Content uploaded by Vitaliy Dorogan
Author content
All content in this area was uploaded by Vitaliy Dorogan
Content may be subject to copyright.
Deep traps in GaAs/InGaAs quantum wells and quantum dots,
studied by noise spectroscopy
Vas. P. Kunets,a兲T. Al. Morgan, Yu. I. Mazur, V. G. Dorogan, P. M. Lytvyn,b兲M. E. Ware,
D. Guzun, J. L. Shultz, and G. J. Salamo
Arkansas Institute for Nanoscale Materials Science and Engineering, University of Arkansas,
Fayetteville, Arkansas 72701, USA
共Received 7 August 2008; accepted 26 September 2008; published online 19 November 2008兲
Remotely doped In0.35Ga0.65As layers of different coverages 6, 9, 11, and 13 ML were grown by
molecular beam epitaxy on 共100兲GaAs. Quantum dot 共QD兲nucleation was observed in situ by
reflection high-energy electron diffraction at 8 ML growth of In0.35Ga0.65As, while for 6 ML, only
two-dimensional 共2D兲growth was observed. Atomic force microscopy, low temperature
photoluminescence, and Hall effect measurements confirmed this transition from 2D to
three-dimensional growth. Low-frequency noise studies have been performed to probe defects in
such heterostructures throughout the transition from a highly strained quantum well to QDs. Results
were compared to a bulk n-type GaAs reference sample. We revealed three main defects in GaAs
with activation energies of 0.8, 0.54, and 0.35 eV. These defects with the same activation energies
were found in all samples. However, structures containing In0.35Ga0.65As QDs show an additional
peak at low temperatures due to the presence of defects which are not observed for reference GaAs
and quantum well samples. Detailed analysis shows that for 9 and 11 ML In0.35Ga0.65As QD samples
this peak corresponds to the well known M1 defect in GaAs with an activation energy of 0.18 eV,
while for a coverage of 13 ML the defect was found to have an activation energy of 0.12 eV. All
defects were characterized quantitatively in terms of their activation energy, capture cross section,
and density. These studies indicate that noise spectroscopy is a very sensitive tool for electronic
material characterization on the nanoscale. © 2008 American Institute of Physics.
关DOI: 10.1063/1.3020532兴
I. INTRODUCTION
During past decade, strain-driven epitaxy of III-V semi-
conductor 共QDs兲quantum dots have attracted the attention of
many researchers due to the unique physical properties of
QDs as a zero-dimensional quantum confined system and the
variety of applications in electronic and optoelectronic de-
vices. Indeed, fabrication of tunable, high efficient QD
lasers,1,2single or multicolored QD photodetectors3–5proved
their importance. There are an enormous number of scientific
reports on the skilful growth of well ordered, uniform, and
controlled density InAs QDs or InGaAs QDs of different
indium compositions. The optical properties of those were
studied extensively. Magnetotransport properties of QD
based heterostructures were studied less intensively and still
are of the great interest, especially for lateral transport. Most
of these reports were focused not only on the pure magne-
totransport properties of QDs themselves but also on elec-
tronic properties of high mobility channels 共remotely doped
quantum wells兲in the presence of QDs.6,7Besides optical
and transport properties of zero-dimensional systems, their
noise properties, and the noise properties of heterosystems
where QDs are utilized as a main functional element are of
the great importance. Better knowledge of the noise proper-
ties would result in improved performance of QD based de-
vices and in understanding of physical limitations dictated by
nature that are difficult to overcome. Recently, the noise
properties of InAs QDs were reported for heterostructures
based on vertical transport.8,9Besides, 1/f共flicker兲,
generation-recombination 共g−r兲, and thermal noise, shot
noise in the InAs QD heterostructures was observed at very
low temperatures as a result of single electron tunneling
through a layer of QDs.10
Even though the formation of InAs or InGaAs self-
organized QDs is well understood by efficient manipulation
of the strain, there are still open questions about defect for-
mation within the volume of the QDs and in the layer cov-
ering them. Even though, current technology in molecular
beam epitaxy 共MBE兲allows for the growth of extremely
high quality dislocation-free QDs, these still need to be
probed on the electronic level. Some defects, such as vacan-
cies, cannot be detected by structural or morphological stud-
ies but can be revealed electronically. In addition, one has to
take into account that QDs are nanoscale objects and tech-
niques allowing for the probing of defects inside of them
must be extremely sensitive.
There were many successful attempts to investigate the
presence of QDs and defects in these QDs or adjusting layers
through space charge spectroscopies. For example, InAs/
GaAs QDs were studied by capacitance-voltage 共C-V兲
spectroscopy,11,12 and InAs/GaAs QDs, InP/GaInP QDs, and
In0.5Ga0.5As/GaAs QDs by deep level transient spectroscopy
共DLTS兲and admittance spectroscopies.13–15 In the present
a兲Electronic mail: vkunets@uark.edu.
b兲Permanent address: V. Lashkaryov Institute of Semiconductor Physics,
NAS of Ukraine, Kiev 03028, Ukraine 1.
