C. Chang, C. H. Chang, Y. J. Lee, P. Chang, J. Kwo, T.
B. Wu, and M. Hong, Appl. Phys. Lett. 87, 252104 ͑2005͒.
3
4
5
1
4
C. H. Chang, Y. K. Chiou, Y. C. Chang, K. Y. Lee, T. D. Lin, T. B. Wu, M.
1
5
+3
only on the fresh n-GaAs surface, not only due to the Ga
oxidation state.
Hong, and J. Kwo, Appl. Phys. Lett. 89, 242911 ͑2006͒.
6
7
C. W. Cheng and E. A. Fitzgerald, Appl. Phys. Lett. 93, 031902 ͑2008͒.
F. Reinhardt, W. Richter, A. B. Műller, D. Gutsche, P. Kurpas, K. Ploska,
K. C. Rose, and M. Zorn, J. Vac. Sci. Technol. B 11, 1427 ͑1993͒.
J. C. Bailar, Jr., H. J. Emeléus, S. R. Nyholm, and A. F. Trotman-
Dickenson, Comprehensive Inorganic Chemistry, Vol. 2, pp. 607–609
͑Academic, New York, 1973͒.
Figure 3͑c͒ shows the D distribution in the bandgap
it
extracted by the conductance-frequency method. The D of
the ex situ sample grown with TMA/IPA is low compared to
that grown with the conventional TMA/H O demonstrating
that the TMA/IPA is superior to a water-based process, and
in situ use of these precursors produces an interface with a
lower D . The D of the ex situ n-GaAs sample is half of that
in the in situ n-GaAs sample and this unexpected result im-
plies the ex situ process helps to improve the interface qual-
ity for the n-GaAs samples. The Dit is around 5ϫ10
8
it
2
9
0
A. G. de Souza, M. G. A. Brasilino, and C. Airoldi, J. Chem. Thermodyn.
28, 1359 ͑1996͒.
1
D. K. Srivastrava, L. K. Krannich, and C. L. Watkins, Inorg. Chem. 29,
it
it
113
502 ͑1990͒.
C. L. Hinkle, M. Milojevic, B. Brennan, A. M. Sonnet, F. S. Aguirre-
Tostado, G. J. Hughes, E. M. Vogel, and R. M. Wallace, Appl. Phys. Lett.
94, 162101 ͑2009͒.
J. F. Moulder, W. F. Stickle, P. E. Sobol, and K. D. Bomben, Handbook of
X Ray Photoelectron Spectroscopy: A Reference Book of Standard Spectra
for Identification and Interpretation of Xps Data ͑Physical Electronics
Division, Perkin-Elmer, Eden Prairie, 1992͒.
D. A. Shirley, Phys. Rev. B 5, 4709 ͑1972͒.
C. L. Hinkle, A. M. Sonnet, M. Milojevic, F. S. Aguirre-Tostado, H. C.
Kim, J. Kim, R. M. Wallace, and E. M. Vogel, Appl. Phys. Lett. 93,
1
1
2
12
͑
1/cm eV͒ for the in situ and AsH -annealed p-GaAs
3
1
6
samples in the measureable trap energy range. Of note, the
measurable trap state energy in the bandgap follows the
equation f =1/2␲͓exp͑⌬E/kT͒/␴v N͔, where f is the mea-
sure frequency, ⌬E is the energy difference between the ma-
jority carrier band edge energy and the trapping state energy
Et, k is the Boltzmann constant, T is the semiconductor tem-
perature, ␴ is the interaction cross section of the trapping
state, v is the thermal velocity of the majority charge carri-
ers, and N is the density of states in the majority carrier band.
Although only 0.27–0.4 eV below and the 0.29–0.38 eV
13
t
1
4
151
13506 ͑2008͒.
M. D. Pashley, K. W. Haberern, R. M. Feenstra, and P. D. Kirchner, Phys.
Rev. B 48, 4612 ͑1993͒.
16
t
G. Brammertz, H. C. Lin, K. Martens, D. Mercier, C. Merckling, J.
Penaud, C. Adelmann, S. Sioncke, W. E. Wang, M. Caymax, M. Meuris,
and M. Heyns, ECS Trans. 16, 507 ͑2008͒.