High mobility In[subscript 0.53]Ga[subscript 0.47]As quantum-well metal oxide semiconductor field effect

advertisement
High mobility In[subscript 0.53]Ga[subscript 0.47]As
quantum-well metal oxide semiconductor field effect
transistor structures
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation
Yang, Li, Cheng-Wei Cheng, Mayank T. Bulsara, and Eugene A.
Fitzgerald. “High Mobility In0.53Ga0.47As Quantum-Well Metal
Oxide Semiconductor Field Effect Transistor Structures.” Journal
of Applied Physics 111, no. 10 (2012): 104511. © 2012 American
Institute of Physics
As Published
http://dx.doi.org/10.1063/1.4721328
Publisher
American Institute of Physics (AIP)
Version
Final published version
Accessed
Thu May 26 10:33:13 EDT 2016
Citable Link
http://hdl.handle.net/1721.1/91922
Terms of Use
Article is made available in accordance with the publisher's policy
and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.
Detailed Terms
High mobility In0.53Ga0.47As quantum-well metal oxide semiconductor
field effect transistor structures
Li Yang, Cheng-Wei Cheng, Mayank T. Bulsara, and Eugene A. Fitzgerald
Citation: J. Appl. Phys. 111, 104511 (2012); doi: 10.1063/1.4721328
View online: http://dx.doi.org/10.1063/1.4721328
View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v111/i10
Published by the American Institute of Physics.
Related Articles
Complementary metal–oxide–semiconductor compatible athermal silicon nitride/titanium dioxide hybrid microring resonators
Appl. Phys. Lett. 102, 051106 (2013)
Thermal analysis of amorphous oxide thin-film transistor degraded by combination of joule heating and hot
carrier effect
Appl. Phys. Lett. 102, 053506 (2013)
Programmable ZnO nanowire transistors using switchable polarization of ferroelectric liquid crystal
Appl. Phys. Lett. 102, 053504 (2013)
Channel access resistance effects on charge carrier mobility and low-frequency noise in a polymethyl
methacrylate passivated SnO2 nanowire field-effect transistors
Appl. Phys. Lett. 102, 053114 (2013)
Tungsten oxide proton conducting films for low-voltage transparent oxide-based thin-film transistors
Appl. Phys. Lett. 102, 052905 (2013)
Additional information on J. Appl. Phys.
Journal Homepage: http://jap.aip.org/
Journal Information: http://jap.aip.org/about/about_the_journal
Top downloads: http://jap.aip.org/features/most_downloaded
Information for Authors: http://jap.aip.org/authors
Downloaded 08 Feb 2013 to 18.82.6.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
JOURNAL OF APPLIED PHYSICS 111, 104511 (2012)
High mobility In0.53Ga0.47As quantum-well metal oxide semiconductor
field effect transistor structures
Li Yang,1 Cheng-Wei Cheng,2 Mayank T. Bulsara,3 and Eugene A. Fitzgerald3
1
Department of Physics, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA
IBM T. J. Watson Research Center, Yorktown Height, New York 10598, USA
3
Department of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge,
Massachusetts 02139, USA
2
(Received 19 January 2012; accepted 21 April 2012; published online 25 May 2012)
In this paper, we demonstrate high electron mobility In0.53Ga0.47As quantum-well metal oxide
semiconductor field effect transistor (MOSFET) structures. The Al2O3 (gate dielectric)/
In0.53Ga0.47As-In0.52Al0.48As (barrier)/In0.53Ga0.47As (channel) structures were fabricated, and the
mobility was obtained by Hall measurements. The structures with in-situ chemical vapor
deposition (CVD) Al2O3 displayed higher mobility than identical structures fabricated with in situ
atomic layer deposition Al2O3, which indicates that CVD process resulted in a lower Al2O3/
In0.53Ga0.47As interfacial defect density. A gate bias was applied to the structure with CVD Al2O3,
and a peak mobility of 9243 cm2/V s at a carrier density of 2.7 1012 cm2 was demonstrated for
the structure with a 4 nm In0.53Ga0.47As-In0.52Al0.48As barrier. A model based on internal phonon
scattering and interfacial defect coulomb scattering was developed to explain the experimental data