JOURNAL OF APPLIED PHYSICS 104, 103709 共2008兲
0021-8979/2008/104共10兲/103709/8/$23.00 © 2008 American Institute of Physics104, 103709-1
Downloaded 26 Nov 2009 to 217.20.172.209. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
work we are using deep level noise spectroscopy 共DLNS兲to
probe defects caused by In0.35Ga0.65As QDs grown between
GaAs barriers. We show that at low temperatures additional
defects appear depending on the In0.35Ga0.65As coverage.
These are a 0.18 eV deep trap for 9 and 11 ML and a 0.12 eV
deep trap for 13 ML coverage. No such defects are revealed
in the same temperature range for reference GaAs and In-
GaAs QW samples. In addition we observe narrowing of full
width at half maximum 共FWHM兲for 0.54 and 0.35 eV re-
lated defects typical for MBE grown GaAs with insertion of
In0.35Ga0.65As layers resulting in better resolution of noise
peaks. The possible mechanisms of this enhancement are dis-
cussed. These results are in good agreement with results of
Fang et al.14 observed by the technique of DLTS.
II. EXPERIMENTAL DETAILS
Using a Riber 32-P solid-source MBE system, we grew
eight different samples on semi-insulating 共100兲oriented
GaAs wafers. After removing the oxide layer at 610 ° C in an
As4flux, an undoped 5000 Å thick GaAs buffer layer was
grown at a rate of 1 ML/s. Then 5000 Å of uniformly Si-
doped GaAs 共Nd=7⫻1016 cm−3兲was grown at a substrate
temperature of 580 °C. A 200 Å thick undoped GaAs spacer
layer was then grown at the same temperature. At this point
growth was stopped and the substrate temperature was de-
creased to 540 °C, where a 200 Å thick undoped GaAs
spacer was grown. Then the substrate temperature was raised
back to 580 °C and the rest of the 1500 Å uniformly Si-
doped GaAs 共Nd=7⫻1016 cm−3兲cap layer was grown. The
growth conditions for next four samples were the same as for
this reference sample, S0, except that 6 ML 共sample S6兲,9
ML 共sample S9兲,11ML共sample S11兲, and 13 ML 共sample
S13兲of In0.35Ga0.65As were grown following the initial
200 Å spacer. The growth temperature for the InGaAs layers
was kept at 540 °C. The growth was monitored by reflection
high-energy electron diffraction 共RHEED兲and the surface
temperature was monitored by band edge measurement using
transmitted light in situ. The nucleation of QDs was observed
at about 8 ML of In0.35Ga0.65As growth. At the end we grew
three additional uncapped QD samples with 9, 11, and 13
ML of In0.35Ga0.65As for atomic force microscopy 共AFM兲
studies. The substrate temperature was rapidly cooled down
after each of the growths of the uncapped QDs.
The photoluminescence 共PL兲measurements were per-
formed at temperature of 10 K in a closed-cycle helium cry-
ostat. The excitation was done with a frequency-doubled yt-
trium aluminum garnet laser at 532 nm. The laser spot
diameter was ⬃20
m and the excitation power density was
20 W/cm2.
Using standard optical photolithography, four-terminal
Greek-cross structures, with square shaped active area of
20⫻20
m2were fabricated. Chemical wet etching was
used for mesa device isolation. Ohmic contacts were formed
by alloying an evaporated AuGe/Ni/Au 共75/20/250 nm兲
metal films. Ohmic contact resistance was optimized using a
fast thermal ramp up to 420 ° C followed by a 2 min anneal-
ing.
The van der Pauw technique was used for resistivity and
Hall effect measurements using a benchtop electromagnet
from MMR technologies at a magnetic field of 0.25 T. The
transverse noise was measured between Hall terminals using
an SR560 differential low-noise preamplifier and a SR785
noise spectrum analyzer in a well-shielded environment on
vibration free table. The frequency range of our measure-
ments covered 3 Hz to 100 kHz. The samples were biased by
a low-noise battery pack 共12/24 V兲in series with a low-noise
metal film load resistor RL. The current in the circuit can be
changed by choosing RL, where RLⰇRsample. The Hall effect
and noise measurements were made in temperature range of
80–400 K.
III. RESULTS AND DISCUSSION
AFM images of the open QD samples with different
In0.35Ga0.65As coverages are shown in Figs. 1共a兲–1共c兲. Our
structural analysis shows that height, density, and size distri-
bution depend on the number of deposited In0.35Ga0.65As
monolayers. For all coverages we observe only single mode
size distribution. For 9 ML of In0.35Ga0.65As we observe QDs
with an average height of 34 Å and areal density of NQD
=3.8⫻1010 cm−2. For 11 ML coverage, the height and areal
density of QDs increase to 47 Å and 8.4⫻1010 cm−2, re-
spectively. Further increase in coverage to 13 ML resulted in
a narrower size distribution 关Fig. 1共c兲兴, a lower density of
QDs NQD=7.2⫻1010 cm−2 compared to 11 ML coverage,
and the height being of 54 Å. It is necessary to mention that
a)
b
)
c) 13 ML
02468
0
20
40
60
80
Count
11 ML
02468
0
20
40
60
Count
9ML
02468
0
20
40
60
Count
Hei
g
ht
(
nm
)
FIG. 1. 共Color online兲AFM images of In0.35Ga0.65As QDs with different
coverages: 共a兲9ML共sample S9兲,共b兲11 ML 共sample S11兲,and共c兲13 ML
共sample S13兲. The height distributions are shown on the right. Enlarged 3D
images of QDs with different coverages are shown on the insets, where
lateral dimensions are 200⫻200 nm2.
103709-2 Kunets et al. J. Appl. Phys. 104, 103709 共2008兲
Downloaded 26 Nov 2009 to 217.20.172.209. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
shape of the dots is elliptical with an elongation along the
关011
¯
兴crystallographic direction as was reported before.16
The PL measurements for covered samples are shown in
Fig. 2. There is a noticeable redshift of the PL line with
increasing coverage. The maxima of the PL lines are mea-
sured at 1.301, 1.249, and 1.207 eV for 9, 11, and 13 ML,
respectively. Each PL line is well fit by a single Gaussian.
This indicates a single size distribution of the capped QDs.
The PL results are in a very good agreement with size dis-
tribution analysis shown in Figs. 1共a兲–1共c兲. From the Gauss-
ian fit of each PL band, the FWHM was determined. We
have measured FWHM of 38, 48, and 45 meV for 9, 11, and
13 ML, respectively. The lower FWHM for the 13 ML
sample agrees with AFM statistics, where the size distribu-
tion for the 13 ML sample also exhibits a smaller FWHM.
Another important feature is the PL line intensity as a func-
tion of coverage. Assuming defect-free QDs with a single
size distribution, one would expect higher integral PL inten-
sity for the samples with higher QD density. However, we
observe the opposite behavior, i.e., increasing the coverage
leads to a decrease in the integral PL intensity even though
the QD density does increase. We assume that such behavior
could be caused by appearing of additional channels of non-
recombinative radiation with coverage increase. This is a
topic of our further discussions.
The Hall effect measurements were performed to study
conductivity in the wide 80–400 Ktemperature range and to
provide required information for noise data analysis. The
Hall mobilities as function of temperature and coverage are
shown in Fig. 3. The reference GaAs sample 共S0兲exhibits a
typical mobility curve versus temperature.17 At high tem-
peratures the absolute value of mobility is limited due to
phonon scattering. By lowering the lattice temperature this
scattering mechanism becomes less efficient leading to an
increase in mobility. At low enough temperatures, scattering
by ionized impurities dominates and mobility is driven
mostly by impurity scattering. The resulting mobility peaks
at temperatures below 150 K and then decreases. In contrast
to the reference GaAs sample, sample S6 consisting of 6 ML
of In0.35Ga0.65As shows behavior typical for a remotely
doped quantum well. At high temperatures the mobility is
about the same as that of bulk GaAs. This is expected be-
cause at high temperatures the InGaAs quantum well is too
narrow to confine the conduction electrons only within In-
GaAs quantum well, and conductivity through the GaAs bar-
riers takes a place. However, at low temperatures the con-
ductivity mostly occurs through the InGaAs channel and less
through the Si-doped GaAs barriers. Thus, for this tempera-
ture range the mobility does not show a peak as in the case of
GaAs, because the electron gas is localized two dimension-
ally in the QW and experiences less scattering from ionized
donors in the barrier. However the absolute value of mobility
is not so high and conductivity in adjacent GaAs layers is
contributing. The temperature behavior of the mobility re-
veals the two-dimensional 共2D兲nature of the growth for 6
ML of In0.35Ga0.65As, which agrees with in situ RHEED ob-
servations. Mobility curves for samples with QDs are similar
to the reference GaAs sample. However the measured values
are lower. Besides, at low temperatures it is obvious that
samples S11 共11 ML兲and S13 共13 ML兲with higher densities
of QDs show lower mobility than the sample S9 共9ML兲.
This result is expected and can be explained if one will con-
sider each QD not only as potential quantum well were an
electron can spend some time but also as a localized charged
center with an ability to scatter the majority of carriers.18
Having the necessary information about the structural
and electronic properties of the samples, we studied their
noise characteristics. As an example, the room temperature
noise spectrum of the reference GaAs sample is shown in
Fig. 4. The noise voltage spectral density versus frequency
shows complex behavior. Besides 1/f共flicker兲noise and
thermal noise, the spectrum is distorted by two additional
Lorentzians at low and high frequencies due to the presence
of local centers in the band gap of the semiconductor. The fit
of the experimental noise spectrum can be done according to
SV,noise =B
f+兺
i
Ai
1
1+共2
f
i兲2+4kBTR,共1兲
where Band Aiare the amplitudes of the 1/fand g−rpro-
cesses in the volume of the semiconductor, fis the fre-
1.10 1.15 1.20 1.25 1.30 1.35 1.4
0
13 ML
11 ML
PL
S
ignal
(
arb. units
)
Photon Energy (eV)
9ML
T=10K
Iex =20W/cm
2
FIG. 2. PL spectra of In0.35Ga0.65As QDs with different coverages of 9 ML
共sample S9兲,11ML共sample S11兲, and 13 ML 共sample S13兲measured at
T=10 K and excitation power of 20 W /cm2.
50 100 150 200 250 300 350 400
4x103
5x103
6x103
7x103
8x103
9x103
11 ML
9ML
13 ML
0ML
6ML
Mobility (cm
2
/Vs)
Temperature
(
K
)
FIG. 3. Temperature dependent measurements of electron mobility for the
given coverages of In0.35Ga0.65As. Here a coverage of 0 ML represents the
reference GaAs sample.
103709-3 Kunets et al. J. Appl. Phys. 104, 103709 共2008兲
Downloaded 26 Nov 2009 to 217.20.172.209. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
quency,
iis the characteristic time constant of the g−rpro-
cess, kBis the Boltzmann constant, Tis the temperature, and
Ris the sample resistance. Fitting the data in Fig. 4to Eq.
共1兲, we found that g−rprocesses are characterized by
1
=4 ms and
2=24
s time constants.
By measuring the temperature dependence of
iand by
plotting ln共
i兲versus 1/T, one can determine the parameters
of the local levels in the band gap causing g−rprocesses in
the semiconductor. Another method, which has an excellent
theoretical background for the analysis of the local centers in
semiconductors, was proposed in Ref. 19. We will follow
these techniques to characterize local centers in our samples.
Figure 5shows the relative noise spectral density S
=SV/V2versus temperature for a frequency of 20 Hz for each
of the samples in this study. Here, Vis the voltage drop
across the sample. These data were taken from noise spectra
similar to that shown in Fig. 4measured through a tempera-
ture range of 80–400 K at steps of 10 K. For the reference
GaAs sample, S0, there are three peaks. These peaks, labeled
A, B, and C, are more clearly found in a linear scale, which
is shown in the inset in Fig. 5. The insertion of an
In0.35Ga0.65As layer into the GaAs leads to a pronounced
sharpening of the peaks. A similar behavior was observed by
Fang et al.,14 who applied DLTS for studies of In0.5Ga0.5As
QDs embedded in GaAs. In addition to this observation for
QD systems, we show here that even fora6MLchannel of
In0.35Ga0.65As where there are no QDs we see these changes
clearly. This can be a good indicator that the sensitivity of
DLNS is good enough for studies on nanoscale materials. As
a new observation, we find that for the 9, 11, and 13 ML
samples which resulted in the formation of QDs an addi-
tional low temperature peak has appeared in the noise spec-
tra. We could not detect this peak for reference sample or the
6MLIn
0.35Ga0.65As QW sample. Thus, we associate this
peak, D, uniquely with the presence of QDs in the GaAs
matrix.
To prove that peaks A, B, C, and D are defect related we
have performed studies of them as a function of temperature
and frequency. The peaks related to the local levels in semi-
conductors shift toward higher temperatures as well as de-
crease in intensity as the frequency is increased.19–21 Such
tendency is represented in Fig. 6for the 11 ML sample.
Noise data for other samples show similar behavior as func-
tion of temperature and frequency.
Following the work done in Ref. 19, and assuming that
10010110210310410
5
g-r
g-r
Johnson Noise, 4kBTR
1/f
i= 893.6 A
T = 294 K
S0
10-
12
10-13
10-14
10-15
10-16
10-17
SV(V
2
/Hz)
f
(
Hz
)
FIG. 4. Noise spectrum from the reference GaAs sample measured at bias
current of 893.6
AandT=294 K across the Hall terminals of the Greek-
cross structure. The fitting was done according to Eq. 共1兲. Each spectral
component participating in conductivity fluctuations is shown separately.
100 150 200 250 300 350 400 45
0
D
C
B
AS13
S11
S9
S6
S0
SV/V
2(arb. units)
Temperature
(
K
)
50 100 150 200 250 300 350 400
1.0x10-14
6.0x10-14
1.1x10-13
S0
C
B
A
Temperature (K)
SV/V2(Hz-1)
FIG. 5. Relative noise spectral density vs temperature for frequency of 20
Hz in a semilogarithmic scale. The curves are shown for all five samples: S0
reference GaAs; S6, S9, S11, and S13 are 6, 9, 11, and 13 ML of
In0.35Ga0.65As, respectively. The curves are shifted for clarity. The inset
shows the relative noise spectral density for sample S0 in a linear scale.
50 100 150 200 250 300 350 40
0
2560 Hz
1280 Hz
640 Hz
320 Hz
160 Hz
80 Hz
40 Hz
20 Hz
10 Hz
5Hz
S11 (11 ML)
D
C
B
A
10-
12
10-13
10-14
10-15
10-16
10-17
SV/V2(Hz-1)
Temperature (K)
FIG. 6. Relative noise spectral density measured as function of temperature
and frequency for 11 ML In0.35Ga0.65As QD sample. Four defects labeled as
A, B, C, and D are distinctly detected.
103709-4 Kunets et al. J. Appl. Phys. 104, 103709 共2008兲
Downloaded 26 Nov 2009 to 217.20.172.209. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
共i兲the capture cross section,
, depends exponentially on
temperature,
=
0exp共−E1/kBT兲, and 共ii兲the time constant
iassociated to the return to the equilibrium of the occupa-
tion of the level is given by
=
cF=
c0exp
冉
E1
kBT
冊
F,共2兲
the relative noise spectral density can be written as19,22
S=SV
V2=A
F共1−F兲
1+共
兲2=A
c0exp共E1/kBT兲F2共1−F兲
1+
2
c0
2exp共2E1/kBT兲F2,共3兲
where
c0=共
0
Tn兲−1 and Fis the occupancy of a level given
by F=1/共1+exp关共EF−E0兲/kBT兴兲. Here E1is the activation
energy, E0is the energy position of the deep level in the band
gap with respect to the bottom of conduction band, EFis the
Fermi energy,
cis the capture time constant,
Tis the ther-
mal velocity, nis the electron concentration,
=2
f, and
A=4Nt/VNd
2, where Ntis the trap density, Vis the sample
volume, and Ndis the concentration of the shallow donors.
To characterize the deep levels, we plotted dependencies
of ln共Smax兲on ln共
兲and 1/kBTmax on ln共
兲for each defect
and for all five samples. Figure 7shows these plots for QW
sample with 6 ML and QD sample with 11 ML of
In0.35Ga0.65As. The generalized results for all samples are
comprised in Table I.
The analysis of ln共Smax兲versus ln共
兲for defect A de-
tected at high temperatures 共see Fig. 5兲shows that this de-
pendence can be very well fitted by a straight line with a
slope of ⬃0.9. According to Ref. 19, we have EFlocated
above the trap level E0and E1⬎E0. For this case, assuming
F2⬇1 and solving the extremum problem for Eq. 共3兲, one
can find an activation energy E1and energy location of the
deep level E0as
E1=1
tan
T
,E0=1 − tan
s
tan
T
,共4兲
where tan
Tand tan
sare the slopes of 1/kBTmax versus
ln共
兲and ln共Smax兲versus ln共
兲, respectively. We found ac-
tivation energy E1to be ⬃0.8 eV and E0located at about
100 meV below the conduction band. For temperature range
of 300–360 K, where defect A is detected, the Fermi level for
GaAs is found to be 0.05 eV 共300 K兲and 0.07 eV 共360 K兲
located below the conduction band. According to our analy-
sis we have not exactly the case E0−EFⰇkBT, however, the
Fermi level is still above the trap over the whole temperature
range. Thus, we used the equations proposed in Ref. 19 for
calculating the trap density, Nt, and capture cross section,
0.
For a frequency, f=320 Hz, measured electron density, n
=6⫻1016 cm−3,
T=4.4⫻107cm/s, and density of states
Nc=5⫻1017 cm−3, we found
0⬃10−10–10−11 cm2and Nt
⬇1014 cm−3.
Trap B was detected for all five samples in the tempera-
ture range of 240–320 K. Analysis of ln共Smax兲versus ln共
兲
gives an average slope close to 1. This situation corresponds
to the case when E1ⰇE0and is unfavorable for the interpre-
tation of deep levels in semiconductors by noise
spectroscopy.19,23–25 Special techniques that inject minority
carriers 共band-to-band illumination or injection through a
forward-biased p-njunction兲were proposed to determine the
local level parameters.24,26,27 Following Refs. 25 and 28, one
can determine the activation energy, E1, using Eq. 共4兲. For
defect B we find that the average activation energy for the
studied samples is ⬃0.54 eV. Then the time constant
共Tmax兲is given by28
共Tmax兲=
0exp
冉
E1
kBT
冊
,共5兲
where the relation
共Tmax兲=2
f
共Tmax兲=1 共6兲
holds true.28
Then
0can be calculated as28
0=
exp
冉
E1
kBTmax
冊
=1
n
T
0
=exp共E1/kBTmax兲
n
T
共Tmax兲.共7兲
For frequency fmax = 320 Hz, n=6⫻1016 cm−3,
T=4.4
⫻107cm/s, and Smax⬇3⫻10−15 Hz−1, we have
0in the
range of 10−12–10−11 cm2. The defect density can be deter-
46810
30
40
50
60
7
0
796 meV
552 meV
388 meV
C
B
A
1/kBT(eV-
1
)
Ln
46810
-38
-36
-34
-32
-30
-28
S6
C
B
A
~0.99
~0.9
~0.87
LnS
max
Ln
(
a)
246810
30
40
50
60
70
80
90
100
110
943 meV
544 meV
323 meV
176 meV
D
C
B
A
1/kBT(eV-
1
)
Ln
246810
-38
-36
-34
-32
-30
-28
S11
D
CB
~0.99
~1.05
~0.98
~0.89
A
LnS
max
Ln
(
b
)
FIG. 7. 1 /kBTmax vs ln共
兲共the Arrhenius plot兲and ln共Smax兲on ln共
兲for 共a兲
6 ML QW and 共b兲11 ML In0.35Ga0.65As QD samples. The activation ener-
gies 共left graph兲and slopes 共right graph兲are depicted for defects A, B, C,
and D.
103709-5 Kunets et al. J. Appl. Phys. 104, 103709 共2008兲
Downloaded 26 Nov 2009 to 217.20.172.209. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
mined for this case only very roughly by using25
Nt=4
fmaxSmaxn2V.共8兲
Assuming the occupancy of the local level to be about 0.5,
we have a trap density of defect B approximately 6
⫻1013 cm−3.
Trap C was detected in all five samples in the tempera-
ture range of 170–210 K. Analysis of the relative noise den-
sity amplitude versus frequency shows that the slope is again
close to 1. Thus, for extracting information about this defect
we used Eqs. 共4兲–共8兲. We have that for defect C with an
average activation energy of E1=0.35 eV at fmax =80 Hz,
n=6⫻1016 cm−3,
T=4.4⫻107cm/s, and Smax=5
⫻10−16 Hz−1 the value of
0is in the range of
10−13–10−12 cm2. The trap density was estimated like in the
previous case and was found to be ⬃3⫻1012 cm−3.
Peak D measured at low temperatures 共100–140 K兲is
the most interesting one observed in this work. Because traps
A, B, and C are present in all samples including the GaAs
reference, we can conclude that they are GaAs related
defects.14,29 Trap D, however, appears to be unique to only
the samples containing QDs as it is not found in the refer-
ence GaAs sample or the 6 ML In0.35Ga0.65As QW sample.
Thus, we assume that at low temperature peak D is caused by
defects related to growth of In0.35Ga0.65As QDs in GaAs and
not simply of strained 6 ML In0.35Ga0.65As 2D layer.
Analysis of ln共Smax兲versus ln共
兲shows that slope of this
dependence is again close to 1 共see Table I兲for samples S9,
S11, and S13 with QDs. The activation energy can be found
from Eq. 共4兲. From an Arrhenius plots similar to one shown
in Figs. 7共a兲and 7共b兲, we found that the D trap activation
energies are 0.18, 0.18, and 0.12 eV for the 9, 11, and 13 ML
samples, respectively. Using n=6⫻1016 cm−3 measured by
Hall effect, and
T=4.4⫻107cm/s, we estimated
0to be
⬃1⫻10−15 cm2for samples S9 共fmax=15 Hz兲and S11
共fmax=20 Hz兲. For sample S13 with 13 ML of In0.35Ga0.65As
with an electron density of n=5⫻1016 cm−3 and
T=4.4
⫻107cm/s, we have
0⬇5⫻10−18 cm2. The defect densi-
ties can be roughly estimated by Eq. 共8兲, where trap densities
of 7⫻1011 cm−3,1⫻1011 cm−3, and 8⫻1010 cm−3 were
found for samples with 9, 11, and 13 ML of In0.35Ga0.65As,
respectively. The large difference in the activation energies
E1and capture cross section
0allows us to assume that
peak D probably has a complex behavior as it was shown in
Ref. 14. It was revealed by DLTS 共Ref. 14兲that in the same
temperature range there are two overlapping peaks 共i兲M1
with activation energy of 0.18 eV which is a typical trap for
MBE-GaAs materials and 共ii兲peak D with activation energy
of 0.1 eV which is related to In0.5Ga0.5As QDs grown on
GaAs. The capture cross sections were found to be 1.6
⫻10−15 cm2for M1 trap and 5.4⫻10−18 cm2for D trap. It
is obvious that values that we measured by noise spectros-
copy are in good agreement with values reported by DLTS.
The question is how activation energies of 0.18 and 0.12 eV
measured by noise can be correlated with DLTS values re-
ported in Ref. 14? Indeed, these two defects are observed
only for samples where QDs are present. Then, it is logical to
assume that these defects are located nearby or inside of
QDs. Their most probable location is the 200 Å thick un-
doped GaAs spacer layer. While the spacer layer is fully
TABLE I. Activation energy E1, energy of the local level E0, trap density Nt, and
0obtained after analysis of
the temperature dependence of noise spectra.
S0
共Ref.兲
S6
共6ML兲
S9
共9ML兲
S11
共11 ML兲
S13
共13 ML兲
Defect A
Slope 0.86 0.87 0.90 0.89 0.90
E1共eV兲0.75 0.79 0.75 0.94 0.76
E0共eV兲0.108 0.103 0.075 0.104 0.077
0共cm2兲4⫻10−11 4⫻10−10 4⫻10−11 3⫻10−8 7⫻10−11
Nt共cm−3兲3.1⫻1014 5.9 ⫻1014 1.1⫻1014 2.3⫻1014 9.2 ⫻1013
Defect B
Slope 1.04 0.90 1.07 0.98 0.94
E1共eV兲0.53 0.55 0.57 0.54 0.52
0共cm2兲1⫻10−12 3⫻10−12 8⫻10−12 4⫻10−12 1⫻10−12
Nt共cm−3兲7.5⫻1013 6.3 ⫻1013 5.9⫻1013 4.9⫻1013 5.4 ⫻1013
Defect C
slope 1.07 0.99 1.06 1.05 1.02
E1共eV兲0.33 0.38 0.32 0.32 0.35
0共cm2兲5⫻10−14 2⫻10−12 5⫻10−14 1⫻10−13 1⫻10−13
Nt共cm−3兲1.6⫻1012 3.2 ⫻1012 3.1⫻1012 2.4⫻1012
Defect D
Slope 1.09 0.99 0.96
E1共eV兲0.18 0.18 0.12
0共cm2兲8⫻10−16 1⫻10−15 5⫻10−18
Nt共cm−3兲6.6⫻1011 1.4 ⫻1011 8.1⫻1010
103709-6 Kunets et al. J. Appl. Phys. 104, 103709 共2008兲
Downloaded 26 Nov 2009 to 217.20.172.209. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
depleted, we have a situation similar to that reported in Ref.
25 where the energy, E1, determined by Eq. 共4兲gives the
level position E0responsible for the observed noise peak.
Then the time constant will be given by the Schockley–Hall–
Read theory and Eq. 共5兲holds true, where
0will be given
by25
0=共
0
TNc兲−1,共9兲
and Ncis the density of states in the conduction band. Then
we have
0equal to 1⫻10−15 cm2for samples S9 and S11
and 5⫻10−18 cm2for sample S13. The densities of traps are
the same and are given by Eq. 共8兲. However, according to
our analysis, we cannot assume that a trap with an energy of
0.12 eV is located in the InGaAs band gap. If this would be
the case, then the Fermi level is located significantly higher
than the trap level E0. Indeed, according to our simple esti-
mations based on the PL data and assuming the ratio of the
InGaAs band offset relative to GaAs to be Ec:Ev
=0.64:0.36, we would have the ground state located at about
165 meV below the GaAs band for 13 ML In0.35Ga0.65As in
approximation of quantum well, where the thickness of the
QW was equal to height of the QDs. For temperature range
of 100–140 K and free carrier density of 6⫻1016 cm−3,we
have the Fermi level located at 3 meV for 100 K and at 10
meV for 140 K below the GaAs conduction band. Compared
to the trap energy E0, the Fermi level is 282 meV above E0
for 100 K and 275 meV above E0for 140 K. The slope of
ln共Smax兲versus ln共
兲dependence would be less than 1.
While experimental data show the slope to be close to 1, we
can only assume that the trap with activation energy of 0.12
eV is physically located in the nearby GaAs spacer and
caused by the presence of the QDs. Besides, if this defect
does not propagate deep into n-type GaAs cap layer than
density of the traps with activation energy of 0.12 eV could
be overestimated 共one has to take correction for the volume
V. The small value of
0for defects caused by growth of
In0.5Ga0.5As 共Ref. 14兲or InAs QDs was reported30 before
and agrees well with our results. Another trap which is re-
vealed in samples with 9 and 11 ML is probably the well
known M1 trap in MBE grown GaAs. The M1 trap with
activation energy of 0.18 eV is related to point defects in
GaAs.14,29 It is quite reasonable to assume that due to the
high lattice mismatch between InGaAs QDs and the GaAs
cap layer, the density of point defects in this region raises
becoming resolved by our technique. In addition, we observe
also a narrowing of peaks B and C. As result, these became
much better resolved by our noise technique. There are two
possible mechanisms of this enhancement 共i兲presence of the
strain field due to the lattice mismatch and/or 共ii兲changes in
electron population in adjusting GaAs barrier layers due to
the presence of the quantum reservoirs such as a quantum
well or QDs. The depopulation of free carriers in GaAs bar-
riers could lead to lowering of the Fermi level and, thus, get
closer to resonance with the deep trap E0. Because we do not
see significant enhancement of the densities for defects B
and C with increasing number of In0.35Ga0.65As monolayers,
the second assumption is the most probable.
IV. CONCLUSIONS
Using noise spectroscopy and MBE grown
In0.35Ga0.65As layers of different thicknesses on GaAs, we
have studied deeps levels at the transition from 2D to three-
dimensional 共3D兲growth. The results were compared to a
reference n-type GaAs sample. We found three main defects
with activation energies of 0.8, 0.54, and 0.35 eV in refer-
ence GaAs. The same defects were detected for 6 ML
In0.35Ga0.65As highly strained quantum well as well as struc-
tures with QDs. At the transition to 3D, QD growth, one
additional defect appears with activation energy of 0.18 eV
which is a well known M1 trap level typical for n-type GaAs
grown by MBE and is most likely related to point defects.
Further increase in the InGaAs coverage above 11 ML leads
to an additional defect that is most probably located in GaAs
spacer layer. This trap is located at 0.12 eV below the GaAs
conduction band and has a small capture cross section. The
results are in agreement with the data received using DLTS
for self-organized In0.5Ga0.5As QDs grown on GaAs.14 This
work demonstrates that noise spectroscopy is a sensitive tool
with the capability to probe nanoscaled materials.
ACKNOWLEDGMENTS
This work was supported by the National Science Foun-
dation under Grant No. DMR-0520550.
1S. Fafard, K. Hinzer, S. Raymond, M. Dion, J. McCaffrey, Y. Feng, and S.
Charbonneau, Science 274,1350共1996兲.
2C. Y. Jin, H. Y. Liu, K. M. Groom, Q. Jiang, M. Hopkinson, T. J. Badcock,
R. J. Royce, and D. J. Mowbray, Phys. Rev. B 76, 085315 共2007兲.
3H. C. Liu, M. Gao, J. McCaffrey, Z. R. Wasilewski, and S. Fafard, Appl.
Phys. Lett. 78,79共2001兲.
4S. Maimon, E. Finkman, G. Bahir, S. E. Schacham, J. M. Garcia, and P.
M. Petroff, Appl. Phys. Lett. 73, 2003 共1998兲.
5Z. Ye, J. C. Campbell, Z. Chen, E. T. Kim, and A. Madhukar, J. Appl.
Phys. 92, 4141 共2002兲.
6H. Sakaki, G. Yusa, T. Someya, Y. Ohno, T. Noda, H. Akiyama, Y. Ka-
doya, and H. Noge, Appl. Phys. Lett. 67,3444共1995兲.
7I. R. Pagnossin, E. C. F. da Silva, A. A. Quivy, S. Martini, and C. S.
Sergio, J. Appl. Phys. 97, 113709 共2005兲.
8A. Tsormpatzoglou, N. A. Hastas, D. H. Tassis, C. A. Dimitriadis, G.
Kamarinos, P. Frigeri, S. Franchi, E. Gombia, and R. Mosca, Appl. Phys.
Lett. 87, 163109 共2005兲.
9N. Arpatzanis, A. Tsormpatzoglou, C. A. Dimitriadis, J. D. Song, W. J.
Choi, J. I. Lee, and C. Charitidis, J. Appl. Phys. 102, 054302 共2007兲.
10P. Barthold, F. Hohls, N. Maire, K. Pierz, and R. J. Haug, Phys. Rev. Lett.
96, 246804 共2006兲.
11P. N. Brounkov, A. Polimeni, S. T. Stoddart, M. Henini, L. Eaves, P. C.
Main, A. R. Kovsh, Yu. G. Musikhin, and S. G. Konnikov, Appl. Phys.
Lett. 73, 1092 共1998兲.
12G. Medeiros-Ribeiro, D. Leonard, and P. M. Petroff, Appl. Phys. Lett. 66,
1767 共1995兲.
13S. Anand, N. Carlsson, M. E. Pistol, L. Samuelson, and W. Seifert, J.
Appl. Phys. 84, 3747 共1998兲.
14Z. Q. Fang, Q. H. Xie, D. C. Look, J. Ehret, and J. E. van Nostrand, J.
Electron. Mater. 28,L13共1999兲.
15P. Krispin, J. L. Lazzari, and H. Kostial, J. Appl. Phys. 84,6135共1998兲.
16S. O. Cho, Zh. M. Wang, and G. J. Salamo, Appl. Phys. Lett. 86, 113106
共2005兲.
17G. E. Stillman and C. M. Wolfe, Thin Solid Films 31,69共1976兲.
18D. C. Look, Electrical Characterization of GaAs Materials and Devices
共Wiley, New York, 1989兲, p. 280.
19M. E. Levinshtein and S. L. Rumyantsev, Semicond. Sci. Technol. 9,1183
共1994兲.
20D. D. Carey, S. T. Stoddart, S. J. Bending, J. J. Harris, and C. T. Foxon,
Phys. Rev. B 54, 2813 共1996兲.
21M. J. Deen, M. E. Levinshtein, S. L. Rumyantsev, and J. Orchard-Webb,
103709-7 Kunets et al. J. Appl. Phys. 104, 103709 共2008兲
Downloaded 26 Nov 2009 to 217.20.172.209. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp
Semicond. Sci. Technol. 14, 298 共1999兲.
22J. A. Copeland, IEEE Trans. Electron Devices 18,50共1971兲.
23M. E. Levinshtein, S. L. Rumyantsev, J. W. Palmour, and D. B. Slater, Jr.,
J. Appl. Phys. 81, 1758 共1997兲.
24J. W. Palmour, M. E. Levinshtein, and S. L. Rumyantsev, Semicond. Sci.
Technol. 11,1146共1996兲.
25S. L. Rumyantsev, N. Pala, M. S. Shur, E. Borovitskaya, A. P. Dmitriev,
M. E. Levinshtein, R. Gaska, M. A. Khan, J. Yang, X. Hu, and G. Simin,
IEEE Trans. Electron Devices 48, 530 共2001兲.
26N. V. Dyakonova, M. E. Levinshtein, and S. L. Rumyantsev, Sov. Phys.
Semicond. 25, 1241 共1991兲.
27M. E. Levinshtein, J. W. Palmour, and S. L. Rumyantsev, Semicond. Sci.
Technol. 9, 2080 共1994兲.
28N. V. Dyakonova, M. E. Levinshtein, and S. L. Rumyantsev, Semicond.
Sci. Technol. 11,177共1996兲.
29N. Chand, A. M. Sergent, J. P. van der Ziel, and D. V. Lang, J. Vac. Sci.
Technol. B 7,399共1989兲.
30M. M. Sobolev, A. R. Kovsh, V. M. Ustinov, A. Yu. Egorov, A. E.
Zhukov, M. V. Maksimov, and N. N. Ledenstsov, Semiconductors 31,
1074 共1997兲.
103709-8 Kunets et al. J. Appl. Phys. 104, 103709 共2008兲
Downloaded 26 Nov 2009 to 217.20.172.209. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp