Metalorganic chemical vapor deposition of high mobility AlGaN/GaN heterostructures

S. Keller, G. Parish, P. T. Fini, S. Heikman, C.-H. Chen, N. Zhang, S. P. DenBaars, U. K. Mishra, and Y.-F. Wu
Citation: Journal of Applied Physics 86, 5850 (1999); doi: 10.1063/1.371602
View online: http://dx.doi.org/10.1063/1.371602
View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/86/10?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
AlGaN/GaN high-electron mobility transistors with low thermal resistance grown on single-crystal diamond (111)
substrates by metalorganic vapor-phase epitaxy
Appl. Phys. Lett. 98, 162112 (2011); 10.1063/1.3574531
Nanostructural characterization and two-dimensional electron-gas properties in high-mobility Al Ga N Al N Ga
N heterostructures grown on epitaxial AlN/sapphire templates
J. Appl. Phys. 98, 063713 (2005); 10.1063/1.2060946
Dislocation and morphology control during molecular-beam epitaxy of AlGaN/GaN heterostructures directly on
sapphire substrates
Appl. Phys. Lett. 81, 1456 (2002); 10.1063/1.1498867
High-mobility AlGaN/GaN heterostructures grown by molecular-beam epitaxy on GaN templates prepared by
hydride vapor phase epitaxy
Appl. Phys. Lett. 77, 2888 (2000); 10.1063/1.1323856
Two-dimensional electron gases in Ga-face and N-face AlGaN/GaN heterostructures grown by plasma-induced
molecular beam epitaxy and metalorganic chemical vapor deposition on sapphire
J. Appl. Phys. 87, 3375 (2000); 10.1063/1.372353

[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:
130.88.0.53 On: Sat, 30 May 2015 16:47:57

JOURNAL OF APPLIED PHYSICS

VOLUME 86, NUMBER 10

15 NOVEMBER 1999

Metalorganic chemical vapor deposition of high mobility

AlGaN/GaN heterostructures
S. Keller,a) G. Parish, P. T. Fini, S. Heikman, C.-H. Chen, N. Zhang, S. P. DenBaars,
and U. K. Mishra
Electrical and Computer Engineering and Materials Departments, University of California,
Santa Barbara, California 93111

Y.-F. Wu
Nitres Inc., 107 S. La Patera Lane, Goleta, California 93117

Received 25 May 1999; accepted for publication 9 August 1999

In this article, we discuss parameters influencing a the properties of thin Alx Ga1x N layers grown
by metalorganic chemical vapor deposition and b the electrical properties of the two-dimensional
electron gas 2DEG forming at the Alx Ga1x N/GaN heterojunction. For x Al0.3, the Alx Ga1x N
layers showed a strong tendency towards defect formation and transition into an island growth
mode. Atomically smooth, coherently strained Alx Ga1x N layers were obtained under conditions
that ensured a high surface mobility of adsorbed metal species during growth. The electron mobility
of the 2DEG formed at the Alx Ga1x N/GaN interface strongly decreased with increasing aluminum
mole fraction in the Alx Ga1x N layer and increasing interface roughness, as evaluated by atomic
force microscopy of the surfaces prior to Alx Ga1x N deposition. In the case of modulation doped
structures (GaN/Alx Ga1x N/Alx Ga1x N:Si/Alx Ga1x N), the electron mobility decreased with
decreasing thickness of the undoped spacer layer and increasing silicon doping. The electron
mobility was only moderately affected by the dislocation density in the films and independent of the
growth temperature of the Alx Ga1x N layers at x Al0.3. For Al0.3Ga0.7N/GaN heterojunctions,
electron mobility values up to 1650 and 4400 cm2/V s were measured at 300 and 15 K, respectively.
1999 American Institute of Physics. S0021-89799901122-6

I. INTRODUCTION

The alloy Al,GaN is of special interest for the fabrication of high mobility electron transistors HEMTs for high
power and high temperature microwave applications.14
With a direct band gap ranging from 3.41 to 6.2 eV for GaN
and AlN, respectively, the alloy system offers unique opportunities in the device design. In addition to this, the chemical
hardness of the material allows device operation in hostile
environments. However, although excellent results for
AlGaN/GaN electronic devices have been reported, only limited information has been published on the crystal growth of
AlGaN/GaN heterostructures and factors influencing the
properties of the two-dimensional 2DEG electron gas forming at the AlGaN/GaN interface.57 Growth studies of bulk
Alx Ga1x N layers revealed difficulties in obtaining atomically smooth surfaces, in particular at higher aluminum
compositions.8,9 Furthermore, local variations in the alloy
composition have been observed.1013 In the case of AlGaN/
GaN heterostructures, additional problems arise from the lattice mismatch between the Alx Ga1x N and the GaN base
layer. The strain in the Alx Ga1x N layer may cause formation of structural defects, and will also influence the electrical properties of the heterojunction via the piezoelectric
effect.14 Thus, a decrease in electron mobility with increasing x Al has been observed in previous experiments.3 To
achieve a high electron mobility of the 2DEG forming at
a

Electronic mail: stacia@ece.ucsb.edu

the AlGaN/GaN heterojunction, the interface must be laterally smooth and the composition transition vertically abrupt.
Theoretical calculations1517 predict an unusually large impact of the interface roughness on the 2DEG mobility in
comparison to GaAs and InP based HEMT structures.18 The
strong influence of the interface, in particular at higher aluminum mole fractions in the AlGaN layer, originates from
the larger band discontinuity and the higher effective mass in
the channel in combination with a higher interface charge
resulting from the strong piezoelectric and spontaneous polarization effects.19 The electrical properties of the 2DEG
also depend on the residual impurity concentration in the
epitaxial layers, and in the case of modulation doped structures, additionally on parameters such as the doping level
and the thickness of the undoped spacer layer separating the
doped layer from the channel,20 and a further factor has to be
considered in the discussion: dislocations, which are formed
during growth of the GaN base film due to the lattice mismatch to the sapphire or silicon carbide substrate, and penetrate through the AlGaN/GaN interface.
In this article, we report on the growth and characterization of thin Alx Ga1x N layers with 0x Al0.6 grown by
metalorganic chemical vapor deposition MOCVD. We systematically studied the effects of aluminum composition,
growth temperature, and V/III ratio on the morphology of the
AlGaN layers as well as on the electrical properties of the
2DEG formed at the AlGaN/GaN interface. In addition, the
influence of silicon doping, spacer layer thickness, disloca-

0021-8979/99/86(10)/5850/8/$15.00
5850
1999 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:
130.88.0.53 On: Sat, 30 May 2015 16:47:57

Keller et al.

J. Appl. Phys., Vol. 86, No. 10, 15 November 1999

5851

FIG. 1. Schematic of a modulation doped Alx Ga1x N/GaN structure.

tion density, and interface roughness on the electrical properties of the 2DEG was investigated.
II. EXPERIMENT

The 230 nm thick Alx Ga1x N layers were grown by

MOCVD on 3 m thick semi-insulating GaN on c-plane
sapphire substrates using the precursors trimethylgallium
TMGa, trimethylaluminum TMA1, and ammonia. The
growth of the GaN base layers was initiated on the sapphire
substrate with an approximately 20 nm thick GaN layer
grown at 525 C while the main GaN layer was grown at
temperatures between 1040 and 1080 C. 21 A wide parameter range was explored for the growth of the AlGaN layers:
the TMGa and the TMA1 flows were varied between 0.6 and
33 mol/min and the ammonia flow between 0.08 and 0.32
mol/min. The growth temperature ranged from 800 to
1125 C. The total gas flow and the reactor pressure were
kept constant at 11 l /min and 76 Torr, respectively. The
properties of the 2DEG forming at the heterojunction were
studied using modulation doped Alx Ga1x N/GaN structures.
Unless otherwise stated, the AlGaN layer was divided into a
2.5 nm thick undoped spacer layer, a 10 nm thick silicon
doped layer Si11019 cm3 , determined by secondary
ion mass spectroscopy and an undoped cap layer of a thickness of 5 nm for x Al0.35 and a thickness of 2 nm for x Al
0.4 see Fig. 1.
The surface morphology of the layers was studied by
atomic force microscopy AFM using a Digital Instruments
Nanoscope III, operated in tapping mode the root-meansquare rms roughness of the surfaces was calculated using
the AFM software. High-resolution x-ray diffraction with a
Phillips Materials Research Diffractometer was used for
composition determination. The electrical properties of the
2DEG formed at the AlGaN/GaN interface were evaluated
by Van der PauwHall measurements.
III. RESULTS AND DISCUSSION
A. AlGaN growth and surface morphology

As part of the detailed exploration of growth parameters,

at first the dependence of the aluminum composition on the
growth rate and the growth temperature was studied. An increase of the growth temperature (T gr) from 1070 to 1125 C
caused the aluminum mole fraction in the Alx Ga1x N layers
to rise from x Al0.37 to x Al0.42, at a constant TMGa and

FIG. 2. Dependence of the aluminum mole fraction x Al in the Alx Ga1x N

layers on a the growth temperature for f TMA1 f TMGa38 mol/min, and
b the TMA1 and TMGa input flow at a T gr1125 C. The TMA1/TMGa
ratio was 0.5.