C 2012 American Institute of
and predict the mobility of In0.53Ga0.47As MOSFET structures. V
Physics. [http://dx.doi.org/10.1063/1.4721328]
I. INTRODUCTION
InxGa1xAs alloys, due to their superior electron mobility, are promising candidates for n-type metal-oxide-semiconductor field effect transistors (MOSFETs). However, for
typical MOSFET structures, the need for doping in the channel and the direct contact of the channel and gate dielectric
lead to carrier scattering that reduces carrier mobility and
overall device speed. Radosavljevic et al. demonstrated a
high performance InxGa1xAs quantum-well MOSFET with
an undoped channel and a barrier layer in between the channel and the gate dielectric.1 GaN based quantum-well MOSFETs were also extensively studied by Ye’s group.2,3 For
such quantum-well MOSFET structures, while the barrier
layer reduces scattering effect and gives higher carrier mobility, it also increases the equivalent oxide thickness, which
is undesirable. Nagaiah et al. reported on their analysis of
channel mobility in In0.77Ga0.23As quantum-well MOSFET
structures.4,5 For III-V MOSFETs, Al2O3 is a commonly
used gate dielectric and researchers have demonstrated high
quality Al2O3/InGaAs interfaces.6–8 In our previous paper,9
we deposited Al2O3 using atomic layer deposition (ALD)
and presented both experimental and modeling results on
how the barrier thickness affects the carrier mobility for
In0.53Ga0.47As quantum-well MOSFET structures. In this paper, we improved the channel carrier mobility by depositing
an Al2O3 gate dielectric using chemical vapor deposition
(CVD). The channel mobility versus carrier density was
studied by gated Hall measurement, and a peak mobility of
9243 cm2/V s at a carrier density of 2.7 1012 cm2 was
demonstrated for an In0.53Ga0.47As MOSFET structure with
a 4 nm In0.53Ga0.47As-In0.52Al0.48As barrier. A model based
on internal phonon scattering and interfacial defect coulomb
scattering was developed to explain the experimental data.
0021-8979/2012/111(10)/104511/5/$30.00
This model can be generally used to predict the mobility of
quantum-well MOSFET structures limited by the gate
dielectric/semiconductor interface quality.
II. EXPERIMENTAL
The In0.53Ga0.47As quantum-well MOSFET structures
were deposited on semi-insulating InP(001) substrates in a
Thomas Swan/AIXTRON low pressure metal-organic chemical vapor deposition (MOCVD) system with a close-coupled
showerhead design. The III-V compound epitaxy was conducted in the CVD mode and the gate dielectric Al2O3 was
deposited in-situ in either the ALD mode10 or the CVD
mode.11 Trimethylgallium (TMGa), trimethylaluminum
(TMAl), trimethylindium (TMIn), arsine (AsH3), and phosphine (PH3) were used as the precursors for the III-V compound epitaxy, disilane (Si2H6) was used for the n-type
modulated doping, and TMAl and isopropyl alcohol (IPA)
were used for the deposition of Al2O3. High-purity N2 was
used for both the carrier gas and the purge gas. III-V compound epitaxy was conducted with a susceptor temperature of
650 C with a chamber pressure of 100 Torr. V/III ratio was
optimized to be 23 to give the highest mobility for the
In0.53Ga0.47As channel deposition. After the III-V compound
epitaxy sequence, under an AsH3 overpressure, the temperature was ramped down to a growth temperature of Al2O3,
which was 370 C for ALD and 500 C for CVD, and a chamber pressure was ramped down to 50 Torr. For the ALD mode
Al2O3 deposition, the pulse time for TMAl (with a partial
pressure of 0.03 Torr) and IPA (partial pressure of 0.20 Torr)
was kept at 4 s, and the N2 purge time in between TMAl and
IPA pulses was also kept at 4 s. For the CVD mode, Al2O3,
TMAl, and IPA were flowed simultaneously into the chamber,
where the TMAl and IPA partial pressures were kept as 0.01
111, 104511-1
C 2012 American Institute of Physics
V
Downloaded 08 Feb 2013 to 18.82.6.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
104511-2
Yang et al.
