Content uploaded by G. E. Cirlin
Author content
All content in this area was uploaded by G. E. Cirlin
Content may be subject to copyright.
1063-7826/03/3702- $24.00 © 2003 MAIK “Nauka/Interperiodica”
0210
Semiconductors, Vol. 37, No. 2, 2003, pp. 210–214. Translated from Fizika i Tekhnika Poluprovodnikov, Vol. 37, No. 2, 2003, pp. 219–223.
Original Russian Text Copyright © 2003 by Makarov, Ledentsov, Tsatsul’nikov, Cirlin, Egorov, Ustinov, Zakharov, Werner.
1. INTRODUCTION
Ongoing interest in Si/Ge nanostructures is due to
considerable progress in the development of new
devices based on nanoheterostructures with size quan-
tization [1]. The successful design of transistors, photo-
detectors and emitters of light, operating on intrasub-
band transitions in quantum wells (QW), should be
noted. At the same time, numerous attempts to derive
effective emitters of light, relying on band-to-band
transitions in QWs in a given system, have failed. In
this case, effective Si/Ge emitters of light, lasers espe-
cially, might potentially ensure the most direct integra-
tion of silicon technology with optoelectronic data
transmission systems, both within a single silicon chip
and in telecommunication applications.
As shown earlier, the use of Si/SiGe QWs does not
noticeably reduce the time of radiative recombination
[2]. Further, owing to specific features of the band
structure and the character of strains in coherent
Si
−
Ge QWs, the Ge–Si heterojunction is of type II [3],
and the overlap of the electron and hole wave func-
tions is reduced not only in
k
-space, but also in real
space. The spatial separation of electrons and holes at
the heterojunction results in a characteristic short-
wavelength shift of the photoluminescence (PL) line
with an increase in excitation density, which is typical
of type-II QWs [2, 3].
In recent years, active studies aimed at enhancing
PL efficiency were devoted to the application of 3D
SiGe and GeSiC/Si QDs obtained in the Stranski–
Krastanow (SK) growth mode on the Si surface [4].
However, the large size of SK QDs, combined with a
high Ge content, results in even stronger spatial separa-
tion of the wave functions of a hole localized in a Ge
QD and an electron localized in an Si matrix. The struc-
tures also exhibit a strong short-wavelength shift of the
luminescence from an SK QD as the pumping intensity
rises, which is typical of type-II QDs [5]. The relatively
large QD size (
ⲏ
10 nm) demands that relatively thick
(
ⲏ
10 nm) Si spacers be used. The surface density of SK
QDs is about 10
9
–10
10
cm
–2
, and the maximum volume
density of SK QDs is also very low (10
15
–10
16
cm
–3
).
Such a low density creates problems for realizing lasing
even for direct-gap QDs in an InAs–GaAs system [6].
Further, the band structure in the
k
-space of Si varies
only slightly, because the characteristic size of hole
localization in real space strongly exceeds the Bohr
radius of a hole.
It should be noted that there exists another class of
QDs produced by ultrathin [7, 8], e.g., submonolayer,
insertions of a narrow-gap material in a wide-gap
matrix [9]. The characteristic lateral size of these QDs
is much smaller and their density is much higher than
those for SK QDs [9]. The possibility of densely stack-
ing these QDs enables ultrahigh modal gain (up to
10
4
−
10
5
cm
–1
) in wide-gap direct-gap materials whose
exciton has a small Bohr radius [9].
LOW-DIMENSIONAL
SYSTEMS
Optical Properties of Structures with Ultradense Arrays
of Ge QDs in an Si Matrix
A. G. Makarov*^, N. N. Ledentsov*, A. F. Tsatsul’nikov*, G. E. Cirlin*,
V. A. Egorov*, V. M. Ustinov*, N. D. Zakharov**, and P. Werner**
* Ioffe Physicotechnical Institute, Russian Academy of Sciences, St. Petersburg, 194021 Russia
^ e-mail: makarov@mail.ioffe.ru
** Max-Planck Institute of Microstructure Physics, Halle/Saale, Germany
Submitted June 10, 2002; accepted for publication June 14, 2002
Abstract
—The structural and optical properties of ultrathin Ge insertions in an Si matrix were studied. Transmis-
sion electron microscopy revealed the spontaneous formation of arrays of disk-shaped quantum dots (QDs) with
a small lateral size (3–10 nm) at a nominal Ge insertion thicknesses, from submonolayer to nearly critical, for the
transition to 3D growth by the Stranski–Krastanow mechanism. Optical study revealed type-I band alignment in
these structures, which results from the strong contribution of the electron–hole Coulomb interaction overpower-
ing the repulsion potential for an electron existing in the Ge conduction band. The small lateral size of QDs lifts
the selection rule prohibiting indirect recombination in the inverse
k
space. At the same time, the high surface den-
sity of QDs (10
12
–10
13
cm
–2
) and the possibility of their stacking with the use of ultrathin Si spacers allows the
obtainment of an ultrahigh volume density of QDs (up to 10
19
cm
–3
), which is necessary to achieve stimulated
emission in Si. A sample with stacked QDs formed by 0.7-nm-thick Ge insertions exhibited a superlinear increase
of the photoluminescence (PL) intensity, accompanied by narrowing of the PL line. The doping of Ge–Si struc-
tures with donors allows for a drastic increase in the PL intensity at high temperatures, which prevents depletion
of the active region in weakly localized electrons.
© 2003 MAIK “Nauka/Interperiodica”.
SEMICONDUCTORS
Vol. 37
No. 2
2003
OPTICAL PROPERTIES OF STRUCTURES WITH ULTRADENSE ARRAYS OF Ge QDs 211
For the Ge–Si system, fabrication of QDs of this
kind, if possible, will resolve all the basic problems of
optoelectronic applications. First, a small (3–5 nm) lat-
eral QD size effectively lifts the momentum selection
rule for radiative recombination with electrons from the
indirect minimum of the conduction band. At the same
time, the repulsive potential in the conduction band
appears to be weak, which allows the localization of an
electron and hole in the same spatial region [10].
As shown earlier in the case of ultrathin type-II lay-
ers, the effective localization of an electron on a hole
can be achieved even on the basis of type-II hetero-
structures with a high potential barrier in the conduc-
tion band [10]. This is so because the Coulomb attrac-
tion of an electron exceeds its repulsive action at a cer-
tain small thickness of the barrier. Indeed, a short-
wavelength shift of the PL line with a rise in pumping
power is absent in the case of ultranarrow Ge insertions
in an Si matrix [2]. In addition, the use of ultrasmall
QDs simplifies the electron localization in contrast to
the QW case, because the barrier strength decreases in
the lateral direction.
In this study, we propose the optoelectronic applica-
tion of ultrasmall QDs grown by depositing Ge layers
with a thickness below the critical one, which is neces-
sary for the transition to 3D growth by the SK mecha-
nism. We demonstrate that, under certain deposition
conditions, ultradense arrays of QDs are obtained, in
which, with account taken of the Coulomb interaction,
both a direct gap structure in real space and a maximum
delocalization of the hole wave function in
k
-space are
realized, which favors radiative recombination. Finally,
an ultrahigh density of QDs can be obtained in these
structures for achieving sufficiently high gain for las-
ing. Dense QD arrays can be closely stacked along the
growth direction, this being another key advantage for
achieving the necessary high gain.
It is also shown that the doping of structures with Sb
suppresses depletion of electrons in the active medium
and significantly enhances the efficiency of radiative
recombination. The possibility of obtaining stimulated
emission from Si–Ge structures is discussed.
2. EXPERIMENTAL
The samples under study are periodical Ge inser-
tions in an Si matrix deposited on a 100-nm-thick
buffer layer grown by MBE at a substrate temperature
of 600
°
C. Two types of superlattices were grown. The
first comprised 20 layers of submonolayer Ge insertions
of varied thickness separated with 4- to 5-nm-thick Si
spacers; the effective thickness of Ge layers in the
structures varied from 0.07 to 0.14 nm. The other type
of superlattice, comprising 10 periods, included 0.5- to
0.7-nm Ge layers separated with 11-nm-thick Si spac-
ers. These spacers consisted of 9 nm of undoped Si and
2 nm of Si doped with 5
×
10
16
cm
–3
Sb at their centers.
The undoped and doped superlattices were grown at
temperatures of 750 an 700
°
C, respectively. To prevent the
segregation of Sb, the spacers were grown at 600
°
C. The
growth rate for Si and Ge was 0.05 and 0.005 nm s
–1
,
respectively. The total vapor pressure in the MBE
chamber was not less than 5
×
10
–9
Torr. Growth was
monitored by recording RHEED patterns. The initial
(2
×
2) reconstruction of the Si surface was preserved
during the growth, with only a slight broadening of the
principal reflections being observed, irrespective of the
growth temperature. Thus, even in upper layers, no
appreciable density of 3D islands was formed by the SK
mechanism. The TEM study was performed using a
JEM 4010 electron microscope with a 400-kV accelerat-
ing voltage. The PL was excited with an Ar-ion laser (
λ
=
514.5 nm) and detected with a cooled Ge photodiode.
3. RESULTS
Figure 1a shows a cross-sectional TEM image of a
structure containing submonolayer Ge insertions with
Ge
Ge
Ge
[001]
Ge + Si
Si
Si
Si
Ge + Si
3 nm
(a) (b)
Fig. 1.
Cross-sectional TEM images of structures containing (a) submonolayer (0.07 nm) and (b) monolayer (0.136 nm) Ge inser-
tions in a Si matrix.
212
SEMICONDUCTORS
Vol. 37
No. 2
2003
MAKAROV
et al
.
an effective thickness of 0.07 nm, grown at a substrate
temperature of 650
°
C. The thickness of Si spacers
between the Ge insertions was 4.4 nm. In order to ana-
lyze the distribution of Ge atoms in each layer, a special
digital analyzer of HRTEM images is necessary. Anal-
ysis shows that the Ge insertions do not constitute a
solid layer; instead, a high density of nanodomain for-
mations 3–5 nm in size with a surface density of
≈
5
×
10
11
cm
–2
is observed [11, 12]. Furthermore, local 3D
islands with a specific size of about 10 nm are formed.
In the case of Ge insertions with a thickness of about
one monolayer (ML) or more, the typical lateral size of
a nanodomain was 7–10 nm.
Figure 2 schematically shows the band diagram of
the structures under study. The Ge insertions form
potential wells in the valence band and potential barri-
ers in the conduction band of the Si–Ge system. In mul-
tilayer structures, minibands are formed in the Si con-
duction band, with the wave function of the electron
having a minimum near the Ge insertions. When non-
equilibrium holes captured by the Ge potential wells
appear in the Si matrix, an additional Coulomb poten-
tial is formed, which attracts an electron to a hole. Since
the Coulomb energy in Si is rather high (14.7 meV) and
the barrier in the conduction band is comparatively
low (<100 meV [3]), an electron can be effectively
localized in the Coulomb potential of a hole in the
Ge region, as shown in the general case for ultranar-
row type-II QWs.
Figure 3 shows typical PL spectra of a sample with
a submonolayer (0.1 nm) Ge insertion in the Si matrix.
PL spectral lines related to acoustical and optical
phonons of the Si matrix are observed, as are lines of
PL from Ge insertions (Ge
NP
, Ge
TO
, and Ge
TO–O
) peaked
at 1.121, 1.064, and 1.004 eV, respectively.
An interesting distinction of submonolayer Ge
insertions is the long-wavelength shift of the PL lines
related to the Ge QD as the excitation density increases.
In this situation, at low excitation densities, the zero-
phonon PL line lies at an energy close to that expected
from the dependence of the PL energy on the thickness
of Ge insertions, which was obtained in [13]. An energy
shift with pumping was not observed in structures with
a Ge insertion thickness slightly exceeding 1 ML,
which was also stated earlier [2]. The lack of a shift
indicates the absence of spatial separation between
electrons and holes and confirms the validity of the
type-I QD model. The long-wavelength shift with an
increase in the excitation density, which is observed for
submonolayer insertions, is evidently due to the forma-
tion of multiple-exciton complexes associated with a
QD. This once again emphasizes the increasing role of
the Coulomb attraction between electrons and holes in
the case when the repulsive action of the Ge potential
barrier in the Si conduction band becomes weaker.
A characteristic feature of the Ge–Si PL spectra is
fast thermal quenching. In our opinion, this feature is
associated with thermal emission of weakly localized
electrons and their subsequent nonradiative recombina-
tion on the surface and in the bulk of the Si substrate.
Electron miniband
Si
CB
Si
VB
GeSi
Electron localized
by hole Coulomb
potential
Localized
hole
Fig. 2.
Schematic band diagram of a multilayer structure
with Ge insertions in a Si matrix.
100
10
1
0.1
0.01 1.00 1.05 1.10 1.15 1.20 1.25
PL intensity, arb. units
Ge
TO–O
Si
TO–O
Ge
NP
Si
TA
Ge
TO
Si
TO
1
2
3
4
5
Photon energy, eV
Fig. 3.
The dependence of PL spectra on the pumping density for a structure with submonolayer QDs (0.74 ML of Ge) at 15 K.
Noteworthy is the long-wavelength shift of the PL lines with increasing excitation density: (
1
) 1000, (
2
) 150, (
3
) 50, (
4
) 25, and (
5
)
15 W cm
–2
.
SEMICONDUCTORS
Vol. 37
No. 2
2003
OPTICAL PROPERTIES OF STRUCTURES WITH ULTRADENSE ARRAYS OF Ge QDs 213
Even relatively weak doping of the active region of a
structure with a donor impurity (with an average con-
centration of
≈
10
16
cm
–3
), which gives rise to a moder-
ate density of equilibrium electrons, dramatically
enhances the PL intensity and enables it to be observed
up to room temperature. Figure 4 shows the tempera-
ture dependence of PL spectra. The long-wavelength
shift of the Ge PL line is evidently weaker than that of
the Si line. This fact, observed in all the samples (with
submonolayer or monolayer insertions, doped or
undoped), and the lack of a short-wavelength shift with
increasing excitation density presumably indicate that
the electron miniband in Si is thermally filled with elec-
trons as the temperature rises. To counteract this effect,
the donor doping level should be raised substantially, to
the point of degeneracy.
The high intensity and temperature stability of the
PL in doped samples with Ge–Si QDs allowed us to
observe a narrowing of the PL line as the observation
temperature decreased at high excitation densities, or as
the excitation density increased at a fixed temperature
(Fig. 5). The narrowing of the PL line is accompanied
by a dramatic rise in the integral PL intensity. This
effect is observed in the vertical direction and only in
the samples with a polished back surface. This may
indicate that stimulated emission is obtained in a verti-
cal Si cavity with an active region of dense stacked
arrays of Ge QDs.
4. CONCLUSIONS
The structural and luminescent properties of struc-
tures with dense arrays of high-density Ge dots have
been studied. As shown, these structures are arrays of
type-I QDs. The doping of QDs allowed us to obtain
high PL intensity at elevated temperatures and superlin-
ear growth of the PL intensity with a rise in the excita-
tion density, which may indicate the occurrence of
stimulated emission in Si–Ge heterostructures. Presum-
ably, the use of ultradense arrays of small-size Ge–Si
QDs heavily doped with donor impurities will enable
lasing in Si–Ge structures at room temperature in the
near future.
REFERENCES
1. D. Bimberg, M. Grundmann, and N. N. Ledentsov,
Quantum Dot Heterostructures
(Wiley, New York,
1998).
Photon energy, eV
1.02 1.06 1.10 1.14
PL intensity, arb. units
Ge
TO
Si
TO
Ge
NP
Si
TA
1
2
3
4
5
6
7
8
Fig. 4.
The dependence of PL spectra of a Si–Ge structure
doped with Sb as the donor impurity (3.7 ML of Ge, the
excitation density 50 W cm
–2
) on the observation tempera-
ture: (
1
) 18, (
2
) 50, (
3
) 130, (
4
) 170, (
5
) 210, (
6
) 230,
(
7
) 250, and (
8
) 300 K.
8
6
4
2
01.00 1.05 1.10
1.04 1.081.00
PL intensity, arb. units
PL intensity, arb. units
Photon energy, eV
Photon energy, eV
1
2
1
2
Ge
TO
Ge
NP
Si
TO
Ge
TO
Si
TO
Ge
NP
Fig. 5.
The dependence of PL spectra of Si–Ge structures
doped with Sb as the donor impurity on the observation
temperature: (
1
) 30 and (
2
) 60 K. Insert: similar spectra at
15 K (5.1 ML of Ge). The excitation density: (
1
) 432 and
(
2
) 186 W cm
–2
.
214
SEMICONDUCTORS
Vol. 37
No. 2
2003
MAKAROV
et al
.
2. L. C. Lenchyshyn, M. L. W. Thewalt, D. C. Houghton,
et
al.
, Phys. Rev. B
47
, 16655 (1993).
3. T. Baier, U. Mantz, K. Thonke,
et al.
, in
Proceedings of
22nd International Conference on the Physics of Semi-
conductors, Vancouver, 1994
, Ed. by D. J. Lockwood
(World Sci., Singapore, 1995), Vol. 2, p. 1568.
4. O. P. Pchelyakov, Yu. B. Bolkhovityanov, A. V. Dvure-
chenski
œ
,
et al.
, Fiz. Tekh. Poluprovodn. (St. Petersburg)
34, 1281 (2000) [Semiconductors 34, 1229 (2000)].
5. M. W. Dashiel, U. Denker, C. Müller, et al., Appl. Phys.
Lett. 80, 1279 (2002).
6. N. N. Ledentsov, M. Grundmann, F. Heinrichsdorff, et
al., IEEE J. Sel. Top. Quantum Electron. 6, 439 (2000).
7. N. N. Ledentsov, I. L. Krestnikov, M. V. Maximov, et al.,
Appl. Phys. Lett. 69, 1343 (1996).
8. N. N. Ledentsov, I. L. Krestnikov, M. V. Maximov, et al.,
Appl. Phys. Lett. 70, 2766 (1997).
9. I. L. Krestnikov, N. N. Ledentsov, A. Hoffmann, and
D. Bimberg, Phys. Status Solidi A 183, 207 (2001).
10. N. N. Ledentsov, J. Böhrer, M. Beer, et al., Phys. Rev. B
52, 14058 (1995).
11. N. D. Zakharov, P. Werner, U. Gösele, et al., Mater. Sci.
Eng. B 87, 92 (2001).
12. N. D. Zakharov, G. E. Cirlin, P. Werner, et al., in Pro-
ceedings of 9th International Symposium on Nanostruc-
tures: Physics and Technology 2001 (St. Petersburg,
2001), p. 21.
13. L. P. Rokinson, D. C. Tsui, J. L. Benton, and Y.-H. Xie,
Appl. Phys. Lett. 75, 2413 (1999).
Translated by D. Mashovets