TMA1 input flow of f TMA1 f TMGa38 mol/min and

TMA1/TMGa0.5 Fig. 2a. A significant increase in the
aluminum mole fraction from x Al0.36 to x Al0.52 was
also observed when the total TMGa and TMA1 input flow as
decreased from 76 to 19 mol/min at TMA1/TMGa0.5
and T gr1125 C Fig. 2b. Investigations of the AlGaN
layer thickness and the growth of GaN itself revealed that
this behavior predominantly originated from variations in the
GaN growth rate caused by changes in the driving force for
0
eq 22
0
eq
p Ga
). Here p Ga
and p Ga
are the input and the
growth, (p Ga
equilibrium partial pressure of gallium, respectively, with
0
being proportional to the TMGa input flow. Typically
p Ga
0
eq
eq
is much larger than p Ga
in MOCVD growth, and p Ga
can
p Ga
be neglected. However with increasing growth temperature
eq
0
p Ga
increases and with decreasing TMGa input flow p Ga
deeq
creases, in which case p Ga cannot be neglected anymore. On
the other hand, the equilibrium partial pressure of aluminum,
eq
is always much smaller than the aluminum input partial
p Al
0
, under the experimental conditions in this
pressure, p Al
study and can be neglected. Hence, with increasing T gr and
0
, the aluminum mole fraction in the
decreasing p Ga
Alx Ga1x N layers:23,24
s
x Al

0
k Al p Al

0
0
eq
k Al p Al
p Ga
k Ga p Ga

increased although the TMA1/TMGa ratio in the gas phase

was kept constant (k Al,k Ga-transport coefficients. The composition of the AlGaN layers was independent of the ammonia flow for 1.5 l /min f NH36 l /min. Prereactions between TMA1 and ammonia2528 were only observed at a
growth pressure higher than 150 Torr in our reactor.

[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:
130.88.0.53 On: Sat, 30 May 2015 16:47:57

5852

Keller et al.

J. Appl. Phys., Vol. 86, No. 10, 15 November 1999

FIG. 3. AFM images of 18 nm thick Alx Ga1x N layers with a x Al0.2, b x Al0.4, and c x Al0.6. gray scale: 3 nm.

Variations in the aluminum composition of the AlGaN

layers had a strong effect on their surface morphology. Figure 3 shows the AFM images of 18 nm thick Alx Ga1x N
layers with a x0.2, b x0.4, and c x0.6. The surface of film a is characterized by a step-like structure comparable to the surface of the GaN base layer see also Fig.
11a. The steps merge at threading dislocations with partial
screw character, seen as dark dots. The smaller dark spots
originate from pure edge dislocations.29 However, with increasing aluminum mole fraction a strong tendency towards
defect formation and transition into an island growth mode is
visible. The defects, seen as dark lines, start to form predominantly at dislocations as shown in Fig. 3b for 18 nm
Al0.4Ga0.6N. 30 Interrupted growth studies revealed that the
defects did not form at the Alx Ga1x N/GaN interface but
developed in a later stage of growth. Thus, the AFM image
of a 2 nm thick Al0.4Ga0.6N layer did not show any dark lines
and looked similar to the image of the 18 nm thick
Al0.2Ga0.8N film Fig. 3a. Preliminary transmission electron microscopy studies suggested that the defects relate to
stacking faults and when present in large densities result in a
partial strain relaxation in the layer.31 No significant effect of
the growth temperature on the surface morphology was
found in the temperature range 1070 CT gr1125 C.
However, the defect formation and transition into an island
growth mode could be largely prevented at low ammonia

flows low V/III ratios during growth as seen in Fig. 4 for

18 nm thick Al0.35Ga0.65N films. In Ref. 17, pseudomorphic
growth of Al0.38Ga0.62N up to a thickness of even 65 nm was
observed. These results imply that the defect formation did
not exclusively originate from the strain in the Alx Ga1x N
layers, but was also influenced by the growth kinetic. Thus,
defect formation and island growth could be suppressed under conditions that ensured a high surface mobility of adsorbed metal species. Consequently, at higher x Al values the
ammonia flow needed to be reduced due to the higher bond
strength of the AlN 2.88 eV compared to the GaN 1.93
eV bond.32 However, further investigations are needed to
fully understand the defect formation mechanism. Since the
growth of AlGaN/GaN heterostructures for device applications requires the growth of low defect density, coherently
strained AlGaN layers, all AlGaN films under discussion in
the following sections were grown with a NH3 flow of 1.5
l /min.
B. Electrical properties of the 2DEG

The properties of the 2DEG forming at the

Alx Ga1x N/GaN interface were investigated using modulation doped Alx Ga1x N layers as described in the experimental section. In Fig. 5, the electron mobility ( 2DEG) and the
sheet carrier density (n s ) of the 2DEG measured at room

FIG. 4. AFM images of 18 nm thick Al0.35Ga0.65N layers grown with a NH3 flow of a 1.5, b 3, and c 6 l /min gray scale: 3 nm.
[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:
130.88.0.53 On: Sat, 30 May 2015 16:47:57

Keller et al.

J. Appl. Phys., Vol. 86, No. 10, 15 November 1999

FIG. 5. Dependence of the electron mobility filled circles and the sheet
carrier density open circles measured at 300 K on the aluminum mole
fraction in the Alx Ga1x N layer.

temperature are plotted versus the aluminum mole fraction in

the Alx Ga1x N layers. For x Al0.2, the electron mobility
dropped quickly with increasing aluminum mole fraction in
the layer. The sheet carrier density n s increased steadily with
increasing x Al . For selected samples, the temperature dependence of 2DEG and n s was measured Fig. 6. At low temperatures, 2DEG showed an even stronger dependence on the
alloy composition. Thus at 15 K, 2DEG dropped from 4400
to 2800 cm2 /V s with only a 2% increase in the aluminum
mole fraction from x Al0.34 to x Al0.36. The mobility of
the Al0.65Ga0.35N/GaN sample was generally very low. Also,
after a slow increase with decreasing temperature, 2DEG decreased again at T170 K and slightly increased at T30
K. This behavior is expected in the presence of band tail
states15 possibly originating from poor interface quality
and/or clustering. For all three samples, the sheet carrier density was independent of the measurement temperature as expected for a 2DEG not shown.
Figure 7 illustrates the effect of the growth temperature
of the Alx Ga1x N layers. The Hall measurements were performed at 77 K, to eliminate corruption of the data by the
underlying GaN base layers, which were slightly n type. Neither nor n s showed any significant dependence on the

5853

FIG. 7. Dependence of the electron mobility filled symbols and the sheet
carrier density open symbols measured at 77 K on the growth temperature
of the Alx Ga1x N for x Al0.35 circles and x Al0.32 triangles.

growth temperature at a constant aluminum composition of

the layers.
The influence of the silicon doping level in the
AlGaN:Si layer on the room temperature mobility and sheet
carrier density is shown in Fig. 8 for samples with x Al
0.35. The Al0.35Ga0.65N layers were grown at two different
growth temperatures, 1125 or 800 C. For both sets of
samples, the mobility decreased only slightly with increasing
doping level and dropped significantly only at a very high
silicon injection. The sheet carrier density increased with increasing doping for films grown at 800 C but saturated for
the samples grown at 1125 C.
The effect of the thickness of the undoped spacer layer
on the electrical properties of the 2DEG at 300 K is seen in
Fig. 9. An increase in the spacer layer thickness resulted in a
significant increase of the electron mobility and a decrease in
the sheet carrier density. For 3.3 nm spacer layer thickness

FIG. 8. Dependence of the electron mobility filled symbols and sheet

carrier density open symbols measured at 300 K on the silicon doping in
the Alx Ga1x N:Si layer (x Al0.35) for growth performed at T gr1125 C
FIG. 6. Temperature dependence of the electron mobility for samples with
different aluminum mole fraction in the Alx Ga1x N layer.
circles and T gr800 C triangles.
[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:

130.88.0.53 On: Sat, 30 May 2015 16:47:57

5854

Keller et al.

J. Appl. Phys., Vol. 86, No. 10, 15 November 1999

FIG. 9. Dependence of the electron mobility filled circles and the sheet
carrier density open circles measured at 300 K on the thickness of the
undoped spacer layer, separating the silicon doped layer from the channel.

and x Al0.33 a 2DEG mobility as high as 1655 cm2 /V s was

measured at 300 K.
In addition to factors related to the growth of the
Alx Ga1x N layers, we also studied influences caused by the
properties of the GaN base layer. In Fig. 10, mobility and
sheet carrier density of the 2DEG measured at 300 K are
plotted against the full width at half maximum FWHM of
the 102 GaN x-ray diffraction peak. Former studies had
shown that the FWHM of the 102 reflection is a better
measure of the dislocation density than that of the 002
reflection, which is insensitive to the density of pure edge
dislocations.33 In this experiment, the dislocation density was
tuned by slight variations of the GaN growth conditions in
the initial stage of growth. The 2DEG mobility decreased
with increasing FWHM of the 102 diffraction peak and
increasing dislocation density in the layers. However, for
samples with a FWHM700 arcsec dislocation density
1010 cm2 ) mobility values of about 1000 cm2/V s were

FIG. 11. AFM images of the surfaces of a 3 m standard GaN grown at

1060 C (rms0.13 nm; b 10 nm In0.04Ga0.96N grown on top of 3 m
standard GaN rms0.19 nm; and c 10 nm GaN grown at 800 C on top
of 3 m standard GaN rms0.49 nm. The image size is 11 m2. The
schematics of the HEMT structures are shown on the right-hand side.

still measured. The sheet carrier density was affected by the

dislocation density too and decreased slightly with increasing
FWHM.
In the final part of our study we conducted experiments
to investigate the impact of the surface roughness of the base
layer and the interface roughness on the mobility of the
2DEG. Since the GaN base layers utilized in all the structures discussed so far showed the smoothest surfaces we can
presently achieve by MOCVD, the surface roughness was
tuned by inserting 10 nm thick GaN and In0.04Ga0.96N layers
grown at 800 C Fig. 11. The indium mole fraction was
kept low to suppress effects caused by an inhomogeneous
indium incorporation.34 Both layers were grown at a very
slow growth rate of 0.02 nm/s. Previously conducted photoluminescence, Hall effect, and secondary ion mass spectroscopy SIMS investigations of GaN and InGaN layers grown
under these conditions had shown that the layers grown at
800 C are of comparable quality to those grown at temperatures above 1000 C. Despite the slow growth rate, the lower
growth temperature and the reduced surface mobility of adsorbed species caused the step density on the surface to increase as seen in previous experiments.35,36 Thus, the surface
roughness increased from rms0.13 nm Fig. 11a to
rms0.19 nm Fig. 11b after In0.04Ga0.96N deposition. The
undoped low temperature LT GaN films showed step
bunching effects, making the surfaces even rougher, as reflected by the rms value of 0.49 nm Fig. 11c. In Fig. 12,
the 300 K mobility of the 2DEG formed at the
Al0.3Ga0.7N/GaInN interface is plotted versus the rms values derived from the AFM images of the surfaces prior to
deposition of the Al0.3Ga0.7N layer. The increase in surface
roughness from rms0.13 nm to rms0.49 nm caused the

FIG. 10. Dependence of the electron mobility filled circles and the sheet
carrier density open circles measured at 300 K on the FWHM of the GaN
102 x-ray reflection. The FWHM is proportional to the dislocation density
in the layer.
[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:

130.88.0.53 On: Sat, 30 May 2015 16:47:57

J. Appl. Phys., Vol. 86, No. 10, 15 November 1999

FIG. 12. Dependence of the electron mobility on the surface roughness prior
to Alx Ga1x N deposition for heterostructures with x Al0.3. The rms values
were calculated using the AFM software.

electron mobility to drop dramatically from 1400 to 800

cm2/V s.
In analyzing the above-mentioned experimental parameters with respect to their impact on the electrical properties
of the 2DEG, three groups can be distinguished: a parameters having a strong impact such as aluminum composition,
surface roughness, and spacer layer thickness; b parameters
with moderate influence such as silicon doping and FWHM/
dislocation density; and c factors with no measurable influence such as the growth temperature. Looking at the parameters of group a both Al composition and surface roughness
had been predicted to be of particular importance to the electrical properties of Alx Ga1x N/GaN heterojunctions. Following the model of Zhang and Singh,15 with increasing x Al the
conduction band discontinuity, the strain, and the piezoelectric field in the Alx Ga1x N layer increase, resulting in a
higher two-dimensional electron charge. However, the
higher the sheet carrier density and the sheet charge, the
more closely the electrons are pushed to the interface, and
the more the electron transport is affected by the surface
roughness. Hence, at a surface roughness 0, the decrease of
2DEG with increasing x Al could be explained by the increase
in sheet charge alone. Thereby, the higher the interface
roughness, the more 2DEG was expected to decline. The
experimental results plotted in Fig. 5 for x Al0.2 follow the
predicted trend. The mobility of the 2DEG at 300 K decreased from 1400 to 1050 cm2/V s as the sheet carrier density increased from 11013 to 1.71013 cm2 by increasing
x Al from 0.25 to 0.45, respectively. At low measurement
temperatures the electron mobility declined even faster with
increasing x Al T15 K, x Al0.34: 2DEG4400 cm2/V s,
x Al0.36: 2DEG2800 cm2/V s, Fig. 6, since at low
temperatures phonon scattering mechanisms, which are superimposed on the behavior at room temperature, are less
efficient.20 The strong impact of the surface/interface roughness on 2DEG was experimentally confirmed by the data
plotted in Fig. 12. At constant x Al in the Alx Ga1x N layer,
2DEG declined from 1400 to 1150 cm2/V s as the surface
roughness increased from 0.13 to 0.19 nm and dropped to a
low value of 800 cm2/V s at a surface roughness of
rms0.49 nm. In addition to the impact of the surface roughness, as x Al increases more difficulties are presented in the
growth of a laterally smooth and vertically abrupt heterojunction. The expansion of structural defects in the near surface region of the Alx Ga1x N layer as shown in Fig. 3c

Keller et al.

5855

may result in local variations in the strain field, which also

contribute to the decrease of 2DEG . For x Al0.2 under our
experimental conditions, the general trend 2DEG1/x Al was
contradicted. The 2DEG mobility of samples with x Al
0.15 ( 2DEG1080 cm2/V s) was considerably lower than
those of samples with x Al0.25 ( 2DEG1350 cm2/V s, Fig.
5. This behavior could be caused by parallel conduction in
the Al0.15Ga0.85N:Si layer, since a steady increase of 2DEG
with decreasing x Al had been observed in former studies with
lower Si doping.3 It is also possible that at low Al-mole
fractions, where the sheet carrier density is lower and the
2DEG less confined, other defects, such as ionized impurities
and/or dislocations, are less efficiently screened and contribute more to the overall scattering.20
The premise that ionized impurities/dopants, if high
enough in concentration, can limit the mobility of the 2DEG,
is seen in the increase of 2DEG with increasing spacer layer
thickness for the modulation doped Al0.33Ga0.67N/GaN structures Fig. 9. At a doping level of SiSIMS11019 cm3 ,
2DEG increased from about 700 to 1600 cm2/V s by increasing the spacer layer thickness from 0.3 to 3.2 nm. The sheet
carrier density dropped from about 1.5 to 0.81013 cm2 .
Again, the increase in sheet carrier density was accompanied
by a decrease in electron mobility. In addition, the influence
of ionized dopants can be evaluated by changing the silicon
doping level in the AlGaN layer Fig. 8. In particular the
series of samples with Al0.35Ga0.65N layers grown at T gr
1125 C may allow a distinction between the general effects of the sheet carrier density and the influence of
impurities/dopants. For this series, n s saturated at a
Si2H6 /TMGaTMA1 ratio of about 2.5104 corresponding to SiSIMS21019 cm3], but the electron mobility still declined slightly from about 1250 to 1150 cm2/V s by
further increasing the Si2H6 /TMGaTMAl ratio. The reasons for the charge saturation are presently not well understood, but similar observations have been made for Si2H6
doping of AlGaN bulk layers.37 SIMS measurements showed
that the chemical Si concentration in the layers increased
proportionally to the Si2H6 /TMGaTMAl ratio.22 Also, n s
still increased with silicon doping when growing the
Al0.35Ga0.65N layers at a much lower temperature of 800 C.
However, in both cases n s was significantly lower than the
value extrapolated from a simple addition of the sheet carrier
density induced by spontaneous polarization and piezoelectric effect, equivalent to n s 81012 cm2 , measured for
the undoped structure and the sheet carrier density expected
from the doping in the AlGaN:Si layer, if all incorporated Si
atoms were electrically active. Further investigations to
clarify these issues are presently under way. For AlGaN
grown at 800 C, 2DEG decreased slightly with increasing
silicon doping as observed before. At the highest
Si2H6 /TMGaTMAl ratio of about 8104 , 2DEG
dropped to a very low value of 680 cm2/V s, possibly caused
by silicon diffusion into the undoped spacer layer occurring
at such a high doping concentration Si61019 cm3 ].
Other reasons for the reduction in 2DEG observed in the case
of high doping and/or very thin spacer layer thickness could
be parallel conduction in the AlGaN:Si layer or electron
spillover into the GaN beneath the channel.7 The experi-

[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:
130.88.0.53 On: Sat, 30 May 2015 16:47:57

5856

Keller et al.

J. Appl. Phys., Vol. 86, No. 10, 15 November 1999

ments at different growth temperature had been performed to

address the residual impurity incorporation, as higher growth
temperatures typically lead to an improved quality of GaN
films.21 Also, interface abruptness and/or alloy clustering
may depend on the growth temperature.38 However, under
our experimental conditions no influence of the growth temperature on the electrical properties of the 2DEG was found.
Regarding the influence of residual impurities, the chosen
aluminum mole fraction and n s ) may have been too high
and the 2DEG properties not sensitive enough to impurities,
as discussed before. Thus, slight differences in the alloy
composition (x Al0.32 vs x Al0.34) had a much stronger
impact on 2DEG than changing the growth temperature from
1070 to 1125 C.
Scattering by dislocations had been shown to affect the
electron mobility in bulk GaN films significantly. Specifically, the 300 K electron mobility decreased from about 600
to 150 cm2/V s, when the dislocation density increased from
4108 to 21010 cm2 , corresponding to an increase in the
FWHM of the 102 GaN x-ray diffraction peak from 413 to
740 arcsec, respectively.39 Theoretical calculations confirmed the importance of dislocation related scattering GaN
bulk films.40,41 In contrast to the observations for bulk films,
the electrical properties of the 2DEG formed at the
Al0.25Ga0.75N/GaN heterojunction in our experiment were
much less influenced by dislocations. Extrapolating 2DEG
using the values of the FHWM of the 102 reflection for the
GaN bulk films discussed above 413 and 740 arcsec, the
2DEG mobility would decrease from about 1550 to 1000
cm2/V s Fig. 9, by a factor of 1.5. This is much smaller than
the factor of 4 in the case of the GaN bulk films. Obviously,
the strong confinement of the 2DEG at the interface and the
high sheet charge result in a screening of the dislocation
related defects, reducing their impact on 2DEG . The sheet
carrier density decreased slightly, from about 1.2 to 1.05
1013 cm2 , with increasing FWHM from 500 to 800 arcsec, most likely due to trapping of electrons into acceptorlike defects created by the dislocations, as described in Ref.
41 and experimentally observed in scanning capacitance
measurements.42 The observed dependence of the 2DEG
properties on the dislocation density supports recent arguments that the higher 2DEG values measured for AlGaN/
GaN heterostructures deposited on SiC instead of sapphire
substrates Al0.2Ga0.8N/GaN: 2DEG 300 K2019 cm2/V s]
are related to a lower dislocation density in GaN on SiC
films.6 On the other hand, a 2DEG electron mobility as high
as 1860 cm2/V s (n s 4.81012 cm2 ) at 300 K has been
recently obtained for Al0.12Ga0.88N/GaN heterostructures
grown by molecular beam epitaxy MBE on MOCVD GaN
on sapphire base layers, in the presence of the same amount
of dislocations as in the entirely MOCVD grown heterostructures discussed here.43 The MBE results reinforce the lower
impact of dislocations on the electron mobility in the case of
2DEG structures compared to bulk films.

studied in relationship to a wide range of experimental parameters. At higher aluminum mole fractions, Alx Ga1x N
showed a strong tendency towards defect formation and island growth. Atomically smooth, coherently strained
Alx Ga1x N layers could be obtained under growth conditions that ensured a high surface mobility of adsorbed metal
species. The electrical properties of the 2DEG formed at the
Alx Ga1x N/GaN heterojunction were most strongly affected
by the aluminum mole fraction in the Alx Ga1x N layer and
the interface roughness, and in the case of modulation doped
structures, by the thickness of the undoped spacer layer. Specifically, 2DEG decreased quickly with increasing x Al and
interface roughness and decreasing spacer layer thickness.
The properties of the 2DEG were only moderately affected
by parameters such as the silicon doping of the AlGaN:Si
layer if sufficiently separated from the channel and the dislocation density, and were independent of the Alx Ga1x N
growth temperature. In agreement with theoretical considerations, the strong confinement of the 2DEG at the interface
and the high sheet charge of the 2DEG reduced the impact of
scattering related to ionized impurities/dopants and dislocations, in particular at higher aluminum mole fractions in the
Alx Ga1x N layers. For Alx Ga1x N/GaN heterojunctions
with x Al0.3 electron mobilities up to 1650 and 4400
cm2/V s at 300 and 15 K, respectively, were obtained.
ACKNOWLEDGMENTS

The authors gratefully acknowledge the support of the

Air Force Office of Scientific Research through Contract No.
F49620-96-1-0398 supervised by Dr. Gerald Witt and the
Office of Naval Research through Contract No. N0014-96-11215 under the direction of Dr. John Zolper.
1

O. Aktas, Z. F. Fan, A. Botchkarev, S. N. Mohammad, M. Roth, T. Jenkins, L. Kehias, and H. Morkoc, IEEE Electron Device Lett. 18, 293
1997.
2
M. S. Shur and M. A. Kahn, MRS Bull. 22, 44 1997.
3
U. K. Mishra, Y.-F. Wu, B. P. Keller, S. Keller, and S. P. DenBaars, IEEE
Trans. Microwave Theory Tech. 46, 756 1998.
4
S. T. Sheppard, K. Doverspike, W. L. Pribble, S. T. Allen, and J. W.
Palmour, High Power Microwave GaN/AlGaN HEMTs on Silicon Carbide, 56th Annual Device Research Conference, June 1998.
5
J. M. Redwing, M. A. Tischler, J. S. Flynn, S. Elhamri, M. Ahoujja, R. S.
Newrock, and W. C. Mitchel, Appl. Phys. Lett. 69, 963 1996.
6
R. Gaska, J. W. Yang, A. Osinsky, Q. Chen, M. A. Kahn, A. O. Orlov, G.
L. Snider, and M. S. Shur, Appl. Phys. Lett. 72, 707 1998.
7
R. Gaska, M. S. Shur, A. D. Bykhovski, A. O. Orlov, and G. L. Snider,
Appl. Phys. Lett. 74, 287 1999.
8
M. D. Bremser, W. G. Perry, T. Zheleva, N. V. Edwards, O. H. Nam, N.
Parikh, D. E. Aspnes, and R. F. Davis, MRS Internet J. Nitride Semicond.
Res. 1, 1996.
9
S. Ruffenach-Clur, O. Briot, B. Gil, R.-L. Aulombard, and J. L. Rouviere,
MRS Internet J. Nitride Semicond. Res. 2, 1997.
10
D. Korakakis, K. F. Ludwig, Jr., and T. D. Moustakas, Appl. Phys. Lett.
71, 72 1997.
11
B. Neubauer, A. Rosenauer, G. Gerthsen, O. Ambacher, and M. Stutzmann, Appl. Phys. Lett. 73, 930 1998.
12
A. V. Buyanov, J. A. Sandberg, B. E. Sernelius, P. O. Holtz, J. P. Bergman, B. Monemar, H. Amano, and I. Akasaki, J. Cryst. Growth 189/190,
758 1998.
13
C. Kisielowski, O. Schmidt, and J. Yang, Mater. Res. Soc. Symp. Proc.
IV. CONCLUSIONS
482, 369 1998.
14
P. M. Asbeck, E. T. Yu, S. S. Lau, G. J. Sullivan, J. Van Hofe, and J.
The morphological and electrical properties of
Redwing, Electron. Lett. 33, 1230 1997.
15
Alx Ga1x N/GaN heterojunctions grown by MOCVD were
Y. Zhang and J. Singh, J. Appl. Phys. 85, 587 1999.
[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:

130.88.0.53 On: Sat, 30 May 2015 16:47:57

J. Appl. Phys., Vol. 86, No. 10, 15 November 1999

R. Oberhuber, G. Zandler, and P. Vogl, Appl. Phys. Lett. 73, 818 1998.
O. Ambacher, J. Smart, J. R. Shealy, N. G. Weimann, K. Chu, M. Murphy, W. J. Schaff, L. F. Eastman, R. Dimitrov, L. Wittmer, M. Stutzmann,
W. Rieger, and J. Hilsenback, J. Appl. Phys. 85, 3222 1999.
18
T. Kikkawa, K. Makiyama, H. Ochimizu, K. Kasai, and J. Komeno, J.
Cryst. Growth 145, 799 1994.
19
F. Bernardini, V. Fiorentini, and D. Vanderbilt, Phys. Rev. B 56, R10024
1997.
20
L. Hsu and W. Walukiewicz, Phys. Rev. B 56, 1520 1997.
21
B. P. Keller, S. Keller, D. Kapolnek, W. N. Yiang, Y.-F. Wu, H. Masui,
X. Wu, B. Heying, J. S. Speck, U. K. Mishra, and S. P. DenBaars, J.
Electron. Mater. 24, 1707 1995.
22
G. Parish, S. Keller, S. P. DenBaars, and U. K. Mishra, J. Electron. Mater.
accepted.
23
G. B. Stringfellow, Organometallic Vapor-Phase Epitaxy: Theory and
Practice Academic, San Diego, 1989, p. 115.
24
A. Koukitu, N. Takahashi, T. Taki, and H. Seki, Jpn. J. Appl. Phys., Part
2 35, L673 1996; A. Koukitu, N. Takahashi, and H. Seki, ibid. 36,
L1136 1997.
25
A. Thon and T. F. Kuech, Appl. Phys. Lett. 69, 55 1996.
26
J. Han, J. J. Figiel, M. H. Craford, M. A. Banas, M. E. Bartram, R. M.
Biefeld, Y. K. Song, and A. V. Nurmikko, J. Cryst. Growth 195, 291
1998.
27
T. G. Mihopoulos, V. Gupta, and K. F. Jensen, J. Cryst. Growth 195, 733
1998.
28
F. Nakamura, S. Hashimoto, M. Hara, S. Imanaga, M. Ikeda, and H.
Kawai, J. Cryst. Growth 195, 280 1998.
29
X. H. Wu, L. M. Brown, D. Kapolnek, S. Keller, B. Keller, S. P. DenBaars, and J. S. Speck, J. Appl. Phys. 80, 3228 1996.
30
These findings are supported by AlGaN growth studies on laterally epitaxial overgrown GaN: almost no defects were observed in the virtually
16
17

Keller et al.

5857

dislocation free overgrown regions; G. Parish, S. Keller, P. T. Fini, R.

Vetury, C.-H. Chen, S. P. DenBaars, U. K. Mishra, and Y.-F. Wu, Proceedings of the 1998 Conference on Optoelectronic and Microelectronic
Materials and Devices, Perth, Australia, 1998.
31
L. J. Zhao unpublished.
32
S. Porowski and I. Grzegory, in Properties of Group III Nitrides, edited by
J. P. Edgar INSPEC, London, 1994.
33
B. Heying, X. H. Wu, S. Keller, Y. Li, D. Kapolnek, B. P. Keller, S. P.
DenBaars, and J. S. Speck, Appl. Phys. Lett. 68, 643 1996.
34
S. Chichibu, K. Wada, and S. Nakamura, Appl. Phys. Lett. 71, 2346
1997.
35
S. Keller, S. F. Chichibu, M. S. Minsky, E. Hu, U. K. Mishra, and S. P.
DenBaars, J. Cryst. Growth 195, 258 1998.
36
S. Keller, U. K. Mishra, S. P. DenBaars, and W. Seifert, Jpn. J. Appl.
Phys., Part 2 37, L431 1998.
37
A. Y. Polyakov, N. B. Smirnov, A. V. Govorkov, M. D. Milvidskii, J. M.
Redwing, M. Shin, M. Skowronski, D. W. Greve, and R. G. Wilson,
Solid-State Electron. 42, 627 1998.
38
T. Suzuki, A. Gomyo, S. Iijima, K. Kobayashi, S. Kawata, I. Hino, and T.
Yuasa, Jpn. J. Appl. Phys., Part 1 27, 2098 1988.
39
S. Keller, B. P. Keller, Y.-F. Wu, B. Heying, D. Kapolnek, J. S. Speck, U.
K. Mishra, and S. P. DenBaars, Appl. Phys. Lett. 68, 1525 1996.
40
H. M. Ng, D. Doppalapudi, T. D. Moustakas, N. G. Weimann, and L. F.
Eastman, Appl. Phys. Lett. 73, 821 1998.
41
D. C. Look and J. R. Sizelove, Phys. Rev. Lett. 82, 1237 1999.
42
P. J. Hansen, Y. E. Strausser, A. N. Erickson, E. J. Tarsa, P. Kozodoy, E.
G. Brazel, J. P. Ibbetson, U. Mishra, V. Narayanamurti, S. P. DenBaars,
and J. S. Speck, Appl. Phys. Lett. 72, 2247 1998.
43
C. R. Elsass, Y. P. Smorchkova, B. Heying, E. Haus, P. Fini, K. Maranowski, J. P. Ibbetson, S. Keller, P. M. Petroff, S. P. DenBaars, U. K.
Mishra, and J. S. Speck accepted.

[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:
130.88.0.53 On: Sat, 30 May 2015 16:47:57