J. Appl. Phys. 111, 104511 (2012)
and 0.15 Torr, respectively. The in-situ Al2O3 deposition
ensures that the III-V compound surfaces were never exposed
to air prior to deposition of the gate dielectric.
The Hall mobility was measured by either the Van der
Paul method or the Hall bar method. For ungated structures,
a Van der Paul configuration was used; for gated hall structures, a Hall bar configuration was used. To form the ohmic
contacts, Al2O3 in the contact area was etched away with a
one minute buffered oxide etch, after which Au/Ge/Ni layers
were deposited by e-beam evaporation followed by a 30 s
450 C rapid thermal annealing. To form the gate contacts,
Al was deposited by thermal evaporation, and no thermal
annealing was applied afterwards. Based on the sample
dimensions, for the Van der Pauw configuration the Hall
measurement error is approximated to be 7.5% (Ref. 12) and
for the Hall bar configuration the error is around 12.7%.13
III. RESULTS AND DISCUSSION
Fig. 1(a) is an illustration of In0.53Ga0.47As quantumwell MOSFET structures fabricated in this study. A thin
(0.3 nm) In0.53Ga0.47As interlayer between the InP epi and
In0.52Al0.48As buffer was deposited to improve the surface
morphology. Fig. 1(b) gives the band diagram of the
quantum-well MOSFET structures. The electrons confined in
the InGaAs channel layer mainly contribute to the transport.
As discussed in our previous paper,9 semiconductor phonon scattering and coulomb scattering from interfacial
charged defects are two dominant scattering processes,
which determine the channel mobility for such quantum-well
structures. In this work, a model, which includes the above
two scattering mechanisms, was developed to calculate the
channel mobility. For the modeling, the temperature condition was assumed to be at 300 K. The charged interfacial
defects were assumed to be precisely at the Al2O3/
In0.53Ga0.47As interface with a two-dimensional (2D) density
of Nitc , which can be approximated as
ð þ1
ð EF
c
d
Dit dE þ
Dait dE;
(1)
Nit ¼
EF
1
where Ddit and Dait are the density of donor- and acceptor-like
interfacial traps, and EF is the Fermi level energy. Therefore,
to get the charged defect density Nitc , one need to know Ddit ,
Dait , and EF . This requires a better understanding of the types
of interfacial traps by techniques such as x-ray photoelectron
spectroscopy or surface potential fluctuation method.14 For
structures without the gate metal, the carriers were assumed
to be in the center of the channel with a 2D density of n,
which can be obtained from Hall measurements. Then, the
separation d between carriers and interfacial defects
becomes: d ¼ tbarrier þ tchannel
2 , where tbarrier is the barrier thickness and tchannel is the channel thickness (tchannel ¼ 15 nm in
our case).
The carrier mobility ðlÞ can be determined using the
Matthiessen’s rule
l¼
1
1
þ
;
lphonon lcoulomb
(2)
where lphonon is the mobility limited by semiconductor phonon scattering (12 000 cm2/V s at 300 K (Ref. 15)) and
lcoulomb is the mobility limited by coulomb scattering from
interfacial charges. lcoulomb can be obtained by the following
steps based on the established methods.16–19
lcoulomb correlates with the transport relaxation time in
the following form:
lcoulomb ¼
ehsi
;
m
(3)
where m is the effective mass of the electrons and hsi is the
mean transport relaxation time, which can be obtained from
Ð
@f
dEsðEÞE @E
;
hsi ¼
(4)
Ð
@f
dEE @E
where f is the room temperature Fermi-Dirac distribution
function, E is the electron energy, and sðEÞ is the transport
relaxation time. Using the Born approximation, the transport
relaxation time sðEÞ is given by
1
2p X
¼
sðEÞ h ~0
k
ð þ1
1
NðzÞjuðk~ k~0 ; zÞj2 ð1 cosðh~k k~0 ÞÞ
dðEk Ek0 Þdz;
(5)
FIG. 1. (a) Schematic and (b) band diagram of the quantum-well MOSFET
structures fabricated for this study. Note
that the distance between the dielectric/
III-V interface and the InGaAs channel
in these experiments is fixed at 4 nm.
Downloaded 08 Feb 2013 to 18.82.6.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
104511-3
Yang et al.
FIG. 2. Channel electron mobility versus the charged interfacial defect density. The blue “^” symbol and the green “~” symbol indicate the measured
mobility for structure with CVD Al2O3/In0.53Ga0.47As interface and ALD
Al2O3/In0.53Ga0.47As interface, respectively. The magenta “” symbols indicate the mobility limited by coulomb scattering, the red “n” symbols indicate the mobility limited by phonon scattering, and the black “þ” symbols
indicate the simulated final mobility. In the model, tbarrier was set as 4 nm
(the barrier thickness determined by the epitaxial structure), n was set as
1.5 1012 cm2 based on the Hall measurement data, and Nitc was extracted
by fitting the modeling data.
where NðzÞ is given by Nitc dðz dÞ, uð~
k~
k 0 ; zÞ is the
screened scattering potential dependent on the separation d
as well as the carrier density n, the angle hk~~k 0 is the scatter~
ing angle, and Ek is the 2D carrier energy for wave vector k.
0
~
~
uðk k ; zÞ can be calculated within the random phase
approximation, which is detailed in Ref. 19.
According to the above model, coulomb scattering
depends on the separation d between interfacial defects and
carriers, the density of charged interfacial traps Nitc , and the
carrier density n. In our previous paper,9 we studied the dependence of channel mobility on the barrier layer thickness
tbarrier by both experiments and modeling, where the experimental results showed a good agreement with the model. In
this work, Al2O3 was deposited in CVD mode to improve
the Nitc and a higher channel mobility was demonstrated. A
gate bias was applied to vary the carrier density n, and the
effect of n on channel mobility was then studied.
We have previously shown that CVD Al2O3/GaAs and
ALD Al2O3/GaAs possess different Dit distribution,10,11 and
therefore, CVD Al2O3/InGaAs and ALD Al2O3/InGaAs
should also have different interfacial properties. Fig. 2 compares the carrier mobility for structures with CVD Al2O3 and
ALD Al2O3. Application of a CVD Al2O3 gate oxide with
our structure improved the carrier mobility to 8883 cm2/V s,
which was much higher than the 6807 cm2/V s mobility
measured for the corresponding structure with an ALD
Al2O3 oxide. By fitting the experimental data with our
model, we were able to determine the interfacial charged
defect density. In the model, the barrier thickness was set as
J. Appl. Phys. 111, 104511 (2012)
4 nm, which gave an 11.5 nm separation between
carriers
and interfacial defects d ¼ tbarrier þ tchannel
;
the
carrier
2
density n was set as 1.5 1012 cm2 based on the Hall
measurement data. The fitted charged defect density for the
CVD Al2O3/In0.53Ga0.47As interface was 8.5 1012 cm2,
which is much lower than that calculated for the ALD
Al2O3/In0.53Ga0.47As interface, 1.7 1013 cm2. Fig. 2 also
shows that with a separation of 11.5 nm and a carrier density
of 1.5 1012 cm2, an interfacial charged defect density
lower than 2.3 1013 cm2 is required to render the effects
of coulomb scattering inconsequential as compared to phonon scattering effects.
According to the mobility model, coulomb scattering is
dependent on the carrier density, because carriers significantly screen the electric field of the interfacial charges and
therefore change the effective potential in the channel. We
intentionally changed the carrier density by applying a gate
bias and the effect of carrier density on the mobility was
then studied. As shown in Fig. 3(a), by applying gate voltages from 1.5 V to 3 V, we were able to tune the carrier
density from 1.2 1012 cm2 to 3.2 1012 cm2. Fig. 3(b)
is a plot of the mobility versus the carrier density for an
FIG. 3. (a) Carrier density versus gate voltage and (b) channel electron mobility versus carrier density for an In0.53Ga0.47As MOSFET structure with
CVD Al2O3 and a 4 nm In0.53Ga0.47As-In0.52Al0.48As barrier.
Downloaded 08 Feb 2013 to 18.82.6.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
104511-4
Yang et al.
In0.53Ga0.47As MOSFET structure with CVD Al2O3 and a
4 nm In0.53Ga0.47As-In0.52Al0.48As barrier. It is clear that
when the carrier density is less than 2.7 1012 cm2, the
mobility increases with increasing carrier density; when the
carrier density is greater than 2.7 1012 cm2, the mobility
starts to decrease with increasing carrier density.
The phenomena can be explained by the coulomb scattering in the model. Applying a gate voltage changes the carrier
density as shown in Fig. 3(a), meanwhile, the gate voltage also
shifts the position of the carriers, which leads to a change of
the separation between the interfacial charges and the carriers.
Sweeping the gate voltage from 1.5 V to 3 V, when the carrier density is less than 2.7 1012 cm2, the effect of the
increase of carrier density with increasing gate voltage dominates and the mobility increases; when the carrier density is
larger than 2.7 1012 cm2, the carrier density starts to saturate with increasing gate voltage, and the effect of the decrease
of the separation between the interfacial charges and the carriers dominates, which leads to the decrease of the mobility.
Based on the experimental results, we find that the mobility increases with increasing barrier thickness,9 decreasing
J. Appl. Phys. 111, 104511 (2012)
interfacial charge density and increasing carrier density. The
mobility model developed in this work quantifies the
effects of these three factors on the channel mobility of
In0.53Ga0.47As quantum-well MOSFET structures as shown
in Fig. 4. The modeling result further verifies that when Nitc is
high (>1012 cm2), a quantum-well (buried channel) structure is needed to give higher channel mobility than that of Si
MOSFET. From that perspective, fundamental advances in
dielectric/III-V interfaces will be required for III-V MOSFETs to compete with Si MOSFET performance. From a IIIV device perspective, these insulating-gate devices may offer
promise as insulating-gate high electron mobility transistors
(HEMTs). A conventional HEMT incorporates a metal gate
directly on the semiconductor, and therefore, such devices
have high gate current. Our results show that it may be possible to design an insulating-gate HEMT with good current
drive and and low gate leakage.
IV. CONCLUSION
In this work, we improved the channel electron mobility of In0.53Ga0.47As quantum-well MOSFET structures by
using CVD Al2O3 as the gate dielectric. This demonstrated
that the gate dielectric/semiconductor interface quality is a
very important factor to determine the mobility of a
MOSFET structure. The channel mobility versus carrier
density was then studied by applying a gate bias and a peak
mobility of 9243 cm2/V s at carrier density of 2.7 1012
cm2 was demonstrated for an In0.53Ga0.47As MOSFET
structure with a 4 nm In0.53Ga0.47As-In0.52Al0.48As barrier.
A mobility model based on internal phonon scattering and
interfacial defect coulomb scattering was developed to
quantify the effects of charged interfacial defect density,
carrier density, and the separation between defects and carriers on the channel mobility of In0.53Ga0.47As MOSFET
structures. According to the model, when the charged interfacial defect density is higher than 1012 cm2, quantumwell (buried channel) structures are needed to give higher
channel mobility than that of Si MOSFET. We conclude
that III-V MOSFETs will require significant fundamental
advances in dielectric/III-V interfaces to compete with Si
MOSFETs. However, our results suggest that an insulatinggate HEMT may have unique performance characteristics
for some applications when competing with conventional
Schottky-gate HEMTs.
ACKNOWLEDGMENTS
This work was supported by the SRC MSD Focus
Center Research Program. The authors would like to thank
Professor Lionel C. Kimerling for access to the Hall setup
and Professor Xi Lin for helpful discussion.
1
FIG. 4. Channel electron mobility versus density of charged interfacial traps
Nitc and the separation d between interfacial traps and carriers for a MOSFET
structure with a carrier density of (a) 1012 cm2 and (b) 2 1012 cm2,
respectively.
M. Radosavljevic, B. Chu-Kung, S. Corcoran, G. Dewey, M. K. Hudait,
J. M. Fastenau, J. Kavalieros, W. K. Liu, D. Lubyshev, M. Metz, K. Millard, N. Mukherjee, W. Rachmady, U. Shah, and Robert Chau, Tech. Dig. Int. Electron Devices Meet. 2009, 319.
2
W. D. Hu, X. S. Chen, Z. J. Quan, C. S. Xia, W. Lu, and P. D. Ye, J. Appl.
Phys. 100, 074501 (2006).
3
W. D. Hu, X. S. Chen, Z. J. Quan, X. M. Zhang, Y. Huang, C. S. Xia,
W. Lu, and P. D. Ye, J. Appl. Phys 102, 034502 (2007).
Downloaded 08 Feb 2013 to 18.82.6.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
104511-5
4
Yang et al.
P. Nagaiah, V. Tokranov, M. Yakimov, S. Koveshnikov, S. Oktyabrsky,
D. Veksier, W. Tsai, and G. Bersuker, J. Vac. Sci. Technol. B 28, C3H5
(2010).
5
S. Oktyabrsky, D. Veksler, P. Nagaiah, T. Chidambaram, V. Tokranov,
M. Yakimov, R. Kambhampati, Y.-T. Chen, G. Bersuker, N. Goel, and
C. Hobbs, ECS Trans. 35, 385–395 (2011).
6
Y. Xuan, T. Shen, M. Xu, Y. Q. Wu, and P. D. Ye, Tech. Dig. - Int. Electron Devices Meet. 2008, 371.
7
W. Wang, J. Deng, J. C. M. Hwang, Y. Xuan, Y. Wu, and P. D. Ye, Appl.
Phys. Lett. 96, 072102 (2010).
8
J. J. Gu, O. Koybasi, Y. Q. Wu, and P. D. Ye, Appl. Phys. Lett. 99,
112113 (2011).
9
L. Yang, C. W. Cheng, M. T. Bulsara, E. A. Fitzgerald, ECS Trans. 41,
219 (2011).
J. Appl. Phys. 111, 104511 (2012)
10
C. W. Cheng and E. A. Fitzgerald, Appl. Phys. Lett. 93, 031902
(2008).
11
C. W. Cheng and E. A. Fitzgerald, Appl. Phys. Lett. 96, 202101 (2010).
12
R. Chwang, B. J. Smith, and C. R. Crowell, Solid-State Electron. 17, 1217
(1974).
13
J. Haeussler and H. Lippmann, Solid-State Electron. 11, 173 (1968).
14
K. Ziegler, Appl. Phys. Lett. 28, 678 (1976).
15
D. P. Docter, K. R. Elliott, A. E. Schmitz, K. Kiziloglu, J. J. Brown, D. S.
Harvey, H. M. Karatnicki, in Proceedings of the 10th International Conference on Indium Phosphide and Related Materials (1998), p. 219.
16
F. Stern, Phys. Rev. Lett. 18, 546 (1967).
17
T. Ando, A. B. Fowler, and F. Stern, Rev. Mod. Phys. 54, 437 (1981).
18
F. Stern, Phys. Rev. Lett. 44, 1469 (1982).
19
S. Das Sarma and E. H. Hwang, Phys. Rev. B 69, 195305 (2004).
Downloaded 08 Feb 2013 to 18.82.6.136. Redistribution subject to AIP license or copyright; see http://jap.aip.org/about/rights_and_permissions
Download