Ex situ Ohmic contacts to n-InGaAs
Ashish Baraskara兲
Department of Electrical and Computer Engineering, University of California, Santa Barbara, Santa
Barbara, California 93106
Mark A. Wistey
Department of Electrical Engineering, University of Notre Dame, Notre Dame, Indiana 46556
Vibhor Jain, Evan Lobisser, Uttam Singisetti, and Greg Burek
Department of Electrical and Computer Engineering, University of California, Santa Barbara, Santa
Barbara, California 93106
Yong Ju Lee
Technology Manufacturing Group, Intel Corporation, Santa Clara, California 95054
Brian Thibeault
Department of Electrical and Computer Engineering, University of California, Santa Barbara, Santa
Barbara, California 93106
Arthur Gossard
Department of Electrical and Computer Engineering and Materials Department, University of California,
Santa Barbara, Santa Barbara, California 93106
Mark Rodwell
Department of Electrical and Computer Engineering, University of California, Santa Barbara, Santa
Barbara, California 93106
共Received 22 February 2010; accepted 24 May 2010; published 19 July 2010兲
The authors report ultralow specific contact resistivity 共c兲 in ex situ Ohmic contacts to n-type
In0.53Ga0.47As 共100兲 layers, with an electron concentration of 5 ⫻ 1019 cm−3. They present the c
obtained for molybdenum 共Mo兲 contacts to n-type In0.53Ga0.47As, with the semiconductor surface
cleaned by atomic H before metal deposition. The authors compare these data with the c obtained
for contacts made without atomic H cleaning. After exposure to air during normal device processing,
the semiconductor surface was prepared by UV-ozone exposure plus a dilute HCl etch and
subsequently exposed to thermally cracked H. Mo contact metal was deposited in an electron beam
evaporator without breaking vacuum after H cleaning. Transmission line model measurements
showed a contact resistivity of 共1.1⫾ 0.9兲 ⫻ 10−8 ⍀ cm2 for the Mo/ In0.53Ga0.47As interface. This
c is equivalent to that obtained with in situ Mo contacts 关c = 共1.1⫾ 0.6兲 ⫻ 10−8 ⍀ cm2兴. Ex situ
contacts prepared by UV-ozone exposure plus dilute HCl 共without any atomic H exposure兲 result in
c = 共1.5⫾ 1.0兲 ⫻ 10−8 ⍀ cm2. © 2010 American Vacuum Society. 关DOI: 10.1116/1.3454372兴
I. INTRODUCTION
Very low resistance metal-semiconductor contacts are required for submillimeter-wave electronics; III-V bipolar and
field-effect transistors require a specific contact resistivity
共c兲 of less than 1 ⫻ 10−8 ⍀ cm2 to achieve simultaneous 1.5
THz current-gain 共f t兲 and power gain 共f max兲 cutoff
frequencies.1,2 Contact resistivity strongly depends on semiconductor surface preparation prior to metal deposition.3 In
situ metal deposition immediately after semiconductor
growth avoids surface oxidation and contamination, and contact resistivity 共c兲 as low as 共1.1⫾ 0.6兲 ⫻ 10−8 ⍀ cm2 has
been so obtained.4 But, transistor fabrication process flows
often require that contacts be formed after the semiconductor
has been exposed to air. To obtain low contact resistivity
with such ex situ contacts, surface preparation requires considerable attention. Ti/Pt/Au contacts deposited on Ar+ sputtered n-InGaAs result in c = 4.3⫻ 10−8 ⍀ cm2.5 c = 4.1
a兲
Electronic mail: ashish.baraskar@ece.ucsb.edu
C5I7
J. Vac. Sci. Technol. B 28„4…, Jul/Aug 2010
⫻ 10−8 ⍀ cm2 共after correcting for metal resistance兲 was obtained with TiW contacts to n-InGaAs; in those samples the
semiconductor surface was oxidized with UV generated
ozone and subsequently treated with dilute HCl prior to
metal deposition.3
III-V semiconductor surfaces contaminated by oxides and
