Content uploaded by Stefano Sanguinetti
Author content
All content in this area was uploaded by Stefano Sanguinetti
Content may be subject to copyright.
Role of the wetting layer in the carrier relaxation in quantum dots
S. Sanguinetti,a) K. Watanabe, T. Tateno, M. Wakaki,b) and N. Koguchi
National Institute for Materials Science, Tsukuba, Ibaraki 305-0047, Japan
T. Kuroda and F. Minami
Department of Physics, Tokyo Institute of Technology, Meguro-ku, Tokyo 152-8551, Japan
M. Gurioli
I.N.F.M. - Department of Materials Science; Via Cozzi 53, I-20125 Milan, Italy and L.E.N.S.,
Via Sansone 1, I–50019, Sesto Fiorentino, Italy
共Received 8 March 2002; accepted for publication 29 May 2002兲
We present picosecond time resolved photoluminescence measurements of GaAs/AlGaAs quantum
dot structures—grown by modified droplet epitaxy—where no wetting layer is connecting the dots.
We find a fast carrier relaxation time 共30 ps兲to the dot ground state, which becomes even faster for
increasing the photogenerated carrier injection. This shows that the two–dimensional character of
the wetting layer is not relevant in determining the quantum dot capture, in contrast with the
conclusions of several models so far presented in literature. We discuss the role of the barrier states
as well as the possibility of Auger processes involving the zero-dimensional levels of the quantum
dots. © 2002 American Institute of Physics. 关DOI: 10.1063/1.1495525兴
The nature of the mechanisms underlying carrier relax-
ation phenomena in quantum dot 共QD兲materials has been
largely debated in the last years.1–5 These phenomena con-
tinue to attract much attention since they involve fundamen-
tal physical aspects of the zero-dimensional semiconductor
systems, and also due to their relevance for device applica-
tions. In QD lasers, the carriers are injected into the barriers
embedding the QDs. After energy dissipation processes, the
carriers are captured by the QDs and then relax to the fun-
damental lasing state. The efficiency of the relaxation cas-
cade directly affects the device performances, such as thresh-
old current, temperature stability, and so on. The
achievement of a very fast capture rate is therefore a relevant
aspect of the device optimization study.
The self-aggregated dots, grown by Stranski–Krastanov
epitaxy on a two-dimensional 共2D兲wetting layer 共WL兲, sat-
isfy this requirement. In general, the characteristic carrier
relaxation time from the barrier states to the fundamental QD
level is on the order of a few tens of picoseconds and be-
comes even shorter for large injection of carriers. The nature
of the fast carrier relaxation in QDs is not yet completely
understood. The typical energy separation of the QD elec-
tronic levels, which matches neither the longitudinal acoustic
phonons nor the longitudinal optical phonons, suggested the
presence of the well known phonon bottleneck effect.1Sev-
eral mechanisms have been invoked to explain the lack of
phonon bottleneck. The following picture has recently
emerged. In the low injection regime, the fast relaxation time
共30–70 ps兲has been interpreted as a consequence of the
presence of a continuum tail of WL defect states to which the
carriers easily relax the excess energy.4In addition, resonant
multiphonon processes can mediate the relaxation between
the localized states.6By increasing the carrier injection, the
carrier–carrier interaction speeds up the energy relaxation
rate, via Auger-type processes.7,8 In particular, the Auger
processes involving carriers localized in the 2D-WL states
have been claimed to increase the carrier relaxation rate in-
side the QD.9,10 The WL is therefore assumed to have a key
role in the QD carrier relaxation.8
In this letter, we intend to test this picture by means of
picosecond time-resolved photoluminescence 共PL兲in a
GaAs/AlGaAs QD grown by modified droplet epitaxy
共MDE兲.11,12 MDE is a nonconventional growth method for
self–assembling semiconductor QDs even in lattice matched
systems, such as GaAs/AlGaAs.12 Among other interesting
features, this alternative method leads to the fabrication of
defect free dots without any WL in the structure.13 Therefore,
we eliminate a priori the role of the WL and investigate the
resulting carrier relaxation processes. We found very similar
results with respect to standard QD with a WL, that is a fast
relaxation time 共30 ps兲which becomes even faster for in-
creasing the photogenerated carrier injection. These results
show that the 2D character of the WL is not relevant in
determining the QD capture. We discuss the role of the bar-
rier states as well as the possibility of Auger processes in-
volving the zero–dimensional levels of the QDs.