carbon compounds can also be cleaned by atomic H.6 After
exposure to atomic H, atomically clean GaAs surfaces have
been confirmed with reflection high energy electron diffraction 共RHEED兲.7,8 Atomic H has been used to clean surfaces
prior to semiconductor regrowth9 by molecular beam epitaxy
共MBE兲. Atomic H cleaning is carried out at lower temperatures than conventional thermal oxide desorption and hence
causes less surface roughening due to less group V desorption. While thermal cleaning under group V flux is usually
carried out at 550 ° C for InP 共100兲 and InGaAs 共100兲, and
580 ° C for GaAs 共100兲, surface cleaning with atomic H can
be achieved at 390 ° C for these semiconductors.10,11 How-
1071-1023/2010/28„4…/C5I7/3/$30.00
©2010 American Vacuum Society
C5I7
C5I8
Baraskar et al.: Ex situ Ohmic contacts to n-InGaAs
C5I8
ever, considerable surface indium loss has been observed after prolonged atomic H cleaning of InGaAs alloys.11
We here report c for ex situ Mo contacts to n-type
In0.53Ga0.47As, with an electron concentration of 5
⫻ 1019 cm−3. We studied two surface cleaning techniques:
one involving UV-ozone/HCl treatment and the other involving UV-ozone/HCl/H treatment. We compare the c obtained
for these two techniques.
II. EXPERIMENT
The experiments used three vacuum chambers connected
under ultrahigh vacuum 共UHV兲, the first containing a H
cracking cell and a substrate heater, the second containing
the RHEED system used for surface characterization, and the
third containing an electron beam evaporator for metal deposition.
The semiconductor epilayers were grown by a Gen II
solid source MBE system. A 150 nm undoped In0.52Al0.48As
layer was grown on a semi-insulating InP 共100兲 substrate,
followed by 100 nm of silicon 共Si兲 doped In0.53Ga0.47As. The
samples were grown at a 420 ° C substrate temperature with
a 1400 ° C Si cell temperature. The active carrier concentration, mobility, and sheet resistance were obtained from Hall
measurements by placing indium 共In兲 contacts on samples.
The samples were removed from vacuum for a long period
共395 days兲; this permits the surface to oxidize, as would
occur during normal device processing. The samples were
then exposed to UV-ozone for 30 min and then treated with
1:10 HCl: H2O and de-ionized 共DI兲 water rinse for 1 min
each. The samples were then immediately loaded into the
vacuum chamber. On a first set of samples, without breaking
vacuum, 20 nm of Mo was deposited in an electron beam
evaporator without any further surface treatment. A second
set of samples was exposed to thermally cracked H for times
ranging from 20 to 40 min and temperatures ranging from
375 to 420 ° C. The filament temperature of the H cracking
cell was maintained at 2200 ° C. The chamber pressure during H cleaning was maintained at 10−6 torr. RHEED patterns
were recorded along the 关110兴 and 关1̄10兴 azimuths after H
cleaning. The samples were then transferred to the electron
beam evaporator where 20 nm Mo was deposited.
To extract the specific contact resistivity, the samples
were processed into TLM structures.12 For the TLM structures 共Fig. 1兲, Ti 共20 nm兲/Au 共500 nm兲/Ni 共50 nm兲 contact
pads were patterned using optical photolithography and liftoff after an e-beam deposition. The Au layer is 500 nm thick
to reduce interconnect resistance. Mo was then dry etched in
a SF6 / Ar plasma using Ni as an etch mask. The TLM structures were then isolated using mesas formed by photolithography
and
a
subsequent
wet
etching
with
1 : 1 : 25: H3PO4 : H2O2 : DI water. A scanning electron microscope 共SEM兲 image of the TLM pattern is shown in Fig. 2.
Resistances were measured using a four-point 共Kelvin兲
probe technique on an Agilent 4155C semiconductor parameter analyzer.4 In the Kelvin probing structure 共Fig. 2兲, the
observed resistance, Rmeasured = 2c / WLT + sLgap / W + Rmetal,
contains a small contribution Rmetal from the sheet resistivity
J. Vac. Sci. Technol. B, Vol. 28, No. 4, Jul/Aug 2010
FIG. 1. Cross-section schematic of the metal-semiconductor contact layer
structure used for contact resistivity 共c兲 measurements.
共m / Tm兲 of the contact metal. Here c is the metalsemiconductor contact resistivity, s is the semiconductor
sheet resistivity, LT = 冑c / s is the transfer length, m is the
bulk metal resistivity, and Tm is the contact metal thickness.
The dimensions W and Lgap are defined in Fig. 2. Rmetal is
determined from separate measurements of m / Tm and from
numerical finite-element analysis of the contact geometry.
Rmetal changes the contact resistivity data by less than 5% for
TLM structures with W = 25 m.
The sheet resistance of the semiconductor between the
contacts does not change after being exposed to SF6 / Ar
plasma etch for removing Mo.4 This validates the extraction
of the contact resistivity 共c兲 from the observed lateral access
resistivity 共H兲 and semiconductor sheet resistivity 共s兲.
III. RESULTS AND DISCUSSION
A diffuse 共1 ⫻ 1兲 RHEED pattern was observed on
samples without any H clean. As the samples were exposed
to atomic H and the exposure time and temperature are increased, a gradual improvement from a 共1 ⫻ 1兲 to a 共2 ⫻ 4兲
reconstruction was observed. Figure 3 shows the RHEED
patterns recorded along the 关110兴 and 关1̄10兴 azimuths after 40
min of atomic H exposure at 420 ° C. The observed 共2 ⫻ 4兲
reconstruction indicates an As-rich or As-terminated
surface.11,13,14
FIG. 2. SEM image of the TLM pattern used for the contact resistivity 共c兲
measurement. Separate pads were used for current biasing and voltage
measurement.
C5I9
Baraskar et al.: Ex situ Ohmic contacts to n-InGaAs
C5I9
the resistance analysis assuming that one-dimensional current flow is appropriate.15 Hall measurements on the samples
indicate an active carrier concentration, a mobility, and a
sheet resistance of 4.8⫻ 1019 cm−3, 984 cm2 / V s, and
13 ⍀, respectively. The sheet resistance obtained with TLM
measurements was 13.5 ⍀, which closely correlates with the
sheet resistance obtained with Hall measurement.
In separate experiments, samples doped at ⬃5
⫻ 1019 cm−3 were exposed to air for 2, 36, and 395 days, and
treated with UV-ozone/HCl/H. Mo was deposited and TLM
structures were fabricated on these samples. Observed c
were 1.1, 1.1, and 1.0 ⍀ m2, respectively. A separate set of
samples, doped at ⬃5 ⫻ 1019 cm−3, exposed to air for 2 and
395 days, was treated with UV-ozone/HCl. Mo contacts were
formed and TLM structures were fabricated on these
samples. Observed c were 1.4 and 1.5 ⍀ m2, respectively.
These data indicate that the duration of air exposure has little
effect on c given the surface preparation procedures employed.
FIG. 3. 共Color online兲 RHEED patterns of the atomic H cleaned sample
along the 关110兴 and 关1̄10兴 azimuths showing 共2 ⫻ 4兲 reconstructed surface.
Figure 4 shows the variation of TLM test structure resistance with contact separation for samples with different surface treatments. From the TLM data, c = 共1.1⫾ 0.9兲 ⫻ 10−8
and 共1.5⫾ 1.0兲 ⫻ 10−8 ⍀ cm2 were determined for samples
with UV-ozone/HCl/H treatment and UV-ozone/HCl treatment, respectively. The contact resistivity here obtained is
the lowest reported to date for ex situ contacts to n-type
In0.53Ga0.47As and is comparable to 共1.1⫾ 0.6兲
⫻ 10−8 ⍀ cm2 obtained for in situ Mo contacts to n-type
In0.53Ga0.47As.4
For the TLM structures, the transfer length was found to
be 280 nm, 2.8:1 larger than the N+ layer thickness; hence,
14
UV-O / HCl/ H clean
3
12
UV-O / HCl
3
Resistance (:)
10
8
6
4
2
0
0
5
10
15
20
25
Contact Separation (Pm)
FIG. 4. Measured TLM resistance as a function of contact separation for ex
situ molybdenum Ohmic contacts to n-In0.53Ga0.47As.
JVST B - Microelectronics and Nanometer Structures
IV. CONCLUSIONS
In summary, we report ultralow contact resistivity with ex
situ contacts to heavily doped n-type In0.53Ga0.47As. The contact resistivities obtained with UV-ozone/HCl cleaned and
UV-ozone/HCl/H cleaned contacts were 共1.5⫾ 1.0兲 ⫻ 10−8
and 共1.1⫾ 0.9兲 ⫻ 10−8 ⍀ cm2, respectively. The contact resistivity obtained here is comparable to that obtained with in
situ Mo contacts, suggesting the effective removal of surface
contaminants. These ultralow resistance ex situ Mo contacts
make them a potential candidate to be applied in highly
scaled HBTs and other devices of near-terahertz bandwidths.
1
M. J. W. Rodwell, M. L. Le, and B. Brar, Proc. IEEE 96, 271 共2008兲.
M. J. W Rodwell et al., Proceedings of the IEEE Compound Semiconductor Integrated Circuit Symposium, 2008.
3
Vibhor Jain, Ashish K. Baraskar, Mark A. Wistey, Uttam Singisetti, Zach
Griffith, Evan Lobisser, Brian J. Thibeault, Arthur C. Gossard, and Mark.
J. W. Rodwell, 21st IEEE International Conference on Indium Phosphide
and Related Materials, 2009 共unpublished兲, pp. 358–361.
4
Ashish Baraskar, Mark A. Wistey, Vibhor Jain, Uttam Singisetti, Greg
Burek, Brian Thibeault, Yong Ju Lee, Arthur Gossard, and Mark Rodwell,
J. Vac. Sci. Technol. B 27, 2036 共2009兲.
5
G. Stareev, H. Künzel, and G. Dortmann, J. Appl. Phys. 74, 7344 共1993兲.
6
G. R. Bell, N. S. Kaijaks, R. J. Dixon, and C. F. McConville, Surf. Sci.
401, 125 共1998兲.
7
T. Sugaya and M. Kawabe, Jpn. J. Appl. Phys., Part 2 30, L402 共1991兲.
8
A. Khatiri, J. M. Ripalda, T. J. Krzyzewski, G. R. Bell, C. F. McConville,
and T. S. Jones, Surf. Sci. 548, L1 共2004兲.
9
U. Singisetti et al., IEEE Electron Device Lett. 30, 1128 共2009兲.
10
T. Kikawa, I. Ochiai, and S. Takatani, Surf. Sci. 316, 238 共1994兲.
11
F. S. Aguirre-Tostado, M. Milojevic, C. L. Hinkle, E. M. Vogel, R. M.
Wallace, S. McDonnell, and G. J. Hughes, Appl. Phys. Lett. 92, 171906
共2008兲.
12
H. H. Berger, Solid-State Electron. 15, 145 共1972兲.
13
W. Braun, Applied RHEED: Reflection High-Energy Electron Diffraction
During Crystal Growth 共Springer, Germany, 1999兲.
14
P. A. Bone, J. M. Ripalda, G. R. Bell, and T. S. Jones, Surf. Sci. 600, 973
共2006兲.
15
E. G. Woelk, H. Krautle, and H. Beneking, IEEE Trans. Electron Devices
33, 19 共1986兲.
2