The samples were grown by MDE12 with the following
procedure.14,15 After the growth of 300 nm GaAs buffer layer
and 500 nm Al0.3Ga0.7As barrier layer at 580 °C, the sub-
strate temperature was lowered to 180 °C, the As valve was
closed and the As pressure in the growth chamber was de-
pleted. 1.75 monolayers of Al0.3Ga0.7 alloy was supplied in
order to form a group-III stabilized surface on the c4⫻4
reconstruction. The subsequent deposition of 3 monolayers
of Ga at this temperature gives rise to the formation of tiny
Ga droplets on the substrate. Following the deposition of the
droplets, an As4molecular beam was irradiated on the sur-
a兲Permanent address: Department of Materials Science, Via Cozzi 53,
I-20125 Milan, Italy; electronic mail: stefano.sanguinetti@mater.unimib.it
b兲Also at: Department of Electro–Photo Optics, Tokai University, Hiratsuka,
Kanagawa 259-1292, Japan.
APPLIED PHYSICS LETTERS VOLUME 81, NUMBER 4 22 JULY 2002
6130003-6951/2002/81(4)/613/3/$19.00 © 2002 American Institute of Physics
Downloaded 16 Apr 2004 to 144.213.253.14. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
face. After the complete change of reflection high-energy
electron diffraction pattern from halo to spots, an
Al0.3Ga0.7As barrier layer of 10 nm was grown by migration
enhanced epitaxy at the same temperature of 180 °C. This
temperature was chosen in order to prevent 2D regrowth of
the naked GaAs microcrystals.12 Then, the growth tempera-
ture was raised back to 580 °C and 90 nm of Al0.3Ga0.7As
and 10 nm of GaAs as cap layer were grown by ordinary
molecular-beam epitaxy. The sample was reloaded in the
growth chamber and annealed in 1.5⫻10⫺5Torr As4atmo-
sphere at 640 °C. This annealing procedure was shown to
improve the sample quality healing the defects formed in the
low-temperature growth of the Al0.3Ga0.7As barrier layer.13,16
Before the deposition of the capping layer, surface and cross
section high-resolution scanning electron microscope images
of the complete annealed structure demonstrate the formation
of pyramidal shape nanocrystals with typical base sizes of
16 nm⫻20 nm and the absence of the WL.15
Continuous wave 共cw兲PL spectra were measured at 17
K and excited with an Ar⫹laser in multiline mode. PL spec-
tra were measured by a grating monochromator operating
with a GaAs photomultiplier. Time-resolved PL measure-
ments, have been performed by using the second-harmonic
(exc⫽400 nm) of a fs mode-locked Ti-Sapphire laser with
76 MHz repetition rate. The PL signal was dispersed by a
polychromator, and time resolved by a streak camera with a
resolution of 1 meV and 9 ps, respectively. All time-resolved
measurements were performed at 5 K.
Figure 1 共a兲shows the cw PL spectra for different exci-
tation power densities. At low-power density, the PL spec-
trum consists in a broad QD band 共160 meV兲, reflecting the
large inhomogeneous distribution of the QD size, and in a
narrower band at ⬇2 eV associated with recombination in
the Al0.3Ga0.7As barriers. This latter consists in a doublet
with intrinsic exciton recombination at 1.97 eV and extrinsic
recombination at 1.94 eV. For Pexc larger that 60 W/cm2, the
QD band shows in the high-energy tail 共at ⬇1.82 eV兲the
appearance of a broad emission associated with the carrier
population of the QD excited levels.
The analysis of the time evolution of the different spec-
tral components in the PL spectra 关Fig. 1 共b兲兴 shows an al-
most constant decay 共400 ps兲and rise time 共35 ps兲of the QD
of different size within the inhomogeneously broadened PL
band. On the wing of the high-energy tail of the QD-PL band
(⬎1.84 eV), the decay time becomes shorter 共200 ps兲and it
is associated with the recombination from the QD excited
states. The comparison between the time evolution of the QD
ground state (Eem⫽1.65 eV) and the time evolution of the
excited-state emission (Eem⫽1.85 eV) is reported in Fig.
1共b兲. Finally, the barrier PL time evolution 共not shown兲is
characterized by a fast rise time (⬍10 ps) and a relatively
long decay time 共250 ps兲. Therefore, the barrier PL kinetics
is not determined by the QD capture mechanisms. This can
be explained by assuming that the carriers which recombine
in the barrier are localized in a spatial region far from the QD
layers.
Figure 2 共a兲reports the PL time evolution of the funda-
mental optical transition in the QDs for three different exci-
tation power densities. The time evolution of the PL band has
been fitted with a phenomenological model based on the dif-
ference between two exponential decays after a convolution
with the experimental response function. The best fits are
reported in Fig. 2 共a兲as solid lines. The decay time
Dis
almost constant for a large range of Pexc 共nearly three orders
of magnitude兲. This shows that
Dstems from the enhanced
optical matrix element of the ground-state transition in
GaAs/Al0.3Ga0.7As QD and not from nonradiative competi-
tive channels. At Pexc higher than 300 W/cm2, a lengthening
is observed due to the QD filling. Aclear reduction of the PL
rise time
Ris also observed when increasing the excitation
power, showing that the carrier capture into the QDs strongly
depends of on the carrier injection. Asummary of the PL rise
time is reported in Fig. 2 共b兲. A sharp decrease of
Ris ob-
served for Pexc larger than ⬇10 W/cm2. At the highest ex-
citation power, the rise time of the PL from the low-lying
transition is almost the same of the rise time of the PL from
the excited state. The solid 共dashed兲line in Fig. 2共b兲refers
to a best fit with 1/Pexc (1/Pexc
2) dependence. Despite the
small amount of data, the comparison indicates that the
1/Pexc dependence gives a better agreement with the experi-
FIG. 1. 共a兲Four different PL spectra taken at increasing Pexc at steps of one
optical density in the range 10 W/cm2–10 kW/cm2.共b兲PL traces recorded
at 1.65 eV 共ground-state emission—circles兲and 1.85 eV 共excited-state
emission—squares兲with Pexc⫽1 kW/cm2.
FIG. 2. 共a兲PL traces of the fundamental QD transition at different Pexc :
3300 W/cm2共squares兲, 330 W/cm2共circles兲, and 3 W/cm2共triangles兲. The
fits are shown by continuous lines. 共b兲PL rise time as a function of Pexc
共circles兲. Continuous and dashed lines show the best 共in ps兲1/Pexc and
1/Pexc
2fits, respectively.
614 Appl. Phys. Lett., Vol. 81, No. 4, 22 July 2002 Sanguinetti
et al.
Downloaded 16 Apr 2004 to 144.213.253.14. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp
mental data. Finally, we note that in the range of Pexc corre-
sponding to the speeding up of
R共Ref. 17兲, the QD PL
spectrum shows the insurgence of the recombination from
excited levels.
Let us now compare our results with the findings re-
ported in literature on standard InAs/GaAs QDs which nucle-
ate on a 2D InAs WL. As already discussed, several recent
papers assume that the presence of the 2D layer connecting
the QDs has relevant consequences on the QD carrier relax-
ation. In particular, Morris et al.8reports a PL rise time
R
⫽35 ps, for Pexc⬍10 W/cm2, and a decrease of
Rtoward a
value of 10 ps when Pexc⬎100 W/cm2共with a 1/Pexc depen-
dence兲. The 1/Pexc rise time behavior has been attributed to
carrier relaxation processes involving two-particle Coulomb
scattering 共pseudo–Auger process兲between carriers in the
QDs and in the WL.9,10,8 The very same results are observed
here in QDs that are not connected by the 2D WL. We,
therefore, conclude that the picture in which the WL states
play a relevant role in speeding up the QD relaxation time is
not completely correct. Our data show that both the fast car-
rier relaxation time, and the dependence of
Ron Pexc , are
general properties of self-aggregated QDs.
In our sample, the continuum of states associated with
the barriers may play the role attributed to the WL in Refs. 4
and 8. That is, the presence of defect states in the
Al0.3Ga0.7As barrier can facilitate the carrier relaxation inside
the QDs.3Similarly, the pseudo–Auger processes could in-
volve the electronic states of the barriers. The most relevant
difference between the barrier and the WL states is, a priori,
the dimensionality. The barrier states have a three–
dimensional character while the electronic states of the WL
are 2D states. Our results show that the 2D character of the
continuum of states over the QD does not play any role in
the carrier capture and relaxation. This can be justified, in
principle, by considering that the electronic wave functions
associated with the WL states largely penetrate into the bar-
riers thus washing out a predominant 2D character of the WL
states.
On the other hand, the observed behavior may stem from
intrinsic relaxation mechanisms in a QD. Fast relaxation pro-
cesses, promoted by polaronic 共electron–phonon兲, anhar-
monic 共phonon–phonon兲,18 and resonant multiphonon
processes6have been proposed. As far as the dependence of
the relaxation rate on the carrier injection is concerned, we
find that the speeding up of the relaxation rate with Pexc
corresponds to a large QD filling, which leads to PL emission
from the QD excited states. We remark that the same effect
was also present in the previous studies,7,8,19 even if the au-
thors did not stress it. The onset of the Auger effect therefore
occurs when the QD is filled by a large number of carriers.
This strongly suggests that the nonlinear mechanisms that
are effective in the QD carrier relaxation processes are pro-
moted by the carrier population inside the QD rather than
involving the carrier population in the continuum of states
above the QD. Note also that the 1/Pexc dependence of the
carrier relaxation time is a common feature to any two-
particle scattering mechanisms. Fast intraband relaxation
processes due to two–particle Coulomb scattering 共intraband
Auger兲, active in a QD for dot carrier population larger than
one, have been recently suggested.5,20 We, therefore, believe
that the Auger-type relaxation in QDs has to be associated
with Coulomb scattering processes involving carriers con-
fined in the QD. This is also supported by the recent obser-
vation of fast relaxation processes in isolated CdSe dots.21
In conclusion, we have studied the carrier relaxation dy-
namics in GaAs QDs embedded in three–dimensional
Al0.3Ga0.7As barriers. The removal of the 2D WL does not
modify the carrier relaxation processes. We believe that the
possibility of growing, and therefore investigate, a defect
free QD without a WL gives a further relevant degree of
freedom for obtaining deeper understanding of the carrier
dynamics and optical properties of self–assembled dots.
One of the authors 共S.S.兲acknowledges partial financial
support from INFM-LENS.
1H. Benisty, C. M. Sotomayor-Torres, and C. Weisbuch, Phys. Rev. B 44,
10 945 共1991兲.
2A. L. Efros, V. A. Kharchenko, and M. Rosen, Solid State Commun. 93,
281 共1995兲.
3D. F. Schroeter, D. J. Griffiths, and P. C. Sercel, Phys. Rev. B 54, 1486
共1996兲.
4Y. Yoda, O. Moriwaki, M. Nishioka, and Y. Arakawa, Phys. Rev. Lett. 82,
4114 共1999兲.
5R. Ferreira and G. Bastard, Appl. Phys. Lett. 74, 2818 共1999兲.
6R. Heitz, M. Veit, N. N. Ledentsov, A. Hoffmann, D. Bimberg, V. M.
Ustinov, P. S. Kop’ev, and Z. I. Alferov, Phys. Rev. B 56, 10435 共1997兲.
7B. Ohnesorge, M. Albrecht, J. Oshinowo, andA. Forchel, Phys. Rev. B 54,
11532 共1996兲.
8D. Morris, N. Perret, and S. Fafard, Appl. Phys. Lett. 75, 3593 共1999兲.
9U. Bockelmann and T. Egeler, Phys. Rev. B 46, 15 574 共1992兲.
10 A. V. Uskov, F. Adler, H. Schweizer, and M. H. Pilkuhn, J.Appl. Phys. 81,
7895 共1997兲.
11 N. Koguchi, S. Takahashi, and T. Chikyow, J. Cryst. Growth 111,688
共1991兲.
12 K. Watanabe, N. Koguchi, and Y. Gotoh, Jpn. J. Appl. Phys., Part 2 39,
L79 共2000兲.
13 K. Watanabe, S. Tsukamoto, Y. Gotoh, and N. Koguchi, J. Cryst. Growth
227, 1073 共2001兲.
14 Y. Horikoshi, M. Kawashima, and H. Yamaguchi, Jpn. J.Appl. Phys., Part
127, 169 共1988兲.
15 T. Tateno, K. Watanabe, S. Sanguinetti, S. Tsukamoto, M. Wakaki, Y.
Gotoh, and N. Koguchi, in Growth of Well Defined Nanostructures, edited
by N. Koguchi 共National Research Institute for Metals, Tsukuba, Japan,
2001兲,p.298.
16 S. Sanguinetti, K. Watanabe, T. Kuroda, F. Minami, Y. Gotoh, and N.
Koguchi, J. Cryst. Growth 共to be published兲.
17 The ratio between the carrier injection in cw and time-resolved 共TR兲con-
ditions is given by Ncw /NTR⫽
rep
PLPcw /PTR , where Pcw (PTR)isthe
excitation power density in the cw the 共TR兲condition,
PL is the PL decay
time and
rep is the laser repetition rate.
18 O. Verzelen, R. Ferreira, and G. Bastard, Phys. Rev. B 62, 4809 共2000兲.
19 S. Raymond, K. Hinzer, S. Fafard, and J. L. Merz, Phys. Rev. B 61, 16331
共2000兲.
20 J. L. Pan, Phys. Rev. B 49, 11272 共1994兲.
21 V. I. Klimov, D. W. McBranch, C. A. Leatherdale, and M. G. Bawendi,
Phys. Rev. B 60, 13740 共1999兲.
615Appl. Phys. Lett., Vol. 81, No. 4, 22 July 2002 Sanguinetti
et al.
Downloaded 16 Apr 2004 to 144.213.253.14